In U.S. Drug Trials Only, the Placebo Started Winning

What We Know

Jeffrey Mogil studies pain at McGill University, in Montreal. In 2015, he and his graduate student, Alexander Tuttle, pulled together eighty-four published trials of neuropathic-pain drugs run between 1990 and 2013 and checked something none of the original trials had been designed to track: how well the placebo pills used as controls were doing their job, year by year. They were doing it better and better. By 2013, placebo-arm patients were reporting an average 30 percent drop in pain, closing in on what the actual drugs were delivering. Then came the part that turned a curious number into a genuine puzzle. Run the identical analysis on trials conducted anywhere else, across Europe and Asia, and the trend vanishes. The placebo had improved. Nowhere else had.

The scale of what else changed over the same twenty-three years is its own kind of evidence. The American trials Mogil and Tuttle studied grew from an average of four weeks to twelve, and from fewer than fifty patients to more than seven hundred. Drug response barely moved during that stretch; it was the placebo arm doing almost all the climbing, and doing it in near lockstep with how long and how large the trials themselves had become.

Robert Dworkin, a pain-trial methodologist at the University of Rochester with no hand in the study, called it an important study, one that could improve how future trials are designed and speed up the search for better analgesics. That leaves the harder question exactly where it was: what, specifically, about running a trial in America teaches a patient’s body to expect relief before the drug ever arrives. Tuttle put it more plainly to reporters at the time: “It remains to be determined why the United States is an outlier.”

The Roadblock

Part of the difficulty is definitional. “Placebo response,” the number a trial actually measures, is not the same claim as “placebo effect,” a real psychobiological phenomenon. A placebo arm’s improvement folds together genuine expectation-driven physiology, patients simply feeling better on their own over time, statistical regression toward the mean after enrolling at their worst, and plain measurement noise, and no standard trial design cleanly separates which of those is doing the work. Two researchers can stare at the same rising line and mean different things by it.

It also isn’t a problem a better dataset can fix after the fact. Mogil and Tuttle’s finding covers trials that already happened, run by companies with no reason to have tracked patient expectations, advertising exposure, or site-level enrollment practices at the time. Data nobody thought to collect in 1990 can’t be recovered in 2026. And because trial data is commercially sensitive, much of what would help, unpublished results, patient-level surveys, site-by-site advertising records, sits behind doors that outside researchers, Mogil included, still can’t open.

Best Guesses

Advertising-primed expectation

Mogil and Tuttle’s own prime suspect is the most distinctly American thing about American trials: direct-to-consumer drug advertising, legal in only two industrialized countries, the US among them. The mechanism has real experimental backing: people shown prescription-drug ads measure a genuine shift in how effective they expect a drug to be, the same machinery a placebo response runs on. What nobody has done is trace it end to end: follow a trial participant from the ad they saw to the pill they swallowed to the relief they reported, and show the line holds.

Trial growth inflates placebo

Trial design and advertising exposure grew up together, in the same industry, over the same twenty-three years — which makes them hard to pull apart, and nobody has found a clean way to do it. But the two curves overlap almost too well to ignore. By Mogil and Tuttle’s own numbers, the American trials changed shape at nearly the rate the placebo response did: four weeks stretched to twelve, fifty patients grew to seven hundred. Regress the placebo curve against trial length and size alone, and it tracks closely, a mechanical explanation that doesn’t require inventing anything about American psychology.

Selective-publication distortion

A trial that fails to beat its placebo arm is one a drug company would rather the public never see, and for years, plenty of them found a way to stay unseen. When researchers compared the published record of antidepressant trials against the FDA’s own internal data on those same studies, 94 percent of published trials read as positive; the FDA’s own accounting put the true figure at 51 percent. None of that proves the placebo response itself is climbing — only that what reaches print has already been filtered by which numbers looked good.

Statistical artifact

Not every researcher looking at this data agrees there is a real trend to explain. In the closely related literature on antidepressant trials, one team reanalyzed the exact datasets that “placebo response is rising” claims rest on — the believers’ own numbers, not outside skepticism — and found the increase disappears once trial-design changes are properly accounted for. Nobody has yet run that rebuttal against Mogil and Tuttle’s own pain data, only against its closest sibling. Their conclusion, stated bluntly: what looks like a strengthening placebo is substantially a confound of design, not a shift in how the body answers a sugar pill.

The Verdict

A trial exists that would settle this, unglamorous and expensive as it is: randomize advertising exposure against none, alongside drug against placebo, with patient-level surveys tracking expectation before a single pill leaves the bottle. Nobody has funded it. It wouldn’t sell a drug — it would only explain why drugs keep failing to beat placebos that keep getting better at impersonating them.

For now, the argument runs on secondhand evidence: registries built for other purposes, reanalyses of reanalyses, a geographic anomaly nobody disputes and nobody has isolated a single cause for. Mogil is still at McGill, studying an effect his own data says doesn’t show up in the European or Asian trials he checked, without an answer for why.

The placebo keeps winning. Only here.