Clinical, ethnic and genetic risk factors associated with postoperative nausea and vomiting in patients undergoing cancer surgery: a case-control study

Abstract

Objectives: To identify the clinical, ethnic, and genetic factors contributing to the varying risks of postoperative nausea and vomiting (PONV) among a Brazilian population undergoing cancer surgery. Methods: A case-control study was conducted involving 152 patients who experienced vomiting and/or retching (cases) and 158 patients who did not report nausea, vomiting, or retching (controls) within 24 h following oncological surgeries. This study is registered as ‘Genetic Polymorphism and postoperative nausea and vomiting (PONV)’ under registration number NCT03627780 (https://clinicaltrials.gov/study/NCT03627780). Thirty-two polymorphisms associated with PONV predisposition and 15 polymorphisms for ancestry analysis were genotyped via real-time polymerase chain reaction (PCR) with customised TaqMan low-density array (TLDA) cards. Results: The C allele of the rs208294 polymorphism (P2RX7 gene) was observed at a significantly higher rate in the control group than in the case group across the genotype (P=0.035), dominant (P=0.010) and allele (0.032) models, thus suggesting a protective effect against PONV. The genotype results for rs208294 were validated via Sanger sequencing, which confirmed the association in the dominant model (P=0.027). In a multivariate regression analysis that included rs208294 and clinical variables that were identified in the univariate analysis, only a prior history of PONV or motion sickness was observed to be a significant predictor of PONV (P<0.05). No association between rs208294 and PONV was detected in an external cohort consisting of 198 cases and 56 controls of Japanese descents (P>0.05). Additionally, ancestry analysis indicated a predominantly European genetic composition in the Brazilian cohort, which differed with the Asian composition of the independent validation cohort. Conclusions: A previous history of PONV or motion sickness was identified as being the strongest predictor of PONV in our analysis. Genetic association, ancestry and external validation analyses suggest that genetic factors for PONV may significantly differ across populations of different continental origins.

Introduction

The incidence of postoperative nausea or vomiting (PONV) in the general surgical population is approximately 30%, and it can increase to as high as 80% in high-risk patients who do not receive prophylaxis [1]. PONV significantly impacts postsurgical recovery, interferes with sleep and oral intake, and is one of the leading causes of unanticipated hospital admissions [2,3].

The predisposition to PONV is multifactorial and involves factors related to anaesthesia, surgery or patient characteristics [4]. According to the Apfel score (which is a well-established predictive tool), factors such as female sex, nonsmoking status, a history of motion sickness and/or PONV and the use of opioids in the postoperative period are linked to an increased incidence of PONV [1]. Although these factors can help in identifying high-risk patients and guiding risk-based therapeutic interventions, they do not fully explain the variability in the occurrence and severity of PONV. Notably, a family history of PONV has been identified as a significant risk factor in paediatric patients, thus suggesting a genetic component of PONV susceptibility. Therefore, genetic factors contribute to susceptibility to PONV [5,6].

Single nucleotide polymorphisms (SNPs) of specific candidate genes are implicated in the systemic pathways of nausea and vomiting, thereby potentially influencing individual responses to antiemetic drugs [7]. Multiple neurotransmitters are involved in the pathophysiology of PONV, and genes encoding membrane receptors, drug transporters, ion channels, metabolic enzymes, or structural proteins may be potential targets that are linked to PONV susceptibility [3]. Inherited factors are believed to play a significant role in the underlying susceptibility to PONV, as evidenced by increased rates of PONV observed across multiple generations within the same family [8] and a higher risk of PONV is observed among children with a family history of the condition [6].

However, the strength of the genetic associations has generally been modest, and these genetic factors alone cannot fully explain interindividual differences in the severity of or baseline sensitivity to PONV [8]. Some studies have identified populations of African descent as exhibiting a reduced risk of PONV [9], whereas individuals of European descent appear to demonstrate a greater prevalence of PONV compared to American or Asian populations [10]. These studies were conducted in relatively homogenous populations, thereby prompting us to investigate whether similar findings could be observed in a more genetically diverse cohort, such as in the Brazilian population.

The present study aimed to explore whether clinical variables, ethnic backgrounds, and genetic polymorphisms are associated with PONV in a cohort of Brazilian patients undergoing oncologic surgery. This study represents the first investigation of genetic polymorphisms and their associations with PONV in a Brazilian population.

