Browse the corpus

Anesthesia Type during Cancer Surgery: Results of the GA-CARES Randomized, Multicenter Trial. BACKGROUND: Surgical resection is a widely used treatment for cancer. Patients can be "seeded" with their own cancer cells during surgery, and it has been postulated that the immune response to these circulating cancer cells can influence recurrence risk. Some preclinical and retrospective studies have suggested that propofol-based general anesthesia may be superior to volatile halogenated ethers with respect to cell-mediated immunity, implantation of circulating tumor cells, and cancer-related outcomes, but there are limited data from large randomized clinical trials. METHODS: The General Anesthetics in Cancer Resection (GA-CARES) trial is a multicenter, pragmatic, investigator-initiated, partially blinded, randomized superiority trial. Adults at five U.S. centers undergoing surgical resection of cancers associated with poor outcomes (pancreas, esophagus, lung, stomach, bile ducts, liver, bladder, or peritoneal surface) were randomized (1:1) to receive exclusive use of either propofol or volatile agent for maintenance of general anesthesia. The intent-to-treat population included all randomized patients (n = 1,766) minus 3 patients who withdrew consent before surgery. The per-protocol population included patients completing surgery, with pathologically confirmed cancer, and receiving the assigned anesthetic drug. The primary endpoint was all-cause mortality (minimum 2-yr follow-up). Secondary endpoints included disease-free survival. RESULTS: Adherence to the protocol was high, with 95.9% of patients who had surgery receiving the assigned anesthetic exclusively. In contrast to the authors' hypothesis, propofol-treated patients did not exhibit better survival (propofol 230 deaths out of 881 [26.1%] vs. volatile 202 deaths out of 882 [22.9%]; hazard ratio, 1.16; 95% CI, 0.96 to 1.41; P = 0.115 by exact stratified log rank test) in the intent-to-treat population (n = 1,763). In the per-protocol population (n = 1,411), significantly more patients randomized to propofol died through 2-yr follow-up (25.5% vs. 20%; hazard ratio, 1.31; 95% CI, 1.05 to 1.64; P = 0.017). Results were similar for disease-free survival (hazard ratio, 1.10; 95% CI, 0.9 to 1.36; P = 0.428) and were consistent across numerous subgroups. CONCLUSIONS: Propofol-based anesthesia is not effective at improving cancer-related outcomes in patients undergoing resection of malignancies.

Basic research has suggested different effects of propofol versus inhaled halogenated anesthetics on cancer cell growth and metastases Retrospective studies suggested improved oncologic outcomes among patients undergoing cancer resections with propofol anesthesia versus inhaled anesthesia Among 1,766 patients enrolled into a multicenter pragmatic randomized trial, patients who received propofol for maintenance during oncologic surgery did not demonstrate better all-cause mortality or cancer-free survival versus those who received volatile inhaled agents Use of propofol for anesthesia maintenance is unlikely to improve cancer-related outcomes compared to using volatile inhaled anesthetics

Among 1,766 patients enrolled into a multicenter pragmatic randomized trial, patients who received propofol for maintenance during oncologic surgery did not demonstrate better all-cause mortality or cancer-free survival versus those who received volatile inhaled agents Use of propofol for anesthesia maintenance is unlikely to improve cancer-related outcomes compared to using volatile inhaled anesthetics According to National Cancer Institute (Bethesda, Maryland) statistics, more than 2 million new cases of cancer are diagnosed annually in the United States, with approximately 600,000 people dying from cancer.1 Cancer will be diagnosed in an estimated 40.5% of individuals during their lifetimes and remains the second leading cause of death.2 Surgical resection is the mainstay of treatment for many solid malignancies. As early as 1889,3 it was postulated that patients can be seeded with their own cancer cells (“seed and soil” hypothesis). The increase of circulating cancer cells along with impaired antitumor activity during oncologic surgery can increase the risk of cancer recurrence.4–6 Support for this theory comes from studies showing that transfusion of allogeneic red blood cells, which is immunosuppressive,7 is associated with worse cancer outcomes after resection of various cancers, including pancreatic,8 bladder,9 liver,10 and colorectal cancers.11

can increase the risk of cancer recurrence.4–6 Support for this theory comes from studies showing that transfusion of allogeneic red blood cells, which is immunosuppressive,7 is associated with worse cancer outcomes after resection of various cancers, including pancreatic,8 bladder,9 liver,10 and colorectal cancers.11 Numerous mechanistic12,13 and animal14 models have suggested that propofol anesthesia may be superior to volatile halogenated ethers with respect to antitumor immune response and inhibition of cell growth and metastases. A previous meta-analysis, comprised almost entirely of retrospective studies, observed higher recurrence-free survival in patients who received propofol versus volatile anesthesia.15 However, due to limited randomized data comparing propofol versus volatile agents in surgery for biologically aggressive cancers, there has been no conclusive evidence to support one practice or the other. Propofol and volatile agent-based anesthesia are both inexpensive and have been shown to result in similar short-term outcomes16; therefore, both are used routinely for maintenance of general anesthesia.

ery for biologically aggressive cancers, there has been no conclusive evidence to support one practice or the other. Propofol and volatile agent-based anesthesia are both inexpensive and have been shown to result in similar short-term outcomes16; therefore, both are used routinely for maintenance of general anesthesia. In 2015, experts representing the British Journal of Anesthesiology Workgroup on Cancer and Anesthesia and the American Society of Anesthesiologists (ASA; Schaumburg, Illinois)17 called for prospective scientific inquiry into whether the anesthetic technique utilized during oncologic resections influences cancer-related outcomes. In response to this call, in 2017, we began an investigator-initiated, pragmatic, multicenter, randomized study in the United States called the General Anesthetics in Cancer Resection (GA-CARES) trial to test the hypothesis that use of propofol (compared with volatile anesthesia) improves overall survival and disease-free survival in patients with biologically aggressive cancers.

