Browse the corpus

An implantable system to restore hemodynamic stability after spinal cord injury. A spinal cord injury (SCI) causes immediate and sustained hemodynamic instability that threatens neurological recovery and impacts quality of life. Here we establish the clinical burden of chronic hypotensive complications due to SCI in 1,479 participants and expose the ineffective treatment of these complications with conservative measures. To address this clinical burden, we developed a purpose-built implantable system based on biomimetic epidural electrical stimulation (EES) of the spinal cord that immediately triggered robust pressor responses. The system durably reduced the severity of hypotensive complications in people with SCI, removed the necessity for conservative treatments, improved quality of life and enabled superior engagement in activities of daily living. Central to the development of this therapy was the head-to-head demonstration in the same participants that EES must target the last three thoracic segments, and not the lumbosacral segments, to achieve the safe and effective regulation of blood pressure in people with SCI. These findings in 14 participants establish the path to designing a pivotal device trial that will evaluate the safety and efficacy of EES to treat the underappreciated, treatment-resistant hypotensive complications due to SCI.

A spinal cord injury (SCI) disrupts the communication between supraspinal vasomotor centers and the regions of the spinal cord that regulate blood pressure1–3. The consequence is pernicious exposure to disabling and potentially life-threatening blood pressure instability that threatens neurological recovery4–9. A prominent feature of this hemodynamic instability is a spectrum of hypotensive complications that include orthostatic hypotension, resting hypotension, postprandial hypotension and exertional hypotension2. These complications are clinically managed with conservative measures, such as uncomfortable abdominal binders, pressure stockings, high sodium diets and slow-acting pharmacological pressor agents2. Despite these measures, the majority of people who present with hemodynamic instability due to SCI remain chronically hypotensive, which impacts their quality of life and reduces engagement in social and professional activities2,10–13. This symptomatology10,11,14, coupled with the unexplained prevalence of stroke and heart disease in this population, compelled health institutions to encourage the development of new approaches to manage hemodynamic instability in people with SCI.

ife and reduces engagement in social and professional activities2,10–13. This symptomatology10,11,14, coupled with the unexplained prevalence of stroke and heart disease in this population, compelled health institutions to encourage the development of new approaches to manage hemodynamic instability in people with SCI. To address the unsatisfying management of hemodynamic instability in people with SCI, we previously developed a preclinical model of hemodynamic instability due to SCI with the aim of studying the mechanisms through which epidural electrical stimulation (EES) applied over the spinal cord can lead to increases in blood pressure1,15. We found that EES depolarizes large-diameter afferents where they bend to enter the spinal cord through the dorsal root entry zones1,16,17. The recruitment of these afferents leads to the activation of splanchnic neurons via the activation of Vsx2-expressing neurons that directly connect to sympathetic preganglionic neurons1,16,18. Notably, the amplitude of pressor responses followed a Gaussian distribution wherein the largest responses were elicited when EES was applied over the lower thoracic spinal cord. This distribution correlated with the density of sympathetic preganglionic neurons, in line with the known anatomy of the autonomic nervous system19. We called this region the hemodynamic hotspot. We translated this understanding into a neuromodulation strategy operating in a closed loop that adjusts EES amplitudes delivered over the hemodynamic hotspot to regulate blood pressure in real time. This strategy maintained blood pressure stability during transient, varying and sustained hemodynamic challenges in rat and nonhuman primate models of SCI, as well as in two human participants1,20.

a closed loop that adjusts EES amplitudes delivered over the hemodynamic hotspot to regulate blood pressure in real time. This strategy maintained blood pressure stability during transient, varying and sustained hemodynamic challenges in rat and nonhuman primate models of SCI, as well as in two human participants1,20. Here we aimed to establish the burden of hypotension and assess the effectiveness of current treatments in large populations of people living with SCI to justify new treatments involving neurosurgical interventions. We then sought to address the ambiguity on the optimal location of the spinal cord to apply EES to treat hemodynamic instability and exploited this understanding to develop a purpose-built implanted system that harnesses the mechanisms through which EES regulates blood pressure. We finally provide preliminary validation on the long-term efficacy and safety of this new system and resulting improvements in quality of life, paving the way toward pivotal clinical trials.

EES is a promising therapy to address the lack of satisfying management options for hemodynamic instability due to SCI. However, EES requires a neurosurgical intervention that must be weighed against the risks and benefits of the procedure. Establishing this balance requires an understanding of the prevalence, symptomatology and effectiveness of current management strategies. While hypotension is a recognized medically refractory complication of SCI2, these epidemiological factors have not been quantified conclusively.

risks and benefits of the procedure. Establishing this balance requires an understanding of the prevalence, symptomatology and effectiveness of current management strategies. While hypotension is a recognized medically refractory complication of SCI2, these epidemiological factors have not been quantified conclusively. To address this knowledge gap, we analyzed the Spinal Cord Injury Community Survey (SCICS)21,22, which includes self-reported information on symptoms of orthostatic hypotension and demographic information in 1,479 individuals living with chronic SCI (Supplementary Fig. 1a–e and Supplementary Note 1). This analysis revealed that 78% of individuals with tetraplegia had been told by a medical practitioner that they had orthostatic hypotension. Of this subset of individuals who were diagnosed, 28% of them were being treated for orthostatic hypotension, yet 91% still experienced symptoms (Fig. 1a).Fig. 1Orthostatic hypotension is a medically refractory condition in people with SCI.a, The prevalence of orthostatic hypotension and management efficacy in 1,479 individuals with tetraplegia (n = 510) from the Rick Hansen Spinal Cord Injury Registry21,22. b, Percentage of individuals with tetraplegia experiencing each symptom scored in the ADFSCI survey across various daily activities (n = 107). c, Scheme of tilt-table test and representative blood pressure response. d, Drop in systolic and diastolic blood pressure during a tilt-table test (n = 17). e, ADFSCI total orthostatic hypotension score in individuals with and without clinically defined orthostatic hypotension (n = 4 with no orthostatic hypotension and n = 13 with confirmed orthostatic hypotension; independent samples two-tailed Welch’s t-test: t = 12.84, ***P = 2.37 × 10−2). f, Kaplan–Meier plot of exposure status to time, segregated by the presence of clinically defined orthostatic hypotension. Bar graph shows the corresponding time to tilt end. See Source Data Fig. 1 for source data and statistics.Source data

independent samples two-tailed Welch’s t-test: t = 12.84, ***P = 2.37 × 10−2). f, Kaplan–Meier plot of exposure status to time, segregated by the presence of clinically defined orthostatic hypotension. Bar graph shows the corresponding time to tilt end. See Source Data Fig. 1 for source data and statistics.Source data a, The prevalence of orthostatic hypotension and management efficacy in 1,479 individuals with tetraplegia (n = 510) from the Rick Hansen Spinal Cord Injury Registry21,22. b, Percentage of individuals with tetraplegia experiencing each symptom scored in the ADFSCI survey across various daily activities (n = 107). c, Scheme of tilt-table test and representative blood pressure response. d, Drop in systolic and diastolic blood pressure during a tilt-table test (n = 17). e, ADFSCI total orthostatic hypotension score in individuals with and without clinically defined orthostatic hypotension (n = 4 with no orthostatic hypotension and n = 13 with confirmed orthostatic hypotension; independent samples two-tailed Welch’s t-test: t = 12.84, ***P = 2.37 × 10−2). f, Kaplan–Meier plot of exposure status to time, segregated by the presence of clinically defined orthostatic hypotension. Bar graph shows the corresponding time to tilt end. See Source Data Fig. 1 for source data and statistics. Source data

a, The prevalence of orthostatic hypotension and management efficacy in 1,479 individuals with tetraplegia (n = 510) from the Rick Hansen Spinal Cord Injury Registry21,22. b, Percentage of individuals with tetraplegia experiencing each symptom scored in the ADFSCI survey across various daily activities (n = 107). c, Scheme of tilt-table test and representative blood pressure response. d, Drop in systolic and diastolic blood pressure during a tilt-table test (n = 17). e, ADFSCI total orthostatic hypotension score in individuals with and without clinically defined orthostatic hypotension (n = 4 with no orthostatic hypotension and n = 13 with confirmed orthostatic hypotension; independent samples two-tailed Welch’s t-test: t = 12.84, ***P = 2.37 × 10−2). f, Kaplan–Meier plot of exposure status to time, segregated by the presence of clinically defined orthostatic hypotension. Bar graph shows the corresponding time to tilt end. See Source Data Fig. 1 for source data and statistics. Source data This analysis compelled us to characterize the pattern of hypotension-related symptoms due to SCI. For this purpose, we surveyed a diverse sample of 254 individuals with SCI, which included responses to the autonomic dysfunction following spinal cord injury (ADFSCI) rating scale. When considering individuals with tetraplegia, we found that every individual who responded to the survey experienced symptoms of hypotension throughout the day. These symptoms were primarily characterized by lightheadedness, dizziness, fatigue, blurred vision and weakness throughout the day (Fig. 1b). In a subset of these individuals, we found that participants who met the criteria for orthostatic hypotension during the tilt-table test reported significantly more hypotensive symptoms (Fig. 1c–f, Supplementary Fig. 2a–i and Supplementary Table 1).

ss, fatigue, blurred vision and weakness throughout the day (Fig. 1b). In a subset of these individuals, we found that participants who met the criteria for orthostatic hypotension during the tilt-table test reported significantly more hypotensive symptoms (Fig. 1c–f, Supplementary Fig. 2a–i and Supplementary Table 1). These results established that (1) the vast majority of individuals living with tetraplegia experience severe and persistent symptoms of hypotension, (2) hypotension is accompanied by quantifiable symptomatology that increases with the level and completeness of injury and (3) people with SCI continue experiencing symptoms of orthostatic hypotension despite medical management. We concluded that these outcomes, combined with the long-term consequences of medically refractory hypotension, justified the consideration of therapeutic strategies involving surgical interventions to manage hypotension due to SCI.

These results established that (1) the vast majority of individuals living with tetraplegia experience severe and persistent symptoms of hypotension, (2) hypotension is accompanied by quantifiable symptomatology that increases with the level and completeness of injury and (3) people with SCI continue experiencing symptoms of orthostatic hypotension despite medical management. We concluded that these outcomes, combined with the long-term consequences of medically refractory hypotension, justified the consideration of therapeutic strategies involving surgical interventions to manage hypotension due to SCI. Extensive preclinical evidence in mice, rats and nonhuman primates, as well as clinical case studies, demonstrated that EES reduces the severity of orthostatic hypotension1,20,23–26. However, the optimal location to deliver EES has remained controversial. We previously showed that EES activates large-diameter afferents where they bend to enter the spinal cord through the dorsal root entry zones and that this recruitment engages sympathetic preganglionic neurons indirectly via Vsx2-expressing excitatory neurons nested in the intermediate lamina of the stimulated region of the spinal cord1,16–18 (Fig. 2a). We also confirmed an enrichment of these sympathetic preganglionic neurons in the lower thoracic spinal cord (Fig. 2a) and showed that this enrichment coincided with the maximal pressor response to EES. This location was thus termed the hemodynamic hotspot1. Despite this evidence, many studies reported unexpected pressor responses during the application of EES over the lumbosacral spinal cord23–26. The incongruence of these observations with the anatomy of the autonomic nervous system19,27 compelled us to perform a head-to-head comparison in the same participants to establish the optimal location to deliver EES for the regulation of blood pressure in humans with SCI.Fig. 2Validation of the location of the hemodynamic hotspot.a, Overview of the neuronal architecture recruited EES to trigger pressor responses. b, Anatomical planning of laminotomies for inserting the paddle leads (left) based on the targeted location of the paddle leads (right) over the thoracic and lumbosacral segments. c, Amplitude of systolic pressor response per spinal segment when delivering EES intraoperatively. Dots (n = 5 participants) denote pressure responses to EES applied over specific segments, the location of which was reconstructed postoperatively. d, Bar graph shows averaged changes in systolic blood pressure in response to EES applied over thoracic or lumbar spinal segments (n = 5, paired samples two-tailed t-test; t = 6.46, **P = 3.00 × 10−3).

ote pressure responses to EES applied over specific segments, the location of which was reconstructed postoperatively. d, Bar graph shows averaged changes in systolic blood pressure in response to EES applied over thoracic or lumbar spinal segments (n = 5, paired samples two-tailed t-test; t = 6.46, **P = 3.00 × 10−3). e, Systolic blood pressure response measured intraoperatively, while EES was applied over the T11 versus L4 spinal segments for a representative participant. f, Sagittal and coronal reconstructions from a postoperative CT scan. g, Changes in blood pressure during a tilt-table test without EES and with EES applied over either the lower thoracic or lumbosacral spinal segments. h, Drop in systolic blood pressure during a 10-min tilt-table test, and corresponding tilt duration (n = 5, repeated-measures ANOVA with Tukey’s HSD, statistics are provided in Source Data Fig. 2). i, Kaplan–Meier plot of exposure status to time until end of tilt, segregated by the location over which EES was applied (n = 37 tilts; mixed model Cox regression with likelihood ratio test estimate = 36.2, P = 2.48 × 10−6). See Source Data Fig. 2 and Extended Data Figs. 1–3 for source data and statistics. NS, not significant; HSD, honestly significant difference.Source data

of tilt, segregated by the location over which EES was applied (n = 37 tilts; mixed model Cox regression with likelihood ratio test estimate = 36.2, P = 2.48 × 10−6). See Source Data Fig. 2 and Extended Data Figs. 1–3 for source data and statistics. NS, not significant; HSD, honestly significant difference.Source data a, Overview of the neuronal architecture recruited EES to trigger pressor responses. b, Anatomical planning of laminotomies for inserting the paddle leads (left) based on the targeted location of the paddle leads (right) over the thoracic and lumbosacral segments. c, Amplitude of systolic pressor response per spinal segment when delivering EES intraoperatively. Dots (n = 5 participants) denote pressure responses to EES applied over specific segments, the location of which was reconstructed postoperatively. d, Bar graph shows averaged changes in systolic blood pressure in response to EES applied over thoracic or lumbar spinal segments (n = 5, paired samples two-tailed t-test; t = 6.46, **P = 3.00 × 10−3). e, Systolic blood pressure response measured intraoperatively, while EES was applied over the T11 versus L4 spinal segments for a representative participant. f, Sagittal and coronal reconstructions from a postoperative CT scan. g, Changes in blood pressure during a tilt-table test without EES and with EES applied over either the lower thoracic or lumbosacral spinal segments. h, Drop in systolic blood pressure during a 10-min tilt-table test, and corresponding tilt duration (n = 5, repeated-measures ANOVA with Tukey’s HSD, statistics are provided in Source Data Fig. 2). i, Kaplan–Meier plot of exposure status to time until end of tilt, segregated by the location over which EES was applied (n = 37 tilts; mixed model Cox regression with likelihood ratio test estimate = 36.2, P = 2.48 × 10−6). See Source Data Fig. 2 and Extended Data Figs. 1–3 for source data and statistics. NS, not significant; HSD, honestly significant difference.

e until end of tilt, segregated by the location over which EES was applied (n = 37 tilts; mixed model Cox regression with likelihood ratio test estimate = 36.2, P = 2.48 × 10−6). See Source Data Fig. 2 and Extended Data Figs. 1–3 for source data and statistics. NS, not significant; HSD, honestly significant difference. Source data Consequently, we initiated two clinical studies in Canada and Switzerland to compare the effect of EES applied over the lower thoracic versus lumbosacral spinal cord on hemodynamic instability (ClinicalTrials.gov registrations: NCT04994886 and NCT05044923). We enrolled six individuals with medically refractory orthostatic hypotension secondary to their SCI (Extended Data Fig. 1a and Supplementary Table 2). The six participants included in these clinical studies met the established clinical criteria for orthostatic hypotension because none of these participants were able to sustain the orthostatic challenge imposed by verticalization on a tilt-table test, showing an average drop of 50 mm Hg in systolic blood pressure within 3 min (Extended Data Fig. 1b–f). This drop far exceeds the international criteria for orthostatic hypotension, set at a 20 mm Hg drop of systolic blood pressure within 3 min28.

he orthostatic challenge imposed by verticalization on a tilt-table test, showing an average drop of 50 mm Hg in systolic blood pressure within 3 min (Extended Data Fig. 1b–f). This drop far exceeds the international criteria for orthostatic hypotension, set at a 20 mm Hg drop of systolic blood pressure within 3 min28. Because only repurposed technologies originally designed to alleviate neuropathic pain were available to target both locations (Fig. 2b and Extended Data Fig. 1g), we sought to optimize the surgical implantation of two separate paddle leads over the lower thoracic spinal cord and the lumbosacral spinal cord. To facilitate this placement, we generated computational models that we personalized with computed tomography (CT) and magnetic resonance imaging (MRI) sequences that we previously optimized to enable the identification of the dorsal root entry zones targeted by EES29 (Fig. 2b and Extended Data Fig. 2). We then injected these coordinates into our preoperative planning software that we consolidated with intraoperative imaging of the spinal column. Under general anesthesia, we performed laminotomies to enable the insertion of both leads over the planned anatomical locations. We then fine-tuned the final position of both leads based on the muscle responses to EES that are expected from the known rostrocaudal distribution of motor neuron pools innervating trunk and leg muscles29 (Extended Data Fig. 1h–j).

minotomies to enable the insertion of both leads over the planned anatomical locations. We then fine-tuned the final position of both leads based on the muscle responses to EES that are expected from the known rostrocaudal distribution of motor neuron pools innervating trunk and leg muscles29 (Extended Data Fig. 1h–j). Once the location of both leads was secured, we applied EES (120 Hz, 1–10 mA and 300 µs) over each pair of electrodes on each row of both paddles to assess pressor responses that we monitored by measuring beat-by-beat blood pressure through an arterial line. We mapped pressor responses to postoperative reconstructions of the spinal cord regions activated by each pair of electrodes. As previously established in rat and nonhuman primate models of SCI1, we found that pressor responses followed a Gaussian distribution, wherein the peak pressor responses were centered over the penultimate segment of the thoracic spinal cord (Fig. 2c–e and Extended Data Fig. 1k–l).

activated by each pair of electrodes. As previously established in rat and nonhuman primate models of SCI1, we found that pressor responses followed a Gaussian distribution, wherein the peak pressor responses were centered over the penultimate segment of the thoracic spinal cord (Fig. 2c–e and Extended Data Fig. 1k–l). Postoperative CT acquisitions confirmed that both leads were positioned over the planned locations (Fig. 2f). Intraoperative assessments confirmed that the optimal location to elicit pressor responses with EES was centered over the lower thoracic spinal cord—the hemodynamic hotspot. Therefore, we next verified whether EES applied over the hemodynamic hotspot was superior to EES applied over the lumbosacral spinal cord to stabilize blood pressure during formal tilt-table testing postoperatively, because previous case studies23–26 reported a mitigation of orthostatic hypotension when EES was applied over the lumbosacral segments (Fig. 2g–i and Extended Data Fig. 3a–g). We conducted comprehensive mapping sessions to identify optimal configurations of electrodes and stimulation parameters that could trigger pressor responses when EES was delivered over the lumbosacral segments (Extended Data Fig. 3a–d). Despite this mapping, we failed to identify parameters of EES that could trigger robust and reproducible pressor responses without eliciting spasms in leg muscles. EES applied over the lumbosacral segments marginally reduced the severity of orthostatic hypotension (Fig. 2g–i and Extended Data Fig. 3e–g).

ta Fig. 3a–d). Despite this mapping, we failed to identify parameters of EES that could trigger robust and reproducible pressor responses without eliciting spasms in leg muscles. EES applied over the lumbosacral segments marginally reduced the severity of orthostatic hypotension (Fig. 2g–i and Extended Data Fig. 3e–g). This failure contrasted with the robust pressor responses elicited when EES was applied over the lower thoracic spinal cord, which enabled all the participants to improve their tolerance to orthostatic challenges induced by the verticalization during tilt-table tests in the absence of undesirable sensations (Fig. 2g–i and Extended Data Fig. 3a–g). These results demonstrate that EES applied over the lumbosacral spinal cord failed to consistently trigger meaningful pressor responses, as expected, based on the incongruence between this location and the anatomy and physiology of the autonomic nervous system. Instead, EES applied over the hemodynamic hotspot, wherein an enrichment of sympathetic preganglionic neurons involved in the modulation of blood pressure resides, consistently triggered robust pressor responses that enabled all six participants to tolerate severe orthostatic challenges.

f the autonomic nervous system. Instead, EES applied over the hemodynamic hotspot, wherein an enrichment of sympathetic preganglionic neurons involved in the modulation of blood pressure resides, consistently triggered robust pressor responses that enabled all six participants to tolerate severe orthostatic challenges. These quantifications provide important results for the ongoing debate on the optimal location of EES for the effective treatment of hemodynamic instability in people with SCI. In addition, recent preclinical experiments demonstrated that long-term application of EES over the lumbosacral spinal cord augmented the density of maladaptive synaptic connections from Vsx2-expressing neurons of the lumbosacral to Vsx2-expressing neurons of the lower thoracic spinal cord, which reinforces the severity of autonomic dysreflexia (AD)16. Given the discrepancy in the safety and efficacy of EES applied over the hemodynamic hotspot versus lumbosacral segments, these results establish a safe path forward on how to deliver EES to treat hemodynamic instability due to SCI. We aimed to develop a clinical-grade implantable neurostimulation platform that addresses the limitations of existing systems and meets the identified requirements for the safe and effective management of hemodynamic instability in people with SCI (Supplementary Note 2).