Methods

Study design

This was a single-centre, observational, case-control, prospective study that was conducted on patients who underwent elective oncological surgeries between June 2015 and March 2020. Participants were recruited as part of the flowchart for recruitment, sample collection, processing, and storage of the biobank of the Academic Biobank for Cancer Research Network at the University of São Paulo (USP), Center for Translational Research in Oncology (LIM24), Instituto do Câncer do Estado de São Paulo (ICESP), São Paulo, Brazil. Written informed consent was obtained from all of the participants before completing the epidemiological and clinical questionnaires. This article adheres to the STREGA Statement [11]. Anonymized data are available at doi.org/10.6084/m9.figshare.24638607.

Ethics

Ethical approval for this Biobank was granted by the Local Research Ethical Committee CEP - N° 031/12 of Universidade de São Paulo, São Paulo, Brazil, as well as the National Ethics Committee CONEP (BIOBANCO) - N° 023/2014 of the Ministry of Brazilian Health, Brasilia, Brazil. The ethics approval report for this projetct has been include in the Supplementary Material II.

Subjects

Patients were recruited after surgery. Patients who presented with vomiting and/or retching within the first 24 hours after surgery were defined as the patients. The controls included patients who did not experience nausea, vomiting or retching during the same time period. All of the patients were assessed by the postoperative pain care team, who interviewed the patients on a daily basis after surgery.

The inclusion criteria were as follows: patients classified as ASA physical status II to IV, ≥18-years-old, undergoing elective cancer surgeries, and with one or more risk factors according to the Apfel score [1].

Patients were excluded from the study if they refused to participate in the study, required prolonged intubation during the first 24 hours after surgery, or exhibited any cognitive impairment, confusion, agitation or delirium after surgery. Patients receiving antiemetic medications prior to surgery were also excluded from the protocol. Furthermore, patients with missing or inconsistent data were excluded from the final analysis.

Data sources

Data were collected from the Electronic Health Record (EHR). Researchers interviewed each patient on the first postoperative day to collect information and complete a data sheet.

Antiemetic administration was intraoperatively and postoperatively performed at the discretion of the anaesthesiologist and surgical team, with different teams selecting the types and dosages of the antiemetic drugs. The utilized antiemetics included dexamethasone, ondansetron, metoclopramide, dimenhydrinate, and droperidol. All of the data were recorded via Research Electronic Data Capture (REDCap v.9.8.5) [12].

Blood sample collection, processing and DNA extraction

Blood samples (15 mL) were collected from all of the patients, processed into buffy coats and stored at -80°C until use. Peripheral leukocyte DNA was extracted using a salting-out procedure [13]. The DNA concentration and purity were measured using a NanoDrop One/OneC Microvolume UV Spectrophotometer^® (Thermo Fisher Scientific, USA). Only DNA samples with an absorption ratio (A260/A280) greater than or equal to 1.8 were used for the experiments. DNA integrity was verified via electrophoretic separation of nucleic acids on a 1% agarose gel. The samples were eluted in Tris-EDTA buffer solution and stored at 4°C until use.

Selection of the genetic variants associated with PONV

To identify the genetic variants associated with PONV, we reviewed the literature for polymorphisms related to drug metabolism, motion sickness, nausea, vomiting, and postoperative pain, with a focus on the variants that could represent a wide range of genes and metabolic pathways. Initially, we selected 96 SNPs in 24 genes based on peer-reviewed studies [7,8,14-21] (Table S1). We further analysed their genomic and clinical information using the ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), Pubmed (https://pubmed.ncbi.nlm.nih.gov/), Ensembl (https://www.ensembl.org/index.html), KEGG (https://www.genome.jp/kegg/), and dbSNP (https://www.ncbi.nlm.nih.gov/snp/) databases. Genes were classified by pathways using the Enrichr server (https://amp.pharm.mssm.edu/Enrichr/). Variants with frequencies lower than 15% in the 1000 Genomes, Genome Aggregation Database (gnomAD), and Online Archive of Brazilian Mutations (ABraOM) databases were excluded from the analysis. We prioritised variants with missense effects, which are more likely to exhibit pathogenic implications, or those located in untranslated regions (UTRs) with potential regulatory impacts. At least one variant was included in each gene or metabolic pathway, and variants without available TaqMan^® assays were excluded from the analysis. Table S2 lists the 32 polymorphisms in 23 genes that were analysed in association with PONV. For missense polymorphisms, we provide additional information on their functional impacts, as predicted via PolyPhen (Polymorphism Phenotyping v2) and SIFT (Sorting Intolerant from Tolerant) software. PolyPhen classifies missense variants as benign, probable, or possibly damaging, whereas SIFT classifies them as tolerated or deleterious.