Basic research has suggested different effects of propofol versus inhaled halogenated anesthetics on cancer cell growth and metastases Retrospective studies suggested improved oncologic outcomes among patients undergoing cancer resections with propofol anesthesia versus inhaled anesthesia

Among 1,766 patients enrolled into a multicenter pragmatic randomized trial, patients who received propofol for maintenance during oncologic surgery did not demonstrate better all-cause mortality or cancer-free survival versus those who received volatile inhaled agents Use of propofol for anesthesia maintenance is unlikely to improve cancer-related outcomes compared to using volatile inhaled anesthetics

The GA-CARES trial is a multicenter, pragmatic, investigator-initiated, partially blinded, randomized superiority trial. The trial design has been described previously.18 After review and approval by the institutional review board at each of the five participating U.S. sites, the trial was registered at ClinicalTrials.gov (NCT03034096; principal investigator, Elliott Bennett-Guerrero, M.D.) on January 27, 2017. This was before enrollment of the first patient on January 31, 2017. All participants gave written informed consent. An independent data and safety monitoring committee, composed of three individuals (appendix 1) not otherwise involved in the trial, met every 6 months to review trial data and recommended continuation without modification after each meeting.

patient on January 31, 2017. All participants gave written informed consent. An independent data and safety monitoring committee, composed of three individuals (appendix 1) not otherwise involved in the trial, met every 6 months to review trial data and recommended continuation without modification after each meeting. See appendix 2 for a detailed list. The study included adult patients with known or suspected cancer, undergoing one of the following surgical oncologic procedures: (1) lung lobectomy or pneumonectomy, (2) esophagectomy, (3) radical (total) cystectomy, (4) pancreatectomy, (5) hepatectomy, (6) gastrectomy (subtotal or total), (7) radical cholecystectomy or bile duct resection for known or suspected cancer, or (8) cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (HIPEC). While we anticipated that the majority of resections would be for primary disease, some of the patients undergoing lung, liver, and cytoreductive surgery and HIPEC did so for metastatic disease. These procedures were selected because they represent surgical oncology cases associated with high recurrence rates and poor outcomes. We postulated that the impact of anesthesia type would be most evident in these cancer surgeries, which are often of long duration and with prolonged exposure to the anesthetic agent. Patients with less aggressive cancers, e.g., neuroendocrine or gastrointestinal stromal tumors, were excluded.

rates and poor outcomes. We postulated that the impact of anesthesia type would be most evident in these cancer surgeries, which are often of long duration and with prolonged exposure to the anesthetic agent. Patients with less aggressive cancers, e.g., neuroendocrine or gastrointestinal stromal tumors, were excluded. Patients were randomized (1:1) using a Research Electronic Data Capture (REDCap, USA) randomization module to propofol or a volatile anesthetic for maintenance of general anesthesia. The clinician determined the volatile agent used (i.e., isoflurane, sevoflurane, or desflurane). Administration of a small amount of propofol with the explicit purpose of induction of general anesthesia was permitted in both study arms, since propofol is the standard of care for induction. As this was a pragmatic trial, there was no standardization of opioid use, regional anesthesia, or other concomitant medications/procedures. These and other potential confounders, however, were recorded and analyzed to ensure balance between groups.

Patients were randomized (1:1) using a Research Electronic Data Capture (REDCap, USA) randomization module to propofol or a volatile anesthetic for maintenance of general anesthesia. The clinician determined the volatile agent used (i.e., isoflurane, sevoflurane, or desflurane). Administration of a small amount of propofol with the explicit purpose of induction of general anesthesia was permitted in both study arms, since propofol is the standard of care for induction. As this was a pragmatic trial, there was no standardization of opioid use, regional anesthesia, or other concomitant medications/procedures. These and other potential confounders, however, were recorded and analyzed to ensure balance between groups. Subjects were not informed of their study group assignment. It was not possible to blind the anesthesia care team to study arm in this trial. The surgical team could obtain this information, but it is unlikely this affected their surgical management. The trial’s primary outcome was all-cause mortality, obtained primarily from the National Death Index (NDI) and secondarily from the New York Cancer Registry, with both sources blinded to study group assignment. For the trial’s secondary endpoint of disease-free survival, New York Cancer Registry staff at each site were functionally blinded, since abstractors did not know of the patients’ study group assignment, and even if they did, it is unlikely this would affect their data abstraction. Main sources of data were the hospital electronic medical record (EMR), New York Cancer Registry, and NDI.

