Anesthetic procedures in the two cohorts
This study was registered at the UMIN Clinical Trials Registry (Identifier: UMIN000022903, principal investigator: Shigekazu Sugino, date of registration: June 27, 2016, retrospectively registered, https://upload.umin.ac.jp/cgi-open-bin/ctr_e/ctr_view.cgi?recptno=R000026392). After registration, 264 Japanese adults (20–85 years of age) with ASA-physical status of 1 or 2 were enrolled in this study. This manuscript adhered to applicable STREGA guidelines [17].
In 179 patients, epidural catheters were inserted through the T9-L2 interspaces before anesthetic induction. General anesthesia was maintained with 0.5–2.0% sevoflurane. Intraoperative use of nitrous oxide was at the discretion of the attending anesthesiologist. Three to five mL of 1.5% lidocaine and 5 μg/mL of epinephrine were administered epidurally at least every 30 min. At the end of the surgery, an epidural analgesia was administered as a continuous 4-mL/h infusion of ropivacaine (0.2%) and fentanyl (2 μg/mL) for postoperative pain control (Epi cohort).
In the other 85 patients, general anesthesia was maintained with 1.5–2.5 % sevoflurane and 0.1–0.5 μg/kg/min of remifentanil. Intraoperative use of nitrous oxide was at the discretion of the attending anesthesiologist. All patients received 25 μg bolus doses of intravenous (i.v.) fentanyl at 10-min intervals in the recovery room after surgery to control early postoperative pain. Intravenous patient-controlled analgesia (IV-PCA) was then initiated with a fentanyl solution of 0.4 μg/kg/min. A PCA pump (Coopdech Syrinjector PCA Set, DAIKEN Medical Co., Japan) was used with the following parameters: demand dose of 0.4 μg/kg of fentanyl, background infusion of 0.4 μg/kg/min, and lockout interval of 10 min (IV-PCA cohort).
Assessment of PONV
Patient demographic information, including age, gender, height, weight, smoking status, history of motion sickness, and type and duration of surgery, was collected. In both cohorts, PONV was assessed 24 h postoperatively. PONV severity was assessed using a 4-point scale: 0 (no nausea), 1 (sensation of discomfort), 2 (severe nausea), and 3 (vomiting or retching). A PONV score of 2 or 3 was defined as an incidence of PONV. Only patients with a PONV score of 3 were treated with 10 mg of metoclopramide via i.v. Patients and anesthesiologists were blinded to the patient’s genotype throughout the postoperative assessment period and data analysis.
Fentanyl blood concentration measurement
Ten mL of blood from each patient was collected 24hr postoperatively in heparinized tubes. Blood samples were immediately placed on ice, centrifuged, and stored at -80°C. Aliquots (270 µL) of serum samples were mixed with 30 μL internal solution (i.e., 0.1 μg/mL flurazepam), 700 μL n-hexane, and 300 μL ethylacetate. These mixtures were vortexed and centrifuged at 13,200 rpm for 10 min at room temperature. The organic phase was transferred to a sterile microtube and evaporated until dry. The residues were redissolved in 500 μL of mobile phase and transferred to an autosampler vial. An aliquot (10 μL) was injected into a high-performance liquid chromatograph (1200 series; Agilent Technologies, Santa Clara, CA, USA) and a tandem mass spectrometer (3200 QTrap; AB Sciex, Framingham, MA, USA). Separation was performed at 30°C on a 2.5 mm × 150 mm, 5 µm C30 analytical column (Develosil RPAQUEOUS; Nomura Chemical, Aichi, Japan) coupled with a 1.5 mm × 10 mm guard column (RP-AR-S; Nomura Chemical). The mobile phase was prepared using a mixture of 50:50 formic acid (0.05%):methanol (99%), and the flow rate was maintained at 0.2 mL/min. In the second quadrupole, fentanyl and flurazepam were monitored by the respective transitions of m⁄z 338.4 to 105.0 and m/z 389.2 to 316.2 with a collision energy of 51 eV. The standard curve was linear from 0.1 to 10 ng/mL.
Genomic DNA extraction and quality control
Genomic DNA was extracted from 2 mL whole blood obtained from patients before induction of anesthesia induction using a Puregene Blood Core Kit B (Qiagen, Hilden, Germany). The quality and quantity of DNA were checked using a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). Samples with an absorbance ratio lower than 1.6 at 260 and 280 nm were excluded from further analysis. The samples were stored at 4°C until use.
