Synergism interaction between genetic polymorphisms in drug metabolizing enzymes and NSAIDs on upper gastrointestinal haemorrhage: a multicenter case-control study

Abstract Background Interindividual genetic variations contribute to differences in patients’ response to drugs as well as to the development of certain disorders. Patients who use non-steroidal anti-inflammatory drugs (NSAIDs) may develop serious gastrointestinal disorders, mainly upper gastrointestinal haemorrhage (UGIH). Studies about the interaction between NSAIDs and genetic variations on the risk of UGIH are scarce. Therefore, we investigated the effect of 16 single nucleotide polymorphisms (SNPs) involved in drug metabolism on the risk of NSAIDs-induced UGIH. Materials and methods We conducted a multicenter case-control study of 326 cases and 748 controls. Participants were sub-grouped into four categories according to NSAID exposure and genetic profile. We estimated odds ratios (ORs) and their 95% confidence intervals (CI) using generalized linear mixed models for dependent binomial variables and then calculated the measures of interaction, synergism index (S), and relative excess risk due to interaction (RERI). We undertook stratified analyses by the type of NSAID (aspirin, non-aspirin). Results We observed an excess risk of UGIH due to an interaction between any NSAID, non-aspirin NSAIDs or aspirin and carrying certain SNPs. The greatest excess risk was observed for carriers of: rs2180314:C>G [any NSAID: S = 3.30 (95%CI: 1.24–8.80), RERI = 4.39 (95%CI: 0.70–8.07); non-aspirin NSAIDs: S = 3.42 (95%CI: 1.12–10.47), RERI = 3.97 (95%CI: 0.44–7.50)], and rs4809957:A>G [any NSAID: S = 2.11 (95%CI: 0.90–4.97), RERI = 3.46 (95%CI: −0.40–7.31)]. Aspirin use by carriers of rs6664:C>T is also associated with increased risk of UGIH [ORaspirin(+),wild-type: 2.22 (95%CI: 0.69–7.17) vs. ORaspirin(+),genetic-variation: 7.72 (95%CI: 2.75–21.68)], yet larger sample size is needed to confirm this observation. Conclusions The joint effect of the SNPs s2180314:C>G and rs4809957:A>G and NSAIDs are more than three times higher than the sum of their individual effects. Personalized prescriptions based on genotyping would permit a better weighing of risks and benefits from NSAID consumption. KEY MESSAGES Multicenter case-control study of the effect of genetic variations involved in drug metabolism on upper gastrointestinal haemorrhage (UGIH) induced by NSAIDs (aspirin and non-aspirin). There is a statistically significant additive synergism interaction between certain genetic polymorphisms and NSAIDs on UGIH: rs2180314:C>G and rs4809957:A>G. The joint effect of each of these single nucleotide polymorphisms and NSAIDs on UGIH is more than three times higher than the sum of their individual effects. Genetic profiling and personalized prescriptions would be useful in managing the risks and benefits associated with NSAIDs.


Introduction
Adverse effects of non-steroidal anti-inflammatory drugs (NSAIDs) were associated with heavy health and economic burdens [1,2]. Upper gastrointestinal haemorrhage (UGIH) is a frequent adverse effect of NSAID treatment that can be life-threatening [3]. Nonetheless, NSAIDs continue to be the most prescribed drugs worldwide [4].
Furthermore, it is well-established that aspirin plays an important prophylactic role against highly incident diseases that are associated with elevated mortality rates, such as several types of cancer and cardiovascular events [5][6][7][8][9][10][11][12]. Nevertheless, the association of aspirin with gastrointestinal bleeding has discouraged the adoption of this drug as a general prophylactic measure against disorders with great public health impact [13]. In addition, gastrointestinal symptoms in patients who used aspirin to protect against cardiovascular events had led to treatment interruption [14], and consequently to an increase in cardiovascular risk [15,16].
Marked interindividual differences with respect to their response to NSAIDs have long been recognized and attributed to many factors including genetic variations in metabolizing enzymes [17][18][19][20]. Several studies also reported a possible relationship between genetic variations in users of NSAIDs and gastrointestinal disorders [21][22][23][24][25][26][27]. In this context, genetic pharmacokinetic factors are of special importance since variations in genes involved in drug metabolism might alter their expression and thus increase the risk of undesirable effects like bleeding and cardiovascular events. Therefore, identifying patients at risk of UGIH based on their genetic background and personalized NSAID prescriptions might help weigh the risks and benefits associated with each type of NSAIDs and thus avoid adverse effects in susceptible individuals.
Currently, there is a lack of knowledge about the effect of variations in genes involved in drug metabolism on the risk of UGIH in general, and in NSAID users in specific. Taking into account the considerable morbidity and mortality rates of UGIH [28,29], and the wide spectrum of NSAID benefits, we carried out a multicenter case-control study that primarily aimed at testing the modification effect of 16 genetic polymorphisms involved in drug metabolism on the risk of NSAIDs-related UGIH. As a secondary objective of this study, we investigated the modification effect of those 16 genetic polymorphisms on the risk of nonaspirin NSAIDs-related UGIH as well as on aspirinrelated UGIH.

