Proton Pump Inhibitors, H₂ Blocker Use, and Risk of Inflammatory Bowel Disease in Children

From 1990 through the 2000s, the burden of pediatric inflammatory bowel disease (IBD) has increased globally.1 In North America, more than 1.3% of the general population has IBD, amounting to over 3.1 million Americans.2 Pediatric IBD can result in adverse consequences, such as delayed puberty and growth failure.3 

Previous longitudinal and cross-sectional studies have shown that IBD is associated with changes in the composition of the gut microbiota,4,5 emphasizing the role of environmental factors in the development and progression of the disease. If microbiome changes influence IBD risk, then it is plausible that drugs that alter the microbiome may also alter the risk of IBD. One drug class that is known to alter the gut microbiome is acid-reducing medications, such as proton pump inhibitors (PPIs) and H₂-receptor agonists (H₂ blockers);6,7 however, whether these drugs alter the microbiome in a way that increases the risk of IBD is still unclear. There has been a significant upsurge in the use of PPIs and H₂ blockers among infants and children in the past decade.8–10 Proton pump inhibitors and H₂ blockers may provoke disease flares in individuals with established IBD;11 however, no study has investigated the relationship between the use of PPIs, H₂ blockers, and incident IBD. Given the higher severity of IBD in the pediatric population,12,13 and the limited evidence of PPI and H₂ blocker safety from clinical trials in younger populations, this gap in knowledge is concerning. The hypothesis was that the use of acid suppression therapy in children might be associated with the development of pediatric IBD.

Study Setting. Kaiser Permanente Northern California (KPNC) is a closed, prepaid, integrated health plan that serves 30% of the San Francisco Bay Area population, with over 4 million currently enrolled members.14,15 The membership of KPNC is representative of the underlying population of the San Francisco Bay Area with respect to race and socioeconomic status (SES).14 All patient encounters, prescription fills, and laboratory results have been recorded in a computerized database since 1996.

Study Design, Subjects, and Data Source. This nested case-control study16,17 used data from KPNC electronic health records as described earlier for the Kaiser Permanente Autoimmune Registry.18 Inpatient and outpatient data were used to identify all children diagnosed with IBD at ≤21 years of age between 1996 and 2016. Cases were selected as those 285 children in the health plan with at least 5 years of continuous membership (with no coverage gaps longer than 60 days) prior to the date of IBD diagnosis (i.e., the index date). Four controls drawn from the general KPNC pediatric population were matched to each case on age (within 1 year), race (Asian/Pacific Islander, black, white, Native American, multiracial, or unknown/other), primary clinic location, and membership status at the case's index date. These data were obtained from membership data and the electronic medical record. Controls were required to have been members at least as long as their matched case, and not to have an IBD diagnosis as of the index date for the case to which they were matched.

Classification of Outcome. IBD diagnosis was defined by International Classification of Disease, Clinical Modifications 9 (ICD-9-CM) codes for IBD (555 for ulcerative colitis [UC] and 556 for Crohn's disease [CD]) and by ICD, Clinical Modifications 10 (ICD-10-CM) codes for IBD (K50 for CD and K51 for UC), using diagnostic codes recorded during inpatient and outpatient encounters. Cases were required to have had at least 2 inpatient or outpatient visits with IBD diagnoses recorded. According to previous research, this case definition has 95% positive predictive value (95% CI, 94%–96%).19 

Classification of Exposure. Use of PPIs and H₂ blockers was assessed from the health plan's outpatient pharmacy database that records all details of prescriptions and dispensing of medications to health plan members. This database was searched for National Drug Codes (NDC) matching PPIs and H₂ blockers. Owing to the possibility that these medications were prescribed for treatment of symptoms attributable to undiagnosed IBD, a 2-year “lag period” was implemented when assessing exposure. Thus, a case or control was only considered to be exposed if there was PPI or H₂ use between 2 and 5 years before the index date. This reduced the possibility of protopathic bias,20 which occurs when a drug of interest is initiated to treat symptoms of the disease under study before it is diagnosed.

Chart Review. A chart review was conducted for all cases and controls in the study who were prescribed PPIs, and a random sample of subjects who were prescribed H2 blockers. The review covered the period 2 to 5 years prior to index date to determine the medical indication for the prescription and whether there was contemporaneous evidence of IBD symptoms (e.g., diarrhea, bloody stools, abdominal pain). The goal was to assess the indication for which the drug was prescribed to determine if prodromal symptoms were plausible as an explanation. The note from the clinic visit preceding the first prescription during the study period was examined for patient-reported symptoms, concomitant diagnostic codes, additional medications, and the primary clinical reason for the prescription.

