Browse the corpus

First genome-wide association study on rocuronium dose requirements shows association with SLCO1A2. BACKGROUND: Rocuronium, a common neuromuscular blocking agent, is mainly excreted unchanged in urine (10-25%) and bile (>70%). Age, sex, liver blood flow, smoking, medical conditions, and ethnic background can affect its pharmacological actions. However, reasons for the wide variation in rocuronium requirements are mostly unknown. We hypothesised that pharmacogenetic factors might explain part of the variation. METHODS: One thousand women undergoing surgery for breast cancer were studied. Anaesthesia was maintained with propofol (50-100 μg kg-1 min-1) and remifentanil (0.05-0.25 μg kg-1 min-1). Neuromuscular block was maintained with rocuronium to keep the train-of-four ratio at 0-10%. DNA was extracted from peripheral blood and genotyped with a next-generation genotyping array. The genome-wide association study (GWAS) was conducted using an additive linear regression model with PLINK software. The FINEMAP tool and data from the Genotype-Tissue Expression project v8 were utilised to study the locus further. RESULTS: The final patient population comprised 918 individuals. Of the clinical variables tested, age, BMI, ASA physical status, and total dose of propofol correlated significantly (all P<0.001) with the rocuronium dose in a linear regression model. The GWAS highlighted one genome-wide significant locus in chromosome 12. The single-nucleotide polymorphisms (SNPs) with the most significant evidence of association were located in or near SLCO1A2. The two top SNPs, rs7967354 (P=5.3e-11) and rs11045995 (P=1.4e-10), and the clinical variables accounted for 41% of the variability in rocuronium dosage. CONCLUSIONS: Genetic variation in the gene SLCO1A2, encoding OATP1A2, an uptake transporter, accounted for 4% of the variability in rocuronium consumption. The underlying mechanism remains unknown.

BrePainGen is a prospective study designed to examine the role of genetics in acute and persistent post-surgical pain, mood, and effects of drugs used in anaesthesia. One thousand women undergoing surgery for breast cancer at the Helsinki University Hospital were recruited between August 1, 2006 and December 31, 2010. The study protocol was approved by the coordinating ethics committee (136/E0/2006) and the ethics committee of the Department of Surgery (Dnro 148/E6/05) of the Hospital District of Helsinki and Uusimaa. Written informed consent was obtained from all patients. A flow chart of patient recruitment is provided in Supplementary Fig. 1. After informed consent, the patients were interviewed for background information about medical conditions, medications, age, height, weight, previous operations, use of alcohol, and smoking. Those using drugs affecting the pharmacology of rocuronium were excluded.

BrePainGen is a prospective study designed to examine the role of genetics in acute and persistent post-surgical pain, mood, and effects of drugs used in anaesthesia. One thousand women undergoing surgery for breast cancer at the Helsinki University Hospital were recruited between August 1, 2006 and December 31, 2010. The study protocol was approved by the coordinating ethics committee (136/E0/2006) and the ethics committee of the Department of Surgery (Dnro 148/E6/05) of the Hospital District of Helsinki and Uusimaa. Written informed consent was obtained from all patients. A flow chart of patient recruitment is provided in Supplementary Fig. 1. After informed consent, the patients were interviewed for background information about medical conditions, medications, age, height, weight, previous operations, use of alcohol, and smoking. Those using drugs affecting the pharmacology of rocuronium were excluded. All patients were premedicated with diazepam 2.5–15 mg and paracetamol 1 g orally. Anaesthesia was induced with propofol 2–3 mg kg−1, and remifentanil infusion of 0.2 mg kg−1 min−1 was started. Tracheal intubation was facilitated with rocuronium 0.6 mg kg−1. During surgery, anaesthesia was maintained with a propofol infusion at 50–100 μg kg−1 min−1 to keep state entropy (M-Entropy S/5TM Module for Anesthesia Monitor; GE Healthcare Finland, Helsinki, Finland) at the level of 50 [5]. Remifentanil infusion was used at 0.05–0.25 μg kg−1 min−1 to keep systolic BP at [15%] of baseline minus 20 mm Hg. The neuromuscular block was maintained throughout the surgery with rocuronium boluses of 10 mg to keep the train-of-four ratio at 0–10% (E-NMT; GE Healthcare Finland). Mechanical ventilation was adjusted to normocapnia with 1:1 oxygen and nitrous oxide. During closure of the skin, remifentanil infusion was stopped, and boluses of fentanyl 1 μg kg−1, ondansetron 4 mg, and droperidol 0.01 mg kg−1 were given intravenously. Neuromuscular block was reversed with neostigmine 2.5 mg and glycopyrrolate 0.5 mg. Before the patient woke from anaesthesia, a blood specimen was drawn for DNA isolation.