Study settings and design
This study represents a continuation of a previous full case-control study (i.e. case-control encompassing exposed and non-exposed patients to NSAIDs, on the contrary to other partial case-control studies that include only exposed patients) [30], published elsewhere, and shares the same protocol [31,32]. Patients were recruited from four hospitals in Spain (Barcelona, Galdakao, Santiago de Compostela, and Valladolid), between January 2004 and November 2007 and between January 2013 and October 2015. The study protocol was approved by the ethics committee of each participating centre (Barcelona: CEIC protocol number: Es38121226Z; Euskadi: CEIC-E protocol number: PI2013101; Galicia: CEIC-G protocol number: 2013/ 263 and Valladolid: CEIC-VA-ESTE-HCUV protocol number: PI-14-142). The participants provided written informed consent before enrolment in the study.

Definition of cases and controls
Cases were patients admitted to the hospital with symptoms of UGIH that were diagnosed surgically or endoscopically. Eligible cases were included irrespective of the grade of UGIH severity.
For each case, controls matched by the hospital, gender, and age (±5 years) were selected. To avoid selection bias due to excessive intake of NSAIDs, controls were either outpatients or patients enrolled from the preoperative unit among subjects who were about to undergo any of the following non-painful mild surgeries which were unrelated to the use of NSAIDs: plastic surgery, inguinal or umbilical hernia (strangulated or programmed), lipoma, varicotomy, prostatic adenoma, prostatic hyperplasia, thyroid nodules and thyroglossal cyst (euthyroid), eye cataract, phimosis, ear pinning, vocal cord cyst, tubal ligation, and septoplasty.
To ensure that all subjects belong to the same source of population, they were recruited from patients and outpatients attended by the same hospitals. All patients were biologically unrelated. The analysis was restricted to European participants to control for the risk of stratification bias [33]. We used the native language of the participants and their parents as a proxy of ethnicity [34][35][36][37]. Patients with a history of neoplasia, liver cirrhosis, or coagulopathy were excluded to control for the risk of Berkson's bias [38].
The inclusion and exclusion criteria of the cases and controls are specified in more detail in Table 1.

Data collection
Both cases and controls were thoroughly interviewed by trained health personnel, using a questionnaire specifically designed for this study. The collected data include participants' sociodemographic characteristics, clinical antecedents, smoking habits, alcohol and caffeine consumption, the motive for hospital admission, underlying symptomatology (for cases), the motive for  to answer or failure to complete the interview  80  0  Impossible to conduct interview within the 15-day period preceding admission  60  0  Date of last admission  0  1  Other  13  0  Non-white patients  15  0  Unavailable biological material  749 0 † Cases and controls were excluded upon presenting one or more exclusion criteria. ‡ Excludable endoscopic diagnosis included gastritis, esophagitis, oesophageal varices, gastric or duodenal neoplasia, Mallory-Weiss syndrome, angiodysplasia, anastomotic ulcers, diverticulitis, acute alcohol intoxication, hiatal hernia, and papule. Bold values represent the total per reason of exclusion group and study period. the scheduled surgery (for controls), previous episodes of gastric diseases, and exposure to pharmaceutical drugs (including the medicine's daily dose and indication). Direct relatives or healthcare assistants, who took care of the patient's medication, could attend and participate in the interview, but only data confirmed by the patient were considered. When the participant was not able to remember any of the requested information, the interview was repeated on a posterior date, or the patient was contacted by telephone if s/he had been discharged from the hospital. In case the patient doubted or was uncertain about specific information, that information was confirmed later by consulting the medical records of the patient.
Index dates were established to ascertain any exposure to NSAIDs. Information on NSAID exposure was extracted from patients' medical records, but the researchers were blind to patients' use of NSAIDs. For the cases, the index date was the day of onset of the first signs or symptoms of UGIH, while for the controls it was the day of the interview. NSAIDs exposure was considered if the consumption took place in the week preceding the index date [39][40][41]. For ease of recall, a catalog of prompt cards of the most consumed NSAID boxes was shown to the participants during the interview.
The reliability of the interview was rated on a scale of 0-10 as perceived by the interviewer, where zero means that the answers provided by the patient were completely unreliable. Patients whose interview was rated by zero were excluded from the study.
A 5 ml blood sample was withdrawn from each participant and stored in EDTA tubes or as spots on IsoCode papers at À80 C until genotyping.