Selection of the ancestry-informative markers (AIMs)

For the ancestry analysis, we searched for polymorphisms with distinct frequencies in four main populations (Admixed American, European, Eastern Asian, and African) [22-25] to assess all of the self-reported racial groups in our study (“white”, “brown”, “black” and “yellow”) [26]. We excluded variants with missense effects, previous associations with clinical conditions, or genotyping frequencies lower than 15% in the 1000 Genomes and gnomAD databases. A total of 15 AIMs were selected for ancestry analysis (Table S4).

SNP Genotyping using TaqMan^® Low-Density Array (TLDA) cards

The samples were genotyped using customised TLDA cards (Thermo Fisher Scientific, USA). Details on card customisation, genotyping, and analysis are provided in the Tables S2, S4.

Sanger sequencing of the rs208294 polymorphism (P2RX7 gene)

The rs208294 polymorphism (P2RX7 gene) genotyping results that were obtained using TLDA cards were technically validated via Sanger sequencing on the same set of samples. The detailed sequencing methodology is described in the Supplementary Material I.

Genotyping of rs208294 (P2RX7 gene) in an independent validation cohort

An additional cohort [27] of 254 Japanese adults (198 controls and 56 patients with PONV at 24 hours) was included for validation of the rs208294-PONV association observed in the Brazilian population. The general cohort characteristics have been previously published [27]. Genotyping details are included in the Supplementary Material I.

Ancestry analysis

We estimated the proportions of ancestral populations for each individual to compare these proportions with self-reported racial identities and case-control groups, as described in the Supplementary Material I.

Study size

For sample size calculation, we assumed a minimum minor allele frequency (MAF) of 15% for all of the genetic variants (based on data from the 1000 Genomes Project, Genome Aggregation Database (gnomAD) and Online Archive of Brazilian Mutations (ABRAOM)), a 30% prevalence of PONV occurrence [28], a potential genotype relative risk of 1.5, a 5% type I error rate, and 80% power. The Genetic Association Study (GAS) Power Calculator by CaTS, University of Michigan (accessed on December 2017 at the following link: https://csg.sph.umich.edu/abecasis/cats/gas_power_calculator/index.html), was used. The adequate sample size calculated a priori was 152 cases and 145 controls.

Statistical analysis

Statistical analyses were performed using SPSS^® v25.0, with a significance threshold of <0.05. The allele and genotype frequencies were calculated via allele counting, and the Hardy-Weinberg equilibrium (HWE) for each polymorphism was tested via the χ² test. The analyses were based on four genetic models: genotype, dominant, recessive, and allele/multiplicative (details of these models are found in the Supplementary Material I). Chi-square tests or Fisher’s exact tests were used to compare genotype and allele distributions between cases and controls, as well as differences in sociodemographic, clinical, and surgical variables. The Shapiro-Wilk test was used to assess data normality, and the nonparametric Mann-Whitney test was used to compare quantitative variables between groups. P values were adjusted for multiple comparisons via the Benjamini-Hochberg method with the “p.adjust” function in R v4.0.2. Univariate and multivariate binary logistic regression models were used to identify the PONV risk factors, and odds ratios (ORs) and 95% confidence intervals (95% CIs) were estimated. Carriers of the wild-type genotype/allele were used as the reference group in the logistic regression models.

Results

We assessed 415 patients for eligibility in this study. Of these patients, 20 patients did not meet the inclusion criteria, five patients refused to participate, 17 patients lacked blood collections, ten patients had low-purity DNA samples, and 53 patients were excluded because they presented with only nausea symptoms. After enrolment, no patients were excluded from the study, thereby resulting in a final sample of 310 patients for analysis. This sample included 152 patients who experienced vomiting and/or retching and 158 controls who did not report of nausea, vomiting, or retching within the first 24 hours following oncological surgeries.

Description of the study population

Significant differences were observed between cases and controls in terms of age, sex, history of PONV or motion sickness, postoperative opioid use, Apfel score and chemotherapy-induced vomiting (P<0.05; Table 1). Compared with the control group, the case group was younger, predominantly female (86.8%), had a greater incidence of prior history of PONV/motion sickness and chemotherapy-induced vomiting, had higher Apfel scores [1] and had greater postoperative opioid use. The primary utilized intraoperative antiemetics were ondansetron and dexamethasone (Table 1). No statistically significant differences were observed in the types or combinations of administered antiemetics (such as dexamethasone, ondansetron, metoclopramide, dimenhydrinate, or droperidol) between the case and control groups (P>0.05). No clinically significant differences were observed in the other analysed variables; therefore, they were not included in the multivariate analysis.