Potential Confounders. Covariates that preceded and may have been associated with both IBD and exposure to PPIs were included in our analysis as potential confounding factors. In addition to the matching factors (age at index date, race, and primary clinic location), these candidate confounders included antibiotic medication use, sex, and SES. Data on antibiotic exposure were obtained by using the same pharmacy databases that were used to identify PPI and H₂ blocker prescriptions. Sex was obtained from the electronic medical record. Two census tract-level measures21 representing SES were identified by using residential address geocodes: proportion of resident adults who are high-school graduates and proportion of family households with below-poverty level income.

Because very few children were prescribed PPI or H₂ blocker medications in this database, traditional statistical adjustment methods were not used. Instead, to control for potential differences in risk of IBD between users of PPI and H₂ blockers and non-users of either type of drug, a disease risk score (DRS) was calculated for all subjects. In general, a DRS estimates the probability of disease for each member of a study population in the absence of the exposure, regardless of true exposure status.22 This method allowed us to account for medication exposure adjusted for a single measure of disease risk, which was critical because of the few children exposed to PPI and H₂ blockers in this population. In addition, the method is compatible with conducting a nested case-control design.23 Both DRSs and propensity scores address statistical problems that arise when there is a large number of covariates and a rare exposure (DRS) or outcome (propensity scores), but a DRS is preferable to propensity scores when the exposure is rare.22 The DRS was estimated from a logistic regression analysis of sex, year of birth, number of antibiotic prescriptions in the 2 to 5 years prior to index date, and the 2 census-level measures described above on the odds of being an IBD case in this study.

Statistical Analysis. Conditional logistic regression16 was used to compare cases and controls with respect to PPI and H₂ blocker prescriptions filled between 2 and 5 years before the diagnosis, adjusting for the DRS as a continuous measure. To consider the possibility that early symptoms might drive an increase in PPI and H₂ blocker prescriptions in the 2 years prior to diagnosis, additional analyses considered exposure within 1 year and 1 to 2 years prior to diagnosis. These analyses allowed for comparison with our a priori exposure period of interest to look for an increase in prescriptions in this time frame (Supplementary Table 1). Odds ratios and 95% CIs were calculated by using conditional logistic regression models that included both PPIs and H₂ blockers. All analyses were conducted by using SAS version 9.4 (SAS Institute, Cary, North Carolina) and proc logistic. This study was approved by the Kaiser Foundation Research Institute Institutional Review Board.

Demographic Characteristics. Two hundred eighty-six cases and 1144 controls were identified (Table 1). One case was dropped after it was discovered that his IBD diagnosis preceded his PPI prescription, for a total of 285 cases. Two controls were missing data for SES measures and thus were dropped from the analysis for a total of 1142 controls. The mean age at index date was 15 years for both cases and controls. More than half (54%) of cases and controls were white. Hispanic was the second most common race/ethnicity category for both cases and controls (14%), followed by multiracial (12%), Asian/Pacific Islander (11%), black (7%), and Native American (1%). Controls were more likely to be female than cases (51% versus 43%). Cases and controls received similar numbers of antibiotic prescriptions in the 2 to 5 years prior to index date. Socioeconomic status measures were also similar between cases and controls.

Association With PPIs and H₂ Blockers. During the 2 to 5 years before the index date, 2.1% of cases (n = 6) and 0.5% of controls (n = 6) were prescribed PPIs; 3.5% of cases (n = 10) and 1.6% of controls (n = 18) were prescribed H₂ blockers (4 children were prescribed both [2 cases and 2 controls]). The OR for the association of receipt of at least 1 prescription with risk of subsequent IBD was 3.6 (95% CI, 1.1–11.7) for PPIs and 1.6 (95% CI, 0.7–3.7) for H₂ blockers after accounting for potential confounders and use of the other medication class (Figure).

There was a substantial increase in the OR for the association of receipt of at least 1 PPI prescription with risk of subsequent IBD during the 2 years prior to diagnosis (Supplementary Table 1).

Chart Review. Among the 12 children prescribed PPIs during the study period, all but 2 (1 case and 1 control) had been diagnosed with gastrointestinal symptoms or conditions at the visit preceding the PPI prescription (Table 2). Two controls and 4 cases were prescribed PPIs for gastroesophageal reflux disease (GERD) or possible GERD. There were no discernible differences in PPI indication between cases and controls. Our analyses of 4 cases and 5 controls with H₂ blocker use similarly did not identify differences between cases and controls (Supplementary Table 2).

Cases were 3.6 times more likely than controls to have been prescribed PPIs in the 2 to 5 years prior to diagnosis. The results from a chart review reveal a variety of indications for the prescription of PPIs with similar indications for initiating the medication class in cases and controls. However, owing to the small number of exposed individuals, further research is needed to assess this potential association.