skin, remifentanil infusion was stopped, and boluses of fentanyl 1 μg kg−1, ondansetron 4 mg, and droperidol 0.01 mg kg−1 were given intravenously. Neuromuscular block was reversed with neostigmine 2.5 mg and glycopyrrolate 0.5 mg. Before the patient woke from anaesthesia, a blood specimen was drawn for DNA isolation. DNA was extracted from peripheral blood using the Autopure LS™ automated DNA purification instrument (Gentra Systems, Inc., Minneapolis, MN, USA). Genotype data were produced at the Wellcome Sanger Institute (Hinxton, UK) on the HumanOmniExpress Illumina BeadChip (Illumina, Inc., San Diego, CA, USA) while blind to phenotypic information. Sample quality control procedures have been described in detail earlier.20 All single-nucleotide polymorphisms (SNPs) were filtered based on minor allele frequency (MAF >0.0005), Hardy–Weinberg equilibrium (P>1 × 10−6), and success rate (>0.97). The mean genotyping success rate was 0.997. After quality control, genotyping data were available for 926 of the 1000 participants. Eight patients were excluded for clinical reasons. The final participant population comprised 918 individuals with both genotype and clinical data available (Table 1, Table 2).Table 1Characteristics of the subjects. n=992 for all included in the association testing, and n=918 for genetic testing based on the number of genetic data available.Table 1All patients, % (n)Final genetic analysesASA physical status 15.6 (56)5.4 (50) 282.5 (818)82.8 (760) 311.9 (118)11.8 (108)Type of surgery in the axilla Sentinel node biopsy55.8 (554)55 (505) Evacuation of axilla44.2 (438)45 (413)Type of surgery in the breast Resection62.4 (619)62.2 (571) Mastectomy37.6 (373)37.8 (347)Smoking habits No82 (813)82.5 (757) Yes18 (179)17.5 (161)Alcohol use No17.5 (173)18.1 (166) Yes82.5 (819)81.9 (752)Table 2Characteristics of the patients, anaesthesia, and total dose of rocuronium; n=918.Table 2Median (inter-quartile range)Age (yr)58 (50–64)Height (cm)164 (160–169)Weight (kg)68 (60–76)BMI (kg m-2)24.8 (22.3–28.2)Total dose of rocuronium (mg)75 (60–90)Time to first additional dose after induction (min)38 (30–46)Time to recovery after last dose (min)47 (38–57)Duration of anaesthesia (min)138 (115–168)Dose of rocuronium (μg kg−1 min−1)7.9 (6.5–9.7)Total dose of remifentanil during anaesthesia (mg)0.8 (0.6–1.1)Dose of remifentanil (μg kg−1 min−1)0.008 (0.007–0.010)Total dose of propofol during anaesthesia (mg)897 (718–1102)Dose of propofol (μg kg−1 min−1)95 (83–110)

47 (38–57)Duration of anaesthesia (min)138 (115–168)Dose of rocuronium (μg kg−1 min−1)7.9 (6.5–9.7)Total dose of remifentanil during anaesthesia (mg)0.8 (0.6–1.1)Dose of remifentanil (μg kg−1 min−1)0.008 (0.007–0.010)Total dose of propofol during anaesthesia (mg)897 (718–1102)Dose of propofol (μg kg−1 min−1)95 (83–110) Characteristics of the subjects. n=992 for all included in the association testing, and n=918 for genetic testing based on the number of genetic data available. Characteristics of the patients, anaesthesia, and total dose of rocuronium; n=918.

