Association between proton pump inhibitor use and the risk of inflammatory bowel disease: A systematic review and meta-analysis

Abstract

BACKGROUND

The worldwide prevalence of inflammatory bowel disease (IBD) continues to rise, motivating extensive exploration into possible contributing factors. Proton pump inhibitors (PPIs) are frequently prescribed medications for managing various gastrointestinal disorders. However, uncertainties persist regarding the implications of chronic PPI therapy and its potential association with the onset of IBD.

AIM

To assess the association between PPI use and incident IBD risk via a systematic review and meta-analysis of observational studies.

METHODS

A comprehensive literature search was performed in databases including PubMed, EMBASE, and Web of Science, covering the period from inception to February 13, 2025. Search terms included “proton pump inhibitors” and “inflammatory bowel disease”. Eligible were observational studies (cohort, case-control, and cross-sectional studies) reporting adjusted effect estimates for the association between PPI use and incident IBD. Study quality was assessed using the Newcastle-Ottawa Scale. Pooled odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using a random-effects model, and heterogeneity was quantified with the I² statistic.

RESULTS

A total of five observational studies, encompassing 48058263 participants across diverse age groups, were included in this meta-analysis. Findings revealed that participants with PPI exposure had a significantly higher risk of developing IBD than those without exposure (OR = 1.85; 95%CI: 1.11-3.09, P < 0.05). However, sensitivity analyses revealed that this association was unstable; excluding one large cross-sectional study reduced the OR to 1.40 (95%CI: 0.98-1.98), which was no longer statistically significant. Moreover, there was notable heterogeneity among the included studies (I² = 97.6%, P < 0.001). Subgroup analyses indicated that variations in participant age and study methodology notably contributed to the observed heterogeneity.

CONCLUSION

The statistically significant association between PPI use and IBD risk is non-robust and unstable, driven by extreme heterogeneity and studies prone to protopathic bias. Well-designed prospective studies are required to validate these findings.

Key Words: Inflammatory bowel disease; Proton pump inhibitor; Risk factor; Observational study; Protopathic bias

Core Tip: Proton pump inhibitors (PPIs) are among the most commonly prescribed medications worldwide, yet their long-term safety remains under debate. This systematic review and meta-analysis demonstrate a statistically significant association between PPI use and an increased risk of incident inflammatory bowel disease. However, substantial heterogeneity and sensitivity analyses suggest that this association may be influenced by study design, age differences, and protopathic bias, thereby limiting causal inference. Clinicians should interpret this association cautiously, and future prospective studies with rigorous bias control are needed to determine whether PPI use contributes to inflammatory bowel disease pathogenesis.

INTRODUCTION

Inflammatory bowel disease (IBD), comprising Crohn’s disease and ulcerative colitis, is defined as a chronic inflammatory disorder driven by immune responses, resulting in recurring inflammation within the gastrointestinal system[1]. Common symptoms typically include unintended weight loss, bloody diarrhea, abdominal discomfort, and cramping. Beyond gastrointestinal manifestations, IBD has the potential to impact a wide range of physiological functions[2]. The prevalence of IBD is currently escalating globally, with an estimated 6.8 million individuals affected worldwide[3]. Although the exact cause of IBD has not been fully elucidated, existing research points toward a complex interplay of genetic susceptibility, environmental triggers, immune system dysregulation, and disruptions in the gut microbiota[4,5].

Proton pump inhibitors (PPIs) primarily exert their therapeutic effect by irreversibly inhibiting H⁺/K⁺-ATPase, thus becoming the cornerstone treatment for various acid-related gastrointestinal conditions, such as Zollinger-Ellison syndrome, gastroesophageal reflux disease, and peptic ulcer disease[6]. By elevating intragastric pH levels above 4 for durations between 15 and 21 hours, PPIs offer substantial therapeutic benefits[7]. Owing to their clinical efficacy, PPIs have emerged among the most frequently prescribed long-term medications globally[8]. For instance, epidemiological studies report that around 8% of United States adults utilized PPIs monthly during the 2011-2012 period, while Denmark experienced a fourfold increase in PPI usage prevalence to 7.4% from 2002 to 2014[9]. Crucially, this widespread prescription pattern frequently exceeds recommendations outlined in clinical practice guidelines, prompting concerns about potential long-term adverse consequences.