Subjects were not informed of their study group assignment. It was not possible to blind the anesthesia care team to study arm in this trial. The surgical team could obtain this information, but it is unlikely this affected their surgical management. The trial’s primary outcome was all-cause mortality, obtained primarily from the National Death Index (NDI) and secondarily from the New York Cancer Registry, with both sources blinded to study group assignment. For the trial’s secondary endpoint of disease-free survival, New York Cancer Registry staff at each site were functionally blinded, since abstractors did not know of the patients’ study group assignment, and even if they did, it is unlikely this would affect their data abstraction. Main sources of data were the hospital electronic medical record (EMR), New York Cancer Registry, and NDI. Per New York state law, the New York Cancer Registry must record and report information about all patients diagnosed and treated with cancer. Data variables for collection include sociodemographic characteristics, disease- or tumor-related characteristics (per American Joint Committee on Cancer (Chicago, Illinois) and Surveillance, Epidemiology, and End Results (SEER) Program and Coding Manuals), and treatment information (e.g., dates of surgical treatment, chemotherapy, and radiation treatments). The New York Cancer Registry follows patients annually for life. The quality, completeness, and timeliness of the registry data are submitted and verified annually by the North American Association of Central Cancer Registries (Springfield, Illinois) certification process. The New York Cancer Registry, which has been collecting information on patients with cancer for more than 50 yr, has consistently received a gold-level certification.

ry data are submitted and verified annually by the North American Association of Central Cancer Registries (Springfield, Illinois) certification process. The New York Cancer Registry, which has been collecting information on patients with cancer for more than 50 yr, has consistently received a gold-level certification. The U.S. government’s National Center for Health Statistics (Hyattsville, Maryland) established the NDI as a resource to aid epidemiologists and other health and medical investigators with their mortality ascertainment activities.19 Of the available national mortality databases, the NDI has been demonstrated to have the highest sensitivity for recording mortality.20 Block randomization with random permuted block sizes was used to ensure a similar number of subjects in each arm over time while minimizing predictability. Randomization was first stratified by site, then stratified by the eight procedure types listed in the inclusion criteria. Randomization was performed using REDCap’s secure interactive web-based randomization system, which provided for allocation concealment.

jects in each arm over time while minimizing predictability. Randomization was first stratified by site, then stratified by the eight procedure types listed in the inclusion criteria. Randomization was performed using REDCap’s secure interactive web-based randomization system, which provided for allocation concealment. The primary and key secondary endpoints are consistent with consensus definitions for standardized endpoints for surgical cancer outcome trials.21 The primary endpoint was all-cause mortality (minimum 2-yr follow-up after randomization). Deaths were obtained by our primary method (“exact match” as defined by NDI’s algorithm) and supplemented with New York Cancer Registry and hospital EMR deaths in patients with no NDI match or an inexact NDI match. Of note, vital status was never inferred/assumed; we only used a known dead or alive date of last contact. An advantage of using mortality as the primary endpoint is that it is still relevant even if patients are never disease-free. A key secondary endpoint was cancer recurrence (for disease-free survival analysis).21 This endpoint is important; however, it cannot be collected in some patients who are never “cancer free,” i.e., patients with gross residual disease after resection or unresectable disease. Therefore, the disease-free survival analysis only included patients undergoing surgery with curative intent, i.e., without gross residual disease after resection or unresectable disease.

ed in some patients who are never “cancer free,” i.e., patients with gross residual disease after resection or unresectable disease. Therefore, the disease-free survival analysis only included patients undergoing surgery with curative intent, i.e., without gross residual disease after resection or unresectable disease. While no differences in short-term outcome were anticipated based on previous studies,16 we recorded postoperative length of hospital stay and in-hospital mortality as reasonable surrogates for immediate postoperative outcome. All data were stored in a study-specific REDCap database. Before unblinding, the primary data managers (S.L.S. and J.L.R.) performed data quality checks. International Classification of Diseases, Tenth Revision, procedure codes were cross-checked with operative reports to determine actual surgery performed (vs. planned). In addition to standard patient data, e.g., age and ASA Physical Status, International Classification of Diseases, Tenth Revision diagnosis codes were used to calculate Elixhauser comorbidity scores.22,23

ion, procedure codes were cross-checked with operative reports to determine actual surgery performed (vs. planned). In addition to standard patient data, e.g., age and ASA Physical Status, International Classification of Diseases, Tenth Revision diagnosis codes were used to calculate Elixhauser comorbidity scores.22,23 The primary analysis population (intent-to-treat [ITT]) included all randomized patients minus patients who withdrew consent to use their data. We assumed a blended mortality rate of at least 15% at 2 yr in the volatile arm and 10% mortality in the propofol arm, with a 5% absolute difference being a small but clinically significant difference in outcome given the ease and low cost of using propofol for these cases if propofol was shown to be better. The estimated study power for 800 and 900 patients per study group was 95.7% and 97.4%, respectively, using a two-sided log-rank test at a significance level of 0.05, with an estimated 2 yr of recruitment of 4 yr total study time.18 No interim analyses were planned or conducted.

ases if propofol was shown to be better. The estimated study power for 800 and 900 patients per study group was 95.7% and 97.4%, respectively, using a two-sided log-rank test at a significance level of 0.05, with an estimated 2 yr of recruitment of 4 yr total study time.18 No interim analyses were planned or conducted. All statistical analyses were performed by the statistical team using the statistical computing software R (version 4.3.1.; R Foundation for Statistical Computing, Vienna, Austria). Time to overall survival was defined as from the date of randomization to the date of death or the date of last known alive follow-up date, whichever occurred first. Kaplan–Meier curves of overall survival were constructed for patients in each arm. The exact stratified log-rank test was applied for the primary objective of comparing overall survival using two stratification factors (study site and procedure type) from the randomization step. Subgroup analyses were performed in a similar fashion. Disease-free survival was analyzed similarly to time to overall survival, as time to disease recurrence or the last known alive follow-up date, whichever occurred first. The population for this disease-free survival analysis only included patients undergoing surgery with curative intent, i.e., without gross residual disease after resection or unresectable disease. Analyses were also performed in a prespecified per-protocol population defined as patients with a confirmed cancer diagnosis who underwent a completed surgery and received the correct anesthesia agent as assigned.