47 (38–57)Duration of anaesthesia (min)138 (115–168)Dose of rocuronium (μg kg−1 min−1)7.9 (6.5–9.7)Total dose of remifentanil during anaesthesia (mg)0.8 (0.6–1.1)Dose of remifentanil (μg kg−1 min−1)0.008 (0.007–0.010)Total dose of propofol during anaesthesia (mg)897 (718–1102)Dose of propofol (μg kg−1 min−1)95 (83–110) Characteristics of the subjects. n=992 for all included in the association testing, and n=918 for genetic testing based on the number of genetic data available. Characteristics of the patients, anaesthesia, and total dose of rocuronium; n=918. Statistical analyses and data management were conducted using IBM SPSS software versions 23.0 and 24.0 and R version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria). To identify possible confounders that should be taken into account when performing the GWAS, we first performed univariate testing (n=992) between clinical variables and the dose of rocuronium needed to maintain adequate neuromuscular block. We took the dose rates of rocuronium (in units of mg kg−1 min−1) using the natural logarithms of their numerical values to ensure normal distribution. The clinical variables to be tested for relevance to rocuronium dose requirements were age, height, BMI, total use of propofol during anaesthesia (mg kg−1 min−1), total use of remifentanil during anaesthesia (mg kg−1 min−1), ASA class, smoking (yes/no), alcohol use (yes/abstinent), type of axillary surgery (sentinel node biopsy/evacuation), and breast surgery (resection/mastectomy). We also tested associations between CYP2D6 copy numbers, CYP2D6-predicted phenotype (poor, intermediate, extensive, or ultra-rapid metabolisers), CYP3A4 variant rs35599367 genotype (CC/CT/TT) and CYP3A5 variant rs776746 genotype (GG/AG/AA), and the natural logarithm of rocuronium dose rates (expressed in mg kg−1 min−1), as the CYP data were available from these patients.21 We tested these associations using t-test, Mann–Whitney U-test, analysis of variance, or Kruskal–Wallis test, depending on the distributions of variables. Pairwise comparisons were adjusted with Bonferroni corrections.

ronium dose rates (expressed in mg kg−1 min−1), as the CYP data were available from these patients.21 We tested these associations using t-test, Mann–Whitney U-test, analysis of variance, or Kruskal–Wallis test, depending on the distributions of variables. Pairwise comparisons were adjusted with Bonferroni corrections. After studying possible associations, we conducted multivariate linear regression modelling based on variables that had statistically significant (P<0.05) associations with rocuronium requirements. We used a stepwise method to construct the final model. The final linear regression model contained only variables that remained significant in multivariate testing. These were used as covariates in the GWAS. In addition, the first five dimensions from multidimensional scaling of genotype data were also used as covariates to take into account a possible hidden population structure. The GWAS was conducted using an additive linear regression model with PLINK software.22 Associations between total dose of rocuronium and 653 034 genetic variants (SNPs) were tested. The standard threshold of genome-wide statistical significance, P<5 × 10−8, was used. After GWAS results became available, we performed another linear regression round, including four lead variants, to examine the impact of these on the total dose of rocuronium.

curonium and 653 034 genetic variants (SNPs) were tested. The standard threshold of genome-wide statistical significance, P<5 × 10−8, was used. After GWAS results became available, we performed another linear regression round, including four lead variants, to examine the impact of these on the total dose of rocuronium. The genomic region showing a significant association with rocuronium dosage was further examined to identify the most likely causal SNPs within the locus. For this, the genomic data were first pre-phased with Eagle software version 2.4.23 Subsequently, the genotypes were imputed using Beagle 4.1 and population-specific Sequencing Initiative Suomi panel as imputation reference.24,25 Poorly imputed variants were excluded (INFO <0.7). The imputation reference panel consisted of 3775 Finns. To identify the number of independent association signals and the lead SNPs within the associated locus, the FINEMAP26 tool was used. The FUMA27 tool was used to examine the potential functional effect of each associated SNP. Genetic effects on gene expression across tissue types were studied using publicly available data from the Genotype-Tissue Expression project v8.28 FINEMAP 1.426 was run, allowing a maximum of K=5 causal variants. The credible sets, those containing the most likely causal variants, were reported, assuming either one or two causal variants.

To test associations with clinical variables, all 992 individuals with complete clinical data were used. For 918 of these, genome-wide genotype data were available and used for the GWAS. The characteristics of the subjects and of rocuronium requirements are shown in Table 1, Table 2. The total dose of rocuronium had a linear relationship (R2=0.412) with the duration of anaesthesia (Fig. 1a). The median number of additional doses was 4 (inter-quartile range [IQR]: 2–5), range 0–20. Twenty-one patients did not receive additional doses of rocuronium after intubation.Fig 1Relationship between rocuronium dose and (a) duration of anaesthesia and (b–d) the two lead SNPs showing genome-wide significant association with rocuronium dose in the GWAS. (a) Rocuronium requirements of the study participants plotted against the duration of anaesthesia. Patients homozygous for the minor alleles of both of the lead SNPs (rs7967354-G and rs11045995-G) are highlighted in blue. Mean rocuronium dosage in this group is lower than the average, and the individual data points are also clearly at the lower end of the requirement distribution. (b) Box and whisker plot showing rocuronium requirements during anaesthesia classified by rs7967354 genotype. The middle line of the box represents median, and the upper and lower edges of the box represent upper and lower quartiles, respectively. Upper whisker represents the minimum between the maximum value and the upper quartile+1.5 ∗ inter-quartile range. Lower whisker represents the maximum between the minimum value and the lower quartile–1.5 ∗ inter-quartile range. Any values outside the whiskers are outliers, presented as circles. (c) Box and whisker plot showing the rocuronium requirements during anaesthesia according to rs11045995 genotype. (d) Box and whisker plot showing rocuronium requirements during anaesthesia classified by number of rs7967354-G and rs11045995-G alleles. Subjects with the homozygous genotype GG+GG in both rs7967354 and rs11045995 (minor allele is G in both variants) needed significantly less rocuronium during anaesthesia than subjects with other genotypes. The same individuals are highlighted in blue in Fig. 2a.

