PPI Deprescribing and the Evidence on Long-Term Harms

Start with the reason to stop, not the reason to fear#

The right question about a proton pump inhibitor is not "could this drug be hurting me?" but "do I still have a reason to take it?" That single reframe organizes the entire debate. The American Gastroenterological Association's 2022 clinical practice update on de-prescribing proton pump inhibitors is built on it: a PPI should be reduced or stopped when a valid indication no longer applies, not because an observational study tied it to a frightening outcome. A reported harm, by itself, is not grounds to withdraw a medicine that a person genuinely needs.

That may sound obvious, yet it runs against how most patients hear the news. Years of headlines have linked long-term PPI use to dementia, kidney disease, fractures, and more, and the natural response is to stop the pill. The update asks clinicians and patients to slow down and separate two different things: the strength of the reason for taking the drug, and the strength of the evidence behind the fear.

Key points#

• Deprescribe a PPI when the indication is gone, not because a study reported a possible harm.
• Most publicized long-term risks come from observational data that can show a link but cannot prove cause.
• A large randomized trial (COMPASS) found no convincing increase in those feared outcomes over about three years.
• Some patients should generally stay on a PPI, including those with Barrett's esophagus, severe erosive esophagitis, or high bleeding risk.
• Stopping can trigger temporary rebound acid symptoms; warn people in advance so they do not misread it as proof the drug was essential.

What the update actually recommends#

The document offers ten best-practice statements for the outpatient setting, and none of them is exotic. The premise is that no one should take a medication without a reasonable expectation of benefit. In practice that means every person on a PPI should have the indication reviewed and written down. Anyone without a clear indication should be offered a trial off the drug. Most people taking it twice daily can drop to once daily.

The statement that carries the real weight is the one about what does not justify stopping. Having a so-called PPI-associated adverse event in the past, or simply having risk factors for one, is named as an inappropriate reason on its own to discontinue the drug. The trigger for deprescribing is the missing indication, and nothing else.

Why "linked to" is not "caused by"#

Long-term PPI use has been associated in the literature with pneumonia, Clostridioides difficile infection, hip fracture, chronic kidney disease, dementia, low magnesium, low vitamin B12, and gastric cancer. Almost every one of those signals comes from observational cohorts, and observational data carry predictable traps.

Consider who tends to stay on a PPI for a decade. They are, on average, older, managing more chronic conditions, and taking more medications than people who never start one. So if that group also has more dementia or kidney disease, the drug is an easy suspect even when the underlying frailty is the real driver. This is confounding by indication. Related problems compound it: protopathic bias, where a drug is prescribed for the earliest symptom of a disease that has not yet been diagnosed, and residual confounding, where the measured variables never fully capture how different the two groups were to begin with. Each can manufacture a link that is not a cause.

The clean way to break the tie is randomization, and one trial does exactly that. Within the COMPASS trial, 17,598 patients with vascular disease were randomly assigned to pantoprazole or placebo and followed for a median of roughly three years (Moayyedi and colleagues, Gastroenterology 2019). Across the whole roster of feared outcomes, pantoprazole showed no convincing increase in risk, with the possible exception of enteric infections. When the groups are balanced by chance, most of the scary associations dissolve, which is precisely what you would expect if confounding, not the drug, produced them.

Weighing a small maybe-harm against a real benefit#

The reasoning becomes concrete when both sides are on the table. Imagine a drug that might slightly raise the risk of some outcome, while it clearly prevents gastrointestinal bleeding in a person taking aspirin plus an anticoagulant. Stopping it to chase away a small and unproven harm can surrender a real benefit for a hypothetical one, and leave the patient worse off.

The update honors that asymmetry by naming groups who generally should not be deprescribed: people with complicated reflux disease (severe erosive esophagitis, an esophageal ulcer, or a peptic stricture), Barrett's esophagus, eosinophilic esophagitis, or idiopathic pulmonary fibrosis, and anyone at high risk of upper gastrointestinal bleeding. For them the indication settles the question, and the adverse-event headlines are noise.

The rebound trap#

Stopping a long-term PPI has a pharmacologic aftereffect that is easy to misinterpret. Suppressing acid raises the hormone gastrin, and when the drug is removed the stomach briefly oversecretes acid. In a double-blind trial, healthy volunteers who had no reflux at baseline developed heartburn, acid regurgitation, or dyspepsia after PPI withdrawal at a rate near 44 percent, roughly three times the 15 percent in the placebo group (Reimer and colleagues, Gastroenterology 2009).

A patient who feels that surge understandably concludes the drug was necessary all along. Usually it was not. The remedy is to set expectations before stopping: transient symptoms are common and they fade as acid output resets over a few weeks. Either a taper or an abrupt stop is acceptable, and a lower-potency bridge such as an H2 blocker or an antacid can smooth the transition.

A sensible step-down, in order#

1. Confirm whether a genuine indication still exists.
2. If dosing is twice daily, cut to once daily.
3. If no indication remains, trial the patient off the drug or onto on-demand use.
4. Warn about rebound and offer a lower-potency bridge if needed.
5. Reassess after a few weeks, not at the first uncomfortable day.

The aim is fewer pills where they add nothing, not zero pills as a slogan.