Table 1.

Comparison of the patient characteristics, postoperative nausea or vomiting (PONV) risk factors and surgical data between controls (N=158) and cases (N=152)

Variables	Controls (N=158)	Cases (N=152)	χ2/U	p	OR (95% CI)	p#
Self-reported ethnicityδ
White	88 (55.7%)	94 (61.8%)	1.921	0.589a	1 (Reference)
Brown	48 (30.4%)	40 (26.3%)			0.78 (0.47-1.30)	0.340
Black	19 (12%)	17 (11.2%)			0.84 (0.41-1.71)	0.628
Yellow	3 (1.9%)	1 (0.7%)			0.31 (0.03-3.06)	0.317
Age (years)
Mean (SD)	52.42 (13.11)	49.13 (13.29)	10,523	0.060b	0.98 (0.97-0.99)	0.030*
BMI
Mean (SD)	26.97 (6.10)	28.01 (5.55)	10,429	0.045b,*	1.03 (0.99-1.07)	0.120
Female Gender	117 (74.1%)	132 (86.8%)	8.020	0.005a,*	2.31 (1.28-4.17)	0.005*
History of PONV or motion sickness	34 (21.5%)	77 (50.7%)	28.62	<0.001a,*	3.74 (2.28-6.14)	<0.001*
Nonsmoking status	142 (89.9%)	142 (93.4%)	1.269	0.260a	1.60 (0.70-3.64)	0.263
Postoperative opioid use	132 (83.5%)	139 (91.4%)	4.400	0.036a,*	2.11 (1.04-4.27)	0.039*
Apfel
1	10 (6.3%)	3 (2.0%)	35.527	<0.001a,*	1 (Reference)
2	53 (33.5%)	24 (15.8%)			1.51 (0.38-5.98)	0.558
3	71 (44.9%)	61 (40.1%)			2.86 (0.75-10.88)	0.122
4	24 (15.2%)	64 (42.1%)			8.89 (2.25-35.08)	0.002*
History of previous chemotherapy	65 (41.1%)	67 (44.1%)	0.274	0.601a	1.13 (0.72-1.77)	0.601
Chemotherapy-induced nausea	37 (59.7%)	44 (67.7%)	0.882	0.348a	1.42 (0.69-2.93)	0.348
Chemotherapy-induced vomiting	20 (32.3%)	33 (50.8%)	4.472	0.034a,*	2.17 (1.05-4.45)	0.036*
Type of surgery
Abdominopelvic surgery	107 (67.7%)	98 (64.5%)	0.780	0.677a	1 (Reference)
Breast surgery	28 (17.7%)	33 (21.7%)			1.29 (0.73-2.28)	0.389
Other surgeries	23 (14.6%)	21 (13.8%)			0.99 (0.52-1.91)	0.993
Duration of surgery (min)
Mean (SD)	238.7 (137.0)	241.5 (135.5)	11,741.50	0.735b	1.00 (0.99-1.02)	0.858
Fluid balance (mL)
Mean (SD)	2,304.4 (1309.1)	2,269.4 (1186.8)	11,993.50	0.985b	1.00 (1.00-1.00)	0.805
Videolaparoscopy surgery	65 (41.1%)	48 (31.6%)	3.057	0.080a	0.66 (0.41-1.05)	0.081
Neuraxial Opioids use	102 (64.6%)	103 (67.8%)	0.356	0.551a	1.15 (0.72-1.85)	0.551
Intraoperative ondansetron use	150 (94.9%)	143 (94.1%)	0.110	0.740a	0.85 (0.32-2.26)	0.740
Intraoperative dexamethasone use	117 (74.1%)	114 (75%)	0.037	0.848a	1.05 (0.63-1.75)	0.848
PCA 24 h use	29 (18.4%)	28 (18.4%)	0	0.988a	1.00 (0.57-1.79)	0.988

^δOfficial Brazilian Census categories [26];

BMI: body mass index; N: Number of individuals; SD: standard deviation; PCA: Patient controlled analgesia; PONV: postoperative nausea or vomiting;

^aChi Square test (χ²);

^bMann-Whitney test (U);

^#Univariate Logistic Regression;

OR: Odds Ratio; 95% CI: 95% confidence interval;

^*P<0.05.