sified by number of rs7967354-G and rs11045995-G alleles. Subjects with the homozygous genotype GG+GG in both rs7967354 and rs11045995 (minor allele is G in both variants) needed significantly less rocuronium during anaesthesia than subjects with other genotypes. The same individuals are highlighted in blue in Fig. 2a. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism.Fig 1

sified by number of rs7967354-G and rs11045995-G alleles. Subjects with the homozygous genotype GG+GG in both rs7967354 and rs11045995 (minor allele is G in both variants) needed significantly less rocuronium during anaesthesia than subjects with other genotypes. The same individuals are highlighted in blue in Fig. 2a. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism.Fig 1 Relationship between rocuronium dose and (a) duration of anaesthesia and (b–d) the two lead SNPs showing genome-wide significant association with rocuronium dose in the GWAS. (a) Rocuronium requirements of the study participants plotted against the duration of anaesthesia. Patients homozygous for the minor alleles of both of the lead SNPs (rs7967354-G and rs11045995-G) are highlighted in blue. Mean rocuronium dosage in this group is lower than the average, and the individual data points are also clearly at the lower end of the requirement distribution. (b) Box and whisker plot showing rocuronium requirements during anaesthesia classified by rs7967354 genotype. The middle line of the box represents median, and the upper and lower edges of the box represent upper and lower quartiles, respectively. Upper whisker represents the minimum between the maximum value and the upper quartile+1.5 ∗ inter-quartile range. Lower whisker represents the maximum between the minimum value and the lower quartile–1.5 ∗ inter-quartile range. Any values outside the whiskers are outliers, presented as circles. (c) Box and whisker plot showing the rocuronium requirements during anaesthesia according to rs11045995 genotype. (d) Box and whisker plot showing rocuronium requirements during anaesthesia classified by number of rs7967354-G and rs11045995-G alleles. Subjects with the homozygous genotype GG+GG in both rs7967354 and rs11045995 (minor allele is G in both variants) needed significantly less rocuronium during anaesthesia than subjects with other genotypes. The same individuals are highlighted in blue in Fig. 2a. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism.

e homozygous genotype GG+GG in both rs7967354 and rs11045995 (minor allele is G in both variants) needed significantly less rocuronium during anaesthesia than subjects with other genotypes. The same individuals are highlighted in blue in Fig. 2a. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism. Of the continuous variables tested, age (P<0.001), BMI (P<0.001), and total doses of propofol (mg kg−1 min−1; P<0.001) and of remifentanil (mg kg−1 min−1; P<0.001) had significant correlations with the dose of rocuronium (in mg kg−1 min−1). Of the dichotomous and ordinal variables, ASA class (P<0.001), use of alcohol compared with abstinence (P=0.001), and breast resection vs mastectomy (P<0.001) were statistically significant at P<0.05.

d of remifentanil (mg kg−1 min−1; P<0.001) had significant correlations with the dose of rocuronium (in mg kg−1 min−1). Of the dichotomous and ordinal variables, ASA class (P<0.001), use of alcohol compared with abstinence (P=0.001), and breast resection vs mastectomy (P<0.001) were statistically significant at P<0.05. Multiple linear regression models were run to understand the effects of the aforementioned variables on the dose of rocuronium (in mg kg−1 min−1). There was homoscedasticity as assessed by visual inspection of a plot of standardised residuals vs standardised predicted values, normality of the residuals being assessed by visual inspection of a normal probability plot: there were no significant outliers, as assessed by Cook's distance. The results of the linear model are presented in Table 3. The clinical variables in the model accounted for 35.3% of the variability in rocuronium dosage.Table 3Linear regression model with statistically significant clinical variables and statistically significant clinical variables and genetic variants with a dose-altering effect. The tested variable was rocuronium dose (in mg kg−1 min−1).Table 3ΒP-value95% CI lower bound95% CI upper boundClinical variablesAge (yr)–0.004<0.001–0.006–0.003BMI–0.029<0.001–0.032–0.025ASA physical status 10.0880.0190.0140.16ASA physical status 20.060.0090.0150.10Propofol (mg kg−1 min−1)2.0<0.0011.32.7Adjusted R2=0.35Clinical variables and genetic variantsAge (yr)–0.005<0.001–0.006–0.003BMI–0.028<0.001–0.032–0.025ASA physical status 10.099<0.0010.0250.17ASA physical status 20.0750.0080.0310.12Propofol (mg kg−1 min−1)1.9<0.0011.22.6rs7967354–0.0430.001–0.069–0.017rs11045995–0.090<0.001–0.12–0.062Adjusted R2=0.41CI, confidence interval.

