Thursday, November 13, 2008

By JUPITER: Slick Drug Marketing, Great Science, or Both?

The JUPITER study, which you have heard about on the news unless you have been in a deep coma this past week, is the sort of thing us Pharma skeptics would love to find fault with. My fellow blogger Merrill Goozner noted its two sins--first, the study was funded by AstraZeneca, maker of the study drug, rosuvastatin (Crestor); second, the first author, Paul Ridker of Harvard, owns a patent on the high-sensitivity test for C-reactive protein, the test that would be widely used if the study results are accepted. This led Goozner to twist himself into pretzels trying to find flaws in the study:

I studied JUPITER and the accompanying editorial in some detail and (while awaiting a better assessment from my friends who know study design and biostatistics better than I do) would offer the view that our pal Gooz went a bit overboard. Still, there is an important lesson here and it is not what the media claimed.

Basically, JUPITER screened about 89,000 patients and found 17,802 of them who met three major inclusion criteria--they had no existing heart disease; they had a normal cholesterol level; and they had an elevated high-sensitivity C-reactive protein test, a test for general bodily inflammation. The subjects were then randomized to the statin drug or a placebo daily. They were supposed to be followed for 4 years but the data monitoring board recommended cessation of the study after 1.9 years because of a significantly lower rate of all major outcomes (heart attack, stroke, unstable angina, need for coronary artery surgery) in the group that got the statin drug.

One point to note is that the JUPITER population was a relatively high-risk because older population, with the median age in each group being 66.

On one point, Gooz got it exactly right. The news media wildly reported that these bad outcomes had been "cut in half" by the statin. In actual fact, while the rate of bad stuff was generally half as much in the statin as in the placebo group, the more accurate statistic, absolute instead of relative risk reduction, was far less impressive. For example, if you added up all the bad outcomes as a composite endpoint, there would be 1.36 bad outcomes per 100 person-years of follow-up in the placebo group and 0.77 per 100 person-years in the statin group. This translates into a number needed to treat (NNT) of 95 over 2 years (that is, you would need to treat 95 patients for 2 years with statins to prevent one bad outcome). It is worth noting that this NNT is quite a bit more favorable than in most previous trials of primary prevention with statins, where it is not unusual to see NNT's in the range of 300 to 400. (A low NNT is good, a high one is bad.)

Another score for JUPITER--the death rate was reduced by statins but not by as much as the other bad outcomes. (There was only about a 20 percent relative risk reduction in deaths vs. the nearly 50 percent relative risk reduction for the other outcomes.) Still, this is very impressive because in many earlier primary prevention trials, while statins reduced the number of heart attacks or strokes, they could not be shown to reduce the overall death rate one bit. (This is partly what has led some to fear that statins increase the risk of other deadly diseases such as cancer at the same time that they reduce cardiac or vascular risks. In JUPITER there was no increase in cancers among the statin group, through the follow-up period was of course very short.)

There was also a very limited downside to taking statins for the duration of the study. The only adverse event that happened more often in the statin group was a new diagnosis of diabetes, and this was only a slight difference (3.0 vs. 2.4 percent). A lot of these subjects were diagnosed at the start with metabolic syndrome so their baseline risk of diabetes was very high.

So, bottom line-- it seem hard to argue that this study is a winner for using statins for primary prevention (before any vascular disease has arisen) in one high-risk population, those with elevated c-reactive protein; and that in turn implies the value of using the CRP test as a routine screening test. The NNT of 95 is quite respectable among primary prevention trials.

The JUPITER authors seem to have been fully aware of the skepticism that would greet their paper due to the obvious conflicts of interest. They went to great lengths to note that AstraZeneca did not even see the unblinded trial data until after the manuscript was submitted for publication. Two committees were said to be completely independent of both the company and the principal investigators--an endpoint adjudication committee and a data monitoring and safety committee. In short, if these statements are all true, the study was squeaky clean despite the commercial sponsorship.

