Proton Pump Inhibitors: Placing Putative Adverse Effects in Proper Perspective

Mitchell L. Schubert


Curr Opin Gastroenterol. 2019;35(6):509-516. 

In This Article

Potential Adverse Effects of Proton Pump Inhibitors

There is now a long list of potential serious adverse effects that have been associated with PPI therapy including alterations in gut microbiome, infection (enteric, peritonitis, pneumonia), drug interactions, microscopic colitis, micronutrient deficiencies, cancer, encephalopathy, chronic kidney disease, cognitive dysfunction, myocardial infarction, stroke, bone fracture, and childhood obesity.[8] The associations with relatively strong evidence are alterations in gut microbiome, enteric infection, drug–drug interactions, microscopic colitis, hypomagnesemia, and fundic gland polyps. It is possible that carcinogenesis (mediated by PPI-induced hypergastrinemia) may also turn out to be a true association. It should be noted that PPIs reside in the bloodstream for only a short period of time (elimination half-life ~1 h) and peak plasma concentrations are relatively low. The reason PPIs cause prolonged inhibition of acid secretion has little to do with blood concentrations but is rather because PPIs become concentrated and trapped within the secretory canaliculus of the parietal cell where they irreversibly bind to and inactivate HKA, the proton pump. Acid secretion resumes only after new pumps are synthesized and inserted into the luminal membrane of the parietal cell. Nevertheless, the sensationalized reporting of potential adverse associations of PPIs in the medical literature and lay press has influenced provider prescribing and patient adherence.

Gut Microbiome

The gut microbiome, constituted of ~40 trillion microorganisms, is important for gut homeostasis, gut immunity, and may protect against enteric infection.[9] The number of bacteria in the stomach is ~101 cells/g and increases to 103 cells/g in small intestine and reaches 1012 cells/g in the colon. It is conceivable that PPIs, by reducing acid secretion, an important defense mechanism against bacteria and fungi, may promote the survival and growth of certain ingested microbes, such as Lactobacillus spp. and Streptococcus spp..[10] It has been postulated that PPIs might also alter the composition of the gut microbiome by selecting the survival of certain microbes by targeting bacterial proton pumps.

In 20 GERD patients treated with PPIs for 8 weeks, there was an increase in abundance of oral bacteria, specifically Lactobacillus spp. and Streptococcus spp., in the feces compared with baseline.[11,12] In a study of 10 healthy volunteers, an increase in Streptococcus spp. was observed in the feces 4 weeks after once daily 20 mg esomeprazole.[13] In a study of 487 participants (77 controls, 193 ward patients, and 217 critically ill patients), PPIs were associated with colonization of the oropharynx with gut flora.[14] Although the most abundant organism isolated from oropharyngeal cultures was Candida spp. (31%), the most frequently isolated gut pathogens were Escherichia coli, Enterobacter spp., Enterococcus spp., Klebsiella spp., and Pseudomonas spp., each ~6%.

The clinical significance of these and other changes in the microbiome induced by PPIs remains to be elucidated. However, in a study of PPI initiation and withdrawal in 343 inpatients with cirrhosis, PPI users had an increase in oral bacteria in the feces as well as higher 30-day and 90-day readmission rates, independent of comorbidities, MELD, medications, and age.[15] Although the study does not prove that PPIs were responsible for the increased hospitalizations, it is, nevertheless, reasonable to prescribe PPIs judiciously in cirrhotic patients and systematically withdraw them, whenever possible.

Infection (Enteric, Peritonitis, and Pneumonia)

As gastric acid kills ingested microorganisms, it is plausible that PPIs, by reducing luminal acidity and elevating gastric pH, may predispose to enteric infection, particularly of acid-sensitive pathogenic organisms, such as Vibrio cholera, Salmonella spp., campylobacter spp., and perhaps, norovirus.[16] In mice, a PPI increased susceptibility to colitis after oral gavage with Citrobacter rodentium, an enteropathogen used experimentally as a model of human enterohemorrhagic Escherichia coli.[17]