Genotype and allele frequency distributions and association with PONV

All 47 genetic markers were genotyped; after quality control (Figure S1, Supplementary Material I), 1.3% of the genotypes were excluded from the final analysis. Table 2 provides detailed information on the 32 selected polymorphisms associated with PONV and their corresponding 23 genes, genomic coordinates (hg19), Hardy-Weinberg equilibrium (HWE) results, and minor allele frequency (MAF) distributions between cases and controls compared with frequencies from the 1000 Genomes, gnomAD and ABraOM databases. Raw p values of the chi-square test comparing frequencies between cases and controls, as well as adjusted p values after multiple testing corrections, are also presented. All of the SNPs conformed to HWE, except for rs33985936 (SCN11A), rs1176744 (HTR3B), rs16947 and rs1065852 (CYP2D6) (P<0.05; Table 2).

Table 2.

Description of the thirty-two selected polymorphisms in 23 genes associated with PONV, comparison of their Minor Allele Frequencies (MAF) between controls (N=158) and cases (N=152), description of the allelic frequencies available in other public databases and results of the Hardy-Weinberg Equilibrium (HWE)

dbSNP ID	Genomic coordinate (hg19)	Gene	Minor allele	χ2 HWE (total sample)	p HWE (total sample)	MAF (%) Controls (N=158)	MAF (%) Cases (N=152)	χ2	p value	adj p value	MAF (%) 1000 Genomes (global)	MAF (%) 1000 Genomes (AMR)	MAF (%) gnomAD (AMR)	MAF (%) ABraOM (Brazil)
rs3766246	chr1:46865671	FAAH	G	0.445	0.505	50	53	0.544	0.461	0.97	43.5	51.3	55.2	53
rs324420	chr1:46870761	FAAH	A	1.054	0.305	29.9	22.7	4.168	0.041*	0.584	26.2	35.2	34.5	26
rs2165870	chr1:239785420	CHRM3	G	0.928	0.335	75.3	76.6	0.15	0.699	0.97	77.5	66.7	60.4	68.6
rs685550	chr1:239924408	CHRM3	A	3.091	0.079	71.3	63.9	3.652	0.056	0.584	66.4	74.1	75.9	72.4
rs3755468	chr2:75382391	TACR1	C	3.148	0.076	62.3	64.6	0.358	0.55	0.97	56.1	57.1	59.4	62.4
rs17641121	chr2:155665752	KCNJ3	C	2.769	0.096	29.1	35.9	3.214	0.073	0.584	19.2	22.3	23.9	31.1
rs901865	chr3:11300707	HRH1	C	0.344	0.557	77.8	77	0.068	0.795	0.97	82.3	86.7	90.1	80.9
rs33985936	chr3:38936134	SCN11A	T	6.9	0.009*	19.9	20.1	0.002	0.968	0.997	15.4	22.5	24	20.1
rs11709492	chr3:38945984	SCN11A	T	0.013	0.91	25.9	24	0.31	0.578	0.97	26.4	13.8	16.8	24.3
rs6280	chr3:113890815	DRD3	T	1.665	0.197	54.5	52.6	0.207	0.649	0.97	51.4	57.3	54.9	55.5
rs6443930	chr3:183754294	HTR3D	C	0.019	0.891	52	52.8	0.044	0.834	0.97	53.4	45.8	46.3	53.7
rs6766410	chr3:183774762	HTR3C	A	0.518	0.472	38.5	39.5	0.057	0.811	0.97	46.5	52.7	53.8	41.9
rs6807362	chr3:183778010	HTR3C	C	0.171	0.679	49	50.4	0.107	0.743	0.97	55.6	67.1	59.5	54.8
rs1799971	chr6:154360797	OPRM1	G	0.082	0.775	13	13.5	0.035	0.851	0.97	22.3	20	21.5	14.9
rs72552763	chr6:160560881	SLC22A1	-	0.898	0.343	17.1	15.7	0.227	0.634	0.97	11.8	28.8	21.7	NA
rs622342	chr6:160572866	SLC22A1	A	1.169	0.28	69.7	67.2	0.429	0.512	0.97	74.1	59.9	64.7	69.6
rs1800795	chr7:22766645	IL6	G	0.13	0.718	73.9	76.3	0.488	0.485	0.97	85.9	81.6	78.6	74.9
rs1045642	chr7:87138645	ABCB1	G	0.018	0.894	59.6	59.5	0	0.997	0.997	60.5	57.2	54.8	58.5
rs1072198	chr7:120327349	KCND2	T	0.171	0.679	66.6	64.8	0.212	0.645	0.97	67.5	68.2	71.9	65.2
rs2545457	chr8:140661285	KCNK9	A	1.485	0.223	62.3	64.9	0.437	0.509	0.97	64.2	68.9	65.8	63
rs1800532	chr11:18047816	TPH1	T	2.432	0.119	31.8	35.7	1.002	0.317	0.97	32.1	37.2	38.3	34.5
rs1800497	chr11:113270828	DRD2	A	1.669	0.196	24.4	27.3	0.697	0.404	0.97	32.6	31.1	44.9	24
rs3758987	chr11:113775275	HTR3B	C	1.263	0.261	28.8	32.2	0.865	0.352	0.97	32.7	35.9	32.8	31
rs1176744	chr11:113803028	HTR3B	C	6.015	0.014*	30	33.8	0.999	0.318	0.97	35.4	40.8	35.5	36.3
rs1062613	chr11:113846006	HTR3A	C	0.208	0.649	68.7	72.5	1.042	0.307	0.97	75.2	83.1	86.3	71.2
rs208294	chr12:121600253	P2RX7	C	0.013	0.909	54.9	46	4.598	0.032*	0.584	53	39	30.3	55.6
rs35364174	chr17:48731392	ABCC3	A	0.076	0.783	45.6	46	0.013	0.909	0.97	40	45.2	45.7	44.7
rs1978153	chr17:48737861	ABCC3	G	1.456	0.228	38	36.5	0.142	0.707	0.97	43.9	37.5	38.5	40.5
rs4633	chr22:19950235	COMT	T	0.002	0.967	37.6	39.5	0.234	0.629	0.97	37.2	38	40.5	41.2
rs4680	chr22:19951271	COMT	A	0.056	0.813	38.2	39.8	0.163	0.686	0.97	36.9	37.8	40.5	40.5
rs16947	chr22:42523943	CYP2D6	A	110.999	<0.001*	36.5	32.4	1.076	0.3	0.97	35.9	32.7	31.2	39.2
rs1065852	chr22:42526694	CYP2D6	A	15.424	<0.001*	16.4	16.1	0.013	0.909	0.97	23.8	14.8	12.3	17.6

