By JUPITER, Part II--A More Skeptical View

When I did my initial post on the JUPITER study (immediately preceding this post), I had not had the benefit of first talking with a true evidence-based-medicine mayvin. I have now just gotten off the phone with Jerry Hoffman of UCLA, who reviewed my post and the attached comments. I thought I had better hurry up and do a new post before my addled brain forgot all of his observations.

Jerry first of all defended Merrill Goozner against some of my skeptical comments about his blog post. Next, Jerry agreed with me that the basic result of JUPITER is to cast more doubt on the "lipid hypothesis" of how statins help in heart disease, and certainly undermine a strategy of widespread testing of cholesterol and trying to achieve a target level of cholesterol by adjusting statin dose.

Jerry concluded by saying that he put JUPITER together with earlier studies in this way. We know that there are basically three groups of folks--high, medium, and low risk for vascular disease. The really high risk people have already had a stroke or heart attack. For them it seems pretty clear that statins help, and that the number needed to treat (NNT), while still on the high side, at least seems to justify the three costs of statin therapy. (These costs are: the monetary cost of these pricey drugs; the known side effects of statins; and the as-yet-unknown side effects among people who may be on the drug for 30-40 years. One reason that the cost-benefit ratio is favorable in the high risk group is that if you already have had a stroke or heart attack, the chances of your hanging around long enough to develop those possible long term side effects becomes quite low.)

Then we have the low-risk people. For them, the NNT is extremely high. It seems hard to show that the costs, in their case, would be low enough to justify these slight benefits.

Finally we come to the moderate-risk group. JUPITER identified a bunch of these folks, who had the elevated C-reactive protein and also were on the older side. Surprise--their NNT turned out to be intermediate between the highest and lowest risk groups. So the question, first, is: does the intermediate NNT justify the still-substantial costs of statin therapy? (This was Goozner's main point.) If this was explained adequately to patients, some might say yes and others would say no.

The second question is trickier and ignoring it was the main defect in my first post. It is true that if you just looked at the people known to have a high CRP test, the NNT is 95 for 2 years. But that is not the best way to look at the data. If a patient comes to you, the doctor, you have a choice of two strategies. You can look at their existing risk factors, place them in a high-or low-risk category, and advise statins or not depending on that risk assessment, using if you want a tool like the Framingham score. Alternatively, you can order a CRP test and then alter your advice to the patient based on whether it is high or normal.

When you consider the true NNT, you cannot just look at those who are already known to have a high CRP. You have to look at the initial population and factor in the uncertainties associated with the test (such as false positives and negatives, etc.) The NNT for the entire strategy (first test, then treat if high) is going to be a higher number than 95 for 2 years to prevent one bad outcome. Jerry's final question is--how much extra value is added by doing the CRP test, compared to just using other knowledge of risk as we would do today? JUPITER did not give us enough data to answer this question. Specifically, it did not tell us how many patients might have been placed on statins anyway due to other risk factors even had the CRP not been done; nor did it tell us how many patients who did not have another reason to be on statins had to be screened with the CRP in order to yield the relatively small percentage of all comers who ended up in the trial.

I will turn Jerry's good observations into a hypothetical patient coming in the door who is perhaps moderate risk for heart disease, and wants to know if he should take a statin and if he should have that new blood test that he heard about on the news. It seems prudent to offer at least three options:

• Stay off statins and focus on lifestyle interventions to prevent heart disease--especially exercise--which for all we know have a lower (more favorable) NNT than statins do
• Don't do a CRP and prescribe statins, or not, based on other risk variables precisely as one would if the CRP did not exist
• Do the CRP and be guided by the results as to whether to give a statin, the cheapest possible one in the lowest reasonable dose (a modified JUPITER protocol)

My point is that all three options make sense to offer the patient. Jerry would add, that if the patient asked which one Jerry would recommend, he's not sure he'd recommend either statin option based on the totality of today's evidence, including JUPITER.

Finally, you might ask--as this blog is not about statins or heart disease prevention, but the pharmaceutical industry and its relationship with medicine, why have I gone off on this long JUPITER tangent? I might reply that we were forced onto this tangent because a major study that got a lot of media play was hampered by obvious conflicts of interest in its sponsorship and authorship (as again, Goozner accurately noted). This is at least the length that one needs to go to, in today's environment, to decide what to believe of what is published, so as to assure that drug company spin does not trump science. The mere fact that the study was discussed at the meeting of the American Heart Association, and published in the New England Journal, is a completely inadequate assurance of scientific reliability, as multiple examples provided in HOOKED attest.