These quantifications provide important results for the ongoing debate on the optimal location of EES for the effective treatment of hemodynamic instability in people with SCI. In addition, recent preclinical experiments demonstrated that long-term application of EES over the lumbosacral spinal cord augmented the density of maladaptive synaptic connections from Vsx2-expressing neurons of the lumbosacral to Vsx2-expressing neurons of the lower thoracic spinal cord, which reinforces the severity of autonomic dysreflexia (AD)16. Given the discrepancy in the safety and efficacy of EES applied over the hemodynamic hotspot versus lumbosacral segments, these results establish a safe path forward on how to deliver EES to treat hemodynamic instability due to SCI. We aimed to develop a clinical-grade implantable neurostimulation platform that addresses the limitations of existing systems and meets the identified requirements for the safe and effective management of hemodynamic instability in people with SCI (Supplementary Note 2). Because the possibility to adjust EES parameters in a closed loop was a desired feature of the platform, we designed an implantable neurostimulation platform that continuously communicates with a wearable hub through near-field magnetic induction (NFMI; Fig. 3a). This hub can acquire external signals, including continuous measurements of blood pressure, and embeds computing power to support closed-loop control algorithms. This configuration enables closed-loop control of EES parameters with latencies as low as 25 ms. We programmed the firmware to support any configuration of anodes and cathodes and to resolve the limited range of programmable EES parameters in current neurostimulation platforms.Fig. 3Validation of the new, purpose-built implantable neurostimulation platform.a, Scheme of the investigational system, including a clinician controller (tablet), patient programmer (external smartwatch), a main controller (hub), IPG and a Medtronic 5-6-5 SureScan Specify paddle lead. b, Changes in blood pressure from a representative participant during an orthostatic challenge without and with EES applied over the hemodynamic hotspot (EESBP). c, Average drop in systolic blood pressure during orthostatic challenge without and with EES (n = 4, paired samples two-tailed t-test; t = 4.27, *P = 2.4 × 10−2). d, Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 24 tilts; mixed model Cox regression with likelihood ratio test estimate = 15.36, P = 4.00 × 10−4) and tilt duration (n = 4, paired samples two-tailed t-test; t = 4.69, *P = 1.8 × 10−2) without EES and with EES applied over the hemodynamic hotspot. e, Closed-loop control of blood pressure consisting of a PID controller that adjusts the amplitude of EES applied over the hemodynamic hotspot to maintain blood pressure within a range of predefined blood pressure targets. f, Stepwise increase in systolic blood pressure in response to gradual increases of EES amplitude (1 mA per min) for a representative participant.

ng of a PID controller that adjusts the amplitude of EES applied over the hemodynamic hotspot to maintain blood pressure within a range of predefined blood pressure targets. f, Stepwise increase in systolic blood pressure in response to gradual increases of EES amplitude (1 mA per min) for a representative participant. Relationship between EES amplitude and systolic blood pressure (n = 4; black line represents model fit of a mixed effects linear regression R2 = 0.97, P = 8.83 × 10−9). g, Changes in systolic blood pressure during a dynamic orthostatic challenge while EES is delivered continuously (top) or controlled in a closed loop (bottom) to maintain the systolic blood pressure within a target range. h, Error between the predefined blood pressure target (top) and variability of error to target (bottom) when EES was delivered continuously or controlled in closed loop (n = 3/4—trial 3, one participant did not complete assessment due to spasticity). See Source Data Fig. 3 and Extended Data Figs. 4 and 5 for source data and statistics. PID, proportional–integral–derivative.Source data

iability of error to target (bottom) when EES was delivered continuously or controlled in closed loop (n = 3/4—trial 3, one participant did not complete assessment due to spasticity). See Source Data Fig. 3 and Extended Data Figs. 4 and 5 for source data and statistics. PID, proportional–integral–derivative.Source data a, Scheme of the investigational system, including a clinician controller (tablet), patient programmer (external smartwatch), a main controller (hub), IPG and a Medtronic 5-6-5 SureScan Specify paddle lead. b, Changes in blood pressure from a representative participant during an orthostatic challenge without and with EES applied over the hemodynamic hotspot (EESBP). c, Average drop in systolic blood pressure during orthostatic challenge without and with EES (n = 4, paired samples two-tailed t-test; t = 4.27, *P = 2.4 × 10−2). d, Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 24 tilts; mixed model Cox regression with likelihood ratio test estimate = 15.36, P = 4.00 × 10−4) and tilt duration (n = 4, paired samples two-tailed t-test; t = 4.69, *P = 1.8 × 10−2) without EES and with EES applied over the hemodynamic hotspot. e, Closed-loop control of blood pressure consisting of a PID controller that adjusts the amplitude of EES applied over the hemodynamic hotspot to maintain blood pressure within a range of predefined blood pressure targets. f, Stepwise increase in systolic blood pressure in response to gradual increases of EES amplitude (1 mA per min) for a representative participant. Relationship between EES amplitude and systolic blood pressure (n = 4; black line represents model fit of a mixed effects linear regression R2 = 0.97, P = 8.83 × 10−9). g, Changes in systolic blood pressure during a dynamic orthostatic challenge while EES is delivered continuously (top) or controlled in a closed loop (bottom) to maintain the systolic blood pressure within a target range. h, Error between the predefined blood pressure target (top) and variability of error to target (bottom) when EES was delivered continuously or controlled in closed loop (n = 3/4—trial 3, one participant did not complete assessment due to spasticity). See Source Data Fig. 3 and Extended Data Figs. 4 and 5 for source data and statistics. PID, proportional–integral–derivative.

op) and variability of error to target (bottom) when EES was delivered continuously or controlled in closed loop (n = 3/4—trial 3, one participant did not complete assessment due to spasticity). See Source Data Fig. 3 and Extended Data Figs. 4 and 5 for source data and statistics. PID, proportional–integral–derivative. Source data The hub communicates wirelessly with a touch-screen tablet and a smartwatch, from which therapeutic programs can be selected and parameterized (Fig. 3a). We optimized these user interfaces to comply with usability assessments and user needs. We also endowed the user interface with practical features that reduce the impact of impaired arm and hand functions for operating the therapy independently and safely. In addition, customizable algorithms enable real-time adjustment of set-point amplitudes based on hemodynamic requirements that are well known to differ throughout the day, such as after meals and during sleep and exercise2,30–32.

ct of impaired arm and hand functions for operating the therapy independently and safely. In addition, customizable algorithms enable real-time adjustment of set-point amplitudes based on hemodynamic requirements that are well known to differ throughout the day, such as after meals and during sleep and exercise2,30–32. After meeting the requirements for all applicable standards for active medical device development, we tested the preliminary safety and efficacy of this new platform under the protocol of a new clinical study conducted in Switzerland (ClinicalTrials.gov registration: NCT05111093; Extended Data Fig. 4a). We enrolled four new participants who presented with debilitating medically refractory hypotension (Extended Data Fig. 4b–f and Supplementary Table 2) and implanted the purpose-built neurostimulation platform but maintained every other aspect of our preoperative and operative procedures, including the implantation of the same repurposed paddle leads over the hemodynamic hotspot as in our previous experiments (Extended Data Fig. 4g,h).

4b–f and Supplementary Table 2) and implanted the purpose-built neurostimulation platform but maintained every other aspect of our preoperative and operative procedures, including the implantation of the same repurposed paddle leads over the hemodynamic hotspot as in our previous experiments (Extended Data Fig. 4g,h). The neurostimulation platform met the requirements to deliver EES waveforms that elicit robust pressor responses (Extended Data Fig. 4i–l), enabling all the participants to improve their tolerance to orthostatic challenges (Fig. 3b–d and Extended Data Fig. 4m–s). As observed in rats and nonhuman primates1, we confirmed in all four participants that the magnitude of pressor responses scaled linearly with increasing amplitudes of EES (Fig. 3f and Extended Data Fig. 5a). This linear relationship suggested that blood pressure could be regulated in real time based on a proportional–integral–differential controller regulating EES amplitudes to maintain blood pressure within user-defined levels.

esponses scaled linearly with increasing amplitudes of EES (Fig. 3f and Extended Data Fig. 5a). This linear relationship suggested that blood pressure could be regulated in real time based on a proportional–integral–differential controller regulating EES amplitudes to maintain blood pressure within user-defined levels. To test this possibility, we leveraged the closed-loop capabilities and versatile software environment of the new neurostimulation platform to implement this proportional–integral–derivative controller (Fig. 3e). We then evaluated the performance of this controller during formal and dynamic tilt tests. As expected, based on previous experiments (Fig. 2g–i), we found that the delivery of continuous EES decreased the severity of orthostatic hypotension during formal tilt tests wherein the amplitude of EES remained constant (Fig. 3b–d). However, dynamic tilt-testing sessions exposed the limitations of continuous EES because pseudo-random changes in body orientation with respect to the direction of gravity led to dynamic perturbations of blood pressure that could not be managed satisfactorily with continuous EES. In contrast, closed-loop control of EES led to rapid and precise updates to EES amplitudes that translated into finely tuned levels of blood pressure during dynamic tilt tests (Fig. 3g and Extended Data Fig. 5b). Closed-loop control of EES maintained blood pressure levels within the desired target range for extended periods of time despite transient and dynamic changes in postural orientation (Fig. 3g, Extended Data Figs. 5b and 9n). Quantifications also revealed the superiority of EES controlled in closed loop over both continuous EES and alternative closed-loop approaches relying on inertial measurement units or manual adjustments of EES (Fig. 3h, Extended Data Figs. 5c and 9o).

es in postural orientation (Fig. 3g, Extended Data Figs. 5b and 9n). Quantifications also revealed the superiority of EES controlled in closed loop over both continuous EES and alternative closed-loop approaches relying on inertial measurement units or manual adjustments of EES (Fig. 3h, Extended Data Figs. 5c and 9o). These results confirmed that the newly developed neurostimulation platform fulfilled the technical requirements to modulate blood pressure with EES and established the possibility of achieving precise control of blood pressure in a closed loop despite transient, varying and sustained hemodynamic challenges in people with orthostatic hypotension due to SCI. We next aimed to develop a clinical-grade paddle lead that combines the necessary features to target the large-diameter afferents projecting to the hemodynamic hotspot while avoiding the undesired side effects of nonselective stimulation29,33–35.

These results confirmed that the newly developed neurostimulation platform fulfilled the technical requirements to modulate blood pressure with EES and established the possibility of achieving precise control of blood pressure in a closed loop despite transient, varying and sustained hemodynamic challenges in people with orthostatic hypotension due to SCI. We next aimed to develop a clinical-grade paddle lead that combines the necessary features to target the large-diameter afferents projecting to the hemodynamic hotspot while avoiding the undesired side effects of nonselective stimulation29,33–35. Our mapping experiments showed that the hemodynamic hotspot is centered over the last three segments of the lower thoracic spinal cord. Moreover, the same experiments revealed that EES applied over multiple segments leads to cumulative increases in the amplitude of pressor responses compared to when the same amplitude of EES is applied over a single spinal segment (Fig. 4a).Fig. 4Purpose-built paddle lead to target the hemodynamic hotspot.a, EES applied over multiple segments of the hemodynamic hotspot leads to a superior increase in blood pressure compared to EES applied over the median of a single segment (paired samples, one-tailed t-test, n = 5, t = 2.69, *P = 2.72 × 10−2). b, Quantification of the length of the lower thoracic spinal cord in the study participants (n = 10; left), and average length of the lower thoracic spinal cord quantified from published data (right). The coverage of the hemodynamic hotspot by the Medtronic 5-6-5 lead and expected from the newly designed paddle lead are reported. c, Illustrative drawing of the anatomical model implemented in silico to predict the relationships between the lateral position of EES electrodes and the relative recruitment of the targeted dorsal root, contralateral dorsal root and dorsal columns, as shown in the plot. d, Photograph of the newly designed purpose-built paddle lead that combines two parallel columns of eight equally spaced electrodes that aim to cover the entire length of the hemodynamic hotspot while avoiding the recruitment of the dorsal columns. See Source Data Fig. 4 and Extended Data Fig. 6 for source data and statistics.Source data

e newly designed purpose-built paddle lead that combines two parallel columns of eight equally spaced electrodes that aim to cover the entire length of the hemodynamic hotspot while avoiding the recruitment of the dorsal columns. See Source Data Fig. 4 and Extended Data Fig. 6 for source data and statistics.Source data a, EES applied over multiple segments of the hemodynamic hotspot leads to a superior increase in blood pressure compared to EES applied over the median of a single segment (paired samples, one-tailed t-test, n = 5, t = 2.69, *P = 2.72 × 10−2). b, Quantification of the length of the lower thoracic spinal cord in the study participants (n = 10; left), and average length of the lower thoracic spinal cord quantified from published data (right). The coverage of the hemodynamic hotspot by the Medtronic 5-6-5 lead and expected from the newly designed paddle lead are reported. c, Illustrative drawing of the anatomical model implemented in silico to predict the relationships between the lateral position of EES electrodes and the relative recruitment of the targeted dorsal root, contralateral dorsal root and dorsal columns, as shown in the plot. d, Photograph of the newly designed purpose-built paddle lead that combines two parallel columns of eight equally spaced electrodes that aim to cover the entire length of the hemodynamic hotspot while avoiding the recruitment of the dorsal columns. See Source Data Fig. 4 and Extended Data Fig. 6 for source data and statistics. Source data

a, EES applied over multiple segments of the hemodynamic hotspot leads to a superior increase in blood pressure compared to EES applied over the median of a single segment (paired samples, one-tailed t-test, n = 5, t = 2.69, *P = 2.72 × 10−2). b, Quantification of the length of the lower thoracic spinal cord in the study participants (n = 10; left), and average length of the lower thoracic spinal cord quantified from published data (right). The coverage of the hemodynamic hotspot by the Medtronic 5-6-5 lead and expected from the newly designed paddle lead are reported. c, Illustrative drawing of the anatomical model implemented in silico to predict the relationships between the lateral position of EES electrodes and the relative recruitment of the targeted dorsal root, contralateral dorsal root and dorsal columns, as shown in the plot. d, Photograph of the newly designed purpose-built paddle lead that combines two parallel columns of eight equally spaced electrodes that aim to cover the entire length of the hemodynamic hotspot while avoiding the recruitment of the dorsal columns. See Source Data Fig. 4 and Extended Data Fig. 6 for source data and statistics. Source data We thus reasoned that the optimal paddle lead to regulate blood pressure with EES must maximize the coverage of the last three thoracic segments of the spinal cord. However, our analysis of the ten individuals who were implanted with the repurposed paddle revealed that the coverage of these segments only reached 56% (Fig. 4b and Extended Data Fig. 6a). Moreover, 6 of the 16 electrodes were located over the midline of the spinal cord. These electrodes elicited undesired muscle spasms in the legs through antidromic volleys elicited along sensory axons ascending the dorsal columns (Extended Data Fig. 6b). This analysis emphasized the necessity to design a new paddle lead with a configuration of electrodes that meets the requirements to modulate blood pressure with EES.

licited undesired muscle spasms in the legs through antidromic volleys elicited along sensory axons ascending the dorsal columns (Extended Data Fig. 6b). This analysis emphasized the necessity to design a new paddle lead with a configuration of electrodes that meets the requirements to modulate blood pressure with EES. To guide the informed design of this lead, we leveraged our digital library of human thoracic spinal cords that were reconstructed from MRI acquisitions and combined these reconstructions with anatomical values from the literature29,33,36–40. These anatomical quantifications uncovered the requirements to arrange two columns of equally distributed electrodes that could cover 100% of the last three segments of the human thoracic spinal cord in more than 95% of the population (Fig. 4b). In addition, we conducted computer simulations to identify the optimal mediolateral location of electrodes that would target the dorsal root entry zones innervating the lower thoracic spinal cord while avoiding the undesired recruitment of the sensory axons in the dorsal columns. These simulations revealed that the optimal tradeoff for the placement of these electrodes within the anatomical constraints of the spinal column was 3.7 mm lateral to the dorsal root entry zones (Fig. 4c and Extended Data Fig. 6c,d). We embedded the identified configuration of electrodes into the design of a new paddle lead that we fabricated with conventional medical-grade technologies and validated for biocompatibility (Fig. 4d).

straints of the spinal column was 3.7 mm lateral to the dorsal root entry zones (Fig. 4c and Extended Data Fig. 6c,d). We embedded the identified configuration of electrodes into the design of a new paddle lead that we fabricated with conventional medical-grade technologies and validated for biocompatibility (Fig. 4d). Having developed the complete implantable system with the required features to regulate blood pressure with EES, we amended our existing clinical study (ClinicalTrials.gov registration: NCT05111093) to interface the new paddle lead with the purpose-built neurostimulation platform (Extended Data Fig. 7a).

straints of the spinal column was 3.7 mm lateral to the dorsal root entry zones (Fig. 4c and Extended Data Fig. 6c,d). We embedded the identified configuration of electrodes into the design of a new paddle lead that we fabricated with conventional medical-grade technologies and validated for biocompatibility (Fig. 4d). Having developed the complete implantable system with the required features to regulate blood pressure with EES, we amended our existing clinical study (ClinicalTrials.gov registration: NCT05111093) to interface the new paddle lead with the purpose-built neurostimulation platform (Extended Data Fig. 7a). We enrolled three new participants who presented with debilitating hypotension (Extended Data Fig. 7b–f and Supplementary Table 2) and implanted them with the complete, new system following the same procedure as described above. The extended length of the lead and optimal coverage of the dorsal root entry zones innervating the hemodynamic hotspot simplified the preoperative planning and intraoperative placement of the lead (Extended Data Fig. 7g,h). Indeed, the extended length avoided compromises on whether the rostral or caudal regions of the hemodynamic hotspot had to be targeted with shorter leads, while the optimized mediolateral electrode locations ensured the selective recruitment of the left versus right dorsal root entry zones (Fig. 4c and Extended Data Fig. 6c,d). These improved features of the paddle lead minimized the need for intraoperative fine-tuning of the paddle lead position. Updates to the personalized computational models of the spinal cord with postoperative imaging data confirmed that the lead covered the entire extent of the hemodynamic hotspot and predicted selective activation of the dorsal root entry zones in all participants (Fig. 5b).Fig. 5Complete purpose-built system to regulate blood pressure.a, Description of the complete system, including the paddle lead, IPG, communication hub and external smartwatch to operate the various programs of the therapy. b, Postoperative reconstruction of the final position of the paddle lead. c, Changes in blood pressure from a representative participant during an orthostatic challenge without EES and with continuous EES applied over the hemodynamic hotspot. The bar plots report the average drop in systolic blood pressure during the orthostatic challenge and average tilt duration without EES and with EES applied over the hemodynamic hotspot (n = 3). Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 3). d, Changes in diastolic and systolic blood pressure during a dynamic orthostatic challenge while EES is applied continuously (left) or in a closed loop (right) to maintain the systolic blood pressure within a target range (red shaded area) in P11.

re status to time, segregated by the presence or absence of EES (n = 3). d, Changes in diastolic and systolic blood pressure during a dynamic orthostatic challenge while EES is applied continuously (left) or in a closed loop (right) to maintain the systolic blood pressure within a target range (red shaded area) in P11. e, Error between the predefined blood pressure target and variability of error to target when EES was delivered continuously or controlled in closed loop (n = 2/3—trial 3, one participant did not complete the assessment due to spasticity). See Source Data Fig. 5 and Extended Data Fig. 7 for source data and statistics.Source data

etween the predefined blood pressure target and variability of error to target when EES was delivered continuously or controlled in closed loop (n = 2/3—trial 3, one participant did not complete the assessment due to spasticity). See Source Data Fig. 5 and Extended Data Fig. 7 for source data and statistics.Source data a, Description of the complete system, including the paddle lead, IPG, communication hub and external smartwatch to operate the various programs of the therapy. b, Postoperative reconstruction of the final position of the paddle lead. c, Changes in blood pressure from a representative participant during an orthostatic challenge without EES and with continuous EES applied over the hemodynamic hotspot. The bar plots report the average drop in systolic blood pressure during the orthostatic challenge and average tilt duration without EES and with EES applied over the hemodynamic hotspot (n = 3). Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 3). d, Changes in diastolic and systolic blood pressure during a dynamic orthostatic challenge while EES is applied continuously (left) or in a closed loop (right) to maintain the systolic blood pressure within a target range (red shaded area) in P11. e, Error between the predefined blood pressure target and variability of error to target when EES was delivered continuously or controlled in closed loop (n = 2/3—trial 3, one participant did not complete the assessment due to spasticity). See Source Data Fig. 5 and Extended Data Fig. 7 for source data and statistics. Source data

a, Description of the complete system, including the paddle lead, IPG, communication hub and external smartwatch to operate the various programs of the therapy. b, Postoperative reconstruction of the final position of the paddle lead. c, Changes in blood pressure from a representative participant during an orthostatic challenge without EES and with continuous EES applied over the hemodynamic hotspot. The bar plots report the average drop in systolic blood pressure during the orthostatic challenge and average tilt duration without EES and with EES applied over the hemodynamic hotspot (n = 3). Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 3). d, Changes in diastolic and systolic blood pressure during a dynamic orthostatic challenge while EES is applied continuously (left) or in a closed loop (right) to maintain the systolic blood pressure within a target range (red shaded area) in P11. e, Error between the predefined blood pressure target and variability of error to target when EES was delivered continuously or controlled in closed loop (n = 2/3—trial 3, one participant did not complete the assessment due to spasticity). See Source Data Fig. 5 and Extended Data Fig. 7 for source data and statistics. Source data The optimized location of the electrodes also eliminated the need for extensive and cumbersome mapping sessions to identify the optimal configurations and parameters to elicit pressor responses41. Instead, minimal adjustments of predetermined electrode configurations (three pairs of electrodes covering the hemodynamic hotspot) and parameters (120 Hz frequency, 0.3 ms pulse duration and up to 20 mA amplitude) were sufficient to trigger robust, safe and well-controlled increases in blood pressure during the first postoperative stimulation session (Extended Data Fig. 7i–l).

igurations (three pairs of electrodes covering the hemodynamic hotspot) and parameters (120 Hz frequency, 0.3 ms pulse duration and up to 20 mA amplitude) were sufficient to trigger robust, safe and well-controlled increases in blood pressure during the first postoperative stimulation session (Extended Data Fig. 7i–l). As observed in the previous ten participants, EES applied over the hemodynamic hotspot led to pressor responses that scaled linearly with increasing amplitudes of EES (Extended Data Fig. 7u) and reduced the severity of orthostatic hypotension during a formal 10-min tilt-table test (Fig. 5c and Extended Data Fig. 7m–t). We next assessed the possibility of controlling blood pressure in a closed loop during formal and dynamic tilt-table testing sessions (Fig. 5d). We found that the complete purpose-built system maintained the desired blood pressure levels with high fidelity despite varying orthostatic challenges (Fig. 5e). We next asked whether the system could be deployed independently by a center that had never been exposed to the therapy. To assess this scalability, we conducted a new clinical study in the Netherlands (ClinicalTrials.gov registration: NCT05941819; Extended Data Fig. 8a).

