I noticed that Ed Silverman picked up on a statin study published in the New England Journal of Medicine (NEJM) that suggests statins such as Pravachol -- the drug used in the study -- help prevent heart attacks for at least a decade after people stop taking them.This caught my attention because I recently stopped taking Pravachol, a Bristol-Myers Squibb (BMS) product, after about 12 years because I started getting leg cramps at night. Since I stopped, no cramps. I wondered if my risk for heart attack would increase because I stopped taking the drug, so I was pleasantly surprised to read this story, especially because it was about Pravachol. [BTW, I will discuss this with my cardiologist on Tuesday.]
The problem is, if I had just read the headlines and skimmed Ed's summary, I may have missed an important fact, which was pointed out in a comment to Ed's post; namely,
"...there was a statistically significant increase in prostate cancer which went up as the years progressed... Prostate cancer developed in 89 on pravastatin versus 59 on placebo, a 51% increase, which is much greater than the 25% decrease in heart attacks. The authors stated that the increase was 'probably due to chance and not causally related.' A similar ruse was given for the finding from PROSPER of a 25% increase in cancer, where the authors said that it was due to 'the play of chance'..."The commenter was Doug Bremner, MD, a physician and researcher at Emory University and author of the book Before You Take That Pill: Why the Drug Industry May be Bad for Your Health.
Clearly, Bremner has a bias against the drug industry. He does, however, make a good point that "somehow" got lost in the original AP story and probably wasn't mentioned in the NEJM editorial about the results.
It turns out that BMS provided at least some of the funding for the study and as ed points out "All but one of the researchers reported receiving consulting fees, lecture fees or research grants from a total of five other drugmakers, four of which sell statins."
NOTE: Pravachol is off-patent and generic versions are available. Therefore, BMS may not benefit so much from this study as would other drug companies marketing statins.It appears that the authors of the study went out of their way to play up the good news and play down the bad news.
According to Dr. Bremner, "I think it is time for these researchers to start playing by the rules of statistics, i.e. don't use statistics only when it goes in your favor."
I would have to agree with the doctor. Now, I am interested in his book and also may invite him as a guest to one of my Pharma Marketing Talk podcasts. If you read this, Dr. Bremner, give me a shout.
You can't directly compare a 25% MI/CHD morality reduction with a 51% increase in incidence of cancer -- it's apples and oranges, and a meaningless and oversimplified percentage comparison.
ReplyDeleteAs I wrote over at Pharmalot:
"Given what we know about the natural histories of the diseases, the overall effect on mortality by the 25% decrease in CV mortality almost certainly outweighs the effect by a 51% increase in prostate cancer incidence. This is supported, at least in part, by the similar rates in non-CV mortality, including cancer mortality, both of which have HRs close to 1, and pretty narrow CIs. What would’ve been helpful would be a mortality rate for prostate cancer alone."
And they certainly didn't ignore the incidence or the statistical significance of it, at least in the paper.
Continuing from Pharmalot:
"I think you (Bremner) give short shrift to the authors’ discussion of the increased incidence. They go well beyond merely saying it’s probably due to chance–they spend a decent portion of 3 paragraphs discussing it and reviewing earlier studies showing no increased incidence with statins.
And I wouldn’t dismiss out-of-hand the issue of multiplicity–it’s a legitimate statistical concern.
I do think, however, that such a large increase is worthy of further investigation."
Also, as a side note, I'm not an expert, but Bremner's hosting the full text article on his site is almost certainly a copyright violation. Actually, unless Bremner has permission it definitely is: http://www.nejm.org/aboutnejm/copyright.asp and http://www.nejm.org/aboutnejm/linkspol.asp. I'm just sayin'...I wouldn't want to be on Jeffrey Drazen's bad side.
It is not a decrease in mortality, it is a decrease in the combined endpoint of heart attack and cardiovascular death.
ReplyDeleteThe reference for the meta-analysis showing an increase in cancer with high dose statins is
Alsheikh-Ali AA, Maddukuri PV, Han H, Karas RH. Effect of the magnitude of lipid lowering on risk of elevated liver enzymes, rhabdomyolysis, and cancer: Insights from large randomized statin trials. Journal of the American Academy of Cardiology. 2007;50(5):409-418.
Since prostate cancer is not rapidly fatal a longer followup would be required to assess mortality from cancer versus heart disease.
Yes, I'm aware of the combined endpoint. However, as the article clearly shows, there were significant decreases in all cause mortality HR 0.88 p=0.03, not to metnion all CV mortality (0.81 p=0.01), and CHD mortality (0.78 p=0.02).
ReplyDeleteI'm also aware of the risks of statin use. But the point you were originally making is that the mortality benefits are outweighed by the incidence in prostate cancer, which is debatable at best. You also greatly simplified the discussion of prostate cancer in the paper.
Yes, I'm aware of the combined endpoint. However, as the article clearly shows, there were significant decreases in all cause mortality HR 0.88 p=0.03, not to metnion all CV mortality (0.81 p=0.01), and CHD mortality (0.78 p=0.02).
ReplyDeleteI'm also aware of the risks of statin use. But the point you were originally making is that the mortality benefits are outweighed by the incidence in prostate cancer, which is debatable at best. You also greatly oversimplified the discussion of prostate cancer in the paper in order to make your point.
John,
ReplyDeleteJust curious if you feel any obligation to correct or amend not only Dr. Bremner's misquoting of the authors (they said "arguably"--and in the context of multiplicy--not "probably", which have two very different meanings), but his insinuation that this is the full extent of the safety discussion in the article? It's not like it was buried in a table -- it took up approximately 1/4 of the entire discussion.
You say that you wouldn't have known about the prostate cancer incidence if you only read the headlines or skimmed Ed's summary. Don't you think it's equally important that people who may have done the same know what the authors actually said in full?
Not to mention the statistically and epidemiologically inappropriate and invalid comparisons of MD/CHD death rate vs. prostate cancer incidence.