Proton Pump Inhibitors: A Long List of Observational Alarms the Most Rigorous Data Keeps Failing to Confirm
PRODOTTO DA
That result lands in a body of literature that splits cleanly along one axis: how well a given study accounts for confounding by indication and reverse causation, patients starting a PPI because of the early, undiagnosed symptoms of the very disease later blamed on the drug. The 2022 ACG clinical guideline on GERD reviewed this literature directly and did not hedge: high-quality studies, it concluded, have found that PPIs do not significantly increase the risk of any of the commonly feared conditions except intestinal infections. Am J GastroenterolThe same guideline is unusually direct about the one interaction most likely to actually change a prescribing decision. For patients on clopidogrel who need a PPI, it concludes that the established benefits of treatment outweigh the drug’s “proposed but highly questionable” cardiovascular risks. Am J Gastroenterol That skepticism has an evidentiary basis, though an imperfect one: the only randomized trial to test the clopidogrel-PPI interaction directly, COGENT, was stopped early when its sponsor lost funding and remains, roughly 15 years later, the sole dedicated RCT on the question. It found no difference in cardiovascular events between omeprazole and placebo (4.9% vs. 5.7%) alongside a highly significant reduction in GI bleeding. N Engl J Med COGENT’s age and its early termination are real limits on how much weight it alone can bear. A second single large randomized trial, COMPASS, tested pantoprazole specifically, not the class as a whole, against placebo in 17,598 patients with stable cardiovascular disease (8,791 pantoprazole, 8,807 placebo) followed for a median of about 3 years, and found the same null pattern for cardiovascular events in the pantoprazole arm: a hazard ratio of 1.04 (95% CI, 0.93–1.15). Moayyedi Extending a single agent’s trial result to the other nine branded products is an extrapolation, one ACG’s guideline makes explicitly and this article follows, resting on the assumption of a shared class effect rather than on independent randomized testing of each agent. Am J Gastroenterol
“Because this study was very rigorous in its methodology and had a huge number of patients, the lack of association is very reassuring,” Daniela Jodorkovsky, MD, who was not involved in the Nordic study, told Medscape Medical News. Medscape The reassurance does not extend evenly across the whole list. Fracture, C. difficile infection, and hypomagnesemia remain plausible, biologically grounded risks that have never had a randomized trial to test them one way or the other, and gastric cancer itself still carries a real, repeatedly observed positive signal in the broader, less-adjusted literature that a single case-control study, however large, does not erase.
The clinical task this year is not to declare PPIs safe or dangerous, and it is not to treat one rigorous 2026 study as having overturned a field of contradictory findings. It is to sort the roster of branded products by which harms track adjustment quality, which have genuinely been tested against confounding and reverse causation and not confirmed, and which patients still need this class enough that the distinction rarely changes the prescription.
Earlier or Less-Adjusted Signal vs. Confounding-Adjusted Verdict
| Earlier or Less-Adjusted Signal | What Confounding-Adjusted or Randomized Data Show | Where the Evidence Stands |
|---|---|---|
| Gastric adenocarcinoma: meta-analyses spanning three decades report pooled relative risks of roughly 1.3 to 2.9, including ~1.8-2.0 in H. pylori-eradicated cohorts and high-incidence East Asian populations Front Pharmacol +2 | A meta-analysis using H2-blockers (not nonusers) as the comparator, to strip out confounding by indication, found no association in the subset of studies with adequate adjustment. Aliment Pharmacol Ther Two 2026 studies modeling exposure timing directly reached the same conclusion: a five-country Nordic case-control study (17,232 cases) found an adjusted OR of 1.01 once long-term use was isolated from recent, symptom-driven use BMJ; an Israeli case-control study found the same signal appeared only in the year before diagnosis and reversed toward protective with adjustment PLoS Med | Not supported after rigorous confounding adjustment, though a real signal persists in less-adjusted studies; no study has specifically examined risk in younger patients BMJ Open Gastroenterol |
| Reduced clopidogrel antiplatelet activity and excess cardiovascular events with concomitant PPI use, based on ex vivo platelet studies and claims-data analyses N Engl J Med | COGENT, the only dedicated RCT (stopped early at 3,761 of a planned 5,000 patients when its sponsor lost funding): cardiovascular event rate 4.9% with omeprazole vs. 5.7% with placebo, no significant difference. N Engl J Med A single large randomized trial, COMPASS, tested pantoprazole specifically (not the class) against placebo and found a similarly null hazard ratio for cardiovascular events in pantoprazole-treated patients, 1.04 (95% CI, 0.93–1.15); ACG’s guideline extrapolates this pantoprazole-specific finding to the class as a whole Am J Gastroenterol | Established benefit is judged to outweigh a risk ACG’s own guideline calls “highly questionable,” though COGENT’s early termination and COMPASS’s single-agent design limit how definitive that class-wide reassurance can be Am J Gastroenterol |
| Incident chronic kidney disease and acute kidney injury with PPI exposure, replicated across two prospective cohorts JAMA Intern Med | The same COMPASS trial, again testing pantoprazole specifically, found no significant association, OR 1.17 (95% CI, 0.94–1.45); ACG’s guideline extrapolates this to the class Am J Gastroenterol | Observational association replicated across cohorts; not confirmed in the pantoprazole trial, whose finding is extrapolated to other agents |
| Hip, wrist, and spine fracture with long-term, high-dose use; a 2025 systematic review found the literature limited by confounding but still urges cautious, judicious prescribing Cureus | The same pantoprazole-specific COMPASS trial found no significant association, OR 0.96 (95% CI, 0.79–1.17), again extrapolated by the guideline to the class Am J Gastroenterol | Weak, inconsistent observational signal; not confirmed in the pantoprazole trial, whose finding is extrapolated to other agents |
| Incident dementia with regular PPI use, from claims-data cohorts in which PPI initiators had markedly higher baseline rates of depression, stroke, and polypharmacy Am J Gastroenterol | The same pantoprazole-specific COMPASS trial found no significant association, OR 1.20 (95% CI, 0.81–1.78), again extrapolated to the class Am J Gastroenterol | Not confirmed in the pantoprazole trial; confounding by indication likely in the observational data |