As observed in the previous ten participants, EES applied over the hemodynamic hotspot led to pressor responses that scaled linearly with increasing amplitudes of EES (Extended Data Fig. 7u) and reduced the severity of orthostatic hypotension during a formal 10-min tilt-table test (Fig. 5c and Extended Data Fig. 7m–t). We next assessed the possibility of controlling blood pressure in a closed loop during formal and dynamic tilt-table testing sessions (Fig. 5d). We found that the complete purpose-built system maintained the desired blood pressure levels with high fidelity despite varying orthostatic challenges (Fig. 5e). We next asked whether the system could be deployed independently by a center that had never been exposed to the therapy. To assess this scalability, we conducted a new clinical study in the Netherlands (ClinicalTrials.gov registration: NCT05941819; Extended Data Fig. 8a). A neurosurgeon who had not been previously involved in the studies followed our procedures to implant the complete system in a participant who presented with severe orthostatic hypotension (Extended Data Fig. 8b–f and Supplementary Table 2). Intraoperative delivery of EES over the hemodynamic hotspot triggered robust pressor responses. Postoperative imaging confirmed the location of the lead over the hemodynamic hotspot (Extended Data Fig. 8g,h).

ticipant who presented with severe orthostatic hypotension (Extended Data Fig. 8b–f and Supplementary Table 2). Intraoperative delivery of EES over the hemodynamic hotspot triggered robust pressor responses. Postoperative imaging confirmed the location of the lead over the hemodynamic hotspot (Extended Data Fig. 8g,h). During the first postoperative session, we asked a physician without prior experience with EES to program the therapy based on well-defined anatomical principles. The nonexpert was able to program the therapy in only a few minutes. As early as the first session postoperatively, EES delivered over the hemodynamic hotspot in a seated position elicited robust pressor responses that increased the systolic blood pressure from 86 mm Hg to 105 mm Hg (Extended Data Fig. 8i). With EES, the participant was able to complete the 10-min tilt-table test, which was not possible without EES (Extended Data Fig. 8j–n). Closed-loop control of EES further improved the maintenance of blood pressure, despite orthostatic challenges of varying severities (Extended Data Fig. 8o–q). We concluded that the safety and efficacy profile of the new system and therapy were sufficient to enable the participants (n = 8 participants) to use the therapy independently at home and thus to evaluate the long-term consequences of this therapy for the management of blood pressure instability after SCI. To support this evaluation, the participants were sent home with stimulation programs that were tailored to their individual pattern of hypotensive symptoms by a trained physician.

ndently at home and thus to evaluate the long-term consequences of this therapy for the management of blood pressure instability after SCI. To support this evaluation, the participants were sent home with stimulation programs that were tailored to their individual pattern of hypotensive symptoms by a trained physician. Although the various participants included in our clinical studies were implanted with successively improved technologies, this diverse cohort of female and male participants spanning 18–74 years of age (Supplementary Tables 2 and 3) across three sites provided the opportunity to conduct a preliminary evaluation of the safety and effectiveness of EES applied over the hemodynamic hotspot to manage hemodynamic instability in people living with chronic SCI.

hort of female and male participants spanning 18–74 years of age (Supplementary Tables 2 and 3) across three sites provided the opportunity to conduct a preliminary evaluation of the safety and effectiveness of EES applied over the hemodynamic hotspot to manage hemodynamic instability in people living with chronic SCI. When evaluated up to 2 years after implantation, participants still exhibited severe orthostatic hypotension during a formal tilt-table test when EES was turned off. They also continued exhibiting robust improvement when EES was turned on, which allowed them to complete the tilt-table test (Fig. 6a,b and Extended Data Fig. 9f–k). Moreover, formal evaluations confirmed the stabilization of blood pressure during constant-varying orthostatic challenges when EES was controlled in a closed loop (Extended Data Fig. 9m–o). Pressor responses during the tilt test with EES were associated with increased concentrations of circulating norepinephrine (Extended Data Fig. 9l), augmented splanchnic volume (Extended Data Fig. 10i) and increased sympathetic nerve activity (Extended Data Fig. 10j), which confirmed the activation of the sympathetic nervous system with EES.Fig. 6Long-term management of blood pressure improves quality of life.a, Changes in systolic blood pressure during a formal tilt-table test before implantation (left), without EES after 3–24 months after system implantation (middle) and at the same timepoint with EES (right). n = 11/11—trials 1, 3 and 4; n = 7 measured after at least 6 months of daily management of blood pressure with the system; n = 4 after at least 3 months. b, Bar plot reporting average changes in systolic blood pressure (left) during a format tilt-table test performed as in a (n = 11, repeated-measures ANOVA with Tukey’s HSD). Corresponding Kaplan–Meier plot of exposure status to time sustained in tilt, segregated by the same condition as in a (n = 11, mixed model Cox regression with likelihood ratio test estimate = 36.36; P = 3.0 × 10−4). c, Changes in systolic blood pressure when delivering EES in a seated position. Each line reports responses from a single participant to EES with default parameters for blood pressure regulation (n = 11/11—trials 1, 3 and 4; n = 9 participants measured after at least 6 months of daily management of blood pressure with the system; n = 2 with at least 3 months).

delivering EES in a seated position. Each line reports responses from a single participant to EES with default parameters for blood pressure regulation (n = 11/11—trials 1, 3 and 4; n = 9 participants measured after at least 6 months of daily management of blood pressure with the system; n = 2 with at least 3 months). Bar plots report the average change in systolic blood pressure in a seated position that was quantified before and at 2 min after the onset of EES (n = 9 postimplant tilt measured with at least 6 months of EES, n = 2 postimplant tilt measured with at least 3 months of EES use, paired, two-tailed t-test; t = 16.63, ***P = 1.29 × 10−8). d, Quantification of the total CBF without and with EES using a Doppler spectrum of the internal carotid without and with EES. The bar graph reports the total change in CBF measured in a seated position without and with EES (n = 10/10—trial 1 and 3; paired, two-tailed, t-test; t = −4.32; **P = 1.90 × 10−3). e, Bar graph reporting ADFSCI orthostatic hypotension score quantified before implantation of the system and after 6 months of management of blood pressure with the system (n = 12/14—trials 1–4, two participants explanted before 6 months; paired, two-tailed t-test; t = 4.54; ***P = 8.45 × 10−4). Symptoms are color-coded by the percentage of participants who experience each symptom. f, Bar graph reporting the performance (n = 8/8—trials 3 and 4, paired, two-tailed t-test; t = −3.92; **P = 5.7 × 10−3) and satisfaction (n = 8/8—trials 3 and 4, paired, two-tailed t-test; t = −5.54; ***P = 8.7 × 10−4) scores of the COPM relative to baseline goals set by occupational therapists before the implantation of the system and after 6 months of management of blood pressure with the system. g, Bar graph reporting scores on the MOS-S—Sleep Problem Index II measured before implantation of the system and after 6 months of management of blood pressure with the system (n = 12/14—trials 1–4, two participants explanted before 6 months; paired, two-tailed t-test; t = 2.58; *P = 2.52 × 10−2). h, Bar graph reporting NBDS before implantation of the system and after 6 months of management of blood pressure with the system (n = 11/11—trials 1–3; paired samples t-test; t = 2.23; *P = 4.95 × 10−2). i, Usage of EES programs to improve blood pressure instability during upright rehabilitation and maintain trunk stability during recreation activities such as sit-skiing. See Source Data Fig. 6 and Extended Data Figs.

od pressure with the system (n = 11/11—trials 1–3; paired samples t-test; t = 2.23; *P = 4.95 × 10−2). i, Usage of EES programs to improve blood pressure instability during upright rehabilitation and maintain trunk stability during recreation activities such as sit-skiing. See Source Data Fig. 6 and Extended Data Figs. 9 and 10 for source data and statistics.Source data

od pressure with the system (n = 11/11—trials 1–3; paired samples t-test; t = 2.23; *P = 4.95 × 10−2). i, Usage of EES programs to improve blood pressure instability during upright rehabilitation and maintain trunk stability during recreation activities such as sit-skiing. See Source Data Fig. 6 and Extended Data Figs. 9 and 10 for source data and statistics.Source data a, Changes in systolic blood pressure during a formal tilt-table test before implantation (left), without EES after 3–24 months after system implantation (middle) and at the same timepoint with EES (right). n = 11/11—trials 1, 3 and 4; n = 7 measured after at least 6 months of daily management of blood pressure with the system; n = 4 after at least 3 months. b, Bar plot reporting average changes in systolic blood pressure (left) during a format tilt-table test performed as in a (n = 11, repeated-measures ANOVA with Tukey’s HSD). Corresponding Kaplan–Meier plot of exposure status to time sustained in tilt, segregated by the same condition as in a (n = 11, mixed model Cox regression with likelihood ratio test estimate = 36.36; P = 3.0 × 10−4). c, Changes in systolic blood pressure when delivering EES in a seated position. Each line reports responses from a single participant to EES with default parameters for blood pressure regulation (n = 11/11—trials 1, 3 and 4; n = 9 participants measured after at least 6 months of daily management of blood pressure with the system; n = 2 with at least 3 months). Bar plots report the average change in systolic blood pressure in a seated position that was quantified before and at 2 min after the onset of EES (n = 9 postimplant tilt measured with at least 6 months of EES, n = 2 postimplant tilt measured with at least 3 months of EES use, paired, two-tailed t-test; t = 16.63, ***P = 1.29 × 10−8). d, Quantification of the total CBF without and with EES using a Doppler spectrum of the internal carotid without and with EES. The bar graph reports the total change in CBF measured in a seated position without and with EES (n = 10/10—trial 1 and 3; paired, two-tailed, t-test; t = −4.32; **P = 1.90 × 10−3). e, Bar graph reporting ADFSCI orthostatic hypotension score quantified before implantation of the system and after 6 months of management of blood pressure with the system (n = 12/14—trials 1–4, two participants explanted before 6 months; paired, two-tailed t-test; t = 4.54; ***P = 8.45 × 10−4). Symptoms are color-coded by the percentage of participants who experience each symptom.

tified before implantation of the system and after 6 months of management of blood pressure with the system (n = 12/14—trials 1–4, two participants explanted before 6 months; paired, two-tailed t-test; t = 4.54; ***P = 8.45 × 10−4). Symptoms are color-coded by the percentage of participants who experience each symptom. f, Bar graph reporting the performance (n = 8/8—trials 3 and 4, paired, two-tailed t-test; t = −3.92; **P = 5.7 × 10−3) and satisfaction (n = 8/8—trials 3 and 4, paired, two-tailed t-test; t = −5.54; ***P = 8.7 × 10−4) scores of the COPM relative to baseline goals set by occupational therapists before the implantation of the system and after 6 months of management of blood pressure with the system. g, Bar graph reporting scores on the MOS-S—Sleep Problem Index II measured before implantation of the system and after 6 months of management of blood pressure with the system (n = 12/14—trials 1–4, two participants explanted before 6 months; paired, two-tailed t-test; t = 2.58; *P = 2.52 × 10−2). h, Bar graph reporting NBDS before implantation of the system and after 6 months of management of blood pressure with the system (n = 11/11—trials 1–3; paired samples t-test; t = 2.23; *P = 4.95 × 10−2). i, Usage of EES programs to improve blood pressure instability during upright rehabilitation and maintain trunk stability during recreation activities such as sit-skiing. See Source Data Fig. 6 and Extended Data Figs. 9 and 10 for source data and statistics. Source data

f, Bar graph reporting the performance (n = 8/8—trials 3 and 4, paired, two-tailed t-test; t = −3.92; **P = 5.7 × 10−3) and satisfaction (n = 8/8—trials 3 and 4, paired, two-tailed t-test; t = −5.54; ***P = 8.7 × 10−4) scores of the COPM relative to baseline goals set by occupational therapists before the implantation of the system and after 6 months of management of blood pressure with the system. g, Bar graph reporting scores on the MOS-S—Sleep Problem Index II measured before implantation of the system and after 6 months of management of blood pressure with the system (n = 12/14—trials 1–4, two participants explanted before 6 months; paired, two-tailed t-test; t = 2.58; *P = 2.52 × 10−2). h, Bar graph reporting NBDS before implantation of the system and after 6 months of management of blood pressure with the system (n = 11/11—trials 1–3; paired samples t-test; t = 2.23; *P = 4.95 × 10−2). i, Usage of EES programs to improve blood pressure instability during upright rehabilitation and maintain trunk stability during recreation activities such as sit-skiing. See Source Data Fig. 6 and Extended Data Figs. 9 and 10 for source data and statistics. Source data Every tested participant continued demonstrating controllable pressor responses when EES was delivered over the hemodynamic hotspot while seated in their wheelchair, corresponding to the conditions of their daily life (Fig. 6c and Extended Data Fig. 9a–e). Pressor responses in seated conditions coincided with an increase in cerebral blood flow (CBF; Fig. 6d).

trating controllable pressor responses when EES was delivered over the hemodynamic hotspot while seated in their wheelchair, corresponding to the conditions of their daily life (Fig. 6c and Extended Data Fig. 9a–e). Pressor responses in seated conditions coincided with an increase in cerebral blood flow (CBF; Fig. 6d). These objective physiological outcomes also concurred with a reduction in self-reported measures of orthostatic hypotension severity (Fig. 6e). These results showed that EES applied over the hemodynamic hotspot led to an immediate and durable reduction in the severity of hypotension. We next asked whether long-term improvement in blood pressure stability translated into an improvement in quality of life and an increased possibility to participate in activities of daily living remained unclear.

These objective physiological outcomes also concurred with a reduction in self-reported measures of orthostatic hypotension severity (Fig. 6e). These results showed that EES applied over the hemodynamic hotspot led to an immediate and durable reduction in the severity of hypotension. We next asked whether long-term improvement in blood pressure stability translated into an improvement in quality of life and an increased possibility to participate in activities of daily living remained unclear. To address this question, we quantified the use of the therapy at home from the stimulation logs and found that the participants turned the stimulation on throughout their waking hours, with increased usage to facilitate verticalization in the morning routine and to reduce postprandial hypotension after meals (Extended Data Fig. 10a). We confirmed the improvement in postprandial hypotension during in-clinic evaluations with controlled food intake (Extended Data Fig. 10b). The daily use of the therapy was accompanied by the medically supervised cessation of all other conservative treatment options that participants had been relying on to manage hypotension, including abdominal binders and midodrine (Extended Data Fig. 10d). However, these improvements only manifested in the presence of the stimulation because none of the participants demonstrated a reduction in the severity of orthostatic hypotension during formal tilt tests when EES was turned off (Fig. 6a,b and Extended Data Fig. 9g–k). Evaluations of system usability42 confirmed that the participants could operate the therapy safely and effectively despite impairments in arm and hand movements (Extended Data Fig. 10c).

in the severity of orthostatic hypotension during formal tilt tests when EES was turned off (Fig. 6a,b and Extended Data Fig. 9g–k). Evaluations of system usability42 confirmed that the participants could operate the therapy safely and effectively despite impairments in arm and hand movements (Extended Data Fig. 10c). These multifaceted ameliorations translated into improvements in quality of life, as measured by formal outcome measures (Fig. 6e), which coincided with other subjective improvements in sleep quality, bowel management, spasticity and engagement in social activities (Fig. 6f–i and Extended Data Fig. 10k,e). Moreover, participants could engage in rehabilitation exercises that involved standing or walking in a motorized exoskeleton43 (Fig. 6i and Extended Data Fig. 10f) as well as improved trunk control with EES programs targeting the trunk musculature (Fig. 6i and Extended Data Fig. 10g,h). We systematically monitored for device- and procedure-related adverse events throughout the clinical studies. All the participants have been operating the system to regulate their hemodynamic instability daily for up to 2 years. During this period, we recorded any serious and nonserious adverse device effects (Supplementary Tables 4–8).

We systematically monitored for device- and procedure-related adverse events throughout the clinical studies. All the participants have been operating the system to regulate their hemodynamic instability daily for up to 2 years. During this period, we recorded any serious and nonserious adverse device effects (Supplementary Tables 4–8). Across the eight participants implanted with the system or components of this system, no serious adverse events related to the device were reported. Only one serious adverse event involving an uncontrolled increase in blood pressure due to the inadvertent removal of a urinary probe occurred while the system was turned off (Supplementary Table 8). Uncontrolled increases in blood pressure during EES occurred only once across hundreds of testing sessions conducted over a period that spanned several months and involved the delivery of EES across a broad range of parameters. This nonserious device-related adverse event occurred while targeting the trunk musculature but ceased as soon as EES was switched off.

Extensive preclinical evidence in mice, rats and nonhuman primates, as well as clinical case studies, demonstrated that EES reduces the severity of orthostatic hypotension1,20,23–26. However, the optimal location to deliver EES has remained controversial. We previously showed that EES activates large-diameter afferents where they bend to enter the spinal cord through the dorsal root entry zones and that this recruitment engages sympathetic preganglionic neurons indirectly via Vsx2-expressing excitatory neurons nested in the intermediate lamina of the stimulated region of the spinal cord1,16–18 (Fig. 2a). We also confirmed an enrichment of these sympathetic preganglionic neurons in the lower thoracic spinal cord (Fig. 2a) and showed that this enrichment coincided with the maximal pressor response to EES. This location was thus termed the hemodynamic hotspot1. Despite this evidence, many studies reported unexpected pressor responses during the application of EES over the lumbosacral spinal cord23–26. The incongruence of these observations with the anatomy of the autonomic nervous system19,27 compelled us to perform a head-to-head comparison in the same participants to establish the optimal location to deliver EES for the regulation of blood pressure in humans with SCI.Fig. 2Validation of the location of the hemodynamic hotspot.a, Overview of the neuronal architecture recruited EES to trigger pressor responses. b, Anatomical planning of laminotomies for inserting the paddle leads (left) based on the targeted location of the paddle leads (right) over the thoracic and lumbosacral segments. c, Amplitude of systolic pressor response per spinal segment when delivering EES intraoperatively. Dots (n = 5 participants) denote pressure responses to EES applied over specific segments, the location of which was reconstructed postoperatively. d, Bar graph shows averaged changes in systolic blood pressure in response to EES applied over thoracic or lumbar spinal segments (n = 5, paired samples two-tailed t-test; t = 6.46, **P = 3.00 × 10−3).

f the autonomic nervous system. Instead, EES applied over the hemodynamic hotspot, wherein an enrichment of sympathetic preganglionic neurons involved in the modulation of blood pressure resides, consistently triggered robust pressor responses that enabled all six participants to tolerate severe orthostatic challenges. These quantifications provide important results for the ongoing debate on the optimal location of EES for the effective treatment of hemodynamic instability in people with SCI. In addition, recent preclinical experiments demonstrated that long-term application of EES over the lumbosacral spinal cord augmented the density of maladaptive synaptic connections from Vsx2-expressing neurons of the lumbosacral to Vsx2-expressing neurons of the lower thoracic spinal cord, which reinforces the severity of autonomic dysreflexia (AD)16. Given the discrepancy in the safety and efficacy of EES applied over the hemodynamic hotspot versus lumbosacral segments, these results establish a safe path forward on how to deliver EES to treat hemodynamic instability due to SCI.

We aimed to develop a clinical-grade implantable neurostimulation platform that addresses the limitations of existing systems and meets the identified requirements for the safe and effective management of hemodynamic instability in people with SCI (Supplementary Note 2).

We aimed to develop a clinical-grade implantable neurostimulation platform that addresses the limitations of existing systems and meets the identified requirements for the safe and effective management of hemodynamic instability in people with SCI (Supplementary Note 2). Because the possibility to adjust EES parameters in a closed loop was a desired feature of the platform, we designed an implantable neurostimulation platform that continuously communicates with a wearable hub through near-field magnetic induction (NFMI; Fig. 3a). This hub can acquire external signals, including continuous measurements of blood pressure, and embeds computing power to support closed-loop control algorithms. This configuration enables closed-loop control of EES parameters with latencies as low as 25 ms. We programmed the firmware to support any configuration of anodes and cathodes and to resolve the limited range of programmable EES parameters in current neurostimulation platforms.Fig. 3Validation of the new, purpose-built implantable neurostimulation platform.a, Scheme of the investigational system, including a clinician controller (tablet), patient programmer (external smartwatch), a main controller (hub), IPG and a Medtronic 5-6-5 SureScan Specify paddle lead. b, Changes in blood pressure from a representative participant during an orthostatic challenge without and with EES applied over the hemodynamic hotspot (EESBP). c, Average drop in systolic blood pressure during orthostatic challenge without and with EES (n = 4, paired samples two-tailed t-test; t = 4.27, *P = 2.4 × 10−2). d, Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 24 tilts; mixed model Cox regression with likelihood ratio test estimate = 15.36, P = 4.00 × 10−4) and tilt duration (n = 4, paired samples two-tailed t-test; t = 4.69, *P = 1.8 × 10−2) without EES and with EES applied over the hemodynamic hotspot. e, Closed-loop control of blood pressure consisting of a PID controller that adjusts the amplitude of EES applied over the hemodynamic hotspot to maintain blood pressure within a range of predefined blood pressure targets. f, Stepwise increase in systolic blood pressure in response to gradual increases of EES amplitude (1 mA per min) for a representative participant.

es in postural orientation (Fig. 3g, Extended Data Figs. 5b and 9n). Quantifications also revealed the superiority of EES controlled in closed loop over both continuous EES and alternative closed-loop approaches relying on inertial measurement units or manual adjustments of EES (Fig. 3h, Extended Data Figs. 5c and 9o). These results confirmed that the newly developed neurostimulation platform fulfilled the technical requirements to modulate blood pressure with EES and established the possibility of achieving precise control of blood pressure in a closed loop despite transient, varying and sustained hemodynamic challenges in people with orthostatic hypotension due to SCI.