Note: HWE: Hardy-Weinberg Equilibrium; N: number of individuals; chr: chromosome; -: deleted; MAF: Minor Allele Frequency (the minor allele was considered as the variant allele, different from the wild-type allele); adj: adjusted; AMR: Ad Mixed American; PONV: postoperative nausea or vomiting; NA: not available;

^*P<0.05.

A significant association was observed between the rs208294 polymorphism (P2RX7 gene) and PONV in the allele, genotype and dominant models (Tables 2, 3), although this association was lost after multiple testing corrections (Table 2).

Table 3.

Association of the rs208294 polymorphism (P2RX7 gene) and PONV by TaqMan Genotyping assay and technical validation by Sanger Sequencing, considering the total sample of controls (N=158) and cases (N=152) in four different genetic models

Genetic Model	Genotypes/Alleles	TLDA genotyping (TaqMan Genotyping assay)						Genotyping by Sanger Sequencing
		Controls N (%)	Cases N (%)	χ2	P	OR (95% CI)	p#	Controls N (%)	Cases N (%)	χ2	p	OR (95% CI)	p#
Genotype	TT	28 (18.3)	43 (31.4)	6.698	0.035*	1 (Ref)		28 (17.7)	43 (28.3)	4.98	0.083	1 (Ref)
	TC	82 (53.6)	62 (45.3)			0.39 (0.13-1.18)	0.1	86 (54.4)	74 (48.7)			0.53 (0.18-1.55)	0.248
	CC	43 (28.1)	32 (23.4)			0.37 (0.11-1.20)	0.1	44 (27.8)	35 (23.0)			0.44 (0.14-1.41)	0.168
Dominant	TT	28 (18.3)	43 (31.4)	6.695	0.010*	1 (Ref)		28 (17.7)	43 (28.3)	4.9	0.027*	1 (Ref)
	TC+CC	125 (81.7)	94 (68.6)			0.38 (0.13-1.09)	0.07	130 (82.3)	109 (71.7)			0.50 (0.18-1.38)	0.179
Recessive	TT+TC	110 (71.9)	105 (76.6)	0.849	0.357	1 (Ref)		114 (72.2)	117 (77.0)	0.949	0.33	1 (Ref)
	CC	43 (28.1)	32 (23.4)			0.73 (0.31-1.71)	0.47	44 (27.8)	35 (23.0)			0.71 (0.31-1.63)	0.415
Allele	T	138 (45.1)	148 (54.0)	4.598	0.032*	1 (Ref)		142 (44.9)	160 (52.6)	3.672	0.055	1 (Ref)
	C	168 (54.9)	126 (46.0)			0.80 (0.41-1.54)	0.5	174 (55.1)	144 (47.4)			0.70 (0.41-1.21)	0.202