nical variables and genetic variantsAge (yr)–0.005<0.001–0.006–0.003BMI–0.028<0.001–0.032–0.025ASA physical status 10.099<0.0010.0250.17ASA physical status 20.0750.0080.0310.12Propofol (mg kg−1 min−1)1.9<0.0011.22.6rs7967354–0.0430.001–0.069–0.017rs11045995–0.090<0.001–0.12–0.062Adjusted R2=0.41CI, confidence interval. Linear regression model with statistically significant clinical variables and statistically significant clinical variables and genetic variants with a dose-altering effect. The tested variable was rocuronium dose (in mg kg−1 min−1). CI, confidence interval.

nical variables and genetic variantsAge (yr)–0.005<0.001–0.006–0.003BMI–0.028<0.001–0.032–0.025ASA physical status 10.099<0.0010.0250.17ASA physical status 20.0750.0080.0310.12Propofol (mg kg−1 min−1)1.9<0.0011.22.6rs7967354–0.0430.001–0.069–0.017rs11045995–0.090<0.001–0.12–0.062Adjusted R2=0.41CI, confidence interval. Linear regression model with statistically significant clinical variables and statistically significant clinical variables and genetic variants with a dose-altering effect. The tested variable was rocuronium dose (in mg kg−1 min−1). CI, confidence interval. The GWAS highlighted one locus on chromosome 12 showing genome-wide significant evidence of association with rocuronium dose (Fig. 2). Eight genotyped SNPs reached the standard threshold of genome-wide statistical significance, P<5 × 10−8 (Table 4). The SNPs with the most significant evidence of association were all located in or near gene SLCO1A2. LocusZoom plots29 of the area are shown in Supplementary Fig. 2a and b. Based on the genetic recombination patterns and the FINEMAP tool, the most likely scenario is that two SNPs are needed to explain the association. The top candidates are rs7967354 (P=5.3e−11; β=–0.143 for allele G; MAF=0.22) and rs11045995 (P=1.4e−10; β=–0.147 for allele G; MAF=0.18), both located in gene SLCO1A2 and in moderate linkage disequilibrium (LD) with each other (r2=0.26). The minor alleles of both variants (rs7967354-G and rs11045995-G) are associated with decreased rocuronium requirements.Fig 2Manhattan plot showing the –log10P-values of all 653 034 SNPs tested in the GWAS, plotted against their respective chromosomal positions. The results highlight one genome-wide significant locus on chromosome 12. The quantile–quantile plot in the upper-right corner of the image shows the deviation of the observed from the expected P-values under the null hypothesis of no association.

ted in the GWAS, plotted against their respective chromosomal positions. The results highlight one genome-wide significant locus on chromosome 12. The quantile–quantile plot in the upper-right corner of the image shows the deviation of the observed from the expected P-values under the null hypothesis of no association. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism.Fig 2Table 4Dose-altering variants found in this genome-wide association analysis.Table 4ChrSNPGeneMinor alleleMAFβP-valueSNP isr2 rs7967354r2 rs1104599512rs7967354SLCO1A2G0.22–0.14255.302e−11Intron variant10.26151612rs4149005SLCO1A2A0.22–0.14176.809e−11Non-coding transcript exon variant0.9969180.26306912rs875234SLCO1A2G0.22–0.141.129e−103′ UTR variant0.9969190.26047812rs11045995SLCO1A2G0.18–0.14731.416e−10Intron variant0.261516112rs10743413SLCO1A2G0.20–0.14242.516e−10Intron variant0.2335910.87346912rs10841798SLCO1A2C0.18–0.14176.672e−10Intron variant0.2599250.98267212rs10770800SLCO1A2G0.18–0.13991.228e−9Intron variant0.2663670.9860312rs10841782SLCO1A2A0.08–0.17578.452e−8Intron variant0.3221230.316403Chr, chromosome; MAF, minor allele frequency in the study sample; r2 squared correlation coefficient measure for linkage disequilibrium, range 0–1; SNP, single-nucleotide polymorphism.