See appendix 2 for a detailed list. The study included adult patients with known or suspected cancer, undergoing one of the following surgical oncologic procedures: (1) lung lobectomy or pneumonectomy, (2) esophagectomy, (3) radical (total) cystectomy, (4) pancreatectomy, (5) hepatectomy, (6) gastrectomy (subtotal or total), (7) radical cholecystectomy or bile duct resection for known or suspected cancer, or (8) cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (HIPEC). While we anticipated that the majority of resections would be for primary disease, some of the patients undergoing lung, liver, and cytoreductive surgery and HIPEC did so for metastatic disease. These procedures were selected because they represent surgical oncology cases associated with high recurrence rates and poor outcomes. We postulated that the impact of anesthesia type would be most evident in these cancer surgeries, which are often of long duration and with prolonged exposure to the anesthetic agent. Patients with less aggressive cancers, e.g., neuroendocrine or gastrointestinal stromal tumors, were excluded.

Patients were randomized (1:1) using a Research Electronic Data Capture (REDCap, USA) randomization module to propofol or a volatile anesthetic for maintenance of general anesthesia. The clinician determined the volatile agent used (i.e., isoflurane, sevoflurane, or desflurane). Administration of a small amount of propofol with the explicit purpose of induction of general anesthesia was permitted in both study arms, since propofol is the standard of care for induction.

As this was a pragmatic trial, there was no standardization of opioid use, regional anesthesia, or other concomitant medications/procedures. These and other potential confounders, however, were recorded and analyzed to ensure balance between groups.

Subjects were not informed of their study group assignment. It was not possible to blind the anesthesia care team to study arm in this trial. The surgical team could obtain this information, but it is unlikely this affected their surgical management. The trial’s primary outcome was all-cause mortality, obtained primarily from the National Death Index (NDI) and secondarily from the New York Cancer Registry, with both sources blinded to study group assignment. For the trial’s secondary endpoint of disease-free survival, New York Cancer Registry staff at each site were functionally blinded, since abstractors did not know of the patients’ study group assignment, and even if they did, it is unlikely this would affect their data abstraction.

Main sources of data were the hospital electronic medical record (EMR), New York Cancer Registry, and NDI. Per New York state law, the New York Cancer Registry must record and report information about all patients diagnosed and treated with cancer. Data variables for collection include sociodemographic characteristics, disease- or tumor-related characteristics (per American Joint Committee on Cancer (Chicago, Illinois) and Surveillance, Epidemiology, and End Results (SEER) Program and Coding Manuals), and treatment information (e.g., dates of surgical treatment, chemotherapy, and radiation treatments). The New York Cancer Registry follows patients annually for life. The quality, completeness, and timeliness of the registry data are submitted and verified annually by the North American Association of Central Cancer Registries (Springfield, Illinois) certification process. The New York Cancer Registry, which has been collecting information on patients with cancer for more than 50 yr, has consistently received a gold-level certification. The U.S. government’s National Center for Health Statistics (Hyattsville, Maryland) established the NDI as a resource to aid epidemiologists and other health and medical investigators with their mortality ascertainment activities.19 Of the available national mortality databases, the NDI has been demonstrated to have the highest sensitivity for recording mortality.20

Block randomization with random permuted block sizes was used to ensure a similar number of subjects in each arm over time while minimizing predictability. Randomization was first stratified by site, then stratified by the eight procedure types listed in the inclusion criteria. Randomization was performed using REDCap’s secure interactive web-based randomization system, which provided for allocation concealment.

The primary and key secondary endpoints are consistent with consensus definitions for standardized endpoints for surgical cancer outcome trials.21 The primary endpoint was all-cause mortality (minimum 2-yr follow-up after randomization). Deaths were obtained by our primary method (“exact match” as defined by NDI’s algorithm) and supplemented with New York Cancer Registry and hospital EMR deaths in patients with no NDI match or an inexact NDI match. Of note, vital status was never inferred/assumed; we only used a known dead or alive date of last contact. An advantage of using mortality as the primary endpoint is that it is still relevant even if patients are never disease-free. A key secondary endpoint was cancer recurrence (for disease-free survival analysis).21 This endpoint is important; however, it cannot be collected in some patients who are never “cancer free,” i.e., patients with gross residual disease after resection or unresectable disease. Therefore, the disease-free survival analysis only included patients undergoing surgery with curative intent, i.e., without gross residual disease after resection or unresectable disease. While no differences in short-term outcome were anticipated based on previous studies,16 we recorded postoperative length of hospital stay and in-hospital mortality as reasonable surrogates for immediate postoperative outcome.

All data were stored in a study-specific REDCap database. Before unblinding, the primary data managers (S.L.S. and J.L.R.) performed data quality checks. International Classification of Diseases, Tenth Revision, procedure codes were cross-checked with operative reports to determine actual surgery performed (vs. planned). In addition to standard patient data, e.g., age and ASA Physical Status, International Classification of Diseases, Tenth Revision diagnosis codes were used to calculate Elixhauser comorbidity scores.22,23

The primary analysis population (intent-to-treat [ITT]) included all randomized patients minus patients who withdrew consent to use their data. We assumed a blended mortality rate of at least 15% at 2 yr in the volatile arm and 10% mortality in the propofol arm, with a 5% absolute difference being a small but clinically significant difference in outcome given the ease and low cost of using propofol for these cases if propofol was shown to be better. The estimated study power for 800 and 900 patients per study group was 95.7% and 97.4%, respectively, using a two-sided log-rank test at a significance level of 0.05, with an estimated 2 yr of recruitment of 4 yr total study time.18 No interim analyses were planned or conducted.

