A multicenter case–control study of the effect of e-nos VNTR polymorphism on upper gastrointestinal hemorrhage in NSAID users

Bleeding in non-steroidal anti-inflammatory drug (NSAID) users limited their prescription. This first multicenter full case–control study (325 cases and 744 controls), explored the association of e-NOS intron 4 variable number tandem repeat (VNTR) polymorphism with upper gastrointestinal hemorrhage (UGIH) in NSAID exposed and unexposed populations and assessed any interaction between this polymorphism and NSAIDs. NSAID users carrying e-NOS intron 4 wild type genotype or VNTR polymorphism have higher odds of UGIH than those unexposed to NSAIDs [Odds Ratio (OR): 6.62 (95% Confidence Interval (CI): 4.24, 10.36) and OR: 5.41 (95% CI 2.62, 11.51), respectively], with no effect modification from VNTR polymorphism-NSAIDs interaction [Relative Excess Risk due to Interaction (RERI): −1.35 (95% CI −5.73, 3.03); Synergism Index (S): 0.77 (95% CI 0.31, 1.94)]. Similar findings were obtained for aspirin exposure. Non-aspirin NSAID users who carry e-NOS intron 4 VNTR polymorphism have lower odds of UGIH [OR: 4.02 (95% CI 1.85, 8.75) than those users with wild type genotype [OR: 6.52 (95% CI 4.09, 10.38)]; though the interaction estimates are not statistically significant [RERI: −2.68 (95% CI −6.67, 1.31); S: 0.53 (95% CI 0.18, 1.55)]. This exploratory study suggests that the odds of UGIH in NSAID or aspirin users does not modify according to patient´s e-NOS intron 4 genotype.

Participants were asked whether they have ever been treated against Helicobacter pylori infection in order to avoid having any false positive result from a previous infection.
Genotyping. The eNOS intron 4 VNTR polymorphism was detected by polymerase chain reaction (PCR) using the following primers: 5′-AGG CCC TAT GGT AGT GCC TTT-3′ and 5′-TCT CTT AGT GCT GTG GTC AC-3′ [41][42][43] . The reaction mixture (a total volume of 25 µl) contained 2.5 µl of 10 × PCR buffer, including 100 ng of genomic DNA, 1.5 mM MgCl2, 0.2 mM of each dNTP, 500 nmol of each primer and 5 U Taq polymerase. PCR was performed with 35 cycles of 30 s at 94 °C for denaturation, 30 s at 63 °C for annealing, and 1 min at 72 °C for extension, followed by final extension for 5 min at 72 °C. The following fragments were produced: 393 bp band (4 copies of the 27 bp repeat), and 420 bp band (5 copies of the 27 bp repeat). The amplified product was size fractionated by the D1000 Screen Tape assay (Agilent Technologies, CA) on Agilent 4200 TapeStation system according to manufacturer's recommendations. For quality control, 5% of the samples were selected to repeat PCR and genotyping, and no discrepancies were found. All cases and controls were genotyped using a phenotype-blind process. Hardy-Weinberg equilibrium was tested in the control group (significance threshold: P-value < 0.001), using SNPassoc Library of the R package 44 , to check for possible bias in the selection of controls 45,46 .
Statistical analysis. The 27 bp e-NOS intron 4 VNTR polymorphism was genotyped for all participants who were then stratified into four categories according to NSAID exposure (exposed/unexposed) and genotype (wild type/genetic variation). The wild type genotype was denoted "b allele" and defined as five copies of the 27 bp repeats in both alleles (5/5). The genetic variant was denoted "a allele" when characterized by four copies of the 27 bp repeats in both alleles (4/4) or "ab allele" in case of possessing four copies of the 27 bp repeats in one allele and five copies of the same repeat in the second allele (4/5). The group that comprised patients who were "unexposed to NSAIDs and carriers of the wild type genotype" was used as a reference category. Adjusted Odds Ratios (OR) of UGIH and their 95% Confidence Intervals (CI) were estimated using the generalized linear mixed models for dependent binomial variables. The statistical models were constructed by considering the following four consecutive levels: patient, strata of cases and controls (each case and its matched controls), health center, and period of patients' recruitment. We used a random-effects model to assess the effect of the period of patients' recruitment and a nested random-effects model for the strata of cases and controls, and health center. The models were estimated using the lmer function of the lme4 R package 47 . Initially, in univariate analysis, we explored the effect of potential confounders including age, Body Mass Index (BMI), sex, tobacco smoking, alcohol intake, caffeine consumption, diet, previous history of arthrosis and/or gastrointestinal disorders other than UGIH, Helicobacter pylori infection, source of information (i.e. whether the interview was answered by the patient independently or with the help of a healthcare assistant / direct relatives), number of undertaken interviews, reliability of the interview as perceived by the interviewer, and exposure to medications that are not NSAIDs. Potential confounders were retained in the model if they changed the OR of the main variable by at least 10%, and provided that the Schwartz's Bayesian Information Criterion was enhanced 48,49 .
Subsequently, the additive interaction effect between e-NOS intron 4 VNTR polymorphism and NSAIDs on UGIH was determined by calculating the Relative Excess Risk due to Interaction (RERI) and the Synergism Index (S) estimates along with their 95% CI.
We also undertook a stratified analysis by type of NSAIDs where we assessed the effect of e-NOS VNTR polymorphism on UGIH in 1) non-aspirin NSAIDs users, and 2) aspirin NSAIDs users. Ethical aspects. Participants signed a written informed consent before their enrolment in the study.