Risk factors associated with UGIH
The following co-variables which were known to affect the risk of UGIH were considered: (1) previous infection with Helicobacter pylori; (2) therapeutic groups, such as proton pump inhibitors or oral anticoagulants; (3) digestive system disorders classified according to the patient's history of ulcer and bleeding (none or dyspepsia; ulcer; or bleeding); and (4) the reliability of the interview.

Helicobacter pylori determination
The presence of anti-H. pylori IgG antibodies in human serum were determined using the commercial ELISA kits: Human Anti-Helicobacter pylori IgG ELISA Kit (ab108736, Abcam, Cambridge, England), and Captia TM H. pylori IgG EIA (ref: 2346400, Trinity Biotech Captia, Co. Wicklaw, Ireland), and following the manufacturer's protocol. The participants were inquired if they had previously been treated against H. pylori infection to avoid any false-positive results caused by old infections.

Single nucleotide polymorphisms (SNPs) selection and genotyping
A comprehensive list of SNPs involved in gastrointestinal disorders (bleeding or ulcer) was retrieved by reviewing research reports published in MEDLINE until April 2017. The reference numbers (rs number) of the selected SNPs were confirmed using PubMed [42]. Subsequently, the function of the corresponding genes and the clinical significance of the genetic variations were identified through a literature review. Finally, SNPs in genes that may influence drug metabolism were selected for genotyping [25,26].
DNA was extracted from blood stored in EDTA tubes using chemagic TM DNA Buffy Coat 200 Kit H96 (PerkinElmer, reference number CMG-713) and from blood spots using chemagic TM DNA Blood 200 Kit H96 (PerkinElmer, reference number CMG-717). Extracted DNA was then quantified using Quant-iT TM PicoGreen TM dsDNA Assay Kits (ThermoFisher Scientific, reference number P7589). DNA concentration was normalized at 10-20 ng/ml in a minimum total volume of 40 ml. Samples were genotyped in a phonotype-blind process. iPlex V R Gold chemistry and MassARRAY platform were used according to the manufacturer's instructions (Agena Bioscience, San Diego, USA). Genotyping assays were designed using the Agena Bioscience MassARRAY Assay Designer 4.1 software. All assays were performed in 384-well plates, including negative controls and a trio of Coriell samples for quality control. The reproducibility of 7% of the samples was also checked between and/or within plates.
The compliance of the SNPs with Hardy-Weinberg equilibrium was checked using the SNPassoc Library of the R package (Version 1.9-2) [43][44][45]. In addition, all cluster plots were manually inspected by trained personnel using MassArray Typer software.

