About as believable as the tooth fairy
It almost seems unnecessary to read any further than the title of
Wald and Law’s paper to realise that it must be deeply flawed. The maxim
‘If it seems too good to be true, then it probably is’ tells us that
anything claiming to reduce cardiovascular disease by more than 80% is
unlikely to stand up to scrutiny.
First, we are expected to believe that the relative risks of the
separate ingredients are multiplicative. What is the evidence for this? It
seems to rely on a failure to find modification of the effect of reducing
one risk factor by the level of other risk factors. However, detecting
such a difference (a statistical interaction term) usually requires
greater statistical power than detecting a main effect. Were the trials
Wald and Law cite in support of their claim adequately powered to detect
the interactions? I doubt it.
Looking closely at their suggestion that a statin can reduce IHD
events by 61% shows more flaws. This seems to be based on the reduction
expected on the basis of cohort studies if the reduction in LDL
cholesterol of 1.8 mmol/l were achieved. First, the figure of 1.8 mmol/l
is based on reductions achieved in patients with high LDL cholesterol, but
we are told the Polypill is suitable for everyone irrespective of their
cholesterol. And why look at cohort studies, with all their inherent
biases and confounding, when we could look at randomised trials? In their
subsequent paper, Wald and Law present the much less impressive data from
randomised trials, which shows reductions in IHD events of about 20–30%.
Only in a subset of trials selected on the basis of showing a large
treatment effect did the reduction in IHD risk reach 50%.
And does the claim of a 46% reduction in IHD from the
antihypertensive components of the pill stand up to scrutiny? Well, no. It
assumes that the antihypertensive effects of the three drugs are additive.
In their third paper, Wald and Law present data showing that combinations
of two drugs are additive (although not convincingly so), and then go on
to make the untested assumption that combinations of three drugs would
also be additive. And again, the expected reduction in blood pressure is
based on patients with high blood pressure at baseline, and reductions in
event rates are based on cohort studies, not on randomised trials.
All this makes me wonder whether this paper was ever meant to be
taken seriously. Perhaps it was simply part of an elaborate experiment
designed to discover what the reaction of the journal’s readership and the
popular media would be if a respectable medical journal published a load
of complete bollocks.
Competing interests: No competing interests