All statistical analyses were performed by the statistical team using the statistical computing software R (version 4.3.1.; R Foundation for Statistical Computing, Vienna, Austria). Time to overall survival was defined as from the date of randomization to the date of death or the date of last known alive follow-up date, whichever occurred first. Kaplan–Meier curves of overall survival were constructed for patients in each arm. The exact stratified log-rank test was applied for the primary objective of comparing overall survival using two stratification factors (study site and procedure type) from the randomization step. Subgroup analyses were performed in a similar fashion. Disease-free survival was analyzed similarly to time to overall survival, as time to disease recurrence or the last known alive follow-up date, whichever occurred first. The population for this disease-free survival analysis only included patients undergoing surgery with curative intent, i.e., without gross residual disease after resection or unresectable disease. Analyses were also performed in a prespecified per-protocol population defined as patients with a confirmed cancer diagnosis who underwent a completed surgery and received the correct anesthesia agent as assigned.

From February 2017 through September 2022, 1,826 patients were consented at five sites in the United States. As shown in figure 1, 60 patients were not randomized, and the remaining 1,766 patients were randomly assigned to receive propofol (n = 883) or a volatile halogenated ether (n = 883) for maintenance of general anesthesia. Three randomized patients withdrew from the study, all before surgery, yielding an ITT population of 1,763 patients. The numbers of ITT patients enrolled per site were 166, 216, 323, 440, and 618. The study population was diverse with 734 (41.6%) women, 389 (22.1%) non-White subjects, and 113 (6.0%) Latino subjects (table 1). Characteristics of the Patients at Baseline (ITT Population) Intent-to-treat (ITT) population, all patients randomized minus three patients who withdrew consent after randomization but before surgery. Comorbidities and Elixhauser scores do not include the 13 randomized patients who never came to the hospital for surgery and thus did not have International Classification of Diseases, Tenth Revision diagnosis codes. IQR, interquartile range.

Intent-to-treat (ITT) population, all patients randomized minus three patients who withdrew consent after randomization but before surgery. Comorbidities and Elixhauser scores do not include the 13 randomized patients who never came to the hospital for surgery and thus did not have International Classification of Diseases, Tenth Revision diagnosis codes. IQR, interquartile range. Enrollment, randomization, and follow-up of patients, according to study group. Shows flow of patients from consent through follow-up. Volatile, volatile inhalational anesthetic agent. Intent-to-treat (ITT), all patients randomized minus three patients who withdrew consent before surgery. Per-protocol population, all patients in the ITT population who received the planned surgical procedure, had a pathologically confirmed eligible cancer, and received the correct anesthetic agent as assigned. *Sum of patients excluded from per-protocol population is greater than 100% as some patients had more than one exclusion reason. GA, general anesthesia.

the ITT population who received the planned surgical procedure, had a pathologically confirmed eligible cancer, and received the correct anesthetic agent as assigned. *Sum of patients excluded from per-protocol population is greater than 100% as some patients had more than one exclusion reason. GA, general anesthesia. The characteristics of the patients were similar in the two groups (tables 1 to 3). Most patients were older than 65 yr and were at higher risk for surgery with approximately 85% classified as ASA Physical Status III or IV; Elixhauser scores (median greater than 10) were consistent with high preoperative comorbidity (table 1). Study groups were well balanced with regard to perioperative characteristics, e.g., planned and actual completed surgical procedure, intravenous fluid volumes, allogenic red cell transfusions, opioids (fentanyl and oral morphine equivalents), steroids, ketamine, and use of neuraxial anesthesia as an adjunct to general anesthesia. Most patients (n = 1,316) received surgery and as part of their first course of treatment defined by the New York Cancer Registry as the initial treatment before disease recurrence or progression; 219 patients received surgery for recurrent or progressive disease (bladder, liver, lung, or peritoneal sites), and 140 patients had surgery canceled or aborted, usually for unresectable disease.

urse of treatment defined by the New York Cancer Registry as the initial treatment before disease recurrence or progression; 219 patients received surgery for recurrent or progressive disease (bladder, liver, lung, or peritoneal sites), and 140 patients had surgery canceled or aborted, usually for unresectable disease. Overall, 1,677 (95.9%) of 1,749 patients who had surgery received the assigned anesthetic exclusively. Only 72 (4.1%) of these patients in the propofol (n = 44) and volatile (n = 28) groups were excluded from the per-protocol analysis for an anesthesia treatment deviation (fig. 1). In some cases, these deviations were due to a medical indication, e.g., administration of volatile agent to treat acute bronchospasm in a patient assigned to the propofol group. The protocol allowed propofol to be administered to both groups for induction of general anesthesia, and dosing totals, median (interquartile range [IQR]), were 2,186 (1,470 to 3,238) and 160 (120 to 200) mg, respectively, in the propofol and volatile groups. Most patients in the volatile group (n = 882 for ITT population) received sevoflurane (n = 580; 65.8%). No significant differences were observed between study groups in our surrogate measures of short-term outcome. In-hospital mortality occurred in the same number of patients (n = 11) in each study group. Postoperative hospital length of stay (median, IQR) was also similar: 5 (3 to 7) and 5 (3 to 8) days, respectively, in the propofol and volatile groups (P = 0.755 from Wilcoxon rank sum test with continuity correction).