2rs10770800SLCO1A2G0.18–0.13991.228e−9Intron variant0.2663670.9860312rs10841782SLCO1A2A0.08–0.17578.452e−8Intron variant0.3221230.316403Chr, chromosome; MAF, minor allele frequency in the study sample; r2 squared correlation coefficient measure for linkage disequilibrium, range 0–1; SNP, single-nucleotide polymorphism. Manhattan plot showing the –log10P-values of all 653 034 SNPs tested in the GWAS, plotted against their respective chromosomal positions. The results highlight one genome-wide significant locus on chromosome 12. The quantile–quantile plot in the upper-right corner of the image shows the deviation of the observed from the expected P-values under the null hypothesis of no association. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism. Dose-altering variants found in this genome-wide association analysis. Chr, chromosome; MAF, minor allele frequency in the study sample; r2 squared correlation coefficient measure for linkage disequilibrium, range 0–1; SNP, single-nucleotide polymorphism.

Manhattan plot showing the –log10P-values of all 653 034 SNPs tested in the GWAS, plotted against their respective chromosomal positions. The results highlight one genome-wide significant locus on chromosome 12. The quantile–quantile plot in the upper-right corner of the image shows the deviation of the observed from the expected P-values under the null hypothesis of no association. GWAS, genome-wide association study; SNP, single-nucleotide polymorphism. Dose-altering variants found in this genome-wide association analysis. Chr, chromosome; MAF, minor allele frequency in the study sample; r2 squared correlation coefficient measure for linkage disequilibrium, range 0–1; SNP, single-nucleotide polymorphism. Patients with two minor alleles (G/G) of the variant rs7967354 needed significantly less rocuronium during anaesthesia (n=47; median dose rate: 6.1 μg kg−1 min−1; IQR: 5.5–7.5) compared with patients with G/A genotype (n=310; median dose rate: 7.7 μg kg−1 min−1; IQR: 6.4–9.5) and A/A genotype (n=561; median dose rate: 8.2 μg kg−1 min−1; IQR: 6.3–9.1) (Fig. 1b) in Kruskal–Wallis testing (P<0.001). Similarly, patients with rs11045995 G/G genotype needed significantly lower doses of rocuronium during anaesthesia (n=37; median dose rate: 6.3 μg kg−1 min−1; IQR: 5.6–7.0; P<0.001) than the patients with G/A (n=254; median dose rate: 7.5 μg kg−1 min−1; IQR: 6.3–9.1) and A/A genotypes (n=627; median dose rate: 8.3 μg kg−1 min−1; IQR: 6.8–10.1) (Fig. 1c). We also found that patients with two minor alleles in both dose-altering variants (rs7967354-GG + rs11045995-GG; n=19; median dose rate: 5.9 [IQR: 5.4–6.5] μg kg−1 min−1) needed even less rocuronium during anaesthesia than with other genotype combinations in the Kruskal–Wallis test (Fig. 1a and d; Supplementary Tables 1 and 2).

ound that patients with two minor alleles in both dose-altering variants (rs7967354-GG + rs11045995-GG; n=19; median dose rate: 5.9 [IQR: 5.4–6.5] μg kg−1 min−1) needed even less rocuronium during anaesthesia than with other genotype combinations in the Kruskal–Wallis test (Fig. 1a and d; Supplementary Tables 1 and 2). The final linear regression model included both the clinical variables and the two lead SNPs rs7967354 and rs11045995 (Table 3). The variables in the model accounted for 41% of the variability (adjusted R2) in the rocuronium dose. In this multivariate analysis, age (P<0.001), BMI (P<0.001), ASA 1 vs ASA 3 (P=0.008), ASA 2 vs ASA 3 (P=0.001), dose of propofol (P<0.001), rs7967354 (P=0.001), and rs11045995 (P<0.001) remained significant at P<0.05. By including the two SNPs in the model, the proportion of the variance explained increased by 4 percentage points.

variate analysis, age (P<0.001), BMI (P<0.001), ASA 1 vs ASA 3 (P=0.008), ASA 2 vs ASA 3 (P=0.001), dose of propofol (P<0.001), rs7967354 (P=0.001), and rs11045995 (P<0.001) remained significant at P<0.05. By including the two SNPs in the model, the proportion of the variance explained increased by 4 percentage points. As both rs7967354 and rs11045995 are located in the intronic regions of SLCO1A2, we analysed the region further using imputed genome data, and FINEMAP and FUMA programs, to locate other variants possibly driving the association signal. On the imputed data, FINEMAP gives a probability of 74% to one causal variant and 26% for two causal variants. Assuming one causal variant, the 95% credible set (the set of variants containing the causal variant with 95% probability) contains 20 variants (Supplementary Table 3). Assuming two causal variants, the lead variants are rs7967354 and rs10743413 (the latter of which is highly correlated with rs11045995; r2=0.88), and the two credible sets contain, respectively, 13 and 5 variants with probability over 1% of being causal (Supplementary Table 4). FINEMAP estimates that the SLCO1A2 region explains 3.75% of the variance of the phenotype (95% credible interval: 1.77–6.29%).