Note: N: Number of individuals;

^#adjusted for gender, age, history of PONV or motion sickness, postoperative opioid use and chemotherapy-induced vomiting;

OR: Odds Ratio; CI: Confidence Interval; Ref: reference;

^*P<0.05.

Via the TLDA methodology, 6.5% of the genotype data for the rs208294 polymorphism were excluded because of quality control criteria. Consequently, we decided to validate our findings by sequencing a 258 bp fragment that included this SNP. Via this method, we accurately genotyped all of the samples without any missing data and confirmed the association between rs208294 and PONV in the dominant model (Table 3).

Age, sex, history of PONV or motion sickness, postoperative opioid use, chemotherapy-induced vomiting, and the rs208294 polymorphism were significantly associated with PONV in the univariate analysis. However, in the multivariate regression models including all of the independently associated variables, only a history of PONV or motion sickness remained a significant predictor of PONV. The rs208294 polymorphism lost statistical significance in both the TLDA and sequencing results (Table 3).

For the other examined polymorphisms, we detected an association between the rs324420 polymorphism (FAAH gene) and PONV in the allele model, with the A allele being more prevalent in controls (29.9%) than in cases (22.7%) (raw P value =0.041). However, this association was not observed in the other genetic models (Table 2). Moreover, no significant associations were observed for the remaining polymorphisms (Table 2 for the allele model; data not shown for the other genetic models).

Validation of the rs208294 (P2RX7 gene) association with PONV in the Japanese Cohort

We did not validate the association of the rs208294 polymorphism with PONV in the Japanese cohort across any of the investigated genetic models (Table S3, Supplementary Material I). The prevalence of the P2RX7 polymorphism was lower in the Japanese sample (42.3%) than in the Brazilian sample (50.7%).

Ancestry-informative marker (AIM) results

We subsequently analysed 15 AIMs, which included intronic or intergenic regions and one synonymous variant (Table S4, Supplementary Material I). All of the analysed markers (except for rs2814778) were in accordance with HWE (Table S5, Supplementary Material I). This table also shows the MAF for the case and control groups and the MAF for each main population of the 1000 Genomes Project (African-AFR, American-AMR, Eastern Asian-EAS, and European-EUR).

Distinct profiles for each ancestral population were identified; however, there were no significant differences observed between the case and control groups in either the supervised or unsupervised approach (P>0.05; Figures S2, S3, Supplementary Material I). We observed slight differences between cases and controls, primarily in the percentage of the EUR component, which was greater in the case group (66% in the case group vs. 62% in the control group), and in the AMR component, which was lower in the case group compared with the control group (Figure S3, Supplementary Material I). Finally, we compared self-reported race with molecular ancestry and observed higher percentages of European, African or Asian components in individuals self-identifying as “white”, “black” or “yellow”, respectively (Figure S4, Supplementary Material I).

Discussion

We conducted the first case-control study aimed at identifying clinical, ethnic, and genetic differences associated with PONV in a sample of the Brazilian population undergoing cancer surgery. Patients were selected based on low-, moderate-, and high-risk factors for PONV, as defined via the Apfel criteria [1,4]. We identified female sex, younger age, a prior history of PONV, postoperative opioid use, higher Apfel scores, and a history of chemotherapy-induced vomiting as significant risk factors for PONV. These findings align with well-established risk factors that have been reported in the literature [1,4,29].

We investigated 32 genetic variants across 23 genes associated with drug metabolism, motion sickness, nausea, vomiting, postoperative pain, and PONV susceptibility, whereby we selected variants with a minimum MAF of 15%, as reported in genetic databases. The inclusion of less common variants would have reduced the statistical power or required a substantially larger sample size. All of the SNPs except for rs33985936 (SCN11A), rs1176744 (HTR3B), rs16947 and rs1065852 (CYP2D6) conformed to HWE, thereby suggesting random allele segregation within the population [30]. Deviations from HWE can arise from genotyping errors, assortative mating, selection, population stratification, or random chance [30].

