The abrupt termination of Pfizer’s ILLUMINATE study with its cholesterol management drug torcetrapib last December illustrates not only the risks and dilemmas attendant on pharmaceutical development but also the “narrow line” walked by data safety monitoring boards, says an editorial in the latest edition of the British Medical Journal (BMJ 2007; 334:326).

The US company scrapped all development of torcetrapib after the independent DSMB for ILLUMINATE recommended stopping the Phase III trial “because of an imbalance of mortality and cardiovascular events.” There were 82 deaths in patients taking torcetrapib and Pfizer’s mainstay cholesterol-reducer atorvastatin (Lipitor), compared with 51 deaths in patients on atorvastatin alone.

It would be “simplistic,” however, to conclude that the ILLUMINATE trial caused 31 unnecessary deaths, write Gorm Boje Jensen of the Department of Cardiology at Copenhagen University Hospital in Denmark and John Hampton, emeritus professor of cardiology at Queen’s Medical Centre in Nottingham, UK. Even though the observed difference in mortality between the treatment groups was statistically significant, this difference could still have been a chance finding.

As clinical trial results accumulate over time, outcomes often vary within treatment groups, Jensen and Hampton point out in the BMJ. The challenge for a DSMB is “to judge whether such differences are statistically and clinically convincing. Only when a sufficient number of deaths have occurred can there be any confidence in the validity of the observation.”

The ILLUMINATE safety monitoring board was right to allow the trial to continue “until harm had been demonstrated with a reasonable degree of confidence,” the authors argue. There was no indication or hypothesis that torcetrapib’s mode of action – inhibiting cholesteryl ester transfer protein – led to serious adverse effects and the “potential benefit of the new treatment to vast numbers of patients cannot be underestimated,” they comment.

Walking a tightrope

DSMBs are essentially walking a tightrope, the editorial says. Clinical trial participants should not be exposed to undue risks, yet many more patients may later be denied a potentially useful drug if the study is halted without compelling evidence of harm or benefit.

Should, then, these boards have to carry the responsibility of observing excess deaths while allowing treatment to continue, Jensen and Hampton ask? In fact, there is no alternative. If the hypothesis for a study is convincing, it can only be tested through a large Phase III trial, they note. And the role of DSMBs in these trials is “mandatory according to binding international guidelines.”

The authors do suggest ways in which the vital role of DSMBs can be optimised. The boards should consist of clinicians and statisticians who “thoroughly understand the clinical area of the trial, who are experienced in the vagaries of trials, and who have no financial or other competing interest in the outcome of the trial,” they say. DSMBs should also be small (five members at most), to allow rapid communication between members. Speedy production and transmission of data from the trial organisation to the board are critical too, so that timely decisions can be made.

Sympathy for Pfizer

“These boards carry heavy responsibilities, and the scientific merits of being a member of one should be recognised as equivalent to co-authorship,” Jensen and Hampton suggest. They also reserve some sympathy for Pfizer, saying the company’s decision to withdraw torcetrapib “appears entirely proper” but must have been difficult. “It must be hoped that the drug industry does not as a result of this and similar events lose the will to engage in the development of innovative drugs,” they comment.