ter of which is highly correlated with rs11045995; r2=0.88), and the two credible sets contain, respectively, 13 and 5 variants with probability over 1% of being causal (Supplementary Table 4). FINEMAP estimates that the SLCO1A2 region explains 3.75% of the variance of the phenotype (95% credible interval: 1.77–6.29%). Next, we performed FUMA analysis to check whether the most likely causal variants were associated with tissue-specific changes in the expression level of SLCO1A2 or other genes. For the 20 potentially causal variants, FUMA analysis detected 39 expression quantitative trait loci (eQTL) for four genes at a false discovery rate of 5% (Supplementary Table 5). For SLCO1A2, three variants were eQTL in the brain and three in the cerebellum. Other eQTL were found for the gene RECQL (expression measured in blood), PYROXD1 (in blood), and C12orf39 (in lymphocytes and in blood). Plots for tissue-wide expression results for rs7967354 and rs11045995 are in Supplementary Fig. 3a and b.

We explored clinical and genetic factors explaining variation in rocuronium requirement during surgery for breast cancer in 918 women. We showed that a locus containing gene SLCO1A2 affects the dose rate needed for maintaining adequate neuromuscular block. Of the clinical variables examined, age, BMI, total dose of propofol, and ASA class were associated with the rocuronium dose. Combined, these factors explained 41% of the dose rate variation. Our study provides further confirmation that neither CYP2D6 nor CYP3A4 plays a role in determining rocuronium requirements. The median rocuronium dose used is in line with previous reports.30,31 In our study, advanced age,12 lower BMI, and higher ASA class decreased the amount of rocuronium needed. Previous reports on the effect of BMI on rocuronium requirements are conflicting.32,33 Our patients were medicated according to their actual body weight, which might explain why lower BMI decreased rocuronium requirements. We observed that higher propofol doses were associated with increased need for rocuronium, whereas some previous studies suggested that propofol would have muscle-relaxing effects,34 reducing the required dose of neuromuscular blockers.35 However, the designs of these studies were very different from ours. There are no previous linear regression models evaluating rocuronium needs during propofol anaesthesia.

whereas some previous studies suggested that propofol would have muscle-relaxing effects,34 reducing the required dose of neuromuscular blockers.35 However, the designs of these studies were very different from ours. There are no previous linear regression models evaluating rocuronium needs during propofol anaesthesia. Our GWAS identified one genome-wide significant association peak, on chromosome 12, in and around the SLCO1A2 gene, which encodes the OATP1A2. The signal was best explained by two lead SNPs, rs7967354 and rs11045995. Higher numbers of the minor alleles of these SNPs were associated with a lesser need for rocuronium. The biggest variation in rocuronium dosage was observed when we compared participants homozygous for both rs7967354 and rs11045995 minor alleles (G) with those homozygous for the major alleles (A) (Fig. 1a and d; Supplementary Tables 1 and 2). Organic anion transporters are cellular transmembrane proteins, important in the distribution, metabolism, and excretion of various drugs and expressed in pharmacokinetically important organs, such as liver, kidney, and intestine.36 Based on immunohistochemical staining, OATP1A2 transporters are located in cholangiocytes, where they have an important role in excretion of drugs into the bile.37 As rocuronium is a known substrate of OATP1A236 and is mainly excreted unchanged in the urine (10–25%) and bile (>70%),1,2 the role of OATPs in its excretion is of interest.

stochemical staining, OATP1A2 transporters are located in cholangiocytes, where they have an important role in excretion of drugs into the bile.37 As rocuronium is a known substrate of OATP1A236 and is mainly excreted unchanged in the urine (10–25%) and bile (>70%),1,2 the role of OATPs in its excretion is of interest. A study with Slco1a/1b–/– knockout mice showed that lack of functioning OATP1A2 leads to accumulation of the substrates of this transporter in plasma.38 Previous studies also indicate that OATPs are a target for drug interactions. Expression of OATPs, especially OATP-A, were significantly increased in patients treated with carbamazepine.39 Carbamazepine use is known to increase the required rocuronium dosage.40 Our results suggest that this would be attributable to induction of OATP1A2 rather than of CYP3A4. Few previous studies have addressed the pharmacogenetics of rocuronium. A candidate gene study based on only 30 patients, by Costa and colleagues,41 showed evidence of association between a variant –189_188InsA (rs3834939), located in the promoter region of the SLCO1A2 gene, and reduced clearance of rocuronium. Neither this SNP rs3834939 (P=0.000026 in our study) nor the other variants tested by Costa and colleagues were amongst the SNPs showing genome-wide significant evidence of association (a standard threshold of genome-wide statistical significance is P<5 × 10−8).

