the simvastatin trial PROSPER
28. Nov 2002
This week Lancet: Shepherd,J, et al Prevastatin in Elderly
individuals at risk of vascular disease (PROSPER): a randomised controlled
trial Vol 360;1623- Nov. 23,2002 in which it is concluded that "Prevastatin
given for 3 years reduced the risk of coronary disease in elderly
individual. PROSPER therefore extends to elderly individuals the treatment
strategy currently used in middle aged people." http://www.thelancet.com/journal/vol360/iss9346/full/llan.360.9346.talking_points.23262.1
There was an editorial comment published in the same issue.
Not only is the editorial comment a bit ambivalent "Given that the
overall reduction in vascular events in PROSPER is not highly
statistically significant,... ", the notorious p value, as is
too commonly the case, is used to test for significant significance.
No careful estimates of systematic bias are given, but most important of
all -if you accept their methodology- while deaths from coronary
disease, stroke, and vascular disease were apparently reduced in the
treatment group, these deaths were balanced by Increased deaths from
cancer and non-vascular causes. See Table 2, Endpoints of Prosper,
I am aware Uffe, Paul Rosch, and others have already
commented extensively on the statin controversy, but I wonder if others of
our group would like to comment on this particular paper, especially
the mathematically and epidemiologically-minded with the idea that
this article is a classic Pro Statin piece, which shares many of the
defects of previous articles on the subject, all of which should be again
reassessed and critiqued in another New article to be submitted by The
Here is the full paper: http://pdf.thelancet.com/pdfdownload?uid=llan.360.9346.original_research.23196.1&x=x.pdf
This is an e-mail I sent to a good friend of mine at Bristol-Myers
Squibb with my take on things re PROSPER: "Few Prosper, Nobody
Not happy news for Pravachol: in high risk for "vascular disease"
[24% of the assessed population] men/women, mean age 75, n=2891 on drug,
3.2 years later overall SURVIVAL in statin group: 89.7%, vs. 89.7% in
placebo. My guess: ~$9m/US in drug cost alone to find only 8 fewer deaths,
and significantly more cancer. From reading the abstract, one would never
guess the outcome since mortality is not mentioned. The "combined
endpoint" masked other effects, including harm in other areas of
In this study LDL was lowered by 34%. This study joins most other
statin studies that are all of short duration. 25% more new
diagnosis of cancer in statin group. There was no assessment if
"patients" liked the overall effect of the drug vs. placebo, e.g.
muscle pain, etc. Unfortunately this is, in my reading of the studies, the
general trend for statins.
Unless it is particularly desirable to die from one cause rather than
another, total mortality is of more interest than is mortality from any
given cause. After all, sky-diving without a parachute will decrease one's
chance of dying from cancer, but its effect on all-cause mortality
precludes recommending it as an intervention.
control group in the PROSPER study contained 805 current smokers among
2913 subjects, while the intervention group contained only 753 smokers
among 2891 subjects. Since 306 members of the control group died from all
causes, compared with 298 members of the (smaller) Pravastatin group, it
seems likely that the difference may be a result of the difference in the
number of smokers. Bear in mind that 60% of excess deaths among smokers
are caused by heart disease- lung cancer causes only 13%, and is so widely
recognized as caused by smoking only because smoking increases its
incidence twentyfold, while heart disease is common even in the absence of
Hello Marshall, Good points. I
find it shocking that 20 authors endorse an abstract that explains away a
significantly greater new cancer incidence in elderly by invoking a
meta-analysis with younger subjects and that the line: "Pravastatin
had no significant effect on cognitive function or disability." did
not go on to read ".. or on overall mortality." The fact that
the difference in mortality was the same [+/-0.1%] was not mentioned in
the abstract despite the stated aim ".. Our aim was to test the
benefits of pravastatin treatment...". The study
would have benefited had the mean levels of creatine kinase been reported
since loss of strength and muscle damage can occur before 10x "higher
than normal" creatine kinase.
The Discussion part of the study does not mention overall mortality either
but it does touch on the observation that the "greatest benefit"
in cardio-endpoints appears in [is limited to] the lower tertile of HDL
[i.e. an area where niacin has a greater numerical effect than statins].
