A 2x mortality risk sounds terrifying. That number, from a Cedars-Sinai study published April 16 in Communications Medicine (2026, n > 650,000), is now ricocheting across health media with predictable results: patients emailing their gastroenterologists in a panic, wellness influencers declaring loperamide a death sentence, and calls to overhaul prescribing guidelines. My position is simple. This study does not justify changing how doctors treat IBS. Not yet.
Confounding by Indication Is Not a Footnote
The study's authors, to their credit, acknowledge the limitation plainly: sicker patients receive stronger drugs. This is confounding by indication, and it is the oldest trap in observational pharmacoepidemiology. A patient taking daily loperamide or diphenoxylate for years likely has more severe, refractory IBS than someone managing symptoms with diet alone. That patient may also have comorbidities, higher healthcare utilization, worse baseline cardiovascular health, and greater exposure to polypharmacy. The drug didn't necessarily cause the outcome. The disease severity did.
No statistical adjustment fully eliminates this problem. You can control for age, sex, BMI, smoking status, and a dozen other variables extracted from electronic health records. You cannot control for the clinical judgment that led a physician to escalate therapy in the first place. That decision encodes information about disease severity that no database captures.
The 35% increased mortality associated with long-term antidepressants tells the same story. Patients prescribed tricyclics or SSRIs for IBS often have concurrent anxiety, depression, or chronic pain syndromes. Each of those conditions independently raises mortality risk. Attributing the excess deaths to the antidepressant rather than the psychiatric comorbidity requires evidence this study cannot provide.
Large Samples Don't Fix Broken Designs
I keep seeing the 650,000-patient figure cited as though sample size alone confers reliability. It doesn't. A large observational cohort amplifies signal and noise equally. With 650,000 records and 20 years of follow-up, you have enormous statistical power to detect associations. You have zero additional power to determine whether those associations are causal.
Ali Rezaie, the study's senior author, said something I agree with completely: most IBS drug trials last less than 1 year, and we lack long-term safety data. He's right. That gap is real. But the solution to inadequate trial data is better trials, not retroactive causal claims from billing codes and prescription records.
The finding that FDA-approved IBS-specific drugs like rifaximin showed no mortality signal is genuinely useful. It suggests these agents deserve more study as long-term options. But notice the asymmetry in how people are interpreting the results: the absence of a mortality signal for rifaximin is treated as evidence of safety, while the presence of a signal for loperamide is treated as evidence of danger. Both conclusions require the same causal inference the study design cannot support.
I'll grant the other side a fair point. When you see a consistent signal across 650,000 patients and 2 decades, ignoring it entirely would be reckless. Clinicians should be aware of this data. But awareness is not the same as action, and action based on confounded associations has a grim track record in medicine. Hormone replacement therapy was abandoned for millions of women based on observational data that later randomized trials partially contradicted. We should have learned something from that.
What should doctors do right now? Exactly what good doctors already do: periodically reassess whether a patient still needs a medication, discuss risks and benefits, and consider alternatives when appropriate. That is standard care. It does not require a headline.
What doctors should not do is pull patients off medications that control their symptoms because a single observational study generated a scary relative risk number. The absolute risk increase for any individual patient remains small. Uncontrolled IBS symptoms carry their own costs: malnutrition, dehydration, lost work, psychiatric deterioration.
The study asks an important question. It does not answer it. Until a prospective trial or at minimum a well-designed propensity-matched cohort replicates these findings, prescribing guidelines should stay where they are.