of the SLCO1A2 gene, and reduced clearance of rocuronium. Neither this SNP rs3834939 (P=0.000026 in our study) nor the other variants tested by Costa and colleagues were amongst the SNPs showing genome-wide significant evidence of association (a standard threshold of genome-wide statistical significance is P<5 × 10−8). The other two earlier rocuronium studies were also candidate gene studies, analysing only a few variants in small patient samples. Based on a sample of 105 Chinese patients, Qi and colleagues42 reported that SNPs rs12720464 and rs1055302 in the ABCB1 gene, coding for an ATP-dependent drug efflux pump, associate with prolonged spontaneous recovery after a single dose of rocuronium. In the study of Mei and colleagues,43 another ABCB1 SNP (rs1128503) and an SNP (rs2306283) in the OATP1B1 transporter gene SLCO1B1 showed association with the clinical action time of rocuronium in 200 patients. Our data do not provide support for the ABCB1 findings, whilst several variants within the SLCO1B1 gene, located right next to SLCO1A2, show almost genome-wide significant evidence of association in our study (Supplementary Fig. 2a). Further analyses showed that these SNPs were not independent from our SLCO1A2 lead variants (data not shown).

support for the ABCB1 findings, whilst several variants within the SLCO1B1 gene, located right next to SLCO1A2, show almost genome-wide significant evidence of association in our study (Supplementary Fig. 2a). Further analyses showed that these SNPs were not independent from our SLCO1A2 lead variants (data not shown). The top SNPs in our study, rs7967354 and rs11045995, and the variants in high LD with them, are located in non-coding parts of the gene. The intronic rs7967354 is in high LD with SNPs rs4149005 (non-coding exon transcript variant) and rs875234 (3′ UTR variant). Our extensive eQTL analyses suggest that a possible mechanism for the genotype–rocuronium dose association is tissue-specific gene expression regulation. As OATP1A2 has a role in excretion of rocuronium into bile,6 changes in the expression of SLCO1A2 caused by polymorphisms in the gene might prolong the effect of rocuronium by reducing clearance. One unexplored option to explain variation in rocuronium dose requirement is that of inter-individual differences in neuromuscular junctions. Here, we can only speculate on the possibility of some underlying variation in neuromuscular junction in otherwise healthy patients that could explain the variation in dose needs. Interestingly, SLCO1A2 is highly expressed in neural tissues, including peripheral nerve tissue, and our eQTL analyses suggested that the lead variants affect the SLCO1A2 expression level in the brain. It is tempting to speculate that neural tissue also plays a role in the impact of SLCO1A2 variants on rocuronium requirements.

ingly, SLCO1A2 is highly expressed in neural tissues, including peripheral nerve tissue, and our eQTL analyses suggested that the lead variants affect the SLCO1A2 expression level in the brain. It is tempting to speculate that neural tissue also plays a role in the impact of SLCO1A2 variants on rocuronium requirements. Our study has some limitations. Creatinine or creatinine clearance values were not available to assess kidney function. However, patients with clinically relevant kidney failure were excluded from the study. Although the study cohort is the largest thus far examined for pharmacogenetic data suitable for rocuronium studies, it is small for the GWAS approach, which usually requires thousands of participants. As our results are based on only 918 participants, all female, it is likely that some of the more subtle genetic effects remain undetected. Our study suggests that genetic variation in the gene SLCO1A2, encoding OATP1A2, is significantly associated with differences in rocuronium requirements. Our discovery offers one explanation for inter-individual differences in the duration of action of rocuronium. This variation was estimated to account for 4% of the variability in rocuronium dose. The most likely underlying mechanism is altered uptake of rocuronium by OATP1A2.

Study design: SA, RJ, KTO, MAK, EK Genetic data design: MAK, MP Recruitment and perioperative management of patients: RJ Whole BrePainGen study: EK, RJ, MAK Data analysis: SA, PB Genetic analysis: SA, LO, AA-O, MP, MAK Interpretation of results: SA, PB, RJ, KTO, MAK, EK Writing of paper: SA, PB, RJ, LO, AA-O, KTO, MAK, EK