Statistical analysis
To determine any interaction between each of the 16 SNPs and NSAID exposure on the risk of UGIH, participants were grouped according to their genotype and NSAID exposure. Stratified analysis by the type of NSAID (any NSAID, non-aspirin NSAIDs, and aspirin) was carried out. In each analysis, the following four groups of participants were obtained: [group 1: drug(þ), wild-type; group 2: drug(þ), genetic-variation; group 3: drug(À), genetic-variation; and group 4: drug(À), wild-type]. Adjusted odds ratios (ORs) of UGIH were calculated in each group and then checked for any potential interaction between the presence of a genetic variation and drug exposure. The group of subjects who were not exposed to the studied drug category (any NSAID, non-aspirin NSAIDs, or aspirin) and who were carriers of the wild-type genotype of the analyzed SNP (group 4) was used as the reference category for the estimations of the interactions.
ORs and their 95% confidence intervals (CI) were estimated by generalized linear mixed models for dependent binomial variables [46]. In the construction of the models, patients were placed at level 1; the strata (each case and its matched controls) at level 2; the hospital at level 3; and the period of patients' recruitment at level 4. A random-effects model was used to examine the effect of the patients' recruitment period, and a nested random-effects model was applied for the strata of cases and controls and health centre. The lmer function of the lme4 R package (version 1.1-21) was applied in the estimation of the models [47]. Potential confounding variables were introduced in the model if they modified the OR of the main variable by at least 10% and provided that the Schwartz's Bayesian Information Criterion improved [48].
The recommendations given by Knol and colleagues were followed to explore any potential interaction between NSAIDs and genetic polymorphisms, whereby we estimated the relative excess risk due to interaction (RERI) and the synergism index (S) along with their 95% CI [49][50][51][52].

Clinical data collection
One thousand and seventy-four patients (326 cases and 748 controls) fulfilled the inclusion criteria and were included in the final analysis. The flow of subjects and the motives of exclusion are presented in Figure 1 and Table 1. The patients' demographic and clinical characteristics are presented in Table 2.

Genotyping
All genotyped samples were included in the analysis. The reproducibility of the 7% replicated random samples was 100%. All SNPs showed an acceptable genotype call rate: !98%. Both the calculations of the Hardy-Weinberg equilibrium (p < .001) and the manual inspection of the cluster plots confirmed that the controls were in equilibrium in terms of the corresponding polymorphisms (Table 3).

Risk estimation and modification of effect
The odds of UGIH varied according to the genotype and NASID (aspirin or non-apsirin) exposure.

Genotypes associated with high excess of risk of UGIH
The presence of certain genetic variations increases the odds of UGIH in users of any NSAID, non-aspirin NSAIDs, or aspirin as compared to users with wildtype genotypes (Table 4).
rs2180314:C>G: Any NSAID use by carriers of rs2180314:C>G is associated with substantially higher odds of UGIH in comparison with NSAID users carrying the wild-type genotype [OR drug(þ),wild-type : 3.17 (95%CI:   (Table 4). This excess in risk is suggested   (Table 4).
The risk associated with rs36079186 could not be determined because it was monovariant (the same genotype was observed in all the study population). Inconclusive results were also obtained for rs4536:C>T and rs58285195:T>C due to the limited number of cases and controls who are on NSAIDs treatment and carriers for these genetic variations.

