Whenever I hear about publication by press conference I get worried. It’s a back door to get your results noticed. And when the actual results are not fully reported, it is potentially misleading.

So I am a bit annoyed with myself at my excitement on hearing that a Thai study had found a significant 31.2% efficacy for an HIV vaccine. True, it wasn’t fully reported, but there was a full and open discussion scheduled on the 20th October at a meeting in Paris.

Last week, before the home game against Olympiacos, I was having a beer with a fellow Arsenal fan who works in an HIV lab. He was complaining that the TV cameras had come to film the lab, as background for a news report on the discovery, and he had been edited out of the footage. “Anyway,” he said “it’s all bollocks. The numbers are really small. It’s seventy something infected versus fifty something.”

This was the cue for me to patronisingly point out that it’s the number of people vaccinated that matters, which was 8000 in both arms of the trial. My friend grumbled and moved onto an entertaining rant about HPV and perianal warts. But I’ve now got to hold my hands up and say it looks like he had a point. About the HIV vaccine trial that is.

According to reports of private briefings, if we exclude people who did not exactly follow the protocols from the trial, then the estimate of efficacy reduces, and statistical significance disappears. This is not altogether surprising, as the significance was borderline anyway and along with those trial participants we lose statistical power.  Given that significance is only arbitrarily defined at the 95% level, this is not necessarily as bad as it might sound, and the data remain interesting. But it does raise the question of why these more equivocal results were not reported. Indeed, given that the results are still not published, why was there any publicity at all? There is, apparently, a paper under review at the NEJM. I guess we just have to hope for rapid review.

Anyway, I bet the Paris meeting will be fun…