Although numerous studies report an association between PPIs and Clostridium difficile, now termed Clostridiodes difficile, the data is not compelling.[18] All studies are observational, retrospective, nonrandomized, and suffer from significant heterogeneity; most have OR less than 3. The Clinical and Laboratory Standard Institutes, based upon scientific evidence, issued a name change from Clostridium difficile to Clostridiodes difficile that has been endorsed by the Centers for Disease Control and Prevention (CDC). Enhanced 16S rRNA gene sequencing and ribosomal protein sequencing revealed that Clostridium difficile is technically not a Clostridium. After much discussion, the organism was reclassified as Clostridiodes (Clostridia-like) in an effort to mitigate confusion from re-naming it entirely and to maintain common abbreviations used in the literature such as C. diff, CDI (C. diff infection), and CDAD (C. diff-associated disease) As spores of C. difficile are resistant to acid, it is more likely that the predisposition to CDI in patients on PPIs, if true, results from dysbiosis rather than to an acid antisecretory effect per se. In the meanwhile, it may be prudent to withhold PPIs, unless necessary, in patients with risk factors for CDI and in patients with recurrent CDI.

A meta-analysis of six studies involving 829 patients reported a small increased risk of enteric peritonitis in end-stage renal disease (ESRD) patients undergoing peritoneal dialysis maintained on antisecretory medications [OR = 1.27; confidence interval (CI) = 1.01–1.57].[19] The conclusions are unlikely to be clinically important as there was much heterogeneity amongst the studies, many confounding factors were not considered (e.g. dose of gastric-acid suppressants, cause of ESRD, and duration of dialysis), and the association held for only less potent acid suppressants (i.e. H2RAs but not PPIs).

Of more than 26 studies performed, only about half have shown a small association between PPIs and community-acquired pneumonia (OR <1.5). Using a large database in England of individuals at least 60 years old in primary care using PPIs for over a year, Zirk-Sadowski et al..[20] report that PPIs were associated with a greater hazard of incident pneumonia (hazard ratio = 1.82; CI = 1.27–2.54) in year 2 of treatment. Study limitations included unmeasured confounders (e.g. neurological disease that might have predisposed to pneumonia) and that PPI use was derived from prescription records. Using the Taiwan National Health Insurance Database that included 1201 cirrhotic patients with pneumonia, Hung et al.[21] report no difference in 30-day mortality between PPI and non-PPI users. The 90-day mortality, however, was slightly increased in those using PPIs (OR = 1.26; CI = 1.05–1.52).

Using the National Health Insurance Service-National Sample Cohort in Korea that included 8319 patients with acute ischemic stroke, Song and Kim[22] report, in a retrospective cohort study, that 25% of patients developed pneumonia over a 4-year period. In the multivariate regression analysis, there was a significantly increased risk for pneumonia with use of PPIs (hazard ratio = 1.56; CI = 1.24–1.96) and H2RAs (hazard ratio = 1.40; CI = 1.25–1.58). Stroke severity and disability, which might be a significant risk factor for pneumonia, were not assessed; neither was use of antiplatelet agents.

Drug–Drug Interactions

It is conceivable that PPIs, by raising luminal gastric pH, could adversely affect the bioavailability and absorption of certain medications. This past year, several studies reported adverse effects of PPIs on the effectiveness of chemotherapeutic agents. Tyrosine kinase inhibitors (TKIs: erlotinib, imatinib, dasatinib, nilotinib, and sorafenib), one of the largest classes of oral chemotherapeutic drugs, are used to treat small cell lung cancer, pancreatic cancer, renal cell cancer, leukemia, and liver cancer. These drugs rely on pH-dependent solubility to be dissolved within the stomach and absorbed. In a retrospective study using the Surveillance, Epidemiology, and End Results-Medicare database for the years 2007–2012, 12 538 patients with lung cancer, renal cancer, pancreatic cancer, chronic myelogenous leukemia, and liver cancer were identified; nearly a quarter of these patients received concomitant PPI and TKI.[23] Concomitant use was associated with a small increased risk of death (hazard ratio = 1.16; CI = 1.05–1.28). In support of this notion, Ohgami et al.[24] report that PPIs decrease the plasma concentration-to-dose and increase the oral clearance of erolotinib in 42 patients treated for nonsmall cell lung cancer. Likewise, in a study of 333 patients with soft-tissue sarcoma treated with pazopanib, a TKI targeting vascular endothelial growth factor receptors, platelet-derived growth factor receptors, and KIT, of which 59 received gastric acid-suppressive agents (PPIs and H2RAs), median progression-free survival, and overall survival were diminished for patients taking gastric acid-suppressive medications.[25] On the other hand, in a randomized cross-over pharmacokinetic trial that included 14 patients, esomeprazole had no significant effect on the geometric means for the area under the curve for regorafenib, an oral multikinase inhibitor currently used to treat colorectal cancer, gastrointestinal stromal tumor, and hepatocellular carcinoma.[26]

