Read the control column: a 288-patient Russian panic disorder trial
- Multicentre, randomised, double-blind, placebo-controlled trial with an open-label active comparator. 288 patients with moderate panic disorder received the domestic anxiolytic Aviandr (maritupirdine) 40 mg/day (n=144), placebo (n=72), or paroxetine 40 mg/day (n=72) for 12 weeks. The paper reports the first stage of the study. The developer is on the byline: ChemDiv Inc. and ChemRar Pharma appear among the affiliations.
- Primary endpoint, change in Panic Disorder Severity Scale total score at week 12: a fall of 4.99 points on Aviandr against 2.28 on placebo, p=0.047. That clears the conventional threshold by 0.003 on a single endpoint with no correction stated. The gap is 2.71 points, and the placebo arm accounts for a little under half the change measured in the drug arm.
- The abstract describes adverse-event frequency as comparable to placebo, giving 44.44% alongside the placebo group and 50% for paroxetine, with no serious adverse events recorded. The paroxetine arm was open-label, so that tolerability comparison is not protected by blinding.
- The abstract also reports superiority on SIGH-A and CGI, significant improvement from week 2, and a significantly lower need for hydroxyzine, so the primary endpoint is not the only result reported in the drug's favour. What the public abstract does not carry is any confidence interval, effect size, baseline PDSS score or separate adverse-event rate for the Aviandr arm, so the claim of clinical significance cannot be checked against those figures.
Russian psychiatry does run placebo-controlled trials, and this is one of them: 288 patients, three arms, twelve weeks, a domestic anxiolytic set against an inert tablet and an SSRI. The authors draw their conclusion about the drug. The more useful reading is the column beside it.
The comparator that was not blinded
The byline spans I.M. Sechenov First Moscow State Medical University, the Mental Health Research Centre, Samara State Medical University and clinics in St Petersburg, Rostov-on-Don and Stavropol, alongside ChemDiv Inc. and ChemRar Pharma. The developer side is on the paper. That is normal for a domestic development programme and it is worth naming rather than discovering later.
The allocation was 144, 72 and 72 – a two-to-one-to-one split that buys statistical power for the drug arm at the cost of a smaller control group. Blinding covered Aviandr and placebo. Paroxetine was given open-label, which the abstract states plainly. So the comparison everyone will quote, 44.44% adverse events against 50% on paroxetine, sits across a blinding boundary. Patients who knew they were taking a familiar SSRI had a public side-effect profile to measure themselves against; patients on the blinded tablets did not.
One earlier placebo-controlled study of this compound, and not the only one, is a 2021 phase II dose-finding pilot in generalised anxiety disorder in the Journal of Psychiatric Research: 129 patients at 17 Russian sites, 40 mg (n=41), 60 mg (n=43) and placebo (n=43) over eight weeks, with HAM-A decreases reported as 53.7%, 47.7% and 16.3%. That abstract does not label what kind of statistic those percentages are. Different disorder, different scale, different duration – not a like-for-like comparison with the panic trial. What carries across is narrower: the pilot was small, its placebo cohort moved little, and the larger panic trial cleared significance at p=0.047.
What nearly half the control group reported
Take the placebo column on its own terms. It improved by 2.28 PDSS points over twelve weeks without pharmacology, and the abstract puts adverse events at 44.44% beside it. The sentence is not perfectly worded, but the reading is clear enough: the parallel figure of 50% unambiguously labels paroxetine, and a rate described as comparable to placebo needs a placebo figure to be comparable to. That makes 44.44% of 72, which is 32 patients. Two decimal places on a group of 72 are a publishing convention, not a red flag. What the abstract never gives anywhere is the Aviandr arm's own adverse-event rate.
An adverse event in a trial is any untoward event during the observation window, drug-attributed or not. Headaches, viral illness and a bad fortnight all count. So 44.44% is not a clean nocebo estimate. It is something more useful: the base rate of new complaints in people who are being watched closely, asked regularly, and expecting a drug to do something.
That is the number to carry into clinic. When a patient calls in week two with dry mouth, headache and unease, the control column tells you roughly how often that happens without any active compound at all. It does not make the complaint imaginary and it does not settle attribution. It does mean two questions come before a dose change: what were they told to expect, and does the symptom track the dose.
An inert tablet produced a little under half the measured improvement, and the abstract puts adverse events at 44.44 percent in that same control column.
The full text is behind a paywall; everything here comes from the structured abstract, which Media Sphera publishes free in Russian and English. That abstract gives no confidence intervals, no baseline scores and no adverse-event rate stated separately for the Aviandr arm, and it does not say whether the Russian-language versions of PDSS, SIGH-A, CGI, the sleepiness visual analogue scale and SDS have published local psychometric validation. Neither a funding statement nor a competing-interests statement appears in the public abstract, so the commercial relationship is established from the author affiliations alone and its terms are unknown. Adverse events in a trial include any untoward event, not only drug-attributed ones.