Discussion
To the best of our knowledge, this is the first study that finds a statistically significant additive synergism interaction between certain genetic polymorphisms in genes involved in drug metabolism (rs2180314:C>G and rs4809957:A>G) and NSAIDs on UGIH. Our results indicate that the joint effect of these SNPs with NSAIDs is more than three times higher [rs2180314:C>G (S ¼ 3.30; 95%CI: 1.24, 8.80), rs4809957:A>G (S ¼ 2.11; 95%CI: 0.90, 4.97)] than the sum of their individual effects. Since (1) UGIH contributes to high mortality and morbidity rates [28], (2) NSAIDs are among the most commonly used medicines worldwide [4], (3) a large fraction of the European population carries the genetic variants rs2180314:C>G (58%) and rs4809957:A>G (21%) [53], and (4) genotyping is a low-cost test, our findings could enable identifying better the individuals at risk and those who are not at risk of UGIH from NSAID exposure.
The interindividual variations in therapeutic responses to NSAIDs were associated earlier with several demographic and clinical factors. Variations in patients' response to aspirin NSAIDs were also suggested to be related to the metabolic rate of this drug and the excretion of aspirin metabolites [54]. Patients are classified as fast and slow acetylators, and these metabolic differences were associated with genetic factors [55].
We are not aware of any previous study that evaluated the interaction between the polymorphisms tested in the present work and NSAID (non-aspirin and aspirin) on UGIH. Consequently, it was not possible to compare our findings to that of other studies. Few studies examined the associations between the genetic variants assessed in the current study and other gastrointestinal disorders (ulcers or small bowel bleeding). Shiotani and colleagues suggested a possible relation between the genetic variants rs2180314:G>C, rs4809957:A>G, rs1060463:C>G/C>T, and rs1695:A>G and the risk of small bowel bleeding in Japanese patients on aspirin therapy, however, no significant association was demonstrated [25]. Another study in Japan also reported a significant association of rs6664:C>T genotype with the risk of ulcer bleeding in aspirin users [26]. Shiotani and colleagues also suggested an association between rs3765070:A>G/A>T, rs4715354:G>A, rs2283458:A>G, and rs3756067:G>A and small bowel bleeding in a genome-wide analysis in the Japanese population; nevertheless, the association disappeared upon the validation of those SNPs [25].
The explanation of the increased odds of UGIH in NSAID consumers who are carriers of the genetic variations rs2180314:C>G or rs4809957:A>G is limited. However, we hypothesize that this excess in risk could be a consequence of an altered function of the corresponding genes due to polymorphisms. rs2180314:C>G belongs to the glutathione S-transferase (GSTA) gene family which plays an important role in the detoxification of electrophilic compounds, including therapeutic drugs, and protects the cells against damage. GSTA genes are highly polymorphic, and their genetic variation may alter the toxicity and efficacy of some drugs [56,57]. rs4809957:A>G belongs to gene members of the P450 family. Enzymes encoded by P450 are monooxygenases that catalyze many drug metabolism reactions [58].
In general, testing a large number of SNPs increases type 1 error, and therefore increments the chance of obtaining false-positive conclusions [59]. Nevertheless, in this study, we attempted to minimize type 1 error by performing a pre-hoc SNP selection whereby we chose specifically those SNPs that belong to genes involved in drug metabolism, and which were suggested to be associated with gastrointestinal disorders [25,26]. Furthermore, we analyzed each SNP in an independent model and reported all the implemented analyses. Our strategy of SNP selection and data reporting exempts from adjusting for multiple testing and leads to fewer errors of interpretation [60,61]. Another strength of this study is the control for all possible biases. The memory bias was reduced by showing prompt cards of the most frequent NSAID commercial boxes to the patients during the interview and by reviewing the medical records. The exclusion of non-white patients also allowed to prevent bias due to racial differences between populations. Moreover, performing the study in biologically unrelated patients, exclusively, avoided the over-representation of the bias of genotype within families. Finally, the measures of effect reported in this study were all adjusted to baseline risk factors that were known to increase the risk of gastrointestinal bleeding.
The main limitation of our study was the sample size. In fact, upon stratification by genotype and types of NSAID, a limited number of observations were left in the subgroups mainly for aspirin, which consequently decreased the statistical power for the associations of many SNPs. Low statistical power is a frequent limitation in candidate gene studies [62,63]. Another consequence imposed by the modest sample size is the curse of dimensionality; i.e. the number of observations was very small when various SNPs were combined [64]. Therefore, it was not feasible to analyze the combined effect of different genetic variations. In addition, it was not possible to undertake a dose-response analysis. Therefore, we believe that further studies with a larger sample size are needed to confirm these results before their implantation in clinical settings. Another limitation of our study is the potential presence of false H. pylori test results. Though we intended to minimize the false positive rate by inquiring about treatment for H. pylori infection in the past, we cannot rule out the possibility of having false negative results since the rate of eradication of H. pylori from a single treatment varies between 60% and 90% [65].
In conclusion, this study revealed that genetic variations might alter the pharmacological and clinical response to NSAIDs. The risk of UGIH in NSAID users with the wild type genotypes of rs2180314 and rs4809957 is significantly lower than that in those users who carry the genetic variants. If our results were confirmed by future studies, they would suggest that simple genetic profiling, a low-cost test, can be used to support a clinical decision towards personalized NSAIDs prescription. These findings are of important clinical relevance since NSAIDs (non-aspirin and aspirin) are among the most frequently prescribed drugs due to their wide spectrum of benefits and cost-efficiency. The medical community needs to carefully weigh the benefits and risks of NSAIDs for each patient and take measures that maximize the benefits of these drugs.