Capecitabine is an oral 5-fluorouracil prodrug used in the adjuvant treatment of gastrointestinal cancer. Recent studies suggest that PPIs may detrimentally affect its efficacy. A retrospective chart review of 389 patients with stage II and III colorectal cancer treated with capecitabine report that PPI recipients had a small but statistically significant decreased 3-year recurrence-free survival (69.5 vs. 82.6%) and were twice as likely to experience cancer recurrence or death as capecitabine-treated non-PPI users.[27]

The data suggests that use of PPIs must be appropriately managed in cancer patients undergoing oral chemotherapy. When possible, therapeutic drug monitoring may be required to improve therapeutic outcomes.

Microscopic Colitis

Risk factor for microscopic colitis include age more than 60 years, female sex, current or past smoking, autoimmune disease, solid organ transplantation, and drugs (e.g. NSAIDs, SSRIs, β-blockers, and PPIs). A systematic review that included 19 publications (5 case control studies and 14 case reports/series) describes a possible association between PPIs and microscopic colitis.[28] All studies were limited by small sample size and none examined the risk according to dose.

A Danish case–control study identified 10 665 patients with a first-time diagnosis of microscopic colitis.[29] All cases were histologically confirmed in the Danish Pathology Register and information on PPI use was obtained from the Danish Prescription Register. The study found a strong association between current PPI use and both collagenous colitis (OR = 6.98; CI = 6.45–7.55) and lymphocytic colitis (OR = 3.95; CI = 3.60–4.33). The association was observed with all PPIs but strongest with lansoprazole for collagenous colitis (OR = 15.74; CI = 14.12–17.55) and lymphocytic colitis (OR = +6.87; CI = 6.00–7.86). No clear dose–response pattern was observed.

Micronutrient Deficiency

Gastric acid is thought to facilitate the absorption of certain minerals (magnesium, iron, and calcium). Active transport of magnesium occurs in the large bowel via dedicated ion channels identified as transient receptor potential melastatin (TRPM) 6 and 7. Although rare, studies strongly suggest that PPIs increase the risk for hypomagnesemia. The Food and Drug Administration issued a warning regarding long-term use of PPIs and hypomagnesemia, even though duration of PPI therapy has been difficult to quantify in retrospective studies and there are no prospective controlled studies to support causation. This year, a cohort study in HIV-positive patients (329 in the PPI cohort and 5718 in the non-PPI cohort), using the Veterans Affairs Informatics and Computing Infrastructure, report that PPI users had a 3.16 (CI = 2.56–3.9) times higher risk of hypomagnesemia than the non-PPI cohort.[30] The mechanism by which PPIs may induce hypomagnesemia is not known but two single nucleotide polymorphisms in the TRPM type 6 gene, present in colonic enterocytes, increase the risk six-fold. Although a more acidic milieu increases TRPM6 activity, it is difficult to envision how a change in gastric pH could elicit a change in colonic luminal pH.


