No explanation given

The Pesario Cervical para Evitar Prematuridad (PECEP) trial

On 12 May The Lancet published an apparently ground breaking trial (click here).  385 pregnant women with a short cervix at 20-23 weeks were randomly allocated to use a vaginal pessary which held the cervix closed (n=192), or to a “no treatment” observation group (n=193).

Spontaneous delivery before 34 weeks occurred in 12 (6%) of the pessary group v. 51 (27%) in the expectant management group, odds ratio 0·18, 95% CI 0·08—0·37. This is a huge treatment effect and highly statistically significant. If true it would be important since pre-term birth remains the single biggest cause of newborn death in the developed world. A breathless editorial gave it a big puff (click here).

The trial was registered here.  The primary endpoint was the same as the published paper but the planned sample size was 2780, i.e.1390 per group! That’s about seven times more than they actually recruited. There’s no explanation for the discrepancy in the paper, in which the predetermined sample size is stated to be 380. I asked the authors for a copy of the protocol which, although they promised to send it, never appeared.  But my letter to the Lancet (here), generated this response (here). They had indeed planned to recruit 2780 but “after the first interim analysis, and in accordance with the advice of our Data Monitoring Committee, the protocol was amended to reach the new and final sample size”.

Does that mean the Data Monitoring Committee did a planned interim analysis and instructed them to stop because the treatment was so clearly effective that it would have been unethical to continue?  That would have been scientifically correct, but was not the case. In the paper they say “Interim analysis was done every 6 months. However, no conditions for stopping the trial were noted by the external data monitoring group.”

It’s a puzzle. Were they peeking at the data, got a significant result, and stopped? I hope not.

The trial was funded by The Instituto Carlos III, Spain’s main public research organisation, i.e. the taxpayer.

I wonder if The Lancet would publish a trial by big pharma where the planned sample size was reduced 7-fold without explanation.

Jim Thornton