On Friday the NHLBI declared victory in the SPRINT trial. As was widely reported here and elsewhere, the NHLBI announced that the trial had been stopped early after significant reductions in cardiovascular events and mortality had occurred in the group of hypertensive patients randomized to a more aggressive blood pressure target than is currently recommended today.
But there was a big problem with the announcement. This sprint to the victory podium may have been premature. The judges– peer reviewers, editors, guideline authors, and other outside experts– haven’t had a chance to scrutinize the race. It’s quite possible that SPRINT will eventually be viewed as a major, game-changing trial. But there’s no way to know for sure until we see the full results.
The only thing the NHLBI told us on Friday was that cardiovascular events were reduced by “almost a third” and mortality by “almost a quarter.” During the news conference I asked the NHLBI twice if they could provide any additional data. In particular, I thought that it was completely inappropriate to release the relative risk reductions without giving some idea of the absolute reduction in risk. It is simply impossible to assess the clinical significance of the trial finding without understanding this part of the story. (And let’s be honest: reporting relative risk reductions and not absolute risk reductions is one of the most common ways to hype a result.)
And there’s much more we still need to know before we can declare with confidence that the study delivers “potentially lifesaving information” (as reported by Gina Kolata in the New York Times). For now we don’t know how robust the results really are, since the NHLBI told us absolutely nothing about the underlying statistics (no p values or confidence intervals).
We don’t know anything about the safety of the more aggressive regimen. This is a big concern because in recent years there’s been increasing awareness that many elderly people are suffering catastrophic falls when their blood pressure falls too low or too fast.
We don’t know whether the lower target was effective in different subgroups. We do know that the SPRINT trial enrolled a high risk group of patients, but we don’t know yet, and it will take a lot of analysis to figure out, whether the results apply to a more general population, especially since there have been earlier trials in different populations that did not achieve the same result as SPRINT. We also don’t know anything about the actual blood pressure reductions in the two treatment groups and we don’t know how these results should be translated for patients who have 24-hour blood pressure monitoring.
The NHLBI also said nothing about the specific hazards that can occur when a trial is stopped early, specifically the increased likelihood of finding an exaggerated effect. It is well understood by statisticians and clinical trial experts that a temporary deviation from the mean can provoke a data and safety monitoring committee to stop a trial. (It’s also worth mentioning that although the preliminary results have been announced the investigators are continuing to follow patients in the study and the trial has not yet closed out. The preliminary numbers could well change by the time of publication.)
I think the NHLBI officials and investigators themselves gave the best reason why they should not have released the relative risk reductions: during the press conference they repeatedly stated that any concrete action based on the trial’s findings needed to await peer-reviewed publication. I think it’s irresponsible for the NHLBI to say they have “potentially lifesaving information” but then go on to say that no immediate changes are warranted until more complete results are available. No one want to wait for “lifesaving” information.
It seems reasonable and clear that if no immediate action is required, and if only partial and preliminary information is available, the best course is simply to release the minimum amount of information. In this case it would have been sufficient for the NHLBI to state that the trial had been halted early after a recommendation by the DSMB based on a positive result in the primary outcome of the study. They could also reasonably state that no safety problems that emerged so far.
But if they do think it’s necessary to release information like the reduction in relative risk, then this must be accompanied by some indication of the absolute reduction as well. We’ve been down this road too many times in the past to accept this kind of thing. The NHLBI should hold itself to a higher standard.
There is one point of comfort. We don’t need to worry too much at all about undue influence from the pharmaceutical industry. Nearly all the most common blood pressure drugs are now off patent. But we do need to remember that blood pressure guidelines have been the subject of enormous controversy and contention in the last few years, and many of the SPRINT investigators have played an important role in these controversies. And we do need to be concerned about the enormous tendency to turn otherwise healthy people into patients.