Patients in our study were prescribed PPIs between 25 and 60 months prior to IBD diagnosis, while the average prodromal phase of IBD in pediatric populations is thought to be 7 to 11 months for CD and 5 to 8 months for UC.24 Despite this, there is a chance that these results reflect protopathic bias, due to treatment for early indications of IBD where the diagnosis process is delayed or complex. However, the association between H₂ blockers and IBD was weak, even though the indications for H₂ blockers were seen to be similar to the indications for PPIs (Supplementary Table 2). This makes it less likely that the association between PPI use and IBD is solely due to protopathic bias and suggests other possibilities.20 However, the large increase in risk during the 2 years prior to diagnosis (Supplementary Table 1) suggests that there may be some role of prodromal IBD symptoms in the use of these medications.

Previous literature has implicated PPIs as a risk factor for several health complications, including gastrointestinal conditions such as Clostridium difficile (C difficile) infection25 and small intestinal bacterial overgrowth.26 Although, to date, no studies have investigated the role of PPIs in the development of IBD, there is evidence that PPI exposure increases the severity of disease when prescribed to adults with a history of IBD.11,27 A large cohort study conducted in Canada found that patients with IBD given a new prescription for PPIs were more likely to experience IBD treatment escalation than patients with IBD who were not prescribed PPIs.11 A nested case-control study conducted within the Veterans Health Affairs system found that PPI prescriptions were associated with an increased risk of IBD-related hospitalization in patients with IBD.27 

There is a link between PPIs and intestinal dysbiosis.28,29 In an age-sex-matched cohort study, Takagi et al.29 analyzed fecal samples from 36 PPI users and 36 non-users. They found significant differences in the microbial composition of the gut, comparing users and non-users. This dysbiosis might be a mechanism by which PPIs increase risk of C difficile and small intestinal bacterial overgrowth. Given that IBD is also associated with disturbances to the gut microbiome,5,30 and that C difficile is very common among pediatric IBD patients,31 this adds plausibility to a potential association between PPIs and pediatric-onset IBD.

This study has several limitations. First, as mentioned above, the number of exposed individuals in our sample was small. This limits the conclusions that can be drawn from our results, although it is notable that there is a safety signal despite the small number of exposed children, which strongly suggests the importance of replication in additional cohorts. In particular, despite efforts to avoid it, the possibility of lengthy protopathic bias cannot be fully eliminated. This will always be a concern in diseases like IBD with long latency periods, especially when the exposure of interest is a drug that targets symptoms that may, at least in part, be similar to manifestations of early disease. However, if the disease is being misdiagnosed for such long periods in pediatric care, this is (in itself) a call for studies with improved power to document this and to improve the diagnostic process for IBD to allow for early and effective treatment.

Another limitation is the inability to capture PPI and H₂ blockers that may have been taken over the counter. Thus, there may be some exposure misclassification in this study, mostly in underascertainment of medication use. However, this bias most likely would be non-differential relative to IBD case status, because it is unlikely that over-the-counter PPI or H₂ blocker use differs by future disease status. It also seems unlikely that, in a population with health insurance, there would be significant self-treatment of children with these medications in the absence of medical visits or advice. The net effect of this misclassification is likely to be an attenuated measure of association by including some exposed children in the reference category and making the risk of this group more similar to that of children prescribed medication.

Pediatric IBD is a growing clinical concern that confers a substantial economic burden on families and health care systems.32,33 In the context of increasing numbers of acid-blocker prescriptions among children,8–10 understanding the adverse effects of these drugs is a public health priority. Short-term treatment of eosinophilic esophagitis and of GERD are the only 2 FDA-approved indications for PPI use in children.34 Only one-half of the children in this study who had been prescribed these drugs had symptoms consistent with these indications, highlighting the need for appropriate prescribing patterns. PPI use in childhood is also associated with allergic disease and bone loss.35,36 Our study adds to the growing body of evidence for prudent use of PPIs by providing evidence of an association between childhood use of PPIs and future IBD diagnoses. Overprescription of PPIs is not limited to pediatric populations37; thus, there is a need to investigate this association in adult populations, as well. Our results have implications for clinical treatment of children with gastrointestinal symptoms, suggesting either a safety signal for common drugs or underdiagnosis of IBD in children, and should be investigated further in a larger cohort of children.

Poster presentation at Digestive Disease Week 2018 – “Proton Pump Inhibitors (PPI), H2 Blocker Use, and Risk of Inflammatory Bowel Disease (IBD) in Children”

DOI: 10.5863/1551-6776-24.6.489.S1; DOI: 10.5863/1551-6776-24.6.489.S2

CD

Crohn's disease

DRS

disease risk score

GERD

gastroesophageal reflux disease

H₂ blockers

H₂-receptor agonists

IBD

inflammatory bowel disease

IBS

irritable bowel syndrome

ICD

International Classification of Disease

KPNC

Kaiser Permanente Northern California

NDC

National Drug Codes

PPI

proton pump inhibitors

PUD

peptic ulcer disease

SES

socioeconomic status

UC

ulcerative colitis