Questioning the scientific validity of the APPROVe study's official interpretation
The manufacturer of rofecoxib recently withdrew the drug from the
market because the APPROVe study demonstrated that the drug was associated
with a significantly increased risk of adverse cardiovascular events,
presumably due to the drug's prothrombotic effect.
The study investigators' official interpretation of the APPROVe study
was that "the study's results primarily reflect a greater number of
myocardial infarctions and ischemic cerebrovascular events in the
rofecoxib group". I think that this particular conclusion may not be
accurate. I personally think that the study's results primarily reflect
the fact that the placebo group had an unexpectedly low control event rate
during the last 18 months of the trial.
During the first 18 months of the study there was little difference
in adverse cardiovascular event rates between the placebo and rofecoxib
treated patients. The rofecoxib treated patients had an adverse event rate
of 1.33/100 patient-years while the placebo patients had an adverse event
rate of 1.13/100 patient-years. The relative risk (RR) was only 1.18,
which is not clinically significant.
During the second half of the 36 month study, the RR increased
dramatically to 4.45. However, the main cause of that dramatic increase in
RR was not due to a marked increase in the adverse event rate in the
rofecoxib treated patients, which only increased from 1.33/100 patient-
years to 1.77/100 patient-years. Rather, it was due to a marked decrease
in the adverse event rate in the placebo patients during that time period
-- the adverse event rate plummeted to 0.38/100 patient-years from a
previous level of 1.13/100 patient-years (a ~300% difference).
I have analysed all these facts in great detail in a critical essay,
called "Questioning the validity of the APPROVe study's official
interpretation". The essay is available online at
essay is situated in the soapbox section of my personal website -- type
"Jeff Mann EM guidemaps" in the US-version of Google's search engine to
locate my website if the link doesn't work).
I concluded that the APPROVE study had a low control event rate, a
small signal, and significant noise (presumably due to patient
heterogeneity and chance effects that possibly created an imbalance in
prognostic variables between the treated and placebo patients during the
second half of the trial trial). Under those circumstances, a RCT is
deemed to have a low signal/noise ratio, and Sackett  states that one
cannot be confident in the conclusion of a randomised controlled trial
that has a low signal/noise ratio. If the APPROVe study is widely deemed
to have a low signal/noise ratio, then public health policy decision
makers need to take that fact into full account.
Jeff Mann. MD.
1. Bresalier RS, Sandler RS, Quan H, Bolognese JA, Oxenius B, Horgan
K, Lines C, Riddell R, Morton D, Lanas A, Konstam MA, Baron JA.
Cardiovascular Events Associated with Rofecoxib in a Colorectal Adenoma
Chemoprevention Trial. NEJM 2005. Online - Feb 15. (due to be published in
the print version of the NEJM in late March 2005).
2. Sackett, David L. Why randomized controlled trials fail but
needn't: 2. Failure to employ physiological statistics, or the only
formula a clinician-trialist is ever likely to need (or understand!) CMAJ
165(9):1226-1237, October 30, 2001.
Available online at http://www.cmaj.ca/cgi/content/full/165/9/1226
Competing interests: No competing interests