Gastric hypoacidity, induced by PPIs, interferes with the feedback pathway whereby luminal acid stimulates somatostatin secretion which, in turn, inhibits gastrin secretion. Patients on PPIs exhibit a decrease in gastric somatostatin secretion that is accompanied by a reciprocal increase in gastrin secretion. Median serum gastrin levels, in response to PPIs, are higher in female patients and positively correlate with PPI dosage.[31] Gastrin is not only a secretagogue but also a growth hormone capable of stimulating proliferation. Although the clinical significance of elevated fasting gastrin levels has not been definitively determined, a theoretical link between hypergastrinemia and various cancers has been proposed: gastrin receptors (termed gastrin/CCK2, CCK2, or CCK-B receptors) have been identified in various human cancers including adenocarcinomas of the stomach and esophagus[32] and hypergastrinemia has been associated with increased risk of gastric adenocarcinoma, neuroendocrine tumors (i.e. enterochromaffin-like cell carcinoids), and debatably esophageal adenocarcinoma and liver cancer.[33,34] In support of this hypothesis, a Korean study of 63 397 H. pylori-eradicated subjects followed for a median of 7.6 years, reports that PPI use, among aspirin nonusers, was associated with a 3.73 (CI = 2.11–6.60) higher risk for development of gastric cancer compared with nonusers;[35] the association was dose-dependent and time-dependent. In a population-based cohort study of 796 492 adults exposed to maintenance therapy with PPI in Sweden from 2005–2012, PPI therapy was associated with a standardized incidence ratio of 3.93 (CI = 3.63–4.24) for development of esophageal adenocarcinoma.[34] An increase in esophageal adenocarcinoma was also observed in the subset of patients without GERD. A United Kingdom-nested case–control study, using the Primary Care Clinical Informatics Unit database, in which 434 liver cancer cases were matched to 2103 controls, PPI use, but not use of H2RAs, was associated with increased liver cancer risk (OR = 1.80; CI = 1.34–2.41).[36]

These findings herald a warning regarding the use of PPIs, especially at high doses, in patients predisposed to develop cancer, such as after H. pylori eradication or after documentation of asymptomatic Barrett's esophagus. The basis by which PPIs may influence progression to cancer may be because of hypergastrinemia as well as hypochlorhydria-induced changes in the microbiome.

Liver Disease (Hepatic Encephalopathy and Spontaneous Peritoneal Peritonitis)

Several studies report an increase in hepatic encephalopathy associated with PPI use in cirrhotic patients. In a meta-analysis of seven studies with 4574 patients, PPI use was associated with an increased risk for hepatic encephalopathy (OR = 1.5; CI = 1.25–1.75).[37] In a retrospective study of 284 patients who underwent transjugular intrahepatic portosystemic shunt (TIPS), PPI use was associated with higher rates of hepatic encephalopathy (OR = 3.19).[38] In multivariate analysis, other significant risk factors for hepatic encephalopathy included age, male sex, higher MELD score, and previous hepatic encephalopathy. The mechanism through which PPI use affects risk of hepatic encephalopathy is not clearly defined but a potential mechanism may be dysbiosis with impaired hepatic fixation of nitrogen metabolism byproducts and other toxins by gut bacteria. It should be emphasized that 40% of patients taking PPIs in this study had no clear indication for acid suppressive therapy, a finding consistent with other studies in cirrhotic patients.

Hung et al.[39] used the Taiwan National Health Insurance Database to identify 858 cirrhotic patients with spontaneous peritoneal peritonitis (SBP) who were taking PPIs. Compared with matched patients not taking PPIs, PPI use was not associated with an increase in 30-day mortality but was associated with a slight increase in 90-day ( hazard ratio = 1.39; CI = 1.15–1.63) and 1-year ( hazard ratio = 1.30; CI = 1.10–1.53) mortality. In an effort to reduce confounding, the study excluded patients with active gastrointestinal bleeding. However, the study did not control for severity or cause of cirrhosis as well as dose and duration of PPI.

Cognitive Decline

Inconsistent and conflicting results have been reported regarding the association of PPIs with cognitive decline. The accumulative data suggests a lack of a causal relationship between PPI use and changes in cognitive function. A recent meta-analysis, that included six cohort studies, concluded that there was no statistical association between PPI use and increased risk of dementia or Alzheimer's disease.[40] Similarly, Park et al.,[41] using a nationwide South Korean database, report that PPIs were not associated with a higher risk of dementia when compared with H2RAs. Comparing PPIs to H2RAs is one approach to reduce confounding by indication.

Myocardial Infarction and Stroke

The evidence linking PPIs to increased risk of major adverse cardiovascular events is weak. In a meta-analysis that included 16 studies (8 randomized controlled trials, 7 observational studies, and 1 retrospective analysis of a randomized controlled trial) with a total of 447 408 participants, the observational studies, but not the randomized controlled trials, showed a slight increased risk of adverse cardiovascular events with PPIs.[42]

Using an administrative claims database from two insurance cohorts constituting over five million patients from 2001 to 2014, a United States study by Landi et al.[43] found no evidence that PPIs increase the risk of myocardial infection (MI) compared with H2RAs. Another United States study of 68 514 women (mean age, 65 ± 7 years) enrolled in the Nurses' Health Study and 28 989 men (mean age, 69 ± 8 years) enrolled in the Health Professionals Follow-up Study reported no increased risk of stroke in PPI users;[44] individuals were followed biennially with detailed questionnaires and follow-up rates exceeded 90%.