We next aimed to develop a clinical-grade paddle lead that combines the necessary features to target the large-diameter afferents projecting to the hemodynamic hotspot while avoiding the undesired side effects of nonselective stimulation29,33–35.

We next aimed to develop a clinical-grade paddle lead that combines the necessary features to target the large-diameter afferents projecting to the hemodynamic hotspot while avoiding the undesired side effects of nonselective stimulation29,33–35. Our mapping experiments showed that the hemodynamic hotspot is centered over the last three segments of the lower thoracic spinal cord. Moreover, the same experiments revealed that EES applied over multiple segments leads to cumulative increases in the amplitude of pressor responses compared to when the same amplitude of EES is applied over a single spinal segment (Fig. 4a).Fig. 4Purpose-built paddle lead to target the hemodynamic hotspot.a, EES applied over multiple segments of the hemodynamic hotspot leads to a superior increase in blood pressure compared to EES applied over the median of a single segment (paired samples, one-tailed t-test, n = 5, t = 2.69, *P = 2.72 × 10−2). b, Quantification of the length of the lower thoracic spinal cord in the study participants (n = 10; left), and average length of the lower thoracic spinal cord quantified from published data (right). The coverage of the hemodynamic hotspot by the Medtronic 5-6-5 lead and expected from the newly designed paddle lead are reported. c, Illustrative drawing of the anatomical model implemented in silico to predict the relationships between the lateral position of EES electrodes and the relative recruitment of the targeted dorsal root, contralateral dorsal root and dorsal columns, as shown in the plot. d, Photograph of the newly designed purpose-built paddle lead that combines two parallel columns of eight equally spaced electrodes that aim to cover the entire length of the hemodynamic hotspot while avoiding the recruitment of the dorsal columns. See Source Data Fig. 4 and Extended Data Fig. 6 for source data and statistics.Source data

Having developed the complete implantable system with the required features to regulate blood pressure with EES, we amended our existing clinical study (ClinicalTrials.gov registration: NCT05111093) to interface the new paddle lead with the purpose-built neurostimulation platform (Extended Data Fig. 7a).

Having developed the complete implantable system with the required features to regulate blood pressure with EES, we amended our existing clinical study (ClinicalTrials.gov registration: NCT05111093) to interface the new paddle lead with the purpose-built neurostimulation platform (Extended Data Fig. 7a). We enrolled three new participants who presented with debilitating hypotension (Extended Data Fig. 7b–f and Supplementary Table 2) and implanted them with the complete, new system following the same procedure as described above. The extended length of the lead and optimal coverage of the dorsal root entry zones innervating the hemodynamic hotspot simplified the preoperative planning and intraoperative placement of the lead (Extended Data Fig. 7g,h). Indeed, the extended length avoided compromises on whether the rostral or caudal regions of the hemodynamic hotspot had to be targeted with shorter leads, while the optimized mediolateral electrode locations ensured the selective recruitment of the left versus right dorsal root entry zones (Fig. 4c and Extended Data Fig. 6c,d). These improved features of the paddle lead minimized the need for intraoperative fine-tuning of the paddle lead position. Updates to the personalized computational models of the spinal cord with postoperative imaging data confirmed that the lead covered the entire extent of the hemodynamic hotspot and predicted selective activation of the dorsal root entry zones in all participants (Fig. 5b).Fig. 5Complete purpose-built system to regulate blood pressure.a, Description of the complete system, including the paddle lead, IPG, communication hub and external smartwatch to operate the various programs of the therapy. b, Postoperative reconstruction of the final position of the paddle lead. c, Changes in blood pressure from a representative participant during an orthostatic challenge without EES and with continuous EES applied over the hemodynamic hotspot. The bar plots report the average drop in systolic blood pressure during the orthostatic challenge and average tilt duration without EES and with EES applied over the hemodynamic hotspot (n = 3). Kaplan–Meier plot of exposure status to time, segregated by the presence or absence of EES (n = 3). d, Changes in diastolic and systolic blood pressure during a dynamic orthostatic challenge while EES is applied continuously (left) or in a closed loop (right) to maintain the systolic blood pressure within a target range (red shaded area) in P11.

igurations (three pairs of electrodes covering the hemodynamic hotspot) and parameters (120 Hz frequency, 0.3 ms pulse duration and up to 20 mA amplitude) were sufficient to trigger robust, safe and well-controlled increases in blood pressure during the first postoperative stimulation session (Extended Data Fig. 7i–l). As observed in the previous ten participants, EES applied over the hemodynamic hotspot led to pressor responses that scaled linearly with increasing amplitudes of EES (Extended Data Fig. 7u) and reduced the severity of orthostatic hypotension during a formal 10-min tilt-table test (Fig. 5c and Extended Data Fig. 7m–t). We next assessed the possibility of controlling blood pressure in a closed loop during formal and dynamic tilt-table testing sessions (Fig. 5d). We found that the complete purpose-built system maintained the desired blood pressure levels with high fidelity despite varying orthostatic challenges (Fig. 5e).

We next asked whether the system could be deployed independently by a center that had never been exposed to the therapy. To assess this scalability, we conducted a new clinical study in the Netherlands (ClinicalTrials.gov registration: NCT05941819; Extended Data Fig. 8a). A neurosurgeon who had not been previously involved in the studies followed our procedures to implant the complete system in a participant who presented with severe orthostatic hypotension (Extended Data Fig. 8b–f and Supplementary Table 2). Intraoperative delivery of EES over the hemodynamic hotspot triggered robust pressor responses. Postoperative imaging confirmed the location of the lead over the hemodynamic hotspot (Extended Data Fig. 8g,h). During the first postoperative session, we asked a physician without prior experience with EES to program the therapy based on well-defined anatomical principles. The nonexpert was able to program the therapy in only a few minutes. As early as the first session postoperatively, EES delivered over the hemodynamic hotspot in a seated position elicited robust pressor responses that increased the systolic blood pressure from 86 mm Hg to 105 mm Hg (Extended Data Fig. 8i). With EES, the participant was able to complete the 10-min tilt-table test, which was not possible without EES (Extended Data Fig. 8j–n). Closed-loop control of EES further improved the maintenance of blood pressure, despite orthostatic challenges of varying severities (Extended Data Fig. 8o–q).

We concluded that the safety and efficacy profile of the new system and therapy were sufficient to enable the participants (n = 8 participants) to use the therapy independently at home and thus to evaluate the long-term consequences of this therapy for the management of blood pressure instability after SCI. To support this evaluation, the participants were sent home with stimulation programs that were tailored to their individual pattern of hypotensive symptoms by a trained physician.

Although the various participants included in our clinical studies were implanted with successively improved technologies, this diverse cohort of female and male participants spanning 18–74 years of age (Supplementary Tables 2 and 3) across three sites provided the opportunity to conduct a preliminary evaluation of the safety and effectiveness of EES applied over the hemodynamic hotspot to manage hemodynamic instability in people living with chronic SCI.

trating controllable pressor responses when EES was delivered over the hemodynamic hotspot while seated in their wheelchair, corresponding to the conditions of their daily life (Fig. 6c and Extended Data Fig. 9a–e). Pressor responses in seated conditions coincided with an increase in cerebral blood flow (CBF; Fig. 6d). These objective physiological outcomes also concurred with a reduction in self-reported measures of orthostatic hypotension severity (Fig. 6e). These results showed that EES applied over the hemodynamic hotspot led to an immediate and durable reduction in the severity of hypotension.

We next asked whether long-term improvement in blood pressure stability translated into an improvement in quality of life and an increased possibility to participate in activities of daily living remained unclear. To address this question, we quantified the use of the therapy at home from the stimulation logs and found that the participants turned the stimulation on throughout their waking hours, with increased usage to facilitate verticalization in the morning routine and to reduce postprandial hypotension after meals (Extended Data Fig. 10a). We confirmed the improvement in postprandial hypotension during in-clinic evaluations with controlled food intake (Extended Data Fig. 10b). The daily use of the therapy was accompanied by the medically supervised cessation of all other conservative treatment options that participants had been relying on to manage hypotension, including abdominal binders and midodrine (Extended Data Fig. 10d). However, these improvements only manifested in the presence of the stimulation because none of the participants demonstrated a reduction in the severity of orthostatic hypotension during formal tilt tests when EES was turned off (Fig. 6a,b and Extended Data Fig. 9g–k). Evaluations of system usability42 confirmed that the participants could operate the therapy safely and effectively despite impairments in arm and hand movements (Extended Data Fig. 10c).

in the severity of orthostatic hypotension during formal tilt tests when EES was turned off (Fig. 6a,b and Extended Data Fig. 9g–k). Evaluations of system usability42 confirmed that the participants could operate the therapy safely and effectively despite impairments in arm and hand movements (Extended Data Fig. 10c). These multifaceted ameliorations translated into improvements in quality of life, as measured by formal outcome measures (Fig. 6e), which coincided with other subjective improvements in sleep quality, bowel management, spasticity and engagement in social activities (Fig. 6f–i and Extended Data Fig. 10k,e). Moreover, participants could engage in rehabilitation exercises that involved standing or walking in a motorized exoskeleton43 (Fig. 6i and Extended Data Fig. 10f) as well as improved trunk control with EES programs targeting the trunk musculature (Fig. 6i and Extended Data Fig. 10g,h).

We conducted two epidemiological analyses and four complementary clinical studies involving implantable medical devices that (1) established the burden of hypotension complications in people with SCI, (2) provided data that supports the lower thoracic spinal cord as the optimal location to apply EES for treating hemodynamic instability, (3) enabled the stepwise development of a purpose-built system harnessing the understanding of the mechanisms through which EES regulates hemodynamics in people with SCI and (4) provided preliminary safety and efficacy data of this system to reduce hypotensive complications throughout the day and thus improved quality of life. Before considering the development of a therapy based on EES to improve hemodynamic instability after SCI, we thought that it was essential to address the controversy on the optimal location to restore hemodynamics with EES23–26,44,45. We found that EES applied to the hemodynamic hotspot is more effective compared to EES applied to the lumbosacral spinal cord. This result was not surprising because attempts to control hemodynamics by stimulating the lumbar region of the spinal cord are incongruent with the basic and well-conserved anatomy of the ancestral autonomic nervous system. To date, the mechanisms underlying the fortuitous observations of pressor responses with EES applied over the lumbosacral spinal cord remain unknown. We surmised that two mechanisms might explain these observations.

ord are incongruent with the basic and well-conserved anatomy of the ancestral autonomic nervous system. To date, the mechanisms underlying the fortuitous observations of pressor responses with EES applied over the lumbosacral spinal cord remain unknown. We surmised that two mechanisms might explain these observations. The first mechanism that we previously exposed in preclinical models is the off-target activation of the dorsal roots innervating the last thoracic segment as they exit the spinal column through the T12 intervertebral foramen1, which roughly coincides with upper lumbar spinal cord segments33. Our mapping experiments demonstrate that the optimal coverage of the last three segments of the thoracic spinal cord, especially the penultimate thoracic segment, is essential to maximize pressor responses to EES. Consequently, the inefficient recruitment of only the last thoracic root is unlikely to be an effective and reproducible methodology to manage hemodynamic instability across the population of people with SCI.

ic spinal cord, especially the penultimate thoracic segment, is essential to maximize pressor responses to EES. Consequently, the inefficient recruitment of only the last thoracic root is unlikely to be an effective and reproducible methodology to manage hemodynamic instability across the population of people with SCI. The second mechanism, which is the most concerning for the safety of people with SCI, is the risk that EES applied over the lower aspect of the lumbosacral spinal cord directly activates the pathways that cause life-threatening AD2,46. Sensory information originating from the overstimulated afferents of the bladder and bowel is the most common trigger of AD in people with SCI. Afferents from these visceral organs project to the spinal cord through the lower lumbar and upper sacral dorsal roots, which are directly recruited when applying EES over the lumbosacral spinal cord, especially with configurations targeting the lower lumbar dorsal roots25,41,45. Because no other neuronal populations that contribute to the control of blood pressure exist in this region of the spinal cord1,19, the only explanation for pressor responses elicited by the application of EES over the lower lumbar and sacral spinal cord is the activation of the pathways that can cause AD. Indeed, preclinical evidence demonstrated that daily application of EES over the lower lumbosacral segments in rodents with complete upper thoracic SCI augments the density of axonal projections from Vsx2-expressing neurons in the lumbosacral spinal cord onto Vsx2-expressing neurons in the lower thoracic spinal cord16. These neurons are embedded in the neuronal architecture that causes AD. Consequently, this aberrant exacerbation of the neuronal architecture responsible for AD doubled the severity of pressor responses that characterize this life-threatening condition16. Although the implication of Vsx2-expressing neurons has not been confirmed in humans, single-cell and spatial transcriptomic technologies47 have advanced sufficiently to enable the exploration of this mechanism, even in human tissue. With this limitation in mind, this mechanism appears to be the most likely explanation for the reported improvements in resting hypotension and orthostatic hypotension in individuals with SCI who underwent extensive stimulation of the lower lumbosacral segments.

the exploration of this mechanism, even in human tissue. With this limitation in mind, this mechanism appears to be the most likely explanation for the reported improvements in resting hypotension and orthostatic hypotension in individuals with SCI who underwent extensive stimulation of the lower lumbosacral segments. Together, consistent results obtained in a total of 14 participants in four clinical studies across three countries conducted at three independent clinical centers with independent clinical teams provide robust preliminary data on the safety and efficacy of EES applied over the hemodynamic hotspot to treat hemodynamic instability in people living with chronic SCI. Monitoring the effectiveness of this therapy in ecological settings will require new technologies suitable to monitor blood pressure regularly throughout the day and concomitantly to changes in postural orientation and ongoing activities of the patients. Because these technologies are not yet available, we were not able to assess the efficacy of EES during the daily activities of the participants. However, we anticipate that these new technologies will open the possibility of quantifying the usage and benefit of EES to reduce the incidence and severity of hypotensive complications in ecological settings. These findings thus establish a path forward to treat the underappreciated, treatment-resistant hypotensive complications due to SCI. Completing this path will require a pivotal device trial that evaluates the safety and efficacy of this therapy.

Ethical approval was obtained from an independent ethics board (Veritas Independent Review Board) and the Research Ethics Board of Université Laval (the principal investigator’s institution). Ethical approval from local research ethics boards was also obtained to recruit from SCI centers across Canada. Individuals with SCI (n = 1,479) across Canada were recruited using a national consumer awareness campaign and provided written informed consent to participate21,22. The survey consisted of a series of variables identified by healthcare and service providers and researchers, as well as individuals with SCI, including demographics, secondary health complications, comorbidities, SCI-related needs, healthcare usage, community participation, quality of life and overall health ratings21,22. Participants were asked how often they experienced symptoms such as lightheadedness and/or dizziness consistent with symptoms of orthostatic hypotension in the past 12 months. Responses were ranked on a six-point ordinal severity scale ranging from 0 (that is, ‘never’) to 5 (that is, ‘every day’). Participants were also asked if they received or sought out treatment concerning these symptoms on a two-point scale (that is, ‘yes’ or ‘no’), along with the degree to which it limited activities from 0 (that is, ‘never’) to 5 (that is, ‘every day’). Participants were also asked if they had experienced specific problems, such as heart disease, in the past 12 months. Participants’ American Spinal Injury Association (ASIA) Impairment Scale (AIS) was estimated using responses to questions about lesion level and sensorimotor/mobility capabilities. A binary approach was used for the evaluation of outcome variables, including the level of injury (that is, cervical SCI versus noncervical SCI), the severity of injury (that is, functionally complete versus incomplete), the presence of orthostatic hypotension (that is, yes versus no) and orthostatic hypotensive symptoms (that is, yes versus no). For variables ranked on a six-point ordinal scale, lower scores (that is, 0–3) were categorized as ‘no’ and higher scores (that is, 4–5) were categorized as ‘yes’.

complete versus incomplete), the presence of orthostatic hypotension (that is, yes versus no) and orthostatic hypotensive symptoms (that is, yes versus no). For variables ranked on a six-point ordinal scale, lower scores (that is, 0–3) were categorized as ‘no’ and higher scores (that is, 4–5) were categorized as ‘yes’. In total, 254 participants were surveyed with the ADFSCI questionnaire (ADFSCI) to measure the severity and frequency of orthostatic hypotension in a diverse sample of individuals with SCI. From this population, 17 individuals (Supplementary Table 1) were pseudo-randomly selected to participate in a single formal, 10-min tilt-table test (‘Orthostatic challenge with tilt-table test’). Hemodynamic measurements were taken via finger plethysmography (‘Postoperative hemodynamic monitoring’), and data were analogously acquired with LabChart Software (ADInstruments) and upsampled to 1,000 Hz.

selected to participate in a single formal, 10-min tilt-table test (‘Orthostatic challenge with tilt-table test’). Hemodynamic measurements were taken via finger plethysmography (‘Postoperative hemodynamic monitoring’), and data were analogously acquired with LabChart Software (ADInstruments) and upsampled to 1,000 Hz. All experiments were carried out as part of four clinical safety (primary objective) and preliminary efficacy (secondary objectives) trials—STIMO-HEMO (ClinicalTrials.gov registration: NCT04994886; Lausanne University Hospital (CHUV)), HEMO (ClinicalTrials.gov registration: NCT05044923; University of Calgary), HemON (ClinicalTrials.gov registration: NCT05111093; CHUV) and HemON-NL (ClinicalTrials.gov registration: NCT05941819; Sint Maartenskliniek). The HemON trial was amended after four participants to include the new thoracic paddle lead. The STIMO-HEMO (Commission cantonale d‘Ethique de la Recherche sur l’être humain CER-VD 2021-00588; Swissmedic, 10000882) and HemON (CER-VD 2021-D0025; Swissmedic, 10000932) studies were approved by the Swiss ethical authorities. The study sponsor submitted an investigation to Swissmedic and Swiss ethical authorities for regulatory approval before the start of the study. The HEMO study was approved by Canadian ethical authorities (REB21-0027). The HemON-NL (Medisch-Ethische Toetsingscommissie Oost Nederlandse - METC, 2023-16316; Centrale Commissie Mensgebonden Onderzoek - CCMO, NL83694.000.23) study sponsor submitted approval before the start of the study to the ethical authorities in the Netherlands and the Central Committee on Research Involving Human participants. The studies were approved by the Dutch ethical authorities. All four trials and corresponding amendments were conducted following the Declaration of Helsinki. A summary of information and approvals related to each study is provided in Supplementary Table 3.

he Central Committee on Research Involving Human participants. The studies were approved by the Dutch ethical authorities. All four trials and corresponding amendments were conducted following the Declaration of Helsinki. A summary of information and approvals related to each study is provided in Supplementary Table 3. All the participants signed a written informed consent at enrollment in the trial. All participants had the option to indicate consent for the publication of identifiable images or videos. All surgical and experimental procedures were performed at the investigational hospital sites, including the neurosurgery department of Lausanne University Hospital (CHUV) in Lausanne, Switzerland, the Neurosurgery Department of Foothills Medical Center in Calgary, Canada, and the Neurosurgery Department of the Radboud University Medical Center in Nijmegen, the Netherlands. The study involved eligibility and baseline assessments conducted before the surgery, the neurosurgical implantation of an investigational device, a postoperative period during which EES protocols were configured and long-term follow-up periods, including independent use of EES at home. To date, 13 individuals have completed the intensive part of the study. Fourteen individuals who had suffered a traumatic SCI participated in one of the five studies. Demographic data and neurological status were evaluated according to the International Standards for Neurological Classification of Spinal Cord Injury48 and are reported in Supplementary Table 2 and Supplementary Data 8.