Long-term administration of PPIs has been associated with alterations in intestinal microbiota, enhanced bacterial translocation, and heightened susceptibility to infections such as Clostridioides difficile[10,11]. Notably, microbial dysbiosis and reduced microbial diversity constitute key characteristics of the intestinal microbiome in IBD patients, and Clostridioides difficile infection is recognized as a well-established trigger for disease exacerbation[12]. Nevertheless, the contribution of PPI-induced microbial disturbances to the pathogenesis of de novo IBD remains to be elucidated. Despite mechanistic insights, epidemiological studies linking PPI use to IBD susceptibility have yielded indeterminate results[13,14], likely attributable to methodological discrepancies and residual confounding[15]. Given the considerable healthcare burden posed by IBD and the extensive prescription of PPIs, a thorough evaluation of existing evidence is warranted to inform clinical practice. Accordingly, this study systematically reviewed and quantitatively synthesized available observational data to clarify the relationship between PPI use and the subsequent development of IBD.

MATERIALS AND METHODS

Literature sources

This systematic review and meta-analysis strictly followed the rules laid out by the PRISMA[16]. With the registration number CRD420251157870, its protocol was prospectively registered in PROSPERO. Electronic databases such as PubMed, EMBASE, and Web of Science were searched, including literature produced from inception to February 13, 2025. Search strategies combined controlled vocabulary terms such as “Inflammatory Bowel Disease” and “Proton Pump Inhibitor” with relevant free-text terms connected via Boolean logic. A full description of search procedures is available in the Supplementary material.

Selection criteria

All retrieved literature was managed using EndNote 20. Screening commenced with the removal of duplicate records, followed by initial screening based on titles, abstracts, and keywords. Articles meeting the eligibility criteria underwent rigorous secondary evaluation against predefined inclusion and exclusion criteria. Final inclusion was confirmed through comprehensive full-text review.

Inclusion criteria: (1) Study design and subjects. Observational studies assessing the risk of IBD in subjects receiving PPIs for any indication, without restrictions on age or sex; (2) Exposure. Individuals with documented PPI exposure within the preceding 12 months, current PPI users, or those receiving long-term PPI therapy for any indication, irrespective of age or sex; and (3) Outcomes. The primary outcome was the incidence of IBD following PPI exposure.

Exclusion criteria: (1) Reviews, descriptive studies, editorials, conference abstracts, and studies lacking sufficient information; and (2) Literature unavailable in full text. Two researchers independently screened the literature according to the predefined criteria. Discrepancies were resolved by discussion with experienced reviewers to achieve consensus.

Data extraction

Two reviewers independently performed data extraction utilizing a structured extraction form. Information gathered included author details, publication year, journal name, geographic location, study design, total sample size, demographic characteristics (age, gender), definitions of PPI exposure, follow-up duration, key covariates adjusted for in each primary study, and risk estimates with 95% confidence interval (CI). If explicit risk estimates were unavailable, they were computed from accessible data. Any disagreements arising during extraction were resolved through discussion, and if unresolved, a third independent reviewer was involved to achieve final consensus.

Assessment of study quality

Two researchers independently evaluated the methodological quality of the chosen studies using the Newcastle-Ottawa Scale (NOS), which includes three components: Selection of participants, comparability of research groups, and determination of results[17]. The total number of points for the NOS ranges from 0 to 9, with studies being categorized as high-quality (≥ 7 points), moderate-quality (5-6 points), or low-quality (≤ 4 points). The quality of one cross-sectional study that was included in the analysis was also evaluated using the Agency for Healthcare Research and Quality criteria, which ranks studies as either low (between 0 and 3 points), moderate (between 4 and 7 points), or high (between 8 and 11 points)[18]. Expert consultation or consensus discussions were used to resolve disagreements as needed.