Is anyone here prepared to state that statins cause cancer? If so, it
may be time to hit the BMJ (they seem to listen to me for some strange
reasons) and say that this is the case. I still feel the need for a causal
biological mechanism, but the evidence is looking pretty clear-cut, and
Gentlemen, I would agree that Lancet is the more appropriate place
for a reaction but BMJ is more open and accessible. I'll try and post this
response on eBMJ shortly after this Thursday's appearance of the BMJ
(See Rapid Response on http://bmj.bmjjournals.com/cgi/eletters/321/7267/983#29001
I agree - this paper should be commented. The scaring fact is that
cancer was significantly (on the figure, but not in the table ?!)
increased. The authors are eager to explain away this finding by
performing a meta-analysis of all clinical statin trials (excluding, of
course and as usual, EXCEL, where the increased mortality in the treatment
group was borderline significant, but the authors didn't mention anything
about the cause of death).
this is a meaningless way of belittling the findings. The crucial question
is: which types of cancer would we expect to see as the first sign of
carcinogenecity. The answer is of course, those that are most easily
detected, eg. skin cancers (4S and HPS), and cancers that already are
present (operated breast cancers in CARE). Next crucial question: which
types of patients would we expect to get cancer in the first place. And
the answer to that question is of course: the elderly (PROSPER).
by Ravnskov U, Rosch PJ, Langsjoen PH, Kauffman JM, McCully KS, entitled Evidence
from the simvastatin trials that cancer is a probable long-term side effect.)
It does seem that statins increase the risk of cancer. But how? Without a
causal hypothesis any association can be more easily dismissed. A t one
time cholesterol lowering was clearly associated with increase death from
accidents and violence. This was dismissed, primarily, because no-one
could see any biological causal link. The link was, possibly, that
cholesteral is used to create synapses in the brain, and without them
brain function could well be altered in a way that may lead to more 'reckless'
behaviour. The synapse connection was not known when the first studies
were done, and now the association seems to have been forgotten about.
for cancer, I think there need to be a clear hypothesis put forward
linking statins with cancer, otherwise people will just keep dismssing it.
(Perhaps there is a hypothesis already, I just don't happen to have heard
Dear Malcolm - I politely disagree that a "without a causal
hypothesis any association can be more easily dismissed."
Without getting into the argument about "what is (scientific) truth?",
I think we are all stuck on the horns of a semantic dilemma: When
does association become causation? Because of poor statistical power
and the insistence of medical editors for the statement of the P-value, a
vast confusion abounds, especially in our Soft Science of medicine, that
statistical is equivalent to clinical signficance. In this
particular study, the fact that the two groups are treated and untreated,
are not consistently matched to avoid strong selection bias (the excess of
smokers alone in the untreated group is noted.)
The medical literature is rife with misunderstandings of the null
hypothesis.Here is a quote on the subject from Introductory Medical
Statistics 2nd Ed. Mould, RF, Adam Hilger publishers1989:" If
H has not been rejected at a given level of probability then we must
assume that is has been accepted. However, this really
implies not proven-because we are working in terms of
probability.What acceptance means is that if the hypotheses is
in fact false, then the experiment was not able to detect it at the
established level of significance."
I'm not much of a mathematician, but have developed over the years a
profound mistrust of the majority of clinical trials, not only the
cholesterol lowering trials, but other drug and clinical trials. If
we could gather some good statisticians and probability specialists
together, I suspect it could be shown that the majority of such trials are
biased, misleading, or worthless. Fertile field for some fellow
"hard scientists." This is where the cholesterol
hypothesis and other hypotheses must be attacked, at their statistical
claims.So much in our journals has given bull feces a bad name.
The statins promote angiogenesis. Cancers need angiogenesis in order to
spread. Elderly people have a much higher incidence of cancers of all
types. A lot of the elderly must already have growing tumours, albeit
still not giving symptoms. It stands to reason that statin treatment in
the elderly must be connected with a higher incidence of cancer of all
types, promoting the growth of tumours already there.