All the participants signed a written informed consent at enrollment in the trial. All participants had the option to indicate consent for the publication of identifiable images or videos. All surgical and experimental procedures were performed at the investigational hospital sites, including the neurosurgery department of Lausanne University Hospital (CHUV) in Lausanne, Switzerland, the Neurosurgery Department of Foothills Medical Center in Calgary, Canada, and the Neurosurgery Department of the Radboud University Medical Center in Nijmegen, the Netherlands. The study involved eligibility and baseline assessments conducted before the surgery, the neurosurgical implantation of an investigational device, a postoperative period during which EES protocols were configured and long-term follow-up periods, including independent use of EES at home. To date, 13 individuals have completed the intensive part of the study. Fourteen individuals who had suffered a traumatic SCI participated in one of the five studies. Demographic data and neurological status were evaluated according to the International Standards for Neurological Classification of Spinal Cord Injury48 and are reported in Supplementary Table 2 and Supplementary Data 8. For the STIMO-HEMO and HEMO clinical trials, participants were eligible if they met the following criteria: were aged between 18 and 70 years; had a radiologically confirmed SCI located between C3 and T6 (inclusive); were neurologically classified as A or B on the AIS; had stable medical, physical and psychological conditions under the discretion of the investigators; had the injury for at least 1 year; did not have any required spinal instrumentation within the past 6 months; were able and willing to attend all scheduled appointments; had confirmed orthostatic hypotension and AD. Exclusion criteria included the following: individuals in any emergency situations; participants with any diseases and conditions that would increase the morbidity and mortality of the SCI surgery; participants who were unable to withhold antiplatelet or anticoagulation agents perioperatively; had history of myocardial infarction or cerebrovascular events within the past 6 months; had a current or anticipated need for opiod pain medications; had pain that would prevent full participation in the rehabilitation program in the judgment of the investigators; had a clinically substantial mental illness in the judgment of the investigators; received botulinum toxin injections within 6 months of enrollment; presence of substantial pressure ulcers, recurrent urinary tract infection refractory to antibiotics, unhealed spinal fractures, known or suspected drug or alcohol abuse; presence of indwelling baclofen or insulin pump; and any other conditions that would make the participant unable to participate in testing in the judgment of the investigators. Women who were pregnant during screening or breastfeeding were also excluded.

ealed spinal fractures, known or suspected drug or alcohol abuse; presence of indwelling baclofen or insulin pump; and any other conditions that would make the participant unable to participate in testing in the judgment of the investigators. Women who were pregnant during screening or breastfeeding were also excluded. For the HemON clinical trial, participants were eligible if they met the following criteria: 18 years of age or older; had an SCI located between C3 and T6 (inclusive); had their injury for at least 1 month; had stable medical, physical and psychological conditions as considered by the investigators; spoke the French or English language to interact with the study team; were available to participate in good faith and attend scheduled appointments per study conditions; had confirmed orthostatic hypotension. Exclusion criteria included the following: SCI related to neurodegenerative disease; diseases and conditions that would increase the morbidity and mortality of the surgery; the inability to withhold antiplatelet or anticoagulation agents perioperatively; history of myocardial infarction or cerebrovascular events within the past 6 months; other conditions that would make the participant unable to participate in testing in the judgment of the investigators; clinically substantial mental illness in the judgment of the investigators; having received botulinum toxin vesical and nonvesical injections within 3 months of enrollment; presence of substantial pressure ulcers, recurrent urinary tract infections refractory to antibiotics; presence of indwelling baclofen or insulin pumps; other clinically substantial concomitant disease states including but not limited to renal failure, hepatic dysfunction, cardiovascular disease, etc.; inability to follow study procedures due to language problems, psychological disorders or dementia; participation in another study with investigational drugs within the 30 days preceding and during the present study; and any relation to the investigator, his/her family members, employees and other dependent persons. Women who were pregnant or breastfeeding, lacked safe contraception if with childbearing capacity or had the intention to become pregnant during the course of the study were excluded.

eding and during the present study; and any relation to the investigator, his/her family members, employees and other dependent persons. Women who were pregnant or breastfeeding, lacked safe contraception if with childbearing capacity or had the intention to become pregnant during the course of the study were excluded. For the HemON-NL trial, participants were eligible if they met the following criteria: 18 years of age or older; had an SCI located between C3 and T6 (inclusive); had a traumatic SCI; had their injury for at least 1 month; were neurologically classified as AIS A, B, C or D; had stable medical, physical and psychological conditions as considered by the investigators; spoke the Dutch or English language to interact and understand the study team; were available to participate in good faith and attend scheduled appointments per study conditions; had, if needed, continuous support from personal caregiver in daily life during the study visit (including independent transport); and had confirmed orthostatic hypotension.

age to interact and understand the study team; were available to participate in good faith and attend scheduled appointments per study conditions; had, if needed, continuous support from personal caregiver in daily life during the study visit (including independent transport); and had confirmed orthostatic hypotension. Exclusion criteria include the following: SCI related to diseases and conditions that would increase the morbidity and mortality of the surgery; diseases and conditions that would require regular MRI; the inability to perform an MRI due to metal, magnetic or electrical device in the body (for example, oral implant with magnet, metal splinter, neurostimulator, artificial heart valve, clips and stents) as assessed by the MRI form of Sint Maartenskliniek; the inability to withhold antiplatelet or anticoagulation agents perioperatively; history of myocardial infarction or cerebrovascular events within the past 6 months; other conditions that would make the participant unable to participate in testing in the judgment of the investigators; clinically substantial mental illness in the judgment of the investigators; having received botulinum toxin vesical and nonvesical injections within 3 months of enrollment; presence of substantial pressure ulcers, recurrent urinary tract infections refractory to antibiotics; presence of indwelling baclofen or insulin pumps; other clinically substantial concomitant disease states including but not limited to renal failure, hepatic dysfunction, cardiovascular disease, etc.; inability to follow study procedures due to language problems, psychological disorders or dementia; participation in another study with investigational drugs within the 30 days preceding and during the present study; and any relation to the investigator, his/her family members, employees and other dependent persons. Women who were pregnant or breastfeeding, lacked safe contraception if with childbearing capacity or had the intention to become pregnant during the course of the study were excluded.

eding and during the present study; and any relation to the investigator, his/her family members, employees and other dependent persons. Women who were pregnant or breastfeeding, lacked safe contraception if with childbearing capacity or had the intention to become pregnant during the course of the study were excluded. All the feasibility clinical trials outlined in Supplementary Table 3 evaluated the safety and preliminary efficacy of the investigational system and/or therapy to restore hemodynamic stability. The primary outcome measures assess the occurrences of serious adverse events and adverse events that are deemed related or possibly related to the investigational system or study procedure during the course of studies. All reportable safety events have been reported per local regulation to the respective authorities (Health Research Ethics Board at the University of Calgary, Ethics Committee CER-VD and competent authorities Swissmedic in Switzerland and METC Oost-Nederland in Nijmegen). Data safety monitoring boards have been appointed in all studies and consulted on a yearly basis or ad hoc in case of reportable events. All adverse device effects for the four clinical trials are summarized in Supplementary Tables 4–8. Two serious adverse events related to the device leading to explantation occurred during the course of the HEMO study due to postoperative infections. Device explanations were reported to local authorities and to the manufacturer. Both were resolved without sequelae following explantation surgery. In studies evaluating the safety of the newly implanted device, one reportable serious adverse event related to the study procedure occurred (subpubic catheter came out during a rehabilitation session while the stimulation was turned off).

anufacturer. Both were resolved without sequelae following explantation surgery. In studies evaluating the safety of the newly implanted device, one reportable serious adverse event related to the study procedure occurred (subpubic catheter came out during a rehabilitation session while the stimulation was turned off). All the participants underwent structural 3T MRI and CT imaging of the thoracolumbar spine to enable the elaboration of a personalized anatomical model of the spine and guide the preoperative planning. MRI acquisitions were conducted in a supine position with both arms on the sides. Before image acquisition, shim boxes were applied to correct for magnetic field inhomogeneities. All sequences were performed without gadolinium-based contrast agent administration. In the STIMO-HEMO and HemON clinical trials, MRI was performed on a Magnetom PrismaFit (Siemens Healthineers) with 18-channel body and 32-channel spine array coils. The standard MRI protocol comprised the following four pulse sequences: (1) two-dimensional sagittal T2-weighted turbo spin-echo (repetition time (TR), 3,080 ms; echo time (TE), 98 ms; voxel size, 0.6 × 0.6 × 3 mm3), (2) T2-weighted sampling perfection with application-optimized contrasts using different flip angle evolution (SPACE) sequence (TR, 1,500 ms; TE, 135 ms; interpolated voxel size, 0.4 × 0.4 × 0.8 mm3), (3) three-dimensional (3D) axial T2-weighted SPACE with ZOOMit (dynamic excitation pulses to achieve selective/zoomed field of view) software (TR, 2,500 ms; TE, 106 ms; interpolated voxel size, 0.3 × 0.3 × 0.5 mm3) and (4) 3D coronal T2-weighted true fast imaging with steady-state precession (TR, 6.04 ms; TE, 3.02 ms; interpolated voxel size, 0.3 × 0.3 × 0.6 mm3). The total scan time was less than 35 min overall. The HemON trial MRI protocol consisted of a subset of these sequences (no 3D axial T2-weighted SPACE with ZOOMit as this trial does not include an implant at the lumbosacral level). In the HEMO trial, a similar protocol was performed on a Discovery MR750 (GE Healthcare) with an eight-channel cervical thoracic lumbar coil. The following two sequences were performed: (1) 3D coronal T2-weighted fast imaging using steady-state acquisition (TR, 7.65 ms; TE, 3.48 ms; voxel size, 0.4 × 0.4 × 0.4 mm3) and (2) 3D sagittal T2-weighted CUBE sequence (TR, 2,000 ms; TE, 90 ms; interpolated voxel size, 0.8 × 1.25 × 0.8 mm3). The total scan time was ~84 min overall.

wo sequences were performed: (1) 3D coronal T2-weighted fast imaging using steady-state acquisition (TR, 7.65 ms; TE, 3.48 ms; voxel size, 0.4 × 0.4 × 0.4 mm3) and (2) 3D sagittal T2-weighted CUBE sequence (TR, 2,000 ms; TE, 90 ms; interpolated voxel size, 0.8 × 1.25 × 0.8 mm3). The total scan time was ~84 min overall. Finally, in the HemON-NL trial, the standard protocol was translated to an Ingenia Omega (Philips Healthcare) with a 32-channel coil. The protocol comprised the following two sequences: (1) sagittal T2-weighted fast field echo (FFE; TR, 3.34 ms; TE, 8.1 ms; interpolated voxel size, 0.75 × 0.75 × 3 mm3) and (2) coronal balanced FFE (TR, 7.0 ms; TE, 3.5 ms; interpolated voxel size, 0.75 × 0.75 × 3 mm3). The total duration of the acquisition was ~20 min overall.

col comprised the following two sequences: (1) sagittal T2-weighted fast field echo (FFE; TR, 3.34 ms; TE, 8.1 ms; interpolated voxel size, 0.75 × 0.75 × 3 mm3) and (2) coronal balanced FFE (TR, 7.0 ms; TE, 3.5 ms; interpolated voxel size, 0.75 × 0.75 × 3 mm3). The total duration of the acquisition was ~20 min overall. CT scans were acquired in a supine position with both arms on the sides. No contrast agents were administered. In the STIMO-HEMO and HemON clinical trials, CT of the trunk, including the pelvis, was performed on a 256-detector row CT system (Revolution Apex; GE Healthcare) using the following acquisition parameters: tube potential, 120 kVp; tube current, 115–200 mA, by enabling automatic tube current modulation; gantry revolution time, 0.35 s; beam collimation, 128 × 0.625 mm; pitch, 0.508. CT images were reconstructed with a field of view of ~150 × 150 and ~400 × 400 mm2 for the spine and pelvis, respectively, at a section thickness/interval of 1.25/0.625 and 1.25/1.25 mm, respectively, with both smooth (standard) and sharp (bone plus) convolution kernels and a deep learning image reconstruction algorithm (TrueFidelity; GE Healthcare; medium strength level) for image noise reduction. The nominal voxel size was ~0.29 × 0.29 × 0.625 and ~0.78 × 0.78 × 1.25 mm3 for the spine and pelvis, respectively. In the HEMO trial, a revolution discovery scanner (GE Healthcare) was used to collect standard and bone plus reconstructions (pitch, 0.531:1; slice thickness, 1.25 mm; interval, 1.25, 120 kV, 115–200 mA; rotation time, 0.4 s) of the thoracolumbar spine. In the HemON trial, an Ingenuity Core 128 scanner (Philips) was used to collect standard iDose and IMR reconstructions.

Healthcare) was used to collect standard and bone plus reconstructions (pitch, 0.531:1; slice thickness, 1.25 mm; interval, 1.25, 120 kV, 115–200 mA; rotation time, 0.4 s) of the thoracolumbar spine. In the HemON trial, an Ingenuity Core 128 scanner (Philips) was used to collect standard iDose and IMR reconstructions. We trained two different nnU-net networks49 on a labeled dataset of 15 healthy participants from a previous study29. The first network was trained on high-resolution MRI to segment the cerebrospinal fluid, white matter and spinal root tissues. The second network was trained on low-resolution MRI data to segment vertebral bodies, intervertebral disks and the spinal canal. We trained the networks on 12 participants and tested them on 3 participants. We achieved high dice scores for each tissue (cerebrospinal fluid, 95.50; white matter, 90.55; roots, 71.88; vertebrae, 89.90; intervertebral disks, 91.22). When applying the networks to a new dataset, we corrected the segmentation manually if needed. The segmentation of the roots was used as guidance to trace each spinal root individually. The rostral location of each root was constrained within the white matter. For each spinal quadrant (dorsal/ventral left/right), the end of root within the white matter points was cut 1 mm within the white matter, and a spline was made with them. These end-of-root lines were then guidelines to force rootlets to attach to them, forcing them to end within the white matter. The procedure to generate the rootlets was described previously29. By forcing each rootlet to end in the end-of-roots guidelines, we made sure the rootlets followed the shape of the spine, thereby increasing robustness to scoliosis. We cut the white matter with respect to the rootlet divisions and obtained, henceforth, the spinal levels.

nerate the rootlets was described previously29. By forcing each rootlet to end in the end-of-roots guidelines, we made sure the rootlets followed the shape of the spine, thereby increasing robustness to scoliosis. We cut the white matter with respect to the rootlet divisions and obtained, henceforth, the spinal levels. We used an anatomically informed framework trained on the VerSE datasets50 to segment the different vertebrae on CT imaging51. The authors of the framework showed a dice score of 91.04 on a large test set. Having both segmentations from MRI and CT spaces, we automatically placed labels on each vertebrae segmentation on MRI data. We then coregistered the CT data to the MRI space automatically using the Spinal Cord Toolbox52. A computer-aided design model of the 16 electrodes composing the paddle lead was positioned over the targeted dorsal root entry zones. Several visualizations of the 3D personalized anatomical model were then generated and shared with the clinical and neurosurgical team to guide preoperative surgical planning. Finally, a virtual paddle lead was placed over the selected location of the spinal cord and then projected onto a large screen in the operation room to guide the neurosurgeon in the insertion of the lead.

el were then generated and shared with the clinical and neurosurgical team to guide preoperative surgical planning. Finally, a virtual paddle lead was placed over the selected location of the spinal cord and then projected onto a large screen in the operation room to guide the neurosurgeon in the insertion of the lead. The participants were put under protocolized propofol-based total intravenous general anesthesia and were placed in a prone position. Preoperative surgical planning informed the neurosurgeon about the vertebral entry level and predicted the optimal paddle lead position. Based on this knowledge, lateral and anteroposterior fluoroscopy X-rays were performed intraoperatively to identify the location of the laminotomies. A midline skin incision of ~5 cm on the back was performed, the fascia opened and the muscles retracted bilaterally. Excision of the midline ligamentous structures and a laminotomy at the desired entry level enabled the insertion of the paddle lead below the vertebra. For participants of the STIMO-HEMO and HEMO trials, a second skin incision or extended opening caudally was made, and a second laminotomy was performed in the lumbar region based on the preoperative planning to allow for the insertion of the paddle lead over the lumbosacral spinal cord. The paddle lead(s) (SureScan Specify 5-6-5 (Medtronic) or ARCIM Thoracic Lead (ONWARD Medical NV)) are inserted and placed over the midline of the dura mater and advanced rostrally to the target position under repeated fluoroscopies. Electrophysiological recordings were acquired using standard neuromonitoring systems (Cascade Elite (Cadwell Industries) or ISIS Xpress (Inomed Medizintechnik)). Single pulses of EES were delivered at 0.5 Hz and increasing amplitude to elicit muscle responses that were recorded from subdermal (Neuroline Twisted Pair Subdermal, 12 × 0.4 mm; Ambu A/S) or intramuscular needle electrodes (Inomed SDN electrodes, 40 × 0.45 mm; Inomed Medizintechnik). The symmetry of the responses and the expected rostrocaudal distribution of muscle responses informed the corrections for lateral and rostrocaudal positioning. When the paddles deviated from a straight midline position, small additional laminotomies were made to remove bony protrusions and guide the paddle to a midline placement. Once the final position was determined, the paddle lead(s) was (were) anchored to the muscular fascia.

ons for lateral and rostrocaudal positioning. When the paddles deviated from a straight midline position, small additional laminotomies were made to remove bony protrusions and guide the paddle to a midline placement. Once the final position was determined, the paddle lead(s) was (were) anchored to the muscular fascia. In the STIMO-HEMO, HemON and HemON-NL trials, the back opening was temporarily closed, and the participants were put in lateral decubitus. Subsequently, the back incision was reopened, and an abdominal incision of about 5 cm was made to implant each pulse generator in a subcutaneous pocket. In the HEMO trial, incisions of about 5 cm were made bilaterally in the upper buttocks region, and subcutaneous pockets were created. The cables of the paddle lead were then tunneled between the back opening and subcutaneous pockets to be connected to the implantable pulse generator (IPG; Intellis, Medtronic or ARCIM IPG (ONWARD Medical NV)). The IPG(s) was (were) implanted in the subcutaneous pocket(s), and all incisions were finally closed. In the STIMO-HEMO trial, both paddle lead cables were tunneled from the same flank side, and tunneling was performed between the two abdominal subcutaneous pockets.

ntellis, Medtronic or ARCIM IPG (ONWARD Medical NV)). The IPG(s) was (were) implanted in the subcutaneous pocket(s), and all incisions were finally closed. In the STIMO-HEMO trial, both paddle lead cables were tunneled from the same flank side, and tunneling was performed between the two abdominal subcutaneous pockets. The configuration of the electrodes for EESXX programs was guided by the understanding of the mechanisms through which EES regulates blood pressure (Fig. 2a). This configuration involved three steps. First, an intraoperative mapping was conducted to identify the relevant rows of electrodes to target the hemodynamic hotspot1 and to elicit the largest pressor response. Second, postoperative imaging was used to update the anatomical models and thus estimate the location of the electrodes that maximized recruitment of the dorsal root entry zones projecting to the hemodynamic hotspot. Third, a single postoperative session was conducted to quantify pressor responses elicited by each row of electrodes. Both monopolar and multipolar configurations of electrodes were tested to ensure the maximal activation of the hemodynamic hotspot. Stimulation frequency was set at 120 Hz because we established that this frequency maximized pressor responses1. Multiple electrode configurations targeted a specific region of the hemodynamic hotspot. Each electrode configuration was delivered with a 2 ms delay between each pulse, forming a traveling wave over the spinal cord. In the STIMO-HEMO and HEMO trials, pulse width was mapped between 100 μs and 500 μs. Five of the six participants had a pulse width of 300 μs, while one participant (P3) had a pulse width of 500 μs. For the HemON and HemON-NL trials, the pulse width was set at 300 μs. The amplitude was set based on the increase in the current delivered through each electrode configuration until the systolic pressure increased by 20 mm Hg, the diastolic pressure increased by 10 mm Hg or the participant reported any discomfort such as muscle contractions or unpleasant sensations such as strong tingling. To validate the combined effect of multiple electrode configurations, we compared the amplitude of the pressor responses when EES was delivered with a configuration targeting a single level of the hemodynamic hotspot versus all the levels embedded in the hemodynamic hotspot. All other stimulation parameters were maintained unchanged.

combined effect of multiple electrode configurations, we compared the amplitude of the pressor responses when EES was delivered with a configuration targeting a single level of the hemodynamic hotspot versus all the levels embedded in the hemodynamic hotspot. All other stimulation parameters were maintained unchanged. Intraoperative beat-to-beat hemodynamic data were obtained from a pressure transducer connected to an arterial line cannula placed in the distal radial artery. The arterial line system was zeroed to atmospheric pressure. Arterial cannulation was performed either by palpation or with ultrasound guidance under sterile conditions. Adjustments in the anesthetic depth, as well as the administration of vasoactive medications or substantial volumes of intravenous fluids, were avoided during hotspot testing. All data were recorded using VitalRecorder software (VitalDB; Department of Anesthesiology and Pain Medicine, Seoul National University College of Medicine) at a sampling frequency of 125 Hz. Alternatively, the arterial line was recorded with LabChart (ADInstruments).

venous fluids, were avoided during hotspot testing. All data were recorded using VitalRecorder software (VitalDB; Department of Anesthesiology and Pain Medicine, Seoul National University College of Medicine) at a sampling frequency of 125 Hz. Alternatively, the arterial line was recorded with LabChart (ADInstruments). Beat-to-beat blood pressure and heart rate were acquired using finger plethysmography (Finometer; Finapres Medical Systems). Beat-by-beat blood pressure was calibrated to brachial artery blood pressure collected using an arm cuff embedded and synchronized with the Finometer53–57. Brachial arterial pressure was sampled at 200 Hz, while the systolic, diastolic and mean arterial pressures were extracted from the calibrated arterial pressure at 1 Hz. The heart rate was also sampled at 1 Hz. Raw data and automatically extracted hemodynamic parameters were saved and exported from the Finometer.

3–57. Brachial arterial pressure was sampled at 200 Hz, while the systolic, diastolic and mean arterial pressures were extracted from the calibrated arterial pressure at 1 Hz. The heart rate was also sampled at 1 Hz. Raw data and automatically extracted hemodynamic parameters were saved and exported from the Finometer. The investigational system used in the STIMO-HEMO and HEMO clinical trials consisted of a set of Certificat Européen-marked, Food and Drug Administration-approved medical devices used off-label. Two IPGs (Intellis with AdaptiveStim; Medtronic) were connected to a paddle lead (Specify 5-6-5 SureScan MRI; Medtronic). These implants are indicated for chronic pain management. A tablet application with a wireless communicator device (Intellis Clinician Programmer; Medtronic) was used by the clinical team to set up the system and optimize the stimulation parameters. A remote control device and transcutaneous charger device (Patient Programmer and Recharger; Medtronic) were used by the participants to charge the implanted pulse generators and to turn EES programs on and off during their daily life, as well as to adapt parameters of EES programs defined by the clinical team.

meters. A remote control device and transcutaneous charger device (Patient Programmer and Recharger; Medtronic) were used by the participants to charge the implanted pulse generators and to turn EES programs on and off during their daily life, as well as to adapt parameters of EES programs defined by the clinical team. The investigational systems used in the HemON and HemON-NL clinical trials were designed and built for treating hemodynamic instability based on the understanding of the mechanisms through which EES regulates blood pressure. A stepwise approach was followed. All participants were implanted with a purpose-built IPG ARCIM (ONWARD Medical NV) that communicates with a purpose-built ecosystem of control devices. The first four participants of the HemON clinical trial were implanted with the off-label paddle lead (Specify 5-6-5 SureScan MRI; Medtronic) as the participants of the STIMO-HEMO and HEMO clinical trials, whereas all other participants were implanted with a newly designed, purpose-built paddle lead (ARC Thoracic Lead; ONWARD Medical NV).

cipants of the HemON clinical trial were implanted with the off-label paddle lead (Specify 5-6-5 SureScan MRI; Medtronic) as the participants of the STIMO-HEMO and HEMO clinical trials, whereas all other participants were implanted with a newly designed, purpose-built paddle lead (ARC Thoracic Lead; ONWARD Medical NV). The purpose-built ARCIM IPG is a new 16-channel neurostimulation platform that was specifically developed to regulate neurological functions with EES. The ARCIM IPG controls and delivers current-controlled stimulation pulses that can be predefined by the user or controlled in real time. The system consists of a hermetically sealed, biocompatible can that surrounds the electrical components and a rechargeable battery that enables the delivery of electrical current. The external envelope is composed of the following two main components: the header containing the connector block that enables connection with two 8-contact lead connectors and two coils for charging and communication, and the can with a rechargeable battery and electronics circuits. The ARCIM IPG was developed according to all applicable standards for medical device development. Conventional biomedical technologies were used to fabricate the system, and extensive bench and in vivo testing was performed to verify its performance.

on, and the can with a rechargeable battery and electronics circuits. The ARCIM IPG was developed according to all applicable standards for medical device development. Conventional biomedical technologies were used to fabricate the system, and extensive bench and in vivo testing was performed to verify its performance. The IPG was implanted subcutaneously in a pocket created in the abdomen. The ARCIM communication hub ensures the exchange of information with the implanted pulse generator using NFMI and also serves as a wireless charging station for the battery embedded in the implanted pulse generator. The communication hub is worn on a belt over or in proximity to the location of the implanted pulse generator. This hub hosts a Bluetooth low-energy chip that enables fast, reliable wireless communication with external sensors or programmers. This versatility enables the acquisition of varying control signals and rapid exchange of information between the communication hub and the implanted pulse generator, which together establish the conditions for closed-loop control of stimulation parameters with latencies as low as 25 ms between the generation of the stimulation command and the execution of this command. The hub communicates with external programmers such as the ARCIM Clinician Programmer, onto which an Android app designed for clinicians allows the configuration of EES programs and evaluation of the system. When EES programs are deemed safe for personal use, the Clinician Programmer provides the possibility to make this stimulation program available to the participant.

as the ARCIM Clinician Programmer, onto which an Android app designed for clinicians allows the configuration of EES programs and evaluation of the system. When EES programs are deemed safe for personal use, the Clinician Programmer provides the possibility to make this stimulation program available to the participant. The participants, or their caregivers, can control the system through the ARCIM Personal Programmer. This Android Watch application allows users to select, start and stop EES programs, as well as to tune stimulation amplitudes within predefined safety limits determined by the clinicians. Device errors, paddle lead impedances and daily stimulation usage were extracted from usage logs across all devices. Furthermore, the Clinician Programmer includes an application programming interface, termed the ARCIM API, that enables other programming software to control the stimulation, for example, for closed-loop control of the stimulation. All devices and software are adherent to the applicable standards, and their performance was extensively tested. The entire system received the equivalent of an investigational device exemption from the competent Swiss authorities.

Furthermore, the Clinician Programmer includes an application programming interface, termed the ARCIM API, that enables other programming software to control the stimulation, for example, for closed-loop control of the stimulation. All devices and software are adherent to the applicable standards, and their performance was extensively tested. The entire system received the equivalent of an investigational device exemption from the competent Swiss authorities. The ARCIM Thoracic Lead developed by ONWARD Medical NV is a new 16-electrode paddle lead that incorporates a configuration of electrodes optimized to achieve the selective recruitment of the dorsal root entry zones projecting to the last three segments of the lower thoracic spinal cord. The configuration of the electrodes was determined based on a digital library of human thoracic spinal cords that we elaborated from CT and MRI acquisitions (Extended Data Fig. 2). The average length and width of spinal segments were calculated from our library and existing literature29,33,36–40. The combination of these two datasets generated global averages and s.d. for each level of the lower thoracic spinal cord, which determined the total length of the hemodynamic hotspot. We created a normal distribution with the calculated global means and s.d. and found that a length of 64.2 mm accurately represents the combined length of the three targeted spinal segments for over 95% of the population. This new paddle lead was designed following all applicable standards for medical implant development and fabricated using conventional biomedical technologies. Extensive bench and in vivo testing was conducted to validate the mechanical, electrical and biocompatibility robustness of the paddled lead. The equivalent of an investigational device exemption was granted by the competent Swiss authorities.

plant development and fabricated using conventional biomedical technologies. Extensive bench and in vivo testing was conducted to validate the mechanical, electrical and biocompatibility robustness of the paddled lead. The equivalent of an investigational device exemption was granted by the competent Swiss authorities. The purpose-built investigational system was embedded in a framework for closed-loop control of systolic blood pressure based on real-time monitoring of hemodynamic parameters. For this purpose, changes in systolic blood pressure were measured with the Finometer (‘Postoperative hemodynamic monitoring’). This signal was acquired using an analog-to-digital converter (Delsys Trigno Avanti Analog; Delsys) that wirelessly streamed the data to a sensor base station (Delsys Trigno Research+; Delsys). The 1 Hz systolic blood pressure measured with the Finometer was upsampled to 1,257 Hz by the Trigno system, which was connected through USB to a Windows computer onto which a medical-grade software developed by Ecole Polytechnique Fédérale de Lausanne (EPFL) was running. This software29 enables closed-loop control of the ARCIM system through the ARCIM API. The specific closed-loop algorithm used for the control of systolic blood pressure was implemented in a multithreaded Python application that interfaces with GDrive+ software via TCP/UDP communication, enabling the reception of systolic blood pressure signals and delivery of stimulation updates.

through the ARCIM API. The specific closed-loop algorithm used for the control of systolic blood pressure was implemented in a multithreaded Python application that interfaces with GDrive+ software via TCP/UDP communication, enabling the reception of systolic blood pressure signals and delivery of stimulation updates. The systolic blood pressure signal was smoothed with an exponentially weighted moving average over a 3-s window with a forgetting factor of 0.3. This elaborated signal was sent as an input to a proportional–integral–derivative controller. To minimize steady-state errors and minimize system saturation, the integral term was penalized over time when the stimulation remained unchanged at the minimum or maximum amplitude of EES. Additionally, a forgetting factor on the integral term reduces error accumulation. A minimum (0 mA) and maximum amplitude (participant dependent) amplitude range was set to ensure the comfort and safety of the participant. The controller target was defined based on average blood pressure measurements from 10-min tilt-table tests with EES (‘Postoperative hemodynamic monitoring’). A target range of ±3 mm Hg was set around the user-defined target to account for intrinsic blood pressure variability. If the change in EES amplitude was greater than 0.5 mA, the controller incrementally ramped in steps of 500 ms per 0.5 mA for additional comfort. To calibrate the closed-loop control of EES, we found a linear operating range of stimulation amplitudes and systolic blood pressure. The minimum was the amplitude at which the first increase in systolic blood pressure occurred, and the maximum was the last amplitude at which the systolic blood pressure stopped increasing or when the participant reported abdominal muscle contractions. Each participant remained in a seated position as the amplitude of EES increased by 1 mA every 1 min to find the minimum (0 mA) and maximum parameters of stimulation (participant dependent). A regression between the average systolic blood pressure per stimulation amplitude resulted in a linear relationship between the systolic blood pressure and the amplitude of stimulation. The closed loop was tuned manually using the heuristic Ziegler–Nichols method58. The initial proportional constant was the slope of the regression derived from the calibration phase. The proportional, integral and derivative constants were tuned over one 70° tilt.

stolic blood pressure and the amplitude of stimulation. The closed loop was tuned manually using the heuristic Ziegler–Nichols method58. The initial proportional constant was the slope of the regression derived from the calibration phase. The proportional, integral and derivative constants were tuned over one 70° tilt. The same parameters were then used across days to test dynamic changes in the severity of the orthostatic challenge. The dynamic changes in tilts consisted of a pseudo-random sequence of eight changes in tilt angle spanning mild (20°) to severe orthostatic challenge (80°). Each tilt position was held for 60–120 s. The same sequence was run with continuous EES to enable comparison. An inertial measurement unit-based control algorithm was also developed using the same framework. The angle of the tilt table was derived from changes in 3D acceleration of an inertial measurement unit placed onto the tilt table (Delsys Trigno Avanti; Delsys). Data from the inertial measurement unit were recorded at 148 Hz. Accelerometer values in three dimensions were used to compute angular changes. Ranges of tilt positions were mapped to supine, seated and standing positions, as well as corresponding amplitudes of EES.

d onto the tilt table (Delsys Trigno Avanti; Delsys). Data from the inertial measurement unit were recorded at 148 Hz. Accelerometer values in three dimensions were used to compute angular changes. Ranges of tilt positions were mapped to supine, seated and standing positions, as well as corresponding amplitudes of EES. Numerical simulations were conducted using the computational life sciences platform Sim4Life (ZMT Zurich MedTech AG). First, a canonical spinal cord model was developed based on average anatomical measurements of the participants’ spinal cord anatomies. It was then used to simulate EES exposure throughout the spinal cord anatomy, as well as induce electrophysiological responses29. The dosimetric simulations involved finite element discretization using Sim4Life, which resulted in ~5 million volume elements (voxels). Each voxel was attributed dielectric properties corresponding to the assigned tissue types59 (that is, spinal cord, spinal roots, cerebrospinal fluid, epidural fat and vertebral bone). Anisotropic conductivity maps were assigned to the white matter and rootlets, leveraging existing Sim4Life functionalities (either diffusion–simulation based or DTI imaging based). By precomputing a suitable basis of electric potential distributions (one per epidural stimulation electrode contact) for the canonical bioelectrical spinal model, the electrical fields resulting from arbitrary EES pulses can rapidly be obtained29. For that purpose, the ohmic-current-dominated electro-quasistatic approximation of Maxwell’s equations was solved60. In addition to providing an average anatomical geometry, the spine model also encompassed trajectories of axons entering the spinal cord through the dorsal roots and ascending in the dorsal columns. Nerve fiber trajectories were generated inside the rootlets using Sim4Life’s IMSafe tool, while the fibers in the dorsal column were uniformly distributed. Sim4Life includes functionality to estimate fiber recruitment using the generalized, Green’s function-based activating function or detailed electrophysiological modeling. For the present work, the classic activating function (proportional to the discrete second derivative of the extracellular potential60) was used to assess relative stimulability for a given pulse shape, while electrophysiological simulations with the NEURON simulator incorporated in Sim4Life61 permitted conversion to absolute thresholds.

k, the classic activating function (proportional to the discrete second derivative of the extracellular potential60) was used to assess relative stimulability for a given pulse shape, while electrophysiological simulations with the NEURON simulator incorporated in Sim4Life61 permitted conversion to absolute thresholds. For the latter, the predefined diameter-dependent spatially extended nonlinear node fiber model underlying low-frequency exposure safety standards was used. By assigning fiber diameters according to histology-based probability distributions, the percentage of recruited fibers per root could be estimated, and the recruitment level across the different spinal segments and within the dorsal column could be predicted. Activation levels were normalized relative to maximal activation (that is, recruitment of all dorsal root fibers). Various methodologies have been developed to optimize selective stimulation, including neural network backpropagation and hybrid surrogate modeling–multiobjective genetic algorithm optimization. Electromyographic activity of selected muscles was acquired at a 1,259 kHz sample rate using the 16-channel wireless Delsys Trigno sensors (Delsys) with bipolar surface electrodes. Skin hairs were shaved, and abrasive gel was used (Nuprep gel) on the area of interest.

For the latter, the predefined diameter-dependent spatially extended nonlinear node fiber model underlying low-frequency exposure safety standards was used. By assigning fiber diameters according to histology-based probability distributions, the percentage of recruited fibers per root could be estimated, and the recruitment level across the different spinal segments and within the dorsal column could be predicted. Activation levels were normalized relative to maximal activation (that is, recruitment of all dorsal root fibers). Various methodologies have been developed to optimize selective stimulation, including neural network backpropagation and hybrid surrogate modeling–multiobjective genetic algorithm optimization. Electromyographic activity of selected muscles was acquired at a 1,259 kHz sample rate using the 16-channel wireless Delsys Trigno sensors (Delsys) with bipolar surface electrodes. Skin hairs were shaved, and abrasive gel was used (Nuprep gel) on the area of interest. Kinematics were captured by Physilog 5 sensors (Mindmaze) that were placed on the vertebral bodies of C3, T1, T6, T10, L1 and S1 of the participants. Each sensor is a stand-alone 10° of freedom. The micro-electro-mechanical system inertial measurement unit includes 3D accelerometer, 3D gyroscope, 3D magnetometer and a barometric pressure sensor at a sampling frequency of 256 Hz. These inertial measurement units were synchronized via a proprietary radiofrequency protocol.

sensor is a stand-alone 10° of freedom. The micro-electro-mechanical system inertial measurement unit includes 3D accelerometer, 3D gyroscope, 3D magnetometer and a barometric pressure sensor at a sampling frequency of 256 Hz. These inertial measurement units were synchronized via a proprietary radiofrequency protocol. Participants were transferred to a supine position on a table capable of head-up tilt. We applied restraint straps to secure the participant on the knees, on the hips and on the chest, with the feet stabilized. Resting supine blood pressure was recorded continuously for ~5 to 10 min to establish baseline values. Following this period, we tilted the participants upright up to a maximum of 70° while recording hemodynamic values and symptoms of orthostatic tolerance. The time to reach the desired tilt angle from supine was achieved in less than 45 s. Participants were tilted until reaching their tolerance threshold or for a maximum duration of 10 min. They were asked not to talk during the test except to inform of and grade symptoms. Participants were asked to report any symptoms every 1–3 min. Participants were asked to rank the intensity of their symptoms between 1 and 10, 1 being no symptoms at all and 10 being maximum feelings of dizziness, lightheadedness62 or nausea26,62. Participants were instructed to notify the research team if they needed to be returned to the supine position.

ymptoms every 1–3 min. Participants were asked to rank the intensity of their symptoms between 1 and 10, 1 being no symptoms at all and 10 being maximum feelings of dizziness, lightheadedness62 or nausea26,62. Participants were instructed to notify the research team if they needed to be returned to the supine position. Blood samples were withdrawn at baseline after a period of rest in the supine position for 15 min and at 2 min (T2) after the onset of the tilt. During baseline testing, a total of three samples of blood were withdrawn. When testing with EES turned on, an additional sample of blood was withdrawn 5 min after the onset of EES in a supine position. Blood samples allowed quantification of plasma catecholamines, including plasma norepinephrine, plasma epinephrine and plasma dopamine. To avoid false positives, the participants were asked to abstain from any β-blocker, dihydropyridine calcium channel blocker, phenoxybenzamine, anxiolytics or decongestant medication. To ensure sample quality, before and immediately after, collection tubes were kept cold via immersion in crushed ice and kept in foil to reduce exposure to light. Samples were analyzed by the local hospital lab services at the University Hospital of Lausanne (Switzerland) and the Foothills Medical Center (Canada). Neuropathic pain prevented blood sampling in P8. Data and tests were not available at the time for P13 and were not part of the clinical protocol for P14.

re to light. Samples were analyzed by the local hospital lab services at the University Hospital of Lausanne (Switzerland) and the Foothills Medical Center (Canada). Neuropathic pain prevented blood sampling in P8. Data and tests were not available at the time for P13 and were not part of the clinical protocol for P14. Blood flow was measured in the seated position using Doppler ultrasound imaging equipment (Epiq Elite; Philips). Measurements were conducted both with and without EES. The left and right proximal internal carotid arteries (ICA) and vertebral arteries (VA) were assessed using a 12-3 MHz broadband linear array probe (L12–3; Philips). Additionally, the right and left middle cerebral arteries were examined using a transcranial 5-1 MHz-pulsed TransCranial Doppler (TCD) probe.

without EES. The left and right proximal internal carotid arteries (ICA) and vertebral arteries (VA) were assessed using a 12-3 MHz broadband linear array probe (L12–3; Philips). Additionally, the right and left middle cerebral arteries were examined using a transcranial 5-1 MHz-pulsed TransCranial Doppler (TCD) probe. On two separate days at the same study timepoint, participants received an identical meal with equal caloric content, and their blood pressure was monitored via an arm cuff before, during and for 30 min to 1 h following food intake. This assessment was performed 1 day without EES and 1 day with EESBP. For the assessment with EESBP, the stimulation was turned on during and after the meal. All points before starting the meal were averaged to determine a baseline value. Available cuff measures between 30 min and 60 min were averaged to quantify postprandial measures. The difference (Δ) between baseline and postprandial averages was reported for the conditions without and with EESBP. Only participants who reported experiencing postprandial hypotension were considered in the analysis (P3, P4, P5, P7 and P10). Patient-reported outcome measure questionnaires were sent to study participants via the electronic data capture system. Baseline questionnaires were sent once before surgery and then repeated every month after surgery. The following questionnaires are reported: Autonomic Dysreflexia Following SCI questionnaire, Medical Outcomes Study Sleep Scale (MOS-S)63, Neurogenic Bowel Dysfunction Score (NBDS) and Usability SUS (System Usability Scale)42.

ne questionnaires were sent once before surgery and then repeated every month after surgery. The following questionnaires are reported: Autonomic Dysreflexia Following SCI questionnaire, Medical Outcomes Study Sleep Scale (MOS-S)63, Neurogenic Bowel Dysfunction Score (NBDS) and Usability SUS (System Usability Scale)42. The study participants were asked to perform a systematic set of breathing and cough tasks. Volume, flow and time data were recorded using the spirometer (Spirobank II Smart Spirometer). During the coughing maneuver, EESCoughing was precisely timed with the expiratory phase after releasing the glottis blockage. All participants from the HemON study performed the respiratory and coughing assessments at 6 months without stimulation and with EESTrunk, except P12 who could not perform the test at this timepoint. The COPM was used to assess the functional performance and perceived satisfaction with activities of daily living of each participant. The COPM was administered by trained occupational therapists to follow standard procedures. At baseline, participants identified up to five activities they perceived as important but difficult to perform and rated their perceived performance and satisfaction on a ten-point scale. The general scores were calculated by simply averaging the performance and satisfaction ratings across activities. The same activities were then rated again by the participants three and 6 months after implantation.

nt but difficult to perform and rated their perceived performance and satisfaction on a ten-point scale. The general scores were calculated by simply averaging the performance and satisfaction ratings across activities. The same activities were then rated again by the participants three and 6 months after implantation. Participants’ upper and lower limb spasticity was assessed using the modified Ashworth scale in the following two conditions: without and with EESSpasticity. Participants were either asked to use their standard-of-care medication or not to use their standard-of-care medication the morning before the assessment. Participants were asked to keep the stimulation off in the mornings before the assessments, which were conducted in the late morning. In the HemON trial, physiotherapists and occupational therapists kept track of the 1-month rehabilitation participants’ sessions and documented whether the different activities were performed without or with EES through a standardized form (Mapping of Rehab Training).