Results
General patients' characteristics. Three hundred twenty-five cases and 744 controls fulfilled the inclusion criteria and were included in the study. Their demographic and clinical characteristics are presented in Table 1.
Genotyping. The genotyping was successful for all 1069 participants. Table 2 represents the distribution of the wild type and the polymorphic genotypes across cases and controls.
Seventy-eight (24%) cases and 194 (26.1%) controls carried genetic variants of e-NOS intron 4 VNTR polymorphism. The controls followed Hardy-Weinberg equilibrium (p-value = 0.025).  Table 3 summarizes the association of NSAID, non-aspirin NSAID and aspirin use with UGIH, stratified by e-Nos intron 4 genotype (wild type/genetic variation).  Table 3, an association was observed between exposure to any NSAID and UGIH. Nonetheless, no substantial difference in the magnitude of this association was detected between NSAIDs users with e-NOS intron 4 (Table 3).

Modification of the effect of NSAIDs on UGIH by e-NOS VNTR polymorphism. As shown in
Our exploratory analysis suggests that there is no interaction between e-NOS intron 4 VNTR polymorphism and any NSAID on UGIH. We also did not observe an interaction between e-NOS intron 4 VNTR polymorphism and aspirin exposure. However, our findings reveal decreased odds of UGIH from a potential interaction between e-NOS intron 4 VNTR polymorphism and non-aspirin NSAID use, though the interaction estimates were not statistically significant [RERI: −2.68 (95% CI −6.67, 1.31); S: 0.53 (95% CI 0.18, 1.55)].  Table 3. Effect of e-NOS intron 4 VNTR polymorphism and NSAIDs exposure on the risk of UGIH. CI, confidence interval; N, number of participants; OR odds ratio; RERI, Relative Excess Risk due to Interaction; S, Synergism Index. a Odds Ratio adjusted for: period of patients' recruitment, previous history of arthrosis, infection with Helicobacter pylori, gastrointestinal disorders (ulcer and bleeding), exposure to inhibitors of the proton pump, exposure to antiaggregant, exposure to anticoagulants, the number of conducted interviews, and the reliability of the interview. Aspirin subgroup analysis was additionally adjusted for exposure to non-aspirin NSAIDs. RERI = 0 and S = 1 imply no interaction, RERI > 0 and S > 1 imply positive interaction and RERI < 0 and S < 1 imply negative interaction.