Statistical analysis

The main objective of the analysis was to determine the total risk of incident IBD linked to PPI use. Adjusted risk assessments were given precedence over crude estimates whenever they were available. The included studies had a low incidence of IBD events, hence the researchers treated the two kinds of effect assessments [odds ratio (OR); hazard ratio (HR)] equally. Summary statistics were 95%CIs and ORs with accompanying graphical representations in Review Manager 5.4 and Stata. P < 0.05 was used to determine statistical significance. Given that Onwuzo et al[19] reported the ORs for Crohn’s disease and ulcerative colitis separately, we merged the ORs using the techniques provided by Hamling et al[20]. The specific steps for this computation are detailed in the Supplementary material. Studies were found to have different levels of heterogeneity using Cochran’s Q test and the I² statistic (P < 0.10); I² > 50% indicated significant heterogeneity[21]. A random-effects model was used to create pooled estimates due to the methodological and demographic diversity. Sensitivity analyses were performed by iteratively excluding individual studies and recalculating the pooled risk estimate to evaluate result stability. Funnel plots were visually inspected to detect potential publication bias; statistical tests for publication bias were omitted due to fewer than 10 studies included in the analysis, consistent with established guidelines[22,23]. Additionally, a high degree of heterogeneity was further investigated through subgroup analyses based on methodological design, sample size, study quality, and participant age.

RESULTS

From database searches, 4587 studies were initially identified, comprising 209 from PubMed, 3259 from EMBASE, and 1119 from Web of Science. After removing 754 duplicates, a total of 3833 unique studies underwent preliminary screening using titles, abstracts, and keywords. This initial phase eliminated non-original publications and irrelevant entries, including reviews, letters, editorials, annotations, patents, meta-analyses, and editorial materials. Subsequently, the eligibility of 18 articles selected for full-text assessment was rigorously evaluated according to predetermined criteria. Eventually, five studies satisfied the inclusion standards and were selected for inclusion in the meta-analysis (Figure 1).

Figure 1  PRISMA flow diagram for the study selection process.

Study characteristics and quality assessment

Data regarding baseline characteristics from these five selected studies were systematically collected using a structured extraction format. Table 1 provides an overview of their methodological details and the results of quality appraisal. These studies, published between 2019 and 2023, included two cohort studies[13,14], two case-control studies[24,25], and one cross-sectional investigation[19]. Altogether, they represented a combined sample size of 48058263 participants, with two studies specifically enrolling pediatric populations, and a predominance of female subjects overall. All studies performed adjustments for multiple confounders and provided adjusted ORs or HRs. Definitions of PPI exposure differed among studies, primarily regarding the criteria used to establish PPI initiation. Methodological quality, assessed via NOS and Agency for Healthcare Research and Quality scales, indicated that all studies possessed moderate to high methodological rigor (Supplementary Table 1).

Table 1 Characteristics and summary of included studies.

Ref.	Country	Study design	Journal	Sample size	Age	Sex (female %)	Definition of PPIs use	Effect estimates	Quality score	Follow-up duration	Key covariates
Räisänen et al[25], 2023	Finland	Case-control study	Front Pediatr	2389	16.7 ± 1.2	57%	At least one PPI purchase record within 6 months prior to diagnosis	aOR = 8.38 (95%CI: 1.09-64.35)	NOS: 5	NA	Age, sex, area of residence, preterm birth, maternal socioeconomic status, number of antibiotic uses before the age of 3, use of systemic glucocorticoids
Onwuzo et al[19], 2023	United States	Cross-sectional study	Cureus	45586150	18-65	NA	PPI related records in the database	UC: AOR = 2.09 (95%CI: 1.98-2.06); CD: AOR = 2.79 (95%CI: 2.75-2.84)	AHRQ: 8	NA	Age, sex, NSAIDs, smoking, alcoholism, GERD, IBS, and metabolic syndrome
Abrahami et al[13], 2023	United Kingdom	Cohort study	Gut	1820890	> 18	54.5%, 56.6%	Receive PPI prescriptions for the first time on record	HR = 1.05 (95%CI: 0.90-1.22)	NOS: 8	4.2 years	Age, sex, year of cohort entry, BMI, smoking, alcohol-related disorders, autoimmune conditions, NSAIDs, HRT, oral contraceptives, DPP-4 inhibitors, antibiotics, statins, evidenced-based indications and off-label indications, faecal occult blood testing or colonoscopy, mammography, PSA testing, influenza and pneumococcal vaccinations, total number of unique drug classes
Schwartz et al[24], 2019	United States	Case-control study	J Pediatr Pharmacol Ther	1427	≤ 21	43.3%, 51.5%	At least one PPI prescription within 2 years to 5 years prior to diagnosis	aOR = 3.60 (95%CI: 1.10-11.74)	NOS: 6	NA	Age, sex, race, clinic location, number of antibiotic prescriptions, proportion of adults who are high-school graduates, and proportion of family households with below-poverty-level income
Xia et al[14], 2021	United States, United Kingdom	Cohort study	Gastroenterology	647407	57.2 ± 11.2, 59.9 ± 7.3	66.6%	Patients reported regular use of PPI in the past	HR = 1.42 (95%CI: 1.22-1.65)	NOS: 8	12 years, 8.1 years	Race, BMI, menopausal status, smoking, alcohol, physical activity, diet quality, fruit and vegetable intake, family history of IBD, GERD, gastric or duodenal ulcer, GI bleeding, multivitamin, NSAIDs, aspirin, statin, oral contraceptive, postmenopausal hormone use, colonoscopy/sigmoidoscopy, IBS, immune-mediated inflammatory diseases, self-reported overall health and longstanding disease