Participants’ upper and lower limb spasticity was assessed using the modified Ashworth scale in the following two conditions: without and with EESSpasticity. Participants were either asked to use their standard-of-care medication or not to use their standard-of-care medication the morning before the assessment. Participants were asked to keep the stimulation off in the mornings before the assessments, which were conducted in the late morning. In the HemON trial, physiotherapists and occupational therapists kept track of the 1-month rehabilitation participants’ sessions and documented whether the different activities were performed without or with EES through a standardized form (Mapping of Rehab Training). In the HEMO trial, efferent postganglionic MSNA was acquired from the left common peroneal (fibular) nerve using microneurography1,64,65. The location of the common fibular nerve was located using palpation. A 2-MΩ tungsten microelectrode (Frederick Haer and Co) was then inserted percutaneously alongside a subdermal low-impedance reference electrode ~2 cm adjacent. The amplifier and head stage (NeuroAmpEx; ADInstruments) were also grounded to the participant with a surface electrode on the patella. The recording electrode was adjusted until entering the nerve. Crossing into the nerve and into different nerve bundles was identified audibly by the microneurographer. Once in the nerve, the recording microelectrode was confirmed to be near nerve fibers directed toward skeletal muscle by auditory feedback during tapping/palpation of the tibialis anterior/peroneal muscles, which has been shown previously to be indicative of microelectrode proximity to efferent postganglionic muscle sympathetic nerves. If auditory feedback was present during light stroking of skin on the dorsal foot/lower leg, this indicated proximity to postganglionic skin sympathetic fibers. The search was stopped when the electrode was in proximity to muscle sympathetic fibers.

ctrode proximity to efferent postganglionic muscle sympathetic nerves. If auditory feedback was present during light stroking of skin on the dorsal foot/lower leg, this indicated proximity to postganglionic skin sympathetic fibers. The search was stopped when the electrode was in proximity to muscle sympathetic fibers. During surgery (‘Neurosurgical intervention’), changes in blood pressure were recorded in response to different locations of EES using the arterial line (‘Postoperative hemodynamic monitoring’). Change in blood pressure or heart rate was defined as the difference in the average of 30-s windows before the start of the stimulation and at the end of the stimulation. For cross-subject analysis, the lower thoracic and lumbosacral stimulation locations in each participant were calculated by their relative distance to the center of the hemodynamic hotspot (T11) normalized by their respective spinal length (measured from T9 to the conus). Due to surgery time limits, P3 blood pressure mapping was not performed.

s, the lower thoracic and lumbosacral stimulation locations in each participant were calculated by their relative distance to the center of the hemodynamic hotspot (T11) normalized by their respective spinal length (measured from T9 to the conus). Due to surgery time limits, P3 blood pressure mapping was not performed. During a tilt-table test (‘Orthostatic challenge with tilt-table test’), changes in blood pressure were recorded without stimulation or in response to different types of stimulation (continuous or closed-loop stimulation) using the Finometer (‘Postoperative hemodynamic monitoring’). Change in blood pressure or heart rate was defined as the difference in the average of a 60-s window before the start of the tilt and a 20-s window at 3 min of the challenge. If the participant could not tolerate at least 3 min of the test due to low blood pressure or other symptoms, an average of a 20-s window before the end of the tilt was used. All measurements in a seated position were measured with stimulation on for 3–5 min. Change in blood pressure or heart rate was defined as the difference in the average of a 20-s window before the start of EES and the average of a 20-s window at 3 min before stop stimulation. All signals were smoothed over a 10-s window for illustration. The same processing was used for postoperative, day 1 quantification. Data from continuous blood pressure monitoring were excluded if technical issues with the Finometer were observed, including loss of the plethysmographic signal, incorrect calibration as measured by comparing continuous measuring to standard measurements from a brachial cuff, spasms or due to study protocol deviations or adverse events.

us blood pressure monitoring were excluded if technical issues with the Finometer were observed, including loss of the plethysmographic signal, incorrect calibration as measured by comparing continuous measuring to standard measurements from a brachial cuff, spasms or due to study protocol deviations or adverse events. The efficacy of continuous, closed loop and inertial measurement unit-based control of EES was measured by (1) target error, defined as the error of systolic blood pressure with respect to a user-defined target on the systolic blood pressure, and (2) error variability, defined as the s.d. of the target error, to measure stability of the stimulation during dynamic orthostatic challenges. Both metrics were calculated from 1 min before the dynamic challenges to 1 min after the end of the challenge.

spect to a user-defined target on the systolic blood pressure, and (2) error variability, defined as the s.d. of the target error, to measure stability of the stimulation during dynamic orthostatic challenges. Both metrics were calculated from 1 min before the dynamic challenges to 1 min after the end of the challenge. Peak systolic and end-diastolic flow velocities were obtained from the Doppler spectrum analysis over three consecutive cycles in all arteries. Mean velocity was derived through the integration of the Doppler curve for each corresponding pair, with the mean value calculated across the three cycles. Arterial diameters were determined in B-mode for the left and right ICAs and VAs. The flow rate (Q) in each vessel, measured in milliliters per minute, was calculated using equation (1), assuming a parabolic velocity profile. The total experimental CBF (CBFexp) was determined by summing the flow of the left and right ICAs and VAs, according to equation (2)66.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q=\frac{{\mathrm{FV}}}{2}\times \frac{\pi {\bar{D}}^{2}}{4}$$\end{document}Q=FV2×πD¯242\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathrm{CBF}}}_{\exp }={Q}_{{\mathrm{ICAs}}}+{Q}_{{\mathrm{VAs}}}.$$\end{document}CBFexp=QICAs+QVAs.

ss[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathrm{CBF}}}_{\exp }={Q}_{{\mathrm{ICAs}}}+{Q}_{{\mathrm{VAs}}}.$$\end{document}CBFexp=QICAs+QVAs. The ADFSCI is a 24-item questionnaire divided into the following four sections: demographics, medication, AD and hypotension. The hypotension section consists of seven items. Each item uses a five-point scale to measure the frequency and severity of symptoms related to hypotension, including headaches, goosebumps, dizziness and lightheadedness, across different situational contexts. The hypotension score corresponds to the sum of all the hypotension items. Participants were categorized as experiencing symptoms if the item score was higher or equal to 2; otherwise, they were categorized as not experiencing symptoms. The acceleration and angular velocity collected by the Physilog 5 sensors were passed through a data-fusion algorithm67 to extract sensor orientation with respect to the Earth in quaternion representation. Roll, pitch and yaw angles were then calculated by using one of two representations (zyx or zyz), depending on the sensor placement with respect to its axis with the greatest motion, such that there would be no discontinuities in the pitch. The ratio of change in back curvature when turning the stimulation on is approximated as the ratio of the angles between the T6 and T10 vertebrae in the sagittal plane with and without stimulation.

sor placement with respect to its axis with the greatest motion, such that there would be no discontinuities in the pitch. The ratio of change in back curvature when turning the stimulation on is approximated as the ratio of the angles between the T6 and T10 vertebrae in the sagittal plane with and without stimulation. The MSNA signal was sampled at 20 kHz, amplified (20,000×) and band-pass filtered (0.3–2.0 kHz) to obtain muscle sympathetic spike activity and then rectified and integrated (0.1-s time constant) to obtain a multiunit (mean voltage) neurogram (Labchart v.8; ADInstruments). After acquiring a stable recording site, a 3-min baseline commenced, followed by step changes in epidural stimulation intensity for 1–2 min at each stage. Muscle sympathetic action potential spikes were identified as waveforms that matched a triphasic morphology, with the main phase being negative. The negative deflections were only assessed if they were larger than 4.5× the s.d. of the neurogram assessed during baseline (stimulator off).

nsity for 1–2 min at each stage. Muscle sympathetic action potential spikes were identified as waveforms that matched a triphasic morphology, with the main phase being negative. The negative deflections were only assessed if they were larger than 4.5× the s.d. of the neurogram assessed during baseline (stimulator off). In the HEMO trial, bioelectrical impedance (BEI) of the lower torso (abdominal cavity) was used to estimate fluid shifts within the splanchnic vascular regions during tilt-table testing68. BEI is measured by driving a small electrical current (0.7 mA with a frequency of 37 kHz) between electrodes and measuring the impedance with voltage-sensing electrodes on each end of the body segment of interest (Model 2994D/4 THRIM; UFI serving science). For the lower torso, surface electrodes were placed at approximately the left sixth rib and around the left inguinal ligament. The impedance, measured in Ω, was low-pass filtered at 0.5 Hz. Because body segment volume is inversely proportional to electrical resistance (V ~ 1/R) such that an increase in resistance reflects a loss of body fluid, the impedance values were used to estimate body segment volume using the following equation: Vgeom = (L2 × reff/R) × 1,000,000, where L is the segment length in meters, reff is the effective resistance 1.0 Wm and R is the segment impedance in W.

~ 1/R) such that an increase in resistance reflects a loss of body fluid, the impedance values were used to estimate body segment volume using the following equation: Vgeom = (L2 × reff/R) × 1,000,000, where L is the segment length in meters, reff is the effective resistance 1.0 Wm and R is the segment impedance in W. All data are reported as mean values and individual data points. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications69. All statistical analysis was performed in R using the base package ‘stats’, with primary implementation through the ‘tidyverse’ and ‘broom’ packages. Tests used included one- or two-tailed paired or independent samples Student’s t-tests, one-way analysis of variance (ANOVA) for evaluations with more than two conditions and one- or two-way repeated-measures ANOVA for assessments, when data were distributed normally, tested using a Shapiro–Wilk test. Post hoc Tukey tests were applied when appropriate. For regressions, mixed model linear regression was used in cases of multiple observations, or else standard linear modeling. In cases where group size was equal to or less than three, null hypothesis testing was not completed. The significance level was set as P < 0.05. Exclusions of data are noted in the relevant subsections of the Methods. Unless stated otherwise, experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

three, null hypothesis testing was not completed. The significance level was set as P < 0.05. Exclusions of data are noted in the relevant subsections of the Methods. Unless stated otherwise, experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

In total, 254 participants were surveyed with the ADFSCI questionnaire (ADFSCI) to measure the severity and frequency of orthostatic hypotension in a diverse sample of individuals with SCI. From this population, 17 individuals (Supplementary Table 1) were pseudo-randomly selected to participate in a single formal, 10-min tilt-table test (‘Orthostatic challenge with tilt-table test’). Hemodynamic measurements were taken via finger plethysmography (‘Postoperative hemodynamic monitoring’), and data were analogously acquired with LabChart Software (ADInstruments) and upsampled to 1,000 Hz.

All experiments were carried out as part of four clinical safety (primary objective) and preliminary efficacy (secondary objectives) trials—STIMO-HEMO (ClinicalTrials.gov registration: NCT04994886; Lausanne University Hospital (CHUV)), HEMO (ClinicalTrials.gov registration: NCT05044923; University of Calgary), HemON (ClinicalTrials.gov registration: NCT05111093; CHUV) and HemON-NL (ClinicalTrials.gov registration: NCT05941819; Sint Maartenskliniek). The HemON trial was amended after four participants to include the new thoracic paddle lead. The STIMO-HEMO (Commission cantonale d‘Ethique de la Recherche sur l’être humain CER-VD 2021-00588; Swissmedic, 10000882) and HemON (CER-VD 2021-D0025; Swissmedic, 10000932) studies were approved by the Swiss ethical authorities. The study sponsor submitted an investigation to Swissmedic and Swiss ethical authorities for regulatory approval before the start of the study. The HEMO study was approved by Canadian ethical authorities (REB21-0027). The HemON-NL (Medisch-Ethische Toetsingscommissie Oost Nederlandse - METC, 2023-16316; Centrale Commissie Mensgebonden Onderzoek - CCMO, NL83694.000.23) study sponsor submitted approval before the start of the study to the ethical authorities in the Netherlands and the Central Committee on Research Involving Human participants. The studies were approved by the Dutch ethical authorities. All four trials and corresponding amendments were conducted following the Declaration of Helsinki. A summary of information and approvals related to each study is provided in Supplementary Table 3.

he Central Committee on Research Involving Human participants. The studies were approved by the Dutch ethical authorities. All four trials and corresponding amendments were conducted following the Declaration of Helsinki. A summary of information and approvals related to each study is provided in Supplementary Table 3. All the participants signed a written informed consent at enrollment in the trial. All participants had the option to indicate consent for the publication of identifiable images or videos. All surgical and experimental procedures were performed at the investigational hospital sites, including the neurosurgery department of Lausanne University Hospital (CHUV) in Lausanne, Switzerland, the Neurosurgery Department of Foothills Medical Center in Calgary, Canada, and the Neurosurgery Department of the Radboud University Medical Center in Nijmegen, the Netherlands. The study involved eligibility and baseline assessments conducted before the surgery, the neurosurgical implantation of an investigational device, a postoperative period during which EES protocols were configured and long-term follow-up periods, including independent use of EES at home. To date, 13 individuals have completed the intensive part of the study.

Fourteen individuals who had suffered a traumatic SCI participated in one of the five studies. Demographic data and neurological status were evaluated according to the International Standards for Neurological Classification of Spinal Cord Injury48 and are reported in Supplementary Table 2 and Supplementary Data 8.

Fourteen individuals who had suffered a traumatic SCI participated in one of the five studies. Demographic data and neurological status were evaluated according to the International Standards for Neurological Classification of Spinal Cord Injury48 and are reported in Supplementary Table 2 and Supplementary Data 8. For the STIMO-HEMO and HEMO clinical trials, participants were eligible if they met the following criteria: were aged between 18 and 70 years; had a radiologically confirmed SCI located between C3 and T6 (inclusive); were neurologically classified as A or B on the AIS; had stable medical, physical and psychological conditions under the discretion of the investigators; had the injury for at least 1 year; did not have any required spinal instrumentation within the past 6 months; were able and willing to attend all scheduled appointments; had confirmed orthostatic hypotension and AD. Exclusion criteria included the following: individuals in any emergency situations; participants with any diseases and conditions that would increase the morbidity and mortality of the SCI surgery; participants who were unable to withhold antiplatelet or anticoagulation agents perioperatively; had history of myocardial infarction or cerebrovascular events within the past 6 months; had a current or anticipated need for opiod pain medications; had pain that would prevent full participation in the rehabilitation program in the judgment of the investigators; had a clinically substantial mental illness in the judgment of the investigators; received botulinum toxin injections within 6 months of enrollment; presence of substantial pressure ulcers, recurrent urinary tract infection refractory to antibiotics, unhealed spinal fractures, known or suspected drug or alcohol abuse; presence of indwelling baclofen or insulin pump; and any other conditions that would make the participant unable to participate in testing in the judgment of the investigators. Women who were pregnant during screening or breastfeeding were also excluded.

All the feasibility clinical trials outlined in Supplementary Table 3 evaluated the safety and preliminary efficacy of the investigational system and/or therapy to restore hemodynamic stability. The primary outcome measures assess the occurrences of serious adverse events and adverse events that are deemed related or possibly related to the investigational system or study procedure during the course of studies. All reportable safety events have been reported per local regulation to the respective authorities (Health Research Ethics Board at the University of Calgary, Ethics Committee CER-VD and competent authorities Swissmedic in Switzerland and METC Oost-Nederland in Nijmegen). Data safety monitoring boards have been appointed in all studies and consulted on a yearly basis or ad hoc in case of reportable events. All adverse device effects for the four clinical trials are summarized in Supplementary Tables 4–8. Two serious adverse events related to the device leading to explantation occurred during the course of the HEMO study due to postoperative infections. Device explanations were reported to local authorities and to the manufacturer. Both were resolved without sequelae following explantation surgery. In studies evaluating the safety of the newly implanted device, one reportable serious adverse event related to the study procedure occurred (subpubic catheter came out during a rehabilitation session while the stimulation was turned off).

All the participants underwent structural 3T MRI and CT imaging of the thoracolumbar spine to enable the elaboration of a personalized anatomical model of the spine and guide the preoperative planning. MRI acquisitions were conducted in a supine position with both arms on the sides. Before image acquisition, shim boxes were applied to correct for magnetic field inhomogeneities. All sequences were performed without gadolinium-based contrast agent administration. In the STIMO-HEMO and HemON clinical trials, MRI was performed on a Magnetom PrismaFit (Siemens Healthineers) with 18-channel body and 32-channel spine array coils. The standard MRI protocol comprised the following four pulse sequences: (1) two-dimensional sagittal T2-weighted turbo spin-echo (repetition time (TR), 3,080 ms; echo time (TE), 98 ms; voxel size, 0.6 × 0.6 × 3 mm3), (2) T2-weighted sampling perfection with application-optimized contrasts using different flip angle evolution (SPACE) sequence (TR, 1,500 ms; TE, 135 ms; interpolated voxel size, 0.4 × 0.4 × 0.8 mm3), (3) three-dimensional (3D) axial T2-weighted SPACE with ZOOMit (dynamic excitation pulses to achieve selective/zoomed field of view) software (TR, 2,500 ms; TE, 106 ms; interpolated voxel size, 0.3 × 0.3 × 0.5 mm3) and (4) 3D coronal T2-weighted true fast imaging with steady-state precession (TR, 6.04 ms; TE, 3.02 ms; interpolated voxel size, 0.3 × 0.3 × 0.6 mm3). The total scan time was less than 35 min overall. The HemON trial MRI protocol consisted of a subset of these sequences (no 3D axial T2-weighted SPACE with ZOOMit as this trial does not include an implant at the lumbosacral level). In the HEMO trial, a similar protocol was performed on a Discovery MR750 (GE Healthcare) with an eight-channel cervical thoracic lumbar coil. The following two sequences were performed: (1) 3D coronal T2-weighted fast imaging using steady-state acquisition (TR, 7.65 ms; TE, 3.48 ms; voxel size, 0.4 × 0.4 × 0.4 mm3) and (2) 3D sagittal T2-weighted CUBE sequence (TR, 2,000 ms; TE, 90 ms; interpolated voxel size, 0.8 × 1.25 × 0.8 mm3). The total scan time was ~84 min overall.

We trained two different nnU-net networks49 on a labeled dataset of 15 healthy participants from a previous study29. The first network was trained on high-resolution MRI to segment the cerebrospinal fluid, white matter and spinal root tissues. The second network was trained on low-resolution MRI data to segment vertebral bodies, intervertebral disks and the spinal canal. We trained the networks on 12 participants and tested them on 3 participants. We achieved high dice scores for each tissue (cerebrospinal fluid, 95.50; white matter, 90.55; roots, 71.88; vertebrae, 89.90; intervertebral disks, 91.22). When applying the networks to a new dataset, we corrected the segmentation manually if needed. The segmentation of the roots was used as guidance to trace each spinal root individually. The rostral location of each root was constrained within the white matter. For each spinal quadrant (dorsal/ventral left/right), the end of root within the white matter points was cut 1 mm within the white matter, and a spline was made with them. These end-of-root lines were then guidelines to force rootlets to attach to them, forcing them to end within the white matter. The procedure to generate the rootlets was described previously29. By forcing each rootlet to end in the end-of-roots guidelines, we made sure the rootlets followed the shape of the spine, thereby increasing robustness to scoliosis. We cut the white matter with respect to the rootlet divisions and obtained, henceforth, the spinal levels.

The participants were put under protocolized propofol-based total intravenous general anesthesia and were placed in a prone position. Preoperative surgical planning informed the neurosurgeon about the vertebral entry level and predicted the optimal paddle lead position. Based on this knowledge, lateral and anteroposterior fluoroscopy X-rays were performed intraoperatively to identify the location of the laminotomies. A midline skin incision of ~5 cm on the back was performed, the fascia opened and the muscles retracted bilaterally. Excision of the midline ligamentous structures and a laminotomy at the desired entry level enabled the insertion of the paddle lead below the vertebra. For participants of the STIMO-HEMO and HEMO trials, a second skin incision or extended opening caudally was made, and a second laminotomy was performed in the lumbar region based on the preoperative planning to allow for the insertion of the paddle lead over the lumbosacral spinal cord. The paddle lead(s) (SureScan Specify 5-6-5 (Medtronic) or ARCIM Thoracic Lead (ONWARD Medical NV)) are inserted and placed over the midline of the dura mater and advanced rostrally to the target position under repeated fluoroscopies. Electrophysiological recordings were acquired using standard neuromonitoring systems (Cascade Elite (Cadwell Industries) or ISIS Xpress (Inomed Medizintechnik)). Single pulses of EES were delivered at 0.5 Hz and increasing amplitude to elicit muscle responses that were recorded from subdermal (Neuroline Twisted Pair Subdermal, 12 × 0.4 mm; Ambu A/S) or intramuscular needle electrodes (Inomed SDN electrodes, 40 × 0.45 mm; Inomed Medizintechnik). The symmetry of the responses and the expected rostrocaudal distribution of muscle responses informed the corrections for lateral and rostrocaudal positioning. When the paddles deviated from a straight midline position, small additional laminotomies were made to remove bony protrusions and guide the paddle to a midline placement. Once the final position was determined, the paddle lead(s) was (were) anchored to the muscular fascia.

The configuration of the electrodes for EESXX programs was guided by the understanding of the mechanisms through which EES regulates blood pressure (Fig. 2a). This configuration involved three steps. First, an intraoperative mapping was conducted to identify the relevant rows of electrodes to target the hemodynamic hotspot1 and to elicit the largest pressor response. Second, postoperative imaging was used to update the anatomical models and thus estimate the location of the electrodes that maximized recruitment of the dorsal root entry zones projecting to the hemodynamic hotspot. Third, a single postoperative session was conducted to quantify pressor responses elicited by each row of electrodes. Both monopolar and multipolar configurations of electrodes were tested to ensure the maximal activation of the hemodynamic hotspot. Stimulation frequency was set at 120 Hz because we established that this frequency maximized pressor responses1. Multiple electrode configurations targeted a specific region of the hemodynamic hotspot. Each electrode configuration was delivered with a 2 ms delay between each pulse, forming a traveling wave over the spinal cord. In the STIMO-HEMO and HEMO trials, pulse width was mapped between 100 μs and 500 μs. Five of the six participants had a pulse width of 300 μs, while one participant (P3) had a pulse width of 500 μs. For the HemON and HemON-NL trials, the pulse width was set at 300 μs. The amplitude was set based on the increase in the current delivered through each electrode configuration until the systolic pressure increased by 20 mm Hg, the diastolic pressure increased by 10 mm Hg or the participant reported any discomfort such as muscle contractions or unpleasant sensations such as strong tingling. To validate the combined effect of multiple electrode configurations, we compared the amplitude of the pressor responses when EES was delivered with a configuration targeting a single level of the hemodynamic hotspot versus all the levels embedded in the hemodynamic hotspot. All other stimulation parameters were maintained unchanged.

Intraoperative beat-to-beat hemodynamic data were obtained from a pressure transducer connected to an arterial line cannula placed in the distal radial artery. The arterial line system was zeroed to atmospheric pressure. Arterial cannulation was performed either by palpation or with ultrasound guidance under sterile conditions. Adjustments in the anesthetic depth, as well as the administration of vasoactive medications or substantial volumes of intravenous fluids, were avoided during hotspot testing. All data were recorded using VitalRecorder software (VitalDB; Department of Anesthesiology and Pain Medicine, Seoul National University College of Medicine) at a sampling frequency of 125 Hz. Alternatively, the arterial line was recorded with LabChart (ADInstruments).

Beat-to-beat blood pressure and heart rate were acquired using finger plethysmography (Finometer; Finapres Medical Systems). Beat-by-beat blood pressure was calibrated to brachial artery blood pressure collected using an arm cuff embedded and synchronized with the Finometer53–57. Brachial arterial pressure was sampled at 200 Hz, while the systolic, diastolic and mean arterial pressures were extracted from the calibrated arterial pressure at 1 Hz. The heart rate was also sampled at 1 Hz. Raw data and automatically extracted hemodynamic parameters were saved and exported from the Finometer.

The investigational system used in the STIMO-HEMO and HEMO clinical trials consisted of a set of Certificat Européen-marked, Food and Drug Administration-approved medical devices used off-label. Two IPGs (Intellis with AdaptiveStim; Medtronic) were connected to a paddle lead (Specify 5-6-5 SureScan MRI; Medtronic). These implants are indicated for chronic pain management. A tablet application with a wireless communicator device (Intellis Clinician Programmer; Medtronic) was used by the clinical team to set up the system and optimize the stimulation parameters. A remote control device and transcutaneous charger device (Patient Programmer and Recharger; Medtronic) were used by the participants to charge the implanted pulse generators and to turn EES programs on and off during their daily life, as well as to adapt parameters of EES programs defined by the clinical team.

The investigational systems used in the HemON and HemON-NL clinical trials were designed and built for treating hemodynamic instability based on the understanding of the mechanisms through which EES regulates blood pressure. A stepwise approach was followed. All participants were implanted with a purpose-built IPG ARCIM (ONWARD Medical NV) that communicates with a purpose-built ecosystem of control devices. The first four participants of the HemON clinical trial were implanted with the off-label paddle lead (Specify 5-6-5 SureScan MRI; Medtronic) as the participants of the STIMO-HEMO and HEMO clinical trials, whereas all other participants were implanted with a newly designed, purpose-built paddle lead (ARC Thoracic Lead; ONWARD Medical NV).

The purpose-built ARCIM IPG is a new 16-channel neurostimulation platform that was specifically developed to regulate neurological functions with EES. The ARCIM IPG controls and delivers current-controlled stimulation pulses that can be predefined by the user or controlled in real time. The system consists of a hermetically sealed, biocompatible can that surrounds the electrical components and a rechargeable battery that enables the delivery of electrical current. The external envelope is composed of the following two main components: the header containing the connector block that enables connection with two 8-contact lead connectors and two coils for charging and communication, and the can with a rechargeable battery and electronics circuits. The ARCIM IPG was developed according to all applicable standards for medical device development. Conventional biomedical technologies were used to fabricate the system, and extensive bench and in vivo testing was performed to verify its performance.

The ARCIM Thoracic Lead developed by ONWARD Medical NV is a new 16-electrode paddle lead that incorporates a configuration of electrodes optimized to achieve the selective recruitment of the dorsal root entry zones projecting to the last three segments of the lower thoracic spinal cord. The configuration of the electrodes was determined based on a digital library of human thoracic spinal cords that we elaborated from CT and MRI acquisitions (Extended Data Fig. 2). The average length and width of spinal segments were calculated from our library and existing literature29,33,36–40. The combination of these two datasets generated global averages and s.d. for each level of the lower thoracic spinal cord, which determined the total length of the hemodynamic hotspot. We created a normal distribution with the calculated global means and s.d. and found that a length of 64.2 mm accurately represents the combined length of the three targeted spinal segments for over 95% of the population. This new paddle lead was designed following all applicable standards for medical implant development and fabricated using conventional biomedical technologies. Extensive bench and in vivo testing was conducted to validate the mechanical, electrical and biocompatibility robustness of the paddled lead. The equivalent of an investigational device exemption was granted by the competent Swiss authorities.

The purpose-built investigational system was embedded in a framework for closed-loop control of systolic blood pressure based on real-time monitoring of hemodynamic parameters. For this purpose, changes in systolic blood pressure were measured with the Finometer (‘Postoperative hemodynamic monitoring’). This signal was acquired using an analog-to-digital converter (Delsys Trigno Avanti Analog; Delsys) that wirelessly streamed the data to a sensor base station (Delsys Trigno Research+; Delsys). The 1 Hz systolic blood pressure measured with the Finometer was upsampled to 1,257 Hz by the Trigno system, which was connected through USB to a Windows computer onto which a medical-grade software developed by Ecole Polytechnique Fédérale de Lausanne (EPFL) was running. This software29 enables closed-loop control of the ARCIM system through the ARCIM API. The specific closed-loop algorithm used for the control of systolic blood pressure was implemented in a multithreaded Python application that interfaces with GDrive+ software via TCP/UDP communication, enabling the reception of systolic blood pressure signals and delivery of stimulation updates.

The systolic blood pressure signal was smoothed with an exponentially weighted moving average over a 3-s window with a forgetting factor of 0.3. This elaborated signal was sent as an input to a proportional–integral–derivative controller. To minimize steady-state errors and minimize system saturation, the integral term was penalized over time when the stimulation remained unchanged at the minimum or maximum amplitude of EES. Additionally, a forgetting factor on the integral term reduces error accumulation. A minimum (0 mA) and maximum amplitude (participant dependent) amplitude range was set to ensure the comfort and safety of the participant. The controller target was defined based on average blood pressure measurements from 10-min tilt-table tests with EES (‘Postoperative hemodynamic monitoring’). A target range of ±3 mm Hg was set around the user-defined target to account for intrinsic blood pressure variability. If the change in EES amplitude was greater than 0.5 mA, the controller incrementally ramped in steps of 500 ms per 0.5 mA for additional comfort. To calibrate the closed-loop control of EES, we found a linear operating range of stimulation amplitudes and systolic blood pressure. The minimum was the amplitude at which the first increase in systolic blood pressure occurred, and the maximum was the last amplitude at which the systolic blood pressure stopped increasing or when the participant reported abdominal muscle contractions. Each participant remained in a seated position as the amplitude of EES increased by 1 mA every 1 min to find the minimum (0 mA) and maximum parameters of stimulation (participant dependent). A regression between the average systolic blood pressure per stimulation amplitude resulted in a linear relationship between the systolic blood pressure and the amplitude of stimulation. The closed loop was tuned manually using the heuristic Ziegler–Nichols method58. The initial proportional constant was the slope of the regression derived from the calibration phase. The proportional, integral and derivative constants were tuned over one 70° tilt.

Numerical simulations were conducted using the computational life sciences platform Sim4Life (ZMT Zurich MedTech AG). First, a canonical spinal cord model was developed based on average anatomical measurements of the participants’ spinal cord anatomies. It was then used to simulate EES exposure throughout the spinal cord anatomy, as well as induce electrophysiological responses29. The dosimetric simulations involved finite element discretization using Sim4Life, which resulted in ~5 million volume elements (voxels). Each voxel was attributed dielectric properties corresponding to the assigned tissue types59 (that is, spinal cord, spinal roots, cerebrospinal fluid, epidural fat and vertebral bone). Anisotropic conductivity maps were assigned to the white matter and rootlets, leveraging existing Sim4Life functionalities (either diffusion–simulation based or DTI imaging based). By precomputing a suitable basis of electric potential distributions (one per epidural stimulation electrode contact) for the canonical bioelectrical spinal model, the electrical fields resulting from arbitrary EES pulses can rapidly be obtained29. For that purpose, the ohmic-current-dominated electro-quasistatic approximation of Maxwell’s equations was solved60. In addition to providing an average anatomical geometry, the spine model also encompassed trajectories of axons entering the spinal cord through the dorsal roots and ascending in the dorsal columns. Nerve fiber trajectories were generated inside the rootlets using Sim4Life’s IMSafe tool, while the fibers in the dorsal column were uniformly distributed. Sim4Life includes functionality to estimate fiber recruitment using the generalized, Green’s function-based activating function or detailed electrophysiological modeling. For the present work, the classic activating function (proportional to the discrete second derivative of the extracellular potential60) was used to assess relative stimulability for a given pulse shape, while electrophysiological simulations with the NEURON simulator incorporated in Sim4Life61 permitted conversion to absolute thresholds.

k, the classic activating function (proportional to the discrete second derivative of the extracellular potential60) was used to assess relative stimulability for a given pulse shape, while electrophysiological simulations with the NEURON simulator incorporated in Sim4Life61 permitted conversion to absolute thresholds. For the latter, the predefined diameter-dependent spatially extended nonlinear node fiber model underlying low-frequency exposure safety standards was used. By assigning fiber diameters according to histology-based probability distributions, the percentage of recruited fibers per root could be estimated, and the recruitment level across the different spinal segments and within the dorsal column could be predicted. Activation levels were normalized relative to maximal activation (that is, recruitment of all dorsal root fibers). Various methodologies have been developed to optimize selective stimulation, including neural network backpropagation and hybrid surrogate modeling–multiobjective genetic algorithm optimization.

Electromyographic activity of selected muscles was acquired at a 1,259 kHz sample rate using the 16-channel wireless Delsys Trigno sensors (Delsys) with bipolar surface electrodes. Skin hairs were shaved, and abrasive gel was used (Nuprep gel) on the area of interest.

Kinematics were captured by Physilog 5 sensors (Mindmaze) that were placed on the vertebral bodies of C3, T1, T6, T10, L1 and S1 of the participants. Each sensor is a stand-alone 10° of freedom. The micro-electro-mechanical system inertial measurement unit includes 3D accelerometer, 3D gyroscope, 3D magnetometer and a barometric pressure sensor at a sampling frequency of 256 Hz. These inertial measurement units were synchronized via a proprietary radiofrequency protocol.

Participants were transferred to a supine position on a table capable of head-up tilt. We applied restraint straps to secure the participant on the knees, on the hips and on the chest, with the feet stabilized. Resting supine blood pressure was recorded continuously for ~5 to 10 min to establish baseline values. Following this period, we tilted the participants upright up to a maximum of 70° while recording hemodynamic values and symptoms of orthostatic tolerance. The time to reach the desired tilt angle from supine was achieved in less than 45 s. Participants were tilted until reaching their tolerance threshold or for a maximum duration of 10 min. They were asked not to talk during the test except to inform of and grade symptoms. Participants were asked to report any symptoms every 1–3 min. Participants were asked to rank the intensity of their symptoms between 1 and 10, 1 being no symptoms at all and 10 being maximum feelings of dizziness, lightheadedness62 or nausea26,62. Participants were instructed to notify the research team if they needed to be returned to the supine position.

Blood samples were withdrawn at baseline after a period of rest in the supine position for 15 min and at 2 min (T2) after the onset of the tilt. During baseline testing, a total of three samples of blood were withdrawn. When testing with EES turned on, an additional sample of blood was withdrawn 5 min after the onset of EES in a supine position. Blood samples allowed quantification of plasma catecholamines, including plasma norepinephrine, plasma epinephrine and plasma dopamine. To avoid false positives, the participants were asked to abstain from any β-blocker, dihydropyridine calcium channel blocker, phenoxybenzamine, anxiolytics or decongestant medication. To ensure sample quality, before and immediately after, collection tubes were kept cold via immersion in crushed ice and kept in foil to reduce exposure to light. Samples were analyzed by the local hospital lab services at the University Hospital of Lausanne (Switzerland) and the Foothills Medical Center (Canada). Neuropathic pain prevented blood sampling in P8. Data and tests were not available at the time for P13 and were not part of the clinical protocol for P14.

Blood flow was measured in the seated position using Doppler ultrasound imaging equipment (Epiq Elite; Philips). Measurements were conducted both with and without EES. The left and right proximal internal carotid arteries (ICA) and vertebral arteries (VA) were assessed using a 12-3 MHz broadband linear array probe (L12–3; Philips). Additionally, the right and left middle cerebral arteries were examined using a transcranial 5-1 MHz-pulsed TransCranial Doppler (TCD) probe.

On two separate days at the same study timepoint, participants received an identical meal with equal caloric content, and their blood pressure was monitored via an arm cuff before, during and for 30 min to 1 h following food intake. This assessment was performed 1 day without EES and 1 day with EESBP. For the assessment with EESBP, the stimulation was turned on during and after the meal. All points before starting the meal were averaged to determine a baseline value. Available cuff measures between 30 min and 60 min were averaged to quantify postprandial measures. The difference (Δ) between baseline and postprandial averages was reported for the conditions without and with EESBP. Only participants who reported experiencing postprandial hypotension were considered in the analysis (P3, P4, P5, P7 and P10).

Patient-reported outcome measure questionnaires were sent to study participants via the electronic data capture system. Baseline questionnaires were sent once before surgery and then repeated every month after surgery. The following questionnaires are reported: Autonomic Dysreflexia Following SCI questionnaire, Medical Outcomes Study Sleep Scale (MOS-S)63, Neurogenic Bowel Dysfunction Score (NBDS) and Usability SUS (System Usability Scale)42.

The study participants were asked to perform a systematic set of breathing and cough tasks. Volume, flow and time data were recorded using the spirometer (Spirobank II Smart Spirometer). During the coughing maneuver, EESCoughing was precisely timed with the expiratory phase after releasing the glottis blockage. All participants from the HemON study performed the respiratory and coughing assessments at 6 months without stimulation and with EESTrunk, except P12 who could not perform the test at this timepoint.

The COPM was used to assess the functional performance and perceived satisfaction with activities of daily living of each participant. The COPM was administered by trained occupational therapists to follow standard procedures. At baseline, participants identified up to five activities they perceived as important but difficult to perform and rated their perceived performance and satisfaction on a ten-point scale. The general scores were calculated by simply averaging the performance and satisfaction ratings across activities. The same activities were then rated again by the participants three and 6 months after implantation.

Participants’ upper and lower limb spasticity was assessed using the modified Ashworth scale in the following two conditions: without and with EESSpasticity. Participants were either asked to use their standard-of-care medication or not to use their standard-of-care medication the morning before the assessment. Participants were asked to keep the stimulation off in the mornings before the assessments, which were conducted in the late morning.

In the HemON trial, physiotherapists and occupational therapists kept track of the 1-month rehabilitation participants’ sessions and documented whether the different activities were performed without or with EES through a standardized form (Mapping of Rehab Training).

In the HEMO trial, efferent postganglionic MSNA was acquired from the left common peroneal (fibular) nerve using microneurography1,64,65. The location of the common fibular nerve was located using palpation. A 2-MΩ tungsten microelectrode (Frederick Haer and Co) was then inserted percutaneously alongside a subdermal low-impedance reference electrode ~2 cm adjacent. The amplifier and head stage (NeuroAmpEx; ADInstruments) were also grounded to the participant with a surface electrode on the patella. The recording electrode was adjusted until entering the nerve. Crossing into the nerve and into different nerve bundles was identified audibly by the microneurographer. Once in the nerve, the recording microelectrode was confirmed to be near nerve fibers directed toward skeletal muscle by auditory feedback during tapping/palpation of the tibialis anterior/peroneal muscles, which has been shown previously to be indicative of microelectrode proximity to efferent postganglionic muscle sympathetic nerves. If auditory feedback was present during light stroking of skin on the dorsal foot/lower leg, this indicated proximity to postganglionic skin sympathetic fibers. The search was stopped when the electrode was in proximity to muscle sympathetic fibers.

During surgery (‘Neurosurgical intervention’), changes in blood pressure were recorded in response to different locations of EES using the arterial line (‘Postoperative hemodynamic monitoring’). Change in blood pressure or heart rate was defined as the difference in the average of 30-s windows before the start of the stimulation and at the end of the stimulation. For cross-subject analysis, the lower thoracic and lumbosacral stimulation locations in each participant were calculated by their relative distance to the center of the hemodynamic hotspot (T11) normalized by their respective spinal length (measured from T9 to the conus). Due to surgery time limits, P3 blood pressure mapping was not performed.

During a tilt-table test (‘Orthostatic challenge with tilt-table test’), changes in blood pressure were recorded without stimulation or in response to different types of stimulation (continuous or closed-loop stimulation) using the Finometer (‘Postoperative hemodynamic monitoring’). Change in blood pressure or heart rate was defined as the difference in the average of a 60-s window before the start of the tilt and a 20-s window at 3 min of the challenge. If the participant could not tolerate at least 3 min of the test due to low blood pressure or other symptoms, an average of a 20-s window before the end of the tilt was used. All measurements in a seated position were measured with stimulation on for 3–5 min. Change in blood pressure or heart rate was defined as the difference in the average of a 20-s window before the start of EES and the average of a 20-s window at 3 min before stop stimulation. All signals were smoothed over a 10-s window for illustration. The same processing was used for postoperative, day 1 quantification. Data from continuous blood pressure monitoring were excluded if technical issues with the Finometer were observed, including loss of the plethysmographic signal, incorrect calibration as measured by comparing continuous measuring to standard measurements from a brachial cuff, spasms or due to study protocol deviations or adverse events.

The efficacy of continuous, closed loop and inertial measurement unit-based control of EES was measured by (1) target error, defined as the error of systolic blood pressure with respect to a user-defined target on the systolic blood pressure, and (2) error variability, defined as the s.d. of the target error, to measure stability of the stimulation during dynamic orthostatic challenges. Both metrics were calculated from 1 min before the dynamic challenges to 1 min after the end of the challenge.

Peak systolic and end-diastolic flow velocities were obtained from the Doppler spectrum analysis over three consecutive cycles in all arteries. Mean velocity was derived through the integration of the Doppler curve for each corresponding pair, with the mean value calculated across the three cycles. Arterial diameters were determined in B-mode for the left and right ICAs and VAs. The flow rate (Q) in each vessel, measured in milliliters per minute, was calculated using equation (1), assuming a parabolic velocity profile. The total experimental CBF (CBFexp) was determined by summing the flow of the left and right ICAs and VAs, according to equation (2)66.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q=\frac{{\mathrm{FV}}}{2}\times \frac{\pi {\bar{D}}^{2}}{4}$$\end{document}Q=FV2×πD¯242\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathrm{CBF}}}_{\exp }={Q}_{{\mathrm{ICAs}}}+{Q}_{{\mathrm{VAs}}}.$$\end{document}CBFexp=QICAs+QVAs.

The ADFSCI is a 24-item questionnaire divided into the following four sections: demographics, medication, AD and hypotension. The hypotension section consists of seven items. Each item uses a five-point scale to measure the frequency and severity of symptoms related to hypotension, including headaches, goosebumps, dizziness and lightheadedness, across different situational contexts. The hypotension score corresponds to the sum of all the hypotension items. Participants were categorized as experiencing symptoms if the item score was higher or equal to 2; otherwise, they were categorized as not experiencing symptoms.

The acceleration and angular velocity collected by the Physilog 5 sensors were passed through a data-fusion algorithm67 to extract sensor orientation with respect to the Earth in quaternion representation. Roll, pitch and yaw angles were then calculated by using one of two representations (zyx or zyz), depending on the sensor placement with respect to its axis with the greatest motion, such that there would be no discontinuities in the pitch. The ratio of change in back curvature when turning the stimulation on is approximated as the ratio of the angles between the T6 and T10 vertebrae in the sagittal plane with and without stimulation.

The MSNA signal was sampled at 20 kHz, amplified (20,000×) and band-pass filtered (0.3–2.0 kHz) to obtain muscle sympathetic spike activity and then rectified and integrated (0.1-s time constant) to obtain a multiunit (mean voltage) neurogram (Labchart v.8; ADInstruments). After acquiring a stable recording site, a 3-min baseline commenced, followed by step changes in epidural stimulation intensity for 1–2 min at each stage. Muscle sympathetic action potential spikes were identified as waveforms that matched a triphasic morphology, with the main phase being negative. The negative deflections were only assessed if they were larger than 4.5× the s.d. of the neurogram assessed during baseline (stimulator off).

In the HEMO trial, bioelectrical impedance (BEI) of the lower torso (abdominal cavity) was used to estimate fluid shifts within the splanchnic vascular regions during tilt-table testing68. BEI is measured by driving a small electrical current (0.7 mA with a frequency of 37 kHz) between electrodes and measuring the impedance with voltage-sensing electrodes on each end of the body segment of interest (Model 2994D/4 THRIM; UFI serving science). For the lower torso, surface electrodes were placed at approximately the left sixth rib and around the left inguinal ligament. The impedance, measured in Ω, was low-pass filtered at 0.5 Hz. Because body segment volume is inversely proportional to electrical resistance (V ~ 1/R) such that an increase in resistance reflects a loss of body fluid, the impedance values were used to estimate body segment volume using the following equation: Vgeom = (L2 × reff/R) × 1,000,000, where L is the segment length in meters, reff is the effective resistance 1.0 Wm and R is the segment impedance in W.

All data are reported as mean values and individual data points. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications69. All statistical analysis was performed in R using the base package ‘stats’, with primary implementation through the ‘tidyverse’ and ‘broom’ packages. Tests used included one- or two-tailed paired or independent samples Student’s t-tests, one-way analysis of variance (ANOVA) for evaluations with more than two conditions and one- or two-way repeated-measures ANOVA for assessments, when data were distributed normally, tested using a Shapiro–Wilk test. Post hoc Tukey tests were applied when appropriate. For regressions, mixed model linear regression was used in cases of multiple observations, or else standard linear modeling. In cases where group size was equal to or less than three, null hypothesis testing was not completed. The significance level was set as P < 0.05. Exclusions of data are noted in the relevant subsections of the Methods. Unless stated otherwise, experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41591-025-03614-w.

Supplementary InformationSupplementary Notes 1 and 2, Supplementary Figs. 1 and 2 and Supplementary Tables 1–8. Reporting Summary Supplementary VideoDevelopment of the implantable system to restore hemodynamic stability after SCI and applications in ecological settings. Supplementary Data 1ASIA Impairment Scale (AIS) score summaries for each study participant. Supplementary Data 2Supporting data and statistics for Supplementary Figs. 1 and 2. Supplementary Notes 1 and 2, Supplementary Figs. 1 and 2 and Supplementary Tables 1–8. Reporting Summary Development of the implantable system to restore hemodynamic stability after SCI and applications in ecological settings. ASIA Impairment Scale (AIS) score summaries for each study participant. Supporting data and statistics for Supplementary Figs. 1 and 2.

Source Data Fig. 1Source data and statistics. Source Data Fig. 2 and Extended Data Figs. 1–3Source data and statistics. Source Data Fig. 3 and Extended Data Figs. 4 and 5Source data and statistics. Source Data Fig. 4 and Extended Data Fig. 6Source data and statistics. Source Data Fig. 5 and Extended Data Fig. 7Source data and statistics. Source Data Fig. 6 and Extended Data Figs. 9 and 10Source data and statistics. Source Data Extended Data Fig. 8Source data and statistics. Source data and statistics. Source data and statistics. Source data and statistics. Source data and statistics. Source data and statistics. Source data and statistics. Source data and statistics.