A social anxiety pill did no better than placebo in a trial of 546 adults

SOCIAL ANXIETY DISORDER is a lasting fear of being watched and judged by other people, strong enough that giving a talk, meeting strangers or speaking up at work becomes distressing. Drugs that calm anxiety quickly tend to make people drowsy, which helps little when the moment calls for a clear head. Vanda Pharmaceuticals hoped to fill that gap with VQW-765, a pill meant to be taken shortly before a stressful event. It acts on the alpha-7 nicotinic receptor, one of the brain receptors that nicotine also binds, and it switches that receptor partly on. On October 14th the company reported that a phase 3 trial of the drug, the large final stage of testing before a drug can be submitted for approval, had failed.

The trial, called Study 3201, enrolled 546 adults aged 18 to 65 at about 30 sites in America. All had a current diagnosis of social anxiety disorder and high scores on two standard questionnaires about social and public-speaking fear. Each took a single capsule and then sat the Trier Social Stress Test, a laboratory exercise in which a person prepares and then gives a job-interview-style talk to a panel that offers no feedback or encouragement. During the talk, participants rated their distress on a scale from 0 to 100, where higher scores mean more distress. Vanda gave a single dose and has not disclosed what that dose was.

Half the participants received a placebo, a capsule identical to the drug but containing no active ingredient. Who got which was decided by chance, in a one-to-one ratio, so that the two groups would be alike in every way except the drug. The trial was also double-blind, meaning that neither the participants nor the staff running the test knew who had taken what. These steps guard against hope and expectation shaping the scores, which matters when people rate their own feelings.

Before the trial began, Vanda fixed its primary endpoint, the single result chosen in advance to decide whether the trial succeeded. According to the release, that endpoint was distress during the talk in the women in the trial. Women on the drug scored 65.6 out of 100, against 66.0 for women on placebo, a gap small enough to arise by chance. A trial that misses its primary endpoint cannot claim that the drug works on that measure. The key secondary result, for men and women together, was no better, at 64.7 on the drug against 63.8 on placebo. Vanda said the drug was safe and well tolerated.

The focus on women came from an earlier phase 2 trial of 230 adults with fear of public speaking, published in the British Journal of Psychiatry in 2025. In that study, a single 10mg dose did not significantly lower distress across everyone who took it. When the researchers looked at women alone, however, the drug appeared to help, and a further analysis, run after the results were in, suggested a benefit only in women whose blood levels of the drug fell in a middle range. A slice of a trial picked out for separate analysis is called a subgroup. Subgroup findings are weak evidence because the more slices a team examines, the more likely one of them shows an effect by chance alone. The usual remedy is the one Vanda chose, which is to name the subgroup in advance and test it again in a larger trial, and in this case the signal did not return.

The size of the trial also says how big an effect it can rule out. The confidence interval is the range of true effects consistent with the data, usually set so that it captures the true value 95 times in 100. For women it ran from 5.3 points in the drug’s favor to 4.6 points against it, on a 100-point scale. A large calming effect in women at this dose is therefore unlikely, though a small one cannot be excluded.

The trial tested one dose, taken once, in a staged laboratory talk, so it says nothing about other doses, daily use or anxiety in ordinary life. Vistagen, another company, has had two phase 3 failures of its own on-demand social anxiety treatment, according to Fierce Biotech. Novartis first developed the drug for schizophrenia, and Vanda licensed it in 2014. Vanda has put the program on hold while it analyzes the full dataset, and the public trial record has no posted results.