PPI: Proton pump inhibitor; aOR: Adjusted odd ratio; CI: Confidential interval; NOS: Newcastle-Ottawa Scale; NA: Not available; UC: Ulcerative colitis; CD: Crohn’s disease; AHRQ: Agency for Healthcare Research and Quality; HR: Hazard ratio; NSAIDs: Non-steroidal anti-inflammatory drugs; GERD: Gastro-oesophageal reflux disease; IBS: Irritable bowel syndrome; BMI: Body mass index; HRT: Hormone replacement therapy; DPP-4: Dipeptidyl peptidase-4; PSA: Prostate-specific antigen; GI: Gastrointestinal.

Risk of IBD after PPI exposure

As shown in Figure 2, the random-effects meta-analysis suggested an increased risk of IBD among PPI users (pooled OR 1.85), albeit with extreme heterogeneity (I² = 97.6%, P < 0.001). Subgroup analyses (Figure 3) indicated that study design and participant age were key contributors to this heterogeneity, with notably higher point estimates observed in retrospective studies and pediatric populations.

Figure 2 Forest plot for the association between proton pump inhibitor use and incident inflammatory bowel disease. OR: Odds ratio; CI: Confidential interval.

Figure 3 Subgroup analysis of proton pump inhibitor use and inflammatory bowel disease risk. A: Subgroup analysis by study design; B: Subgroup analysis by sample size; C: Subgroup analysis by study quality; D: Subgroup analysis by participant age. OR: Odds ratio; CI: Confidential interval.

Publication bias and sensitivity analysis

Visual examination of the funnel plot suggested the possibility of publication bias, as evidenced by apparent asymmetry (Supplementary Figure 1). Nevertheless, as recommended by the Cochrane Collaboration Handbook, formal statistical assessments for publication bias are generally discouraged when fewer than ten studies are included in the analysis, owing to insufficient statistical power to reliably differentiate between genuine asymmetry and chance effects[23]. Furthermore, the trim-and-fill analysis demonstrated stability in the pooled estimate, with no additional studies being imputed into the analysis (Supplementary Figure 2). Further quantitative tests were not conducted due to the limited number of included studies.

Sensitivity analyses conducted using the leave-one-out method evaluated the impact of individual studies on the pooled result (Supplementary Figure 3). Removal of each study separately resulted in variations of the pooled OR ranging from 1.40 (95%CI: 0.98-1.98) to 2.17 (95%CI: 1.35-3.49) (Table 2). Notably, excluding the cross-sectional study conducted by Onwuzo et al[19], which accounted for 28.3% of the total weight, substantially attenuated the pooled estimate to a non-significant level (OR = 1.40, 95%CI: 0.98-1.98), underscoring that the main findings were considerably influenced by a single study.

Table 2 Sensitivity analysis on proton pump inhibitor use and risk of inflammatory bowel disease.

Study omitted	OR	95%CI
		Lower limit	Upper limit
Räisänen et al[25], 2023	1.71	1.01	2.88
Onwuzo et al[19], 2023	1.40	0.98	1.98
Abrahami et al[13], 2023	2.17	1.35	3.49
Schwartz et al[24], 2019	1.70	0.99	2.94
Xia et al[14], 2021	2.14	1.07	4.30

OR: Odds ratio; CI: Confidential interval.

DISCUSSION

Current evidence concerning the safety of PPIs and their possible role in triggering IBD remains inconsistent. In this systematic review and meta-analysis, data from five studies were synthesized to clarify the relationship between exposure to PPIs and the risk of developing IBD. Findings indicated a statistically significant association between PPI use and elevated IBD incidence (OR = 1.85; 95%CI: 1.11-3.09), whereas marked heterogeneity among included studies was evident.

Despite the inconsistent nature of existing epidemiological findings, several mechanisms grounded in biological plausibility support a potential link between PPIs and increased IBD risk. The primary hypothesis involves dysbiosis induced by PPIs. Specifically, by inhibiting gastric acid secretion, PPIs potentially compromise gastric barrier functions, facilitating colonization of the lower gastrointestinal tract by oral microbes and altering intestinal microbiome composition[26-31]. This microbiota imbalance, analogous to that observed in IBD patients, may result in decreased short-chain fatty acid production, subsequently impairing epithelial barrier integrity and regulatory T-cell function[32]. Simultaneously, PPIs might directly impair intestinal epithelial barrier function by upregulating myosin light chain kinase, increasing tight junction permeability[33-35]. This structural impairment enables microbial antigen (e.g., liposaccharide) translocation, subsequently activating innate immune pathways such as the Toll-like receptor 4-nuclear factor κB signaling cascade, and promoting T helper 17-mediated mucosal inflammation. Metabolic dysregulation is another potential factor. PPI therapy alters bile acid profiles and suppresses farnesoid X receptor signaling, disrupting the crosstalk between microbiota and host immune homeostasis. Additionally, PPIs may interfere with neutrophil-derived reactive oxygen species production and cellular autophagy, further exacerbating autoimmune pathology[36]. Nevertheless, the observed association may partly reflect reverse causation and drug co-exposure effects, as PPIs are frequently prescribed for prodromal IBD symptoms before diagnosis.

Several systematic reviews have suggested an association between PPI use and IBD development without quantitatively synthesizing overall effect sizes[37]. Our primary findings are consistent with previous studies indicating increased IBD risk among PPI users[14,24]. However, a notable exception is the long-term study by Abrahami et al[13], which reported no increased IBD risk associated with PPIs (HR = 1.05; 95%CI: 0.90-1.22), maintaining robust findings across dose-response and subgroup analyses after bias adjustment. The discrepancy between studies likely arises from differing methodological approaches to controlling protopathic bias and other covariates. Abrahami et al[13] employed a rigorous new-user, active-comparator design (PPI vs histamine-2 receptor antagonist) with a 2-year lag period, effectively balancing gastrointestinal symptom-related confounding and eliminating protopathic bias due to PPIs prescribed for prodromal IBD symptoms. Their study demonstrated that including early events yielded a higher HR (HR = 1.39), consistent with results from Xia et al[14] (HR = 1.42) and Onwuzo et al[19] (OR = 2.43)[14]. After excluding early events, the association diminished to null (HR = 1.05; 95%CI: 0.90-1.22), suggesting that stringent bias control provides more reliable estimates.

