Are long-term acid-reflux drugs as risky as observational studies imply?

The long-term safety of proton pump inhibitors — the acid-suppressing drugs millions take for heartburn and reflux — looks considerably better in randomised trial data than a decade of worrying headlines implies. When 17,598 people were randomly assigned to take pantoprazole or a placebo for three years and monitored for a long list of feared harms, only one, a modest increase in gut infections, held up [s1].

That matters because almost everything the public has heard about PPI danger rests on a weaker kind of evidence. Observational studies — which follow people who happen to take a drug and compare them with people who do not — have linked long-term PPI use to pneumonia, bone fractures, chronic kidney disease, dementia, heart disease and early death [s1]. Those associations drove alarming coverage and prompted many patients to abandon drugs they needed. But observational data cannot cleanly separate the drug from the reasons people take it: someone on years of acid suppression is, on average, older, heavier and sicker than someone who is not, and those differences alone can manufacture a false signal.

What the trial did

The safety analysis was built into COMPASS, a large cardiovascular trial, and its design is what gives it weight: a double-blind, placebo-controlled randomisation, the same method used to license drugs [s1]. Participants with stable cardiovascular and peripheral artery disease were assigned to pantoprazole 40 mg daily (8,791 people) or matching placebo (8,807 people), then followed for a median of 3.01 years — 53,152 patient-years in total [s1]. Every six months, investigators collected data on pneumonia, Clostridioides difficile and other enteric infections, fractures, gastric atrophy, chronic kidney disease, diabetes, chronic lung disease, dementia, cardiovascular disease, cancer, hospitalisations and death from any cause [s1].

What it found

Across that entire list, there was no statistically significant difference between the pantoprazole and placebo groups — with a single exception [s1]. Enteric infections were more common on pantoprazole, 1.4% versus 1.0% on placebo (odds ratio 1.33; 95% confidence interval 1.01 to 1.75) [s1]. That is a real but small effect, and a biologically plausible one: stomach acid is a barrier to swallowed microbes, so blunting it can let a few more through. C. difficile infection was roughly twice as common in the pantoprazole group, but there were only 13 cases in total, too few to reach statistical significance [s1]. The outcomes that generated the headlines — kidney disease, dementia, fractures, pneumonia, cancer — did not separate from placebo [s1].

What the trial cannot say

This is strong evidence, not a blanket clearance. Three years is not a lifetime, and some proposed harms are argued to take longer to appear [s1]. The participants all had established vascular disease, so they are not a perfect stand-in for a young person taking a PPI for ordinary reflux [s1]. And the C. difficile count, while not statistically significant, points the same direction as the enteric-infection finding and is worth watching rather than dismissing [s1]. What the trial does do is set a ceiling on how large the other risks can plausibly be: had PPIs truly caused dementia or kidney failure at the rates some observational papers suggested, a trial this size and this long would very likely have detected it.

Where that leaves prescribing

Professional guidance has moved to match the evidence. The American Gastroenterological Association's expert review concluded that PPIs are appropriate long-term for genuine indications — erosive oesophagitis, Barrett's oesophagus, and ulcer prevention in high-risk users of anti-inflammatory drugs — but that the dose should be periodically reassessed so the lowest effective one is used, and that people with uncomplicated reflux who respond should attempt to stop or step down [s2]. It also advised, pointedly, that long-term PPI users should not routinely take probiotics to prevent infection, nor load up on calcium, vitamin B12 or magnesium beyond the recommended dietary allowance to counter supposed deficiencies [s2] — the very countermeasures the risk scare had encouraged.

The honest summary is that PPIs are neither harmless nor as dangerous as the observational literature made them look. Reserving them for people who need them, at the lowest dose that works, is sensible for reasons that predate the safety panic. But for someone with a real indication, the randomised evidence is reassuring — a pattern seen elsewhere when trials finally test what observational data only hinted at, as with the many side effects listed on statin labels that blinded trials do not support. Readers weighing whether they still need the drug are better served by what actually drives reflux than by a list of dreaded but unproven harms.