Not only should physicians and patients not be concerned about starting a PPI for an indicated condition for fear of MI or stroke but also there are studies in laboratory animals that indicate that PPIs might even be beneficial for these conditions. In experimental rats, in which MIs were induced by isoproterenol, PPIs minimized myocardial necrosis by reducing reactive oxygen species (ROS), indicated by an increase in endogenous antioxidants and a decrease in nitric oxide and malondialdehydre.[45]

PPI-induced hypochlorhydria elicits hypergastrinemia. In rats, administration of gastrin improves myocardial function and reduces infarct size and cardiomyocyte apoptosis induced by ischemia/reperfusion injury.[46] The protective effect was mediated via CCK2 receptors. Thus, experimental laboratory data suggests that PPIs and gastrin may exert protective effects in cardiomyocytes.

Bone Fracture

There is a debatable, but unproven, relationship, between PPIs and fracture risk.[47] Some studies indicate a small increased risk for fracture with PPIs whereas others indicate no increased risk or even a protective effect. The biological plausibility for increased risk is based upon the ability of PPIs to increase parathyroid hormone (PTH) secretion in laboratory animals. PTH influences bone remodeling and may weaken its microstructure and strength. The increase in PTH has been hypothesized to be secondary to hypochlorhydria-induced reduction in calcium absorption and/or hypergastrinemia. It should be noted that data supporting this hypothesis is limited and it has not been convincingly demonstrated that PPIs increase PTH secretion in humans.

A meta-analysis of 24 observational studies concluded that PPI use was significantly associated with a small increased risk of hip fracture (relative risk = 1.20; CI = 1.14–1.28).[48] Nine of the studies, however, showed no significant risk. In a more recent population-based propensity-matched retrospective cohort study, using the National Health Insurance Research Database in Taiwan, PPI use after stroke was associated with a small but significant increased risk of hip ( hazard ratio = 1.18; CI = 1.00–1.38) and vertebral fracture ( hazard ratio = 1.33; CI = 1.14–1.54).[49]

In contrast, a Finnish nested case–control study using a nationwide database of elderly patients with Alzheimer's disease, half of which used PPIs, found no association between PPIs used for longer than a year and hip fracture.[50] Similar results were reported in three other studies;[51–53] interestingly, the relative risk of fracture was lower with PPI use, in both young and old adults, suggesting that PPIs could potentially exert a protective effect.[51,53] In a prospective, multicenter, double-blind study of 115 healthy postmenopausal women randomly assigned to dexlansoprazole (60 mg), esomeprazole (40 mg), or placebo daily for 26 weeks, there were no significant changes in markers of bone turnover (bone mineral density, true fractional calcium absorption, serum and urine levels of minerals, plasma levels of procollagen type 1 N-terminal propeptide, C-terminal telopeptide of type 1 collagen, and plasma levels of PTH) within the groups between baseline and week 26.[54] The PPI groups, however, showed an increase in markers of bone turnover; bone formation and resorption both increased but remained within normal range and were appropriately coupled.

In summary, there is no convincing evidence that PPIs induce osteoporosis or bone fracture. The reason some studies report increased bone fractures in PPI users may be related to the underlying comorbidities including advanced age, stroke, and Alzheimer's disease that increase the risk for falls.


Obesity and PPIs have been linked to changes in the gut microbiota. A cohort study of US Department of Defense TRICARE beneficiaries born between 2006 and 2013 (333 353 children with 241 502 prescribed an antibiotic, 39 488 a H2RA, and 11 089 a PPI) found that antibiotics, acid suppressants, and the combination, given within the first 2 years of life, were associated with a diagnosis of childhood obesity.[55] The individual risks associated with H2RAs and PPIs were extremely small (H2RA: OR = 1.03, CI = 1.02–1.03; PPI: OR = 1.02, CI = 1.01–1.03).