This methodological difference directly influences the interpretation of our pooled estimate. Compared with the other four studies, the cross-sectional study by Onwuzo et al[19] did not include a lag period, highlighting protopathic bias. Sensitivity analysis confirmed this; excluding Onwuzo et al[19], which contributed 28.3% of the weight, decreased the pooled OR from 1.85 (95%CI: 1.20-2.70) to 1.40 (95%CI: 0.98-1.98). Thus, the positive association heavily relies on studies vulnerable to reverse causation. Protopathic bias occurs when medications treat early, undiagnosed disease symptoms[38]. In IBD, gastrointestinal symptoms such as abdominal pain or reflux often precede diagnosis[39]. A retrospective analysis by Singh et al[40] reported individuals were three times more likely to receive PPIs in the three years before IBD diagnosis compared to matched controls, irrespective of age. Consequently, the four studies incorporating a lag period accounted for reverse causation, resulting in conservative risk estimates. In contrast, the cross-sectional study, defining exposure as any prior PPI prescription without adjustment, likely captured prodromal prescriptions, exaggerating the association. Given its large sample size (> 60% of total person-years), its influence on pooled estimates is amplified. Similarly, our subgroup analyses indicated that study design (prospective vs non-prospective) significantly contributed to heterogeneity. Therefore, while this meta-analysis demonstrates an overall statistical association, the pronounced heterogeneity and sensitivity to a single study lacking protopathic bias safeguards necessitate cautious interpretation.

Subgroup analysis indicated an elevated IBD risk in pediatric populations compared with adults. This increased risk in children is clinically significant, as PPIs are commonly prescribed for pediatric conditions such as gastroesophageal reflux disease and Helicobacter pylori eradication, despite uncertain indications in infants[41]. The age-dependent risk may be attributable to developmental susceptibility in the pediatric gut. Gastrointestinal microbiome alterations, reduced gastric mucus viscosity, and enhanced bacterial invasion disproportionately affect children, whose immune systems, intestinal barriers, and microbial ecosystems are still maturing[42,43]. These vulnerabilities might synergize with PPI-mediated ecological shifts, predisposing pediatric patients to IBD.

PPIs rank among the most frequently prescribed medications, yet their overuse imposes substantial economic burdens[44]. However, concerns about their long-term safety could lead to inappropriate discontinuation when clinically indicated[45]. Given their importance for acid-associated diseases, cautious evaluation of their risks and benefits is essential. This meta-analysis, pooling data from several observational studies, suggests a potential association between PPIs and increased IBD risk, especially in pediatric populations. Although the risk increase appears small, and observational data have inherent limitations, promoting rational PPI use remains essential given accumulating evidence of adverse effects[46]. Strict adherence to established guidelines is recommended[47]. Additionally, regularly reassessing treatment necessity, individualizing dosage, and considering therapeutic alternatives may be beneficial strategies[48].

Several limitations should be acknowledged when interpreting the findings of this review. Due to the observational design of the included studies, the review is inherently subject to biases such as selection bias and residual confounding. Furthermore, the relatively small number of eligible studies assessing PPI use and incident IBD also limits the generalizability of these conclusions. Although adjusted risk estimates were used, residual confounding may persist. Additionally, significant heterogeneity across studies prevents definitive conclusions. Furthermore, insufficient data precluded exploration of variations among different PPI agents and dose-response relationships. Finally, reliance on prescription records and ICD codes potentially introduced misclassification bias for exposure and outcomes. This systematic review and meta-analysis highlight variations in reported IBD risk based on study methodology, patient demographics, and latency periods following initiation of PPI treatment. Given these constraints, future high-quality prospective cohort studies are warranted to accurately quantify the potential risk associated with PPI use and to identify specific patient populations that may be at increased risk of PPI-related IBD.

Despite these limitations, this review has notable strengths. It provides the most current and comprehensive synthesis of evidence on the association between PPI use and IBD risk, following rigorous systematic review methodology. Moreover, the inclusion of original studies with moderate-to-high quality and large sample sizes enhances the statistical power and reliability of pooled estimates. Our work identifies sources of heterogeneity and highlights gaps regarding dosing and duration data, laying the foundation for future research on vulnerable subgroups and underlying biological mechanisms.

CONCLUSION

In conclusion, while a meta-analytic pooling of available studies yields a statistically significant association, this finding is not robust. It is characterized by extreme heterogeneity and is entirely dependent on the inclusion of studies highly susceptible to protopathic bias. Therefore, the current body of observational evidence is of very low certainty and is insufficient to support a causal link between PPI use and IBD risk. Future high-quality prospective studies with rigorous designs to minimize protopathic and confounding bias are essential.