outcome. In-hospital mortality occurred in the same number of patients (n = 11) in each study group. Postoperative hospital length of stay (median, IQR) was also similar: 5 (3 to 7) and 5 (3 to 8) days, respectively, in the propofol and volatile groups (P = 0.755 from Wilcoxon rank sum test with continuity correction). Median follow-up for survivors was 55 (IQR, 39 to 60) months. In the ITT population, 1,713 (97.2%) patients had at least 2 yr of follow-up. As expected, overall mortality in ITT patients through 2 yr after randomization was highest in aborted cases (82 of 140; 58.6%) followed by surgical resection for recurrent disease (50 of 219; 22.8%) and surgery as part of the first course of therapy, i.e., resection of primary tumor (293 of 1,316; 22.3%). Mortality rates (ITT population, through 2 yr) for the eight planned surgical procedures were pancreas (38.7%), esophagus (34.3%), liver (22.7%), biliary (50.0%), stomach (27.8%), lung (13.6%), bladder (29.2%), and HIPEC (44.0%).

rt of the first course of therapy, i.e., resection of primary tumor (293 of 1,316; 22.3%). Mortality rates (ITT population, through 2 yr) for the eight planned surgical procedures were pancreas (38.7%), esophagus (34.3%), liver (22.7%), biliary (50.0%), stomach (27.8%), lung (13.6%), bladder (29.2%), and HIPEC (44.0%). In contrast to our primary hypothesis, propofol-treated patients did not exhibit better survival (propofol, 230 deaths of 881 [26.1%] vs. volatile, 202 deaths of 882 [22.9%]; hazard ratio, 1.16; 95% CI, 0.96 to 1.41; P = 0.115 by exact stratified log rank test; fig. 2A) in the ITT population (n = 1,763). In the per-protocol population (n = 1,411; fig. 2B), significantly more patients randomized to propofol died through 2-yr follow-up (25.5% vs. 20%; hazard ratio, 1.31; 95% CI, 1.05 to 1.64; P = 0.017). Results were similar for disease-free survival (fig. 2C) and even through the potential maximum 5 yr of follow-up (fig. 2D). These results were also generally consistent across the eight surgery type subgroups as well as for different case durations (fig. 3).

. 20%; hazard ratio, 1.31; 95% CI, 1.05 to 1.64; P = 0.017). Results were similar for disease-free survival (fig. 2C) and even through the potential maximum 5 yr of follow-up (fig. 2D). These results were also generally consistent across the eight surgery type subgroups as well as for different case durations (fig. 3). Overall survival and disease-free recurrence. (A) Kaplan–Meier estimates of the time to death by study group (intent-to-treat [ITT] primary analysis). (B) Kaplan–Meier estimates of the time to death by study group (per protocol). (C) Kaplan–Meier estimates of the time to disease recurrence or death by study group (disease-free survival [DFS] per protocol). (D) Kaplan–Meier estimates of the time to death by study group through the potentially maximum 5-yr duration of follow-up (ITT). ITT included all randomized patients minus three patients who withdrew after randomization but before surgery. P values are results from exact stratified log-rank tests. Per-Protocol, excluded patients with a benign or low-grade cancer histology, canceled or aborted surgery, usually due to extensive unresectable disease, or protocol drug deviation. DFS per-protocol population (C) further limited to those undergoing surgery as first course of treatment and who were disease-free after surgery.

otocol, excluded patients with a benign or low-grade cancer histology, canceled or aborted surgery, usually due to extensive unresectable disease, or protocol drug deviation. DFS per-protocol population (C) further limited to those undergoing surgery as first course of treatment and who were disease-free after surgery. Treatment effect on overall survival in subgroups. The forest plot shows hazard ratios for death and disease-free survival (cancer recurrence or death) and 95% CI in a comparison of propofol versus volatile anesthesia. The size of the diamond is proportional to the amount of statistical information in that category (i.e., the inverse of the variance). Intent-to-treat (ITT; n = 1,763) included all randomized patients minus three patients who withdrew after randomization but before surgery. Per protocol (PP, n = 1,411) excluded patients with a benign or low-grade cancer histology, canceled or aborted surgery, usually due to extensive unresectable disease, or protocol drug deviation. Disease-free survival (DFS) per-protocol population further limited to those undergoing surgery as first course of treatment and who were disease-free after surgery. Lung (lobectomy or pneumonectomy); pancreas (pancreaticoduodenectomy or distal pancreatectomy); esophagus (esophagectomy); liver (hepatectomy); biliary (radical cholecystectomy, bile duct resection); stomach (gastrectomy); bladder (radical cystectomy); peritoneal/hyperthermic intraperitoneal chemotherapy (HIPEC; cytoreductive and HIPEC).

Overall, 1,677 (95.9%) of 1,749 patients who had surgery received the assigned anesthetic exclusively. Only 72 (4.1%) of these patients in the propofol (n = 44) and volatile (n = 28) groups were excluded from the per-protocol analysis for an anesthesia treatment deviation (fig. 1). In some cases, these deviations were due to a medical indication, e.g., administration of volatile agent to treat acute bronchospasm in a patient assigned to the propofol group. The protocol allowed propofol to be administered to both groups for induction of general anesthesia, and dosing totals, median (interquartile range [IQR]), were 2,186 (1,470 to 3,238) and 160 (120 to 200) mg, respectively, in the propofol and volatile groups. Most patients in the volatile group (n = 882 for ITT population) received sevoflurane (n = 580; 65.8%).