Genetic association analyses were performed using genotype, dominant, recessive and allele genetic models, which complement each other in reducing the likelihood of false negatives. This approach allowed us to better explore the influence of SNP alleles and their modes of inheritance on PONV [30]. We observed that the C allele of the rs208294 polymorphism in the P2RX7 gene was more prevalent among controls than in cases across genotype, dominant, and allele models, thus indicating a potential protective effect against PONV. The genotypes of rs208294 were validated via Sanger sequencing, thereby confirming the association in the dominant model. However, after adjusting for clinical variables in a multivariate regression model, only a prior history of PONV or motion sickness remained a significant predictor of PONV.

P2RX7 is a purinergic receptor for ATP that plays a crucial role in gastrointestinal inflammation by increasing the production of proinflammatory mediators [31]. Surgical procedures [32] and subsequent postoperative ileus [33] can induce inflammation, thereby potentially exacerbating PONV. Interestingly, antiemetic agents such as dexamethasone, 5-HT3 receptor antagonists [34] and NK1 receptor antagonists [35] exhibit anti-inflammatory properties. As a result, these medications may alleviate PONV not by directly affecting the vomiting centre but by reducing inflammation [31].

The rs208294 polymorphism, which is a missense variant of the P2RX7 gene, may influence PONV. The T allele encodes a tyrosine in place of a histidine (C allele), thus resulting in a moderate gain-of-function effect [36]. In contrast, the C allele of rs208294 has been linked to reduced levels of IL-12p40 in patients with localised aggressive periodontitis [37], which is a pattern that is similarly observed with dexamethasone administration [38]. Therefore, patients with the C allele may exhibit a lower incidence of PONV. Additional studies are needed to further elucidate the involvement of this polymorphism in PONV.

Despite the lack of statistical significance for rs208294 in the multivariate model, we proceeded to validate the association in an independent cohort of Asian patients comprising 198 cases and 56 controls, as previously described [27]. However, no association between rs208294 and PONV was identified in the Japanese cohort. This cohort was previously explored regarding genetic associations with PONV using a genome-wide association study (GWAS) designed for the Asian population [39]. Moreover, a recent GWAS [40] involving another Japanese sample similarly revealed no association between rs208294 and PONV. Importantly, the genetic backgrounds of the Asian samples differ from those of our Brazilian cohort, which displayed a high proportion of European ancestry; this observation is consistent with prior studies on the Brazilian population [26,41]. Given the high rate of genetic admixture in Brazil (but with a predominant European influence), PONV risk factors may significantly vary between populations with different ancestral compositions. This scenario underscores the importance of further research in diverse cohorts worldwide.

In addition to the P2XR7 gene (rs208294), we also detected an association between the FAAH gene (rs324420) and PONV in the allele model, with further details available in the Supplementary Material I.

This study is the first to investigate 32 polymorphisms associated with PONV alongside 15 ancestry markers in a Brazilian population. Our selection included at least one SNP from the most relevant genes that have been previously implicated in PONV susceptibility. Recently, a large-cohort GWAS of over 60,000 individuals demonstrated a homogeneous distribution of potential PONV-related variants across all chromosomes, including genes involved in inflammatory and immune response pathways, such as P2RX7 [42]. Our findings are valuable for understanding the genetic underpinnings of PONV among Brazilian patients with cancer.

The clinical significance of the investigation of genetic variants associated with PONV predisposition is based on their potential to improve personalised anaesthesia and postoperative care. The identification of genetic markers for PONV risk can help clinicians to predict which patients are more likely to experience PONV, thus allowing for customised preventive strategies and improved management. This approach could lead to a reduced incidence of PONV, increased patient comfort, faster recovery times, and lower health care costs. Moreover, an understanding of the genetic basis of PONV may facilitate the development of targeted antiemetic therapies.

A limitation of our study was the relatively small sample size, which was partly due to our decision to include only patients who experienced vomiting or retching. Although this choice reduced the sample size, it enhanced the robustness of our findings by excluding patients with only nausea, which is a subjective symptom with distinct pathophysiological mechanisms from vomiting [32]. Additionally, our study did not record the constant use of postoperative antiemetics, which could have influenced the incidence of PONV.

In conclusion, a history of prior PONV or motion sickness emerged as the strongest predictor of PONV in our study. Further research is needed to explore the rs208294 polymorphism in the P2XR7 gene and its association with PONV in larger, ethnically diverse case-control studies. Our genetic analyses, ancestry assessments, and external validation emphasise the idea that PONV risk factors may vary across population-specific risk factors.

Disclosure of conflict of interest

None.