Discussion
Numerous studies associated the risk of gastrointestinal bleeding with genetic polymorphisms. However, there is a lack of knowledge about the influence of genetic variation on UGIH from NSAID and aspirin exposure. eNOS gene is known to exert fundamental activities in protecting gastric mucosa and repairing NSAIDs/aspirin-caused damage. Nevertheless, there is a shortage of information about the effect of variations in this gene on drug response 26 . In a recent study on the effect of polymorphisms on aspirin-related UGIH, we suggested that aspirin users who carry the inherited allele of rs1799983 polymorphism of NOS3 gene have lower odds of UGIH 21 . Therefore, in this multicentre case-control study, we conducted an exploratory analysis of the effect of eNOS intron 4 VNTR polymorphism on UGIH in users and non-users of NSAIDs (any NSAID, non-aspirin and aspirin NSAIDs). The findings of this study reveal that genetic variants of e-NOS intron 4 VNTR polymorphism might not interact with aspirin and consequently, they might not modify the odds of UGIH in aspirin users. Our results also suggest a potentially decreased odds of UGIH in users of non-aspirin NSAIDs who carry the polymorphic genotype of e-NOS intron 4 VNTR. The interaction between non-aspirin NSAIDs and e-NOS intron 4 VNTR polymorphism was not statistically significant, hence, these secondary findings should be considered exploratory and interpreted with caution until replicated in a larger study. If these findings were confirmed, it would mean that the odds of UGIH in any NSAID users carrying e-NOS intron 4 VNTR polymorphism could be overestimated due to the inclusion of aspirin users. They would also allow better management of non-aspirin NSAID personalized medicine.
At the molecular level, the production of NO, a mediator of various physiological functions in the gastrointestinal tract 27 , seems to be affected by the genetic variation of 27-bp VNTR of e-NOS intron 4. The microRNA derived from this VNTR polymorphism suppresses e-NOS expression, and in specific the ancestral allele "4b" generates more microRNA and thus inhibits further the production of NO 24,50,51 . On the contrary, the intake of low-dose aspirin increases NO production by blood vessels and stimulates e-NOS enzymatic activity 52 . As e-NOS is the main isoform responsible for the regulation of vasodilation in the gastrointestinal system, it is expected that carrying the allelic variant 4a of e-NOS may reduce the risk of gastrointestinal bleeding. However, our exploratory analysis revealed that e-NOS intron 4 VNTR polymorphism is not associated with reduced odds of bleeding in aspirin users. This finding could be influenced by the limited number of individuals with e-NOS intron 4 VNTR polymorphism who were on aspirin treatment in our study. Due to the lack of full-case control studies (i.e., studies that encompass users and non-users of aspirin), we cannot compare our findings with those of other reports. However, it is crucial to highlight that our results differ from that of Piazuelo and colleagues who found an association between the genetic variant "4a" of e-NOS gene and lower odds of UGIH in aspirin users 32 . The study of Piazuelo involved aspirin users exclusively, and thus the effect of 4a e-NOS allele on aspirin-related UGIH could not be determined 53 . Accordingly, future larger studies are required to guarantee a comprehensive assessment of the effect of e-NOS intron 4 VNTR polymorphism on UGIH in users of aspirin and NSAIDs.
In general, the use of VNTR polymorphism in NSAID or aspirin personalized medicine has been very limited so far, although the high polymorphic characteristics of VNTR make them very informative in determining the loci for diseases 31 . Research studies have mainly focused on investigating the effect of SNPs in complex diseases or studying drug responses 54 . To date, only five studies have examined the association of VNTR polymorphism with aspirin and/or NSAIDs. These studies were carried out in different populations and targeted distinct health conditions, highlighting therefore the need for more research on the influence of VNTR polymorphism on NSAIDs-and aspirin-related UGIH (Table 4).
Though our study is exploratory, it is characterized by various points of strength that might help designing future larger studies. Including both groups of participants "exposed and unexposed to NSAIDs" and stratifying by the type of NSAIDs (i.e., aspirin and non-aspirin) permitted providing a comprehensive assessment of these drugs. Adjusting the measures of effect for a large number of potentially confounding factors decreased the risk of confounding in these estimates 59 . Limiting the analysis to a specific ethnic group (i.e., using the native tongue of the participant or the original mother/father language as a proxy of ethnicity) allowed avoiding bias due to ethnic differences. The inclusion of only biologically unrelated participants prevented bias from the over-representation of ancestral alleles 60 . The use of catalogue of NSAIDs prompt cards during the interview aided the participants www.nature.com/scientificreports/ to recall and identify better the use of these drugs, and reviewing the medical records of the patients assisted in the reduction of the risk of recall bias. Our findings need to be confirmed by larger studies and in different populations. Stratifying the study population by drug exposure and genotype resulted in a limited number of UGIH cases and controls in each stratum, especially in aspirin subgroup analysis, which was then reflected in the width of the estimated confidence intervals. The sample size limitation also precluded carrying out the initially planned dose-response analysis. Therefore, our findings correspond to any exposure to NSAIDs/aspirin. Reproducing the study in a larger population would allow for (1) increasing the number of cases and controls per subgroup and consequently improving the statistical power of the observed associations and (2) undertaking further analysis such as evaluating a dose-response relationship. Furthermore, our results cannot be generalized to non-white European populations due to ethnic differences. Replicating the study in different ethnic groups would permit testing the association between e-NOS intron 4 VNTR polymorphism, NSAIDs and UGIH in genetically different populations.
In conclusion, this exploratory study suggests that exposure to NSAIDs or aspirin does not modify the odds of UGIH according to patient's e-NOS intron 4 genotype. It also suggests that e-NOS intron 4 VNTR polymorphism might reduce the odds of UGIH in non-aspirin NSAID users. More research is needed to understand the pharmacogenetics of NSAIDs, especially aspirin, which possess a large spectrum of protective and prophylactic properties, and more room should be given to VNTR polymorphism in studies oriented towards personalized medicine.

Data availability
The datasets generated and analyzed for this study can be found in the FigShare