Genotyping of SNPs by Japanese-specific designed DNA microarray
Twenty-four patients were selected to explore SNPs associated with the incidence of PONV. Fourteen patients (female, undergoing gynecological open abdominal surgery, 7 of 14 patients had PONV) were collected from the Epi cohort, and ten patients (female, undergoing mastectomy, 5 of 10 patients had PONV) were collected from the IV-PCA cohort. Therefore, in the selected cohort, 12 patients experienced PONV, and the other 12 patients did not. This cohort was modeled as a clinically typical situation with frequent PONV occurrence.
Two hundred nanograms of genomic DNA from each individual was used to genotype 659,636 SNPs with a custom-made DNA microarray (Japonica Array; Toshiba Corporation, Tokyo, Japan). Genotype calling was conducted using the apt-probeset-genotype program in Affymetrix Power Tools (ver. 1.18.2; Thermo Fisher Scientific Inc., Waltham, MA). All samples passed the recommended sample quality control metric (dish quality control > 0.82 and sample call rate > 97%). Quality control for SNPs was conducted using the Ps classification function in the SNPolisher package (version 1.5.2, Thermo Fisher Scientific Inc.). SNPs that were classified "recommended" by the Ps function were retained. SNPs with a call rate of less than 99.0%, significant deviation from Hardy-Weinberg equilibrium (HWE; p < 0.0001), or a minor allele frequency (MAF) of less than 0.5% were excluded from downstream analyses.
Genotype imputation following genome-wide association study
Imputation of missing genotype data from the 24 patients was performed using IMPUTE2 (ver. 2.3.1) [18]. Pre-phasing was first conducted with these SNPs using SHAPEIT (v2.r644) with options --burn 10, --prune 10, and --main 25 [19]. Genotype imputation was performed on phased genotypes with IMPUTE2 using the 1KJPN panel. For IMPUTE2, the applied options were -Ne 2000, -k hap 1000, -k 120, -burnin 15, and -iter 50.
After genome-wide imputation with the Japonica array™, we cleaned the control sample data by applying quality-control parameters of SNP call rate ≥ 95%, a MAF ≥ 1%, and HWE p ≥ 0.001. The 6,714,496 SNPs and insertions and deletions (indels) on autosomes that passed these quality control filters were used for the genome-wide association study. Associations between the PONV incidence and alleles of each SNP were analyzed by PLINK (ver. 1.9) [20]. A Manhattan plot was created to visualize genome-wide associations with qqman.
Variant annotation and determination of candidate SNPs
SNPs with statistically significant associations with PONV incidence were ranked for variant annotation (p < 0.0001, see Results sections). High-confidence SNPs with a p value of less than 10-4 were annotated using the Variant Annotation Integrator of the UCSC Genome Browser [21]. After the annotation, SNPs in intergenic regions were excluded, whereas SNPs in genes were retained. When more than two annotated SNPs were in the same gene, the constant of linkage disequilibrium strength (Gabriel’s D’) was calculated among SNPs using the 1KJPN reference panel dataset. One SNP was selected from SNPs in strong linkage disequilibrium with the D’ values of more than 90 in the same gene. In this selection by Gabriel’s D’, SNPs in exons were preferred to SNPs in introns. One SNP in an exon was ultimately selected in each gene to yield the candidate SNPs.
Genotyping by qPCR in the two cohorts
The final set of candidate SNPs was genotyped in all patients in both cohorts using the TaqMan SNP Genotyping Assay (Thermo Fisher Scientific Inc., Waltham, MA) according to the manufacturer’s instructions. Quantitative PCR conditions were 1 cycle of 95°C for 10 min; 40 cycles of 95°C for 15 sec and 60°C for 1 min.
In silico prediction of SNP function
Transcription factor binding sites in SNP-containing sequences were predicted with the TRANSFAC program of RegulomeDB [22]. Phenotypes associated with identified transcription factors and their encoding genes were investigated in the Human Phenotype Ontology database [23].
Statistical analysis
Fentanyl blood concentration data were compared between the patents with and without PONV using a Mann–Whitney test. In the genome-wide association study, associations of the incidence of PONV with the alleles of each SNP were tested using a Fisher’s exact test. In the genotyping by qPCR analysis, deviation of observed genotypes frequencies from HWE was tested at each SNP using a chi-square test. Associations of PONV incidence with SNP genotypes or alleles were tested using a chi-square test. We also performed a multivariate analysis combining genetic and clinical factors in a logistic regression model. Differences were considered significant at p < 0.05 unless otherwise noted.
A power analysis was performed to estimate the necessary sample size to achieve statistical power of greater than 0.8 when detecting a statistically significant (p < 0.05) genetic effect of a 15% difference in incidence of PONV between alleles using an additive genetic model. The minor allele frequency was presumptively set to 0.20. We calculated the minimum sample size as 218 patients.