Monday, July 13, 2009

More on Statins: New BMJ Meta-Analysis, Old Numbers

At intervals I have posted in some detail about statins and cholesterol, primarily to make a point about drug industry marketing. It is my claim, following experts such as Dr. John Abramson in OverDo$ed America, that one sign of the power of drug industry marketing is that we hear one story about statins, so loudly that it drowns out any other story. If science rather than money ruled, we'd hear a second story that would compete with the first story for our attention. The two stories are:
  • The Story We Hear: Statins are very effective in saving lives by preventing heart disease. The smart patient keeps an eye on cholesterol and at the first sign of its going up, asks the doc for a prescription for a statin. The doc and patient then monitor the cholesterol levels carefully to keep them as low as possible to get maximum benefit. Practice guidelines based on compelling recent studies are regularly revising downward the threshhold level of cholesterol that should trigger statin prescription.
  • The Story We Don't Hear: Statins are pretty good for one group of patients--those with existing heart/vessel disease--to prevent future events such as heart attacks. We cannot be sure that they achieve their benefits in this group by lowering cholesterol and indeed a different mechanism, such as anti-inflammation effects, may be responsible. There is much less evidence that statins work for primary prevention--preventing heart/vessel disease where it does not yet exist. As a rule, hundreds of patients must be treated in a primary prevention study to have one who benefits. There's no good evidence that measuring cholesterol levels or monitoring cholesterol levels after treatment adds anything useful to statin therapy.

As somebody who argues for the second story at least getting equal time, if not actually being accepted as the more correct account, I naturally wonder about what newer data will reveal about the statin/cholesterol hypothesis. As I previously posted (, it was intriguing how the JUPITER study actually blew a huge hole in the standard hypothesis, yet was covered by the media as if it was just one more reason why everyone and his duck should be taking statins.

So now-- thanks to the tip from Doug Bremner's blog-- comes the online advance publication in BMJ of a meta-analysis on stains for primary prevention--

--which aims to answer just about exactly the question I pose above--how do statins perform in patients without existing heart/vessel disease, considering not surrogate endpoints like cholesterol measurements, but hard patient-oriented outcomes like death, heart attack, and stroke?

The first thing to note is that the list of financial conflicts of interest among the authors occupies a long paragraph, so this is clearly a group used to seeing the green of drug company money.

The authors then analyzed ten randomized trials of statins for primary prevention (or primary prevention subgroups within other trials), including JUPITER. They concluded that when the trials were all summed, the statin group showed statistically greater improvements in all cause mortality; major coronary events; and major cerebrovascular events.

Based on the data provided, one can calculate the critical statistic, number needed to treat (NNT). You would have to treat 167 people for 4.1 years to prevent one death; 77 people for 4.1 years (or 316 people for 1 year) to prevent one major coronary event; or 250 people for 4.1 years to prevent one major cerebrovascular event.

Here is what may be the most important take-home message--the authors admit in their discussion that the overall baseline mortality rate across all ten studies is 1.4 %. They note that this is a high number--indeed, it is nearly the same mortality rate found in many secondary prevention trials of patients with known heart disease.

Now, if the story we usually don't hear is the correct story, what would we expect such a meta-analysis to show? First, we should not be surprised if the statin group does statistically better than the control group. Second, we'd expect that the NNT would be high--generally in the 100-300 range. Third, we'd expect that the NNT would get lower in proportion to the patient's risk status for later heart/vessel disease going up.

And what did this meta-analysis by Brugts et al. demonstrate? Exactly that. Statins offer a statistical advantage but the advantage in terms of absolute risk reduction is low, so it yields a high NNT--but a lower NNT than in some studies taken individually. But the lower NNT is not unexpected when we realize that this population of patients is actually a very high risk population--indeed, in just about the same risk category as those with known heart disease, who clearly show benefit from statins. Finally, there is yet once again no evidence at all that using any particular statin, or using any particular dose of statin, or achieving any particular cholesterol target level, makes the slightest bit of difference (though to be fair those variables were not the major outcomes looked at in the study design).

Brugts JJ, Yetgin T, Hoeks SE, Gotto AM, et al. The benefits of statins in people without established cardiovascular disease but with cardiovascular risk factors: meta-analysis of randomized controlled trials. BMJ 2009:338:b2376, doi:10.1136/bmj.b2376


Marilyn Mann said...

"there is yet once again no evidence at all that using any particular statin, or using any particular dose of statin, or achieving any particular cholesterol target level, makes the slightest bit of difference"

What about PROVE IT? Pravastatin v. atorvastatin, atorvastatin was superior.

Howard Brody said...

Marilyn, excellent question, as always! PROVE IT (Cannon CP et al., N Engl J Med 350: 1495-1504, 2004) compared 40 milligrams of pravastatin (normal dose) with 80 milligrams of atorvastatin (high dose) in patients with acute coronary syndrome, that is, secondary prevention trial in patients with existing coronary disease. The claim is that the combined endpoint showed the superiority of the intensive dose regimen. We do not know if the cholesterol-lowering effects vs. the anti-inflammatory or other effects of the two statins might have caused the different outcomes. The combined endpoint is a very suspicious feature of the trial--just about the only endpoints that CLEARLY showed the superiority of the intensive dosage were perceived need for revascularization surgery and repeat acute coronary syndrome requiring hospitalization. None of the harder endpoints like heart attack or death were significant. So they had to lump all the endpoints together to get the "combined" endpoint that gave them the answer they wanted. Finally there was something fishy about the statistics, which they reported as "The study did not meet the prespecified criteria for equivalence but did identify the superiority of the more intensive regimen. However it was also stated, "The study was designed to test the noninferiority of pravastatin as compared to atorvastatin..." I'm not enough of a statistician to make sense of all that, but I am concerned that the conclusion drawn from the study was not actually the conclusion the study was designed to test. Bottom line-- I am not sure how much you can conclude at all from PROVE IT, but you cannot conclude anything about primary prevention. Thanks as always, Howard

Marilyn Mann said...

Howard, PROVE IT was funded by the maker of pravastatin. The sponsor of pravastatin was trying to show that pravastatin was noninferior to atorvastatin, despite less LDL-lowering.

I was aware that PROVE IT was a secondary prevention trial. I was not trying to say anything about primary prevention. Perhaps I misunderstood, but I thought you were trying to make a general statement that all statins were equivalent.

I readily concede that statins are overprescribed for primary prevention, and that the absolute benefit of taking a statin is relatively low in most people without known cardiovascular disease. However, some people have a high baseline risk, and some people are interested in tiny reductions in risk.

In science, there is no absolute proof of anything, only greater or lesser quality and amount of evidence. There is an substantial amount of evidence that LDL is causal in heart disease. One of those pieces of evidence is the extremely high incidence of heart disease in people with familial hypercholesterolemia. Another piece of evidence is the low incidence of heart disease in people with loss-of-function PCSK9 mutations, who have lifelong low LDL levels. There is also epidemiological evidence and evidence from animal models. In addition, not only statins, but a number of other LDL-lowering interventions have been shown to reduce the risk of heart attacks. In view of the fact that these other drugs or interventions do not have the pleiotropic effects of statins, how do you explain their ability to reduce heart attacks?

How do you explain people with heterozygous familial hypercholesterolemia suffering heart attacks as early as their 30s and 40s if LDL is not causal? This is not an abstract question for me, since my husband's grandfather and uncle suffered fatal MIs at 35 and 40.

I happen to be a big believer in the pleiotropic effects of statins, but it is very hard to separate the pleiotropic effects from the LDL-lowering effects. In JUPITER some of the benefit presumably came from LDL-lowering.