No significant differences were observed between study groups in our surrogate measures of short-term outcome. In-hospital mortality occurred in the same number of patients (n = 11) in each study group. Postoperative hospital length of stay (median, IQR) was also similar: 5 (3 to 7) and 5 (3 to 8) days, respectively, in the propofol and volatile groups (P = 0.755 from Wilcoxon rank sum test with continuity correction).

Median follow-up for survivors was 55 (IQR, 39 to 60) months. In the ITT population, 1,713 (97.2%) patients had at least 2 yr of follow-up. As expected, overall mortality in ITT patients through 2 yr after randomization was highest in aborted cases (82 of 140; 58.6%) followed by surgical resection for recurrent disease (50 of 219; 22.8%) and surgery as part of the first course of therapy, i.e., resection of primary tumor (293 of 1,316; 22.3%). Mortality rates (ITT population, through 2 yr) for the eight planned surgical procedures were pancreas (38.7%), esophagus (34.3%), liver (22.7%), biliary (50.0%), stomach (27.8%), lung (13.6%), bladder (29.2%), and HIPEC (44.0%). In contrast to our primary hypothesis, propofol-treated patients did not exhibit better survival (propofol, 230 deaths of 881 [26.1%] vs. volatile, 202 deaths of 882 [22.9%]; hazard ratio, 1.16; 95% CI, 0.96 to 1.41; P = 0.115 by exact stratified log rank test; fig. 2A) in the ITT population (n = 1,763). In the per-protocol population (n = 1,411; fig. 2B), significantly more patients randomized to propofol died through 2-yr follow-up (25.5% vs. 20%; hazard ratio, 1.31; 95% CI, 1.05 to 1.64; P = 0.017). Results were similar for disease-free survival (fig. 2C) and even through the potential maximum 5 yr of follow-up (fig. 2D). These results were also generally consistent across the eight surgery type subgroups as well as for different case durations (fig. 3).

In this multicenter, pragmatic, randomized trial of 1,763 patients undergoing resection of biologically aggressive malignancies, we found no evidence that propofol-based anesthesia improves cancer-related outcomes compared to volatile anesthesia. Overall survival was not significantly different between the two groups (P = 0.115), but in the prespecified per-protocol analysis, more patients receiving propofol died (P = 0.017). These findings contradict preclinical studies and a previous meta-analysis of retrospective studies suggesting that propofol improves cancer-related outcomes.15 Our results align with the growing recognition that clinical translation of preclinical findings in oncology is often challenging.24 Numerous mechanistic12,13 and animal14 models have suggested that propofol-based anesthesia may be superior to volatile anesthesia with respect to immune-mediated clearing of tumor cells and development of metastases. A previous meta-analysis observed higher recurrence-free survival and overall survival in patients who received propofol anesthesia versus volatile anesthesia.15 Indeed, this potential benefit was observed in cancer surgery patients with metastatic as well as nonmetastatic disease (fig. 2c in Wigmore et al.25). Given the significant clinical implication of these findings, experts called for prospective scientific inquiry into whether anesthetic technique utilized during oncologic resections influences cancer-related outcomes.17

er surgery patients with metastatic as well as nonmetastatic disease (fig. 2c in Wigmore et al.25). Given the significant clinical implication of these findings, experts called for prospective scientific inquiry into whether anesthetic technique utilized during oncologic resections influences cancer-related outcomes.17 Several large trials26 and translational scientists are exploring this putative cancer-related beneficial effect of propofol. The only published multicenter trial that is somewhat similar to ours was conducted in China.27 This trial, which randomized patients to sevoflurane or propofol, initially (2018) published results for a 7-day cognitive function endpoint (single center, n = 392),28 and later (2023) reported results for overall survival and recurrence-free survival in 1,195 patients at 14 sites in China. Patients in this trial were generally lower-risk for morbidity and cancer recurrence. For example, they excluded ASA Physical Status IV patients, and 74.6% of their patients were ASA Physical Status I or II, compared with 15.9% ASA Physical Status I or II patients in our trial. They also enrolled many patients (44.4%) undergoing surgical resections for cancers with generally favorable outcomes, e.g., breast,27 colorectal,27 and prostate,27 which we specifically excluded from our study. In contrast to our trial, their study also enrolled some patients with less aggressive histology such as neuroendocrine and gastrointestinal stromal tumors, which we excluded. Additionally, they excluded patients receiving neoadjuvant therapy or neuraxial analgesia, both common practices in the United States. Finally, cancer diagnosis, cancer biology, and treatment may differ in China,29 further limiting the generalizability of their results. Notwithstanding the above, it is interesting to note that Cao et al.27 observed no benefit to propofol, and similar to our trial saw a trend toward worse cancer-related outcomes in patients randomized to propofol.

ncer biology, and treatment may differ in China,29 further limiting the generalizability of their results. Notwithstanding the above, it is interesting to note that Cao et al.27 observed no benefit to propofol, and similar to our trial saw a trend toward worse cancer-related outcomes in patients randomized to propofol. The observation that propofol was not superior to volatile anesthesia in our ITT as well as per-protocol analyses is important. These two separate analyses can be useful for different reasons. The ITT analysis, which is the accepted standard for many large randomized controlled trials, minimizes several sources of potential bias that might be introduced by investigators/clinicians; however, it can include patients with factors, e.g., no confirmed cancer at surgery, that could dilute the effect size. In contrast, the per-protocol analysis is an attempt to give the intervention, i.e., propofol, the best chance of showing a benefit if it does exist. For this reason, it made sense for us to exclude patients from the per-protocol analysis if their pathology was benign, their surgery was aborted, or they did not receive the assigned anesthetic drug. In addition, while volatile anesthesia was statistically significantly better in our secondary (per-protocol) analysis, this was statistically fragile; therefore, our interpretation is that the per-protocol analysis (similar to the ITT analysis) demonstrates that propofol anesthesia does not confer a beneficial effect.

In addition, while volatile anesthesia was statistically significantly better in our secondary (per-protocol) analysis, this was statistically fragile; therefore, our interpretation is that the per-protocol analysis (similar to the ITT analysis) demonstrates that propofol anesthesia does not confer a beneficial effect. Our study has several limitations. We were not able to blind the anesthesiologist in the operating room to study group. However, the delivered intraoperative care did not appear to be affected by this knowledge as reflected by the high degree of balance in potential confounders between study arms (table 2). Furthermore, our outcome measures were determined many months or years after surgery by individuals (NDI, New York Cancer Registry) who were not aware of study group assignment. Since our study was focused on clinical outcomes, another potential limitation was the lack of mechanistic data that could help explain the observation of no benefit with propofol despite a well-powered study/sample size and many events. In addition, while the groups were well balanced between neoadjuvant and adjuvant therapies (table 3), we did not record or analyze specific adjuvant cycle counts or chemotherapy agents, or subsequent cancer treatments after recurrence. Given the heterogeneity of the cancers in our study population and the widely varying treatment options for these different cancers, analysis of this would be underpowered and limit any statistical inferences. Moreover, our study was not specifically powered to detect differences by surgical procedure type. Nonetheless, we believe these data are important to share with readers but encourage caution in their interpretation as the sample size was small in several of these groups. Finally, we acknowledge that prognostic factors could have differed for the small number of participants who did versus did not adhere to treatment assignment, or whose procedure was versus was not aborted, and that lack of adjustment for such factors in the per-protocol analysis could introduce some bias into the estimates reported.

nowledge that prognostic factors could have differed for the small number of participants who did versus did not adhere to treatment assignment, or whose procedure was versus was not aborted, and that lack of adjustment for such factors in the per-protocol analysis could introduce some bias into the estimates reported. Perioperative Characteristics (ITT Population) Intent-to-treat (ITT) population, all patients randomized minus three patients who withdrew consent after randomization but before surgery. Neuraxial, epidural or spinal. HIPEC, hyperthermic intraperitoneal chemotherapy; IQR, interquartile range. Cancer Staging and Related Treatments and Surgical Characteristics for First Course of Treatment First course of treatment, defined by New York Cancer Registry as initial treatment plan before disease recurrence or progression. Does not include hepatectomy and hyperthermic intraperitoneal chemotherapy surgeries. AJCC, American Joint Committee on Cancer; SEER, Surveillance, Epidemiology, and End Results.

Cancer Staging and Related Treatments and Surgical Characteristics for First Course of Treatment First course of treatment, defined by New York Cancer Registry as initial treatment plan before disease recurrence or progression. Does not include hepatectomy and hyperthermic intraperitoneal chemotherapy surgeries. AJCC, American Joint Committee on Cancer; SEER, Surveillance, Epidemiology, and End Results. Our study has numerous strengths. Its pragmatic design allowed us to test our hypothesis in a real-world setting, with very few exclusion criteria, thus increasing external validity. We recorded numerous potential confounders, and these were well-balanced between study groups due to effective randomization. In addition, we enrolled patients undergoing a diverse group of surgical procedures, with a focus on cancers with high recurrence rates and poor outcomes. This is in contrast to other studies that focused on patients with fewer comorbidities27 and/or undergoing surgical resections for cancers with generally more favorable outcomes, e.g., breast,27,30 colorectal,27 prostate,27 which we specifically excluded from our study. Furthermore, we observed high adherence to the protocol with 95.9% of patients receiving the assigned anesthetic. Finally, the study had high rates of follow-up with 97.2% of patients receiving at least 2 yr of the minimum follow-up. In conclusion, our trial does not support the hypothesis that propofol-based anesthesia improves cancer-related outcomes in patients undergoing resection of biologically aggressive malignancies.

Our study has numerous strengths. Its pragmatic design allowed us to test our hypothesis in a real-world setting, with very few exclusion criteria, thus increasing external validity. We recorded numerous potential confounders, and these were well-balanced between study groups due to effective randomization. In addition, we enrolled patients undergoing a diverse group of surgical procedures, with a focus on cancers with high recurrence rates and poor outcomes. This is in contrast to other studies that focused on patients with fewer comorbidities27 and/or undergoing surgical resections for cancers with generally more favorable outcomes, e.g., breast,27,30 colorectal,27 prostate,27 which we specifically excluded from our study. Furthermore, we observed high adherence to the protocol with 95.9% of patients receiving the assigned anesthetic. Finally, the study had high rates of follow-up with 97.2% of patients receiving at least 2 yr of the minimum follow-up. In conclusion, our trial does not support the hypothesis that propofol-based anesthesia improves cancer-related outcomes in patients undergoing resection of biologically aggressive malignancies. Support was provided solely from institutional and departmental sources. The authors declare no competing interests. Full protocol available at: Elliott.Bennett-Guerrero@StonyBrookMedicine.edu. Raw data available at: Elliott.Bennett-Guerrero@StonyBrookMedicine.edu.