Now I come to the point that the media seems completely to have missed. Skeptics of the widespread use of statins (like yours truly) have for a long time insisted that the bulk of the research literature to date does not support the "lipid hypothesis"--that statins work by lowering "bad cholesterol" levels; and that it is critically important to screen patients by testing their cholesterol, and then doing repeated tests to be sure that the response to statin treatment is a low enough level. We have insisted instead that the research indicates that whether or not statins are good for you depends almost completely on whether you have certain risk factors. The biggest risk factor, and the groups where statins have been shown most clearly to work, is already existing vascular disease (that is, secondary prevention). By contrast, very little benefit from statins has been seen in most primary prevention trials, except in men at high risk in the roughly 40-65 age group, and even there with a very high NNT. Since you don't need a cholesterol test to know if you have those risk factors, we have argued that the massive use of cholesterol screening leads to the inappropriate overuse of statins. Moreover, there is no evidence that hitting a lower target cholesterol level by increasing the statin dose lowers your risk to any significant degree.

One way to summarize all these data is to suggest that it is quite possible that statins prevent heart attacks and strokes, but not by lowering cholesterol. They are known to have an anti-inflammatory effect, and it is thought that inflammation plays a role in turning arterial plaque into a major blockage of a vessel that leads to bad disease. It may be their anti-inflammatory effect and not the lipid-lowering effect that explains why they work in the highest-risk groups.

Well, if this hypothesis was worthy of consideration yesterday, it has now received a huge boost with JUPITER. Based on lipid screening, these subjects did not need any statin. Yet they did better on a statin for primary prevention. They were a very high risk group (mainly due to age) and also had signs suggestive of an inflammatory process.

The media never picked up on this. All the coverage I heard simply took for granted that it was wise to treat elevated cholesterol with statins, so now we had yet another huge group of patients who needed a statin in addition to those with high cholesterol. On NPR (of all places) this morning, I heard a physician quoted that JUPITER showed that we need to reduce levels of cholesterol much lower than anyone had previously realized--when the real message is to test CRP and forget all about cholesterol. Even if the lipid hypothesis of statin treatment is all wet, we still seem completely hooked on it.

Ridker PM, Danielson E, Fonseca FAH, et al. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. New Engl J Med 359:2195-2207, Nov. 20, 2008.

Hlatky MA. Expanding the orbit of primary prevention--moving beyond JUPITER [editorial]. New Engl J Med 359:2280-2282, Nov. 20, 2008.


Anonymous said...

Dr. Rich at the Covert Rationing blog has a somewhat different take on the statistics and the NNT,at am so naive in matters statistical I am typically convinced by the last spin (interpretation) that I read.Do you have any comments re Rich's way of looking at it (I am referring mainly to his footnote with the K-M curve approach)?

Howard Brody said...

Thanks for the helpful comment and link. I read DrRich's post, and I personally am equally unskilled in matters such as the K-M curve. However, taking DrRich's post as a whole, I think it makes many of the same points that I do in my post. He argues that the NNT for Crestor in JUPITER is actually pretty good, and is better than the statistic that the NEJM editorial claims. I used the NNT data from JUPITER itself, not from the editorial, and I also said that the NNT's are impressive compared to earlier primary prevention trials.

My main disagreement with DrRich is that he claims that solid application of EBM principles would demand that we prescribe Crestor only, and no cheaper statin, since it was that one drug only that was tested. I believe that's an overly rigid understanding of what EBM requires, and that the evidence that a statin is a statin is a statin is well enough established to allow us to conclude that this is a class effect and not an individual drug effect.

Anonymous said...


There was a reduction in total mortality in WOSCOPS. I believe at the end of the trial the difference was not significant, but it became significant in long-term followup. I am not of the school that total mortality must be reduced in order to show a benefit, in any case. To me, avoiding a heart attack or stroke is a benefit. Maybe because I watched my father's last few miserable years of existence after his last stroke, in a wheelchair, incontinent, one arm useless . . . He would have been better off dead.

You are never going to get a high absolute benefit in a primary prevention trial. However, some people have high baseline risk, and their absolute risk reduction from a statin will be higher than what you see in clinical trials. My husband is one of those people.

I am not sure CRP should be measured routinely if by routinely you mean the whole adult population. Based on what I know now, I would limit CRP testing to those who are at intermediate risk.

I agree that much of the benefit of statins is due to effects unrelated to LDL-lowering, but it seems unlikely to me that none of the benefits are related to LDL-lowering, since other interventions that lowered LDL have lowered CVD risk (e.g., ileal bypass, cholestyramine). Unfortunately, these other interventions have typically not lowered CVD risk by all that much, and in some cases there has been a distressing increase in all cause mortality.

All the more reason to avoid non-statins whenever possible in those persons who need medication.

I was glad to see a clear benefit for women in this trial.


Anonymous said...

Ver mis comentarios en: