On March 31, 2008, pharmaceutical giant AstraZeneca trumpeted the early closing of its so-called JUIPITER trial of a cholesterol-lowering drug (statin), Crestor. The results after only two years yielded "unequivocal evidence" of the drug's effectiveness, the trial concluded, and the company argued that it could not be withheld from anyone who was well and had normal cholesterol levels but had an elevation in another normal blood constituent, the C-reactive protein (CRP).
I am the skeptical physician who is unwilling to let anyone test my cholesterol until I see unequivocal data that taking a statin yields meaningful benefit for me. Now AstraZeneca wants me to get my CRP measured so that I can swallow Crestor if it's elevated.
On Nov. 9, 2008, the results of JUPITER were published in the online edition of the New England Journal of Medicine.
I knew there was a devil in the details. Let me flush it out for you.
AstraZeneca invested a great deal in this Herculean drug trial. They contracted with physicians in over 1,300 centers in 26 countries to recruit subjects. Some 90,000 were screened and nearly 18,000 enrolled.
At each center, half the recruits were randomly assigned to swallow a placebo pill, the other half Crestor. The intent was to monitor this army of volunteers for five years to see if the groups differed in their incidence of any of the following: heart attack, stroke, hospitalization for unstable angina or for surgery on their coronary arteries and death from cardiovascular causes.
JUPITER, as is true for all modern trials, had an oversight committee charged with breaking the code periodically to see if the volunteers on Crestor were fairing better or worse than the volunteers on the placebo. The JUPITER oversight committee comprised luminaries in the world of cardiology who, like nearly all the principal JUPITER trial investigators, had declared financial involvements with the industry that serves the cardiovascular enterprise, many with AstraZeneca.
After 1.9 years, the oversight committee sounded the alarm when they noted a highly statistically significant 56 percent reduction in the incidence of these feared outcomes. The trial was terminated; AstraZeneca trumpeted the benefit of Crestor and stockholders took notice.
A reduction of 56 percent is hard to ignore -- at first blush. It conjures up an image of marshalling 100 soldiers armed with Crestor and 100 not so armed to assault the cardiovascular monster for two years, at the end of which 56 of the Crestor soldiers are the only ones left standing.
If that were true, I'd have my CRP tested today. But that's not even close to truth.
At the end of two years, about 2 percent of study participants suffered a cardiovascular event. On Crestor, 1.6 percent suffered one of the cardiovascular events, whereas it was 2.8 percent of those not afforded Crestor -- a difference of 1.2 percent.
However, not all these people were in the trial all of the first two years; they entered at different times reflecting the vagaries of recruitment. A more accurate reflection that takes this into account is to calculate for every 100 how many would have suffered one of the cardiovascular outcomes in a year in the trial. This event rate for any of the events (the "composite outcome") is 0.77 on Crestor and 1.36 without Crestor.
That's the 56 percent reduction that is being trumpeted. That means I'd have to treat a hundred or more people with Crestor for a year to spare one of them a cardiovascular event that they would not have otherwise had. I'd have to treat several hundred for a year to spare one a heart attack, and perhaps hundreds more to spare one a stroke. I am unwilling to even suggest a life-saving benefit.
So the reduction of 56 percent may be hard to ignore, but it calls for reflection rather than prescribing zeal. It is a reduction in a very small outcome to an even smaller outcome. Consider these two questions:
Are you convinced this small effect is real, that it will reproduce if one were to repeat the JUPITER trial?
I am not. I am reflexively skeptical of effects of this magnitude. My main reason relates to the nature of the randomized controlled trials we rely on for evidence. There are many factors vying to seal a well person's cardiovascular fate.
For example, there are the so-called cardiovascular risk factors such obesity and tobacco abuse. By assigning volunteers randomly to Crestor or placebo, one hopes that the number of smokers and obese folks are equal in the two groups.
When the JUPITER investigators checked, indeed such measurable risk factors were distributed 50-50. One has to have faith that the factors that cannot be safely measured (such as the degree to which the blood vessels are diseased) also distribute 50-50. And one has to have faith that the factors that JUPITER was designed to ignore distribute 50-50.
Socioeconomic status, job security, education level are even more important risk factors that are independent of those measured and likely to vary widely across the research sites in these 26 countries. Slight imbalances between the Crestor and placebo groups could result in effects of the magnitude touted by JUPITER.
I never leap to act on the basis of such small effects. It's why this year if you feed your family margarine, you're not a caring person and last year it was butter that was bad for you.
If you're convinced these small effects are real, are they meaningful to you?
Are you willing to swallow Crestor every day for two years in the hopes you're the one in hundreds who just might be spared a non-fatal heart attack? Does it bother you that more of the volunteers on Crestor were diagnosed with diabetes?
This possible association aside, there is nothing to suggest that the volunteers for JUPITER were harmed in the two years. But that does not mean the drug is risk-free. Does it bother you that the occasional person on Crestor develops a muscle disease, or that some have liver or kidney irritation?
I am not tormented by such uncertainties as I doubt the small effects are real and therefore have no interest in taking Crestor. You and your prescribing physician should take pause, at the very least.
However, the JUPITER investigators and AstraZeneca do not share my concerns. Rather they take refuge in several of the tenets of contemporary small-effect epidemiology. They believe that these small effects are real.
Furthermore, they believe that the small effects recognized in the first two years are likely to prove cumulative and therefore grow as the years pass. It's this belief that triggered the halting of the trial.
And finally they believe that the small likelihood of a good effect for an individual translates into a major public health benefit; benefiting one in a hundred means benefiting 1,000 in every million.
Therein lays a heated academic debate and an important philosophical conundrum. As for me, I won't let you check my CRP either.
Dr. Nortin Hadler is professor of medicine and microbiology/immunology at the University of North Carolina at Chapel Hill, and an attending rheumatologist at University of North Carolina Hospitals. He is the author of Worried Sick: A Prescription for Health in an Overtreated America and The Last Well Person.