Interpreting and Evaluating the Details of the JUPITER Study

to the editor: I would like to question some of the assumptions and conclusions of the editorial on the JUPITER (Justification for the Use of Statins in Prevention: An Intervention Trial Evaluating Rosuvastatin) study. The authors start with an unproven assumption: a study of 1,315 community sites cannot have high-level quality control. As the chief investigator at one of the study sites, I can assure you that the monitoring and control were thorough, careful, and vigorous. The importance of community-based studies should not be belittled—they come much closer to an interface with real practice conditions than research institution sites. The long list of exclusion criteria is necessary to assure safety and a valid study effect. If anything, the community-based studies are biased toward the less-ill patients, which would minimize the study effect.

It is misleading to consider death as the only relevant endpoint when the study also used cerebrovascular and cardiovascular events as endpoints, both of which are important. It is also misleading to criticize the tight exclusionary data as a weakness. These data made recruitment for the study difficult, but minimized the potential impact of confounding variables. The issue is not one of comparing lifestyle management with a medication, but to see if there is a clinical benefit from treating inflammation as a cardiac risk factor independent of other interventions.

Physicians are tasked with preventing cardiovascular disease when possible. There are multiple risk detection tools available: family history, smoking, diet, lack of exercise, hypertension, and lipid levels. I believe we can now add inflammation to this list. Similarly, we can possibly add to our armamentarium of lifestyle interventions, blood pressure reduction, and lipid reduction, which can affect inflammation reduction. We have a responsibility to select and use all available tools.

The arguments about numbers needed to treat are fraught with misleading assumptions. JUPITER data do not argue for screening patients who meet the study exclusion criteria, or the general population. The data do not argue for considering C-reactive protein measurements in patients with a normal lipid profile, but who are otherwise at risk of cardiovascular disease. The study authors calculated that the number of patients who would need to be treated with rosuvastatin (Crestor) for four years to prevent one adverse event would be 31.¹

The findings of the JUPITER trial were dramatic. The study was terminated after an average of 1.9 years of follow-up because of marked divergence between the treatment and placebo groups. Such early termination of a study powered to detect endpoints such as cardiovascular events or death in favor of treatment is nearly unprecedented. The P value arguing in favor of a meaningful difference was .000001.¹ A wise physician would not ignore such dramatic findings.

in reply: Dr. Neft disagrees with several elements of our editorial. First, he claims we stated that a study with 1,315 community sites cannot have high-level quality control. Actually, we said it is more difficult to achieve quality control when only a handful of patients are enrolled per site. The average site enrolled 13 patients over 3.5 years, or about one patient every three months.

He then argues that the community setting improves relevance and more closely resembles the real world, but then states that a long list of exclusion criteria are necessary to assure safety and a valid study effect. But, study participants either look like the typical patient or they are highly selected to have the best chance of responding to the treatment in question—not both. He argues that the extensive exclusion criteria are necessary to “minimize the potential impact of confounding variables.” However, if done properly, randomization takes care of that by assuring that known and unknown confounding variables are evenly distributed between treatment and control groups.

Dr. Neft states that we ignore the benefit of fewer strokes (0.18 versus 0.34 per 100 person years) and fewer myocardial infarctions (0.17 versus 0.37 per 100 person years). The corresponding numbers needed to treat (NNT) for the two-year study period were approximately 300 and 250 for these conditions, respectively. Remember, these small benefits were seen in a fairly high-risk group of patients. Despite their normal lowdensity lipoprotein (LDL) cholesterol, their average age was 66 years; 40 percent had metabolic syndrome; most were hypertensive; one in six smoked; only one in six took aspirin; and one half had at least a 10 percent 10-year risk of heart disease. These modest benefits, achieved at a very high cost, will be much less impressive when applied to typical primary care patients and younger patients who are at lower risk.

The JUPITER trial's primary endpoint included “all bad things” (a much criticized practice)¹ and had an NNT of 85. Approximately eight patients must be screened to identify one with the combination of normal LDL and elevated C-reactive protein studied in the JUPITER trial. Dr. Neft says the JUPITER trial does not argue for screening, but how else are we to apply the results in practice?

The size of the P value is a reflection of the enormous size of the study, not the magnitude of the effect. With almost 18,000 patients, even a small and clinically unimportant difference can generate an “impressive” P value. It is important to look at the NNT or harm, not just the P values.

We stand by our primary assertions: the benefit demonstrated by the JUPITER trial is modest and comes at a very high cost, and non-drug approaches have been shown to have more important benefits at a lower cost.This trial is a good example of what happens when the drug industry studies questions of interest to them, rather than when physicians and scientists study questions of importance to patients. For example, comparisons of rosuvastatin (Crestor) with less expensive generic statins, with low dose aspirin, and with diet and exercise would provide valuable information for our practices. Unfortunately, we are unlikely to see those comparisons in an industry-sponsored clinical trial.