Surrogate outcomes miss the point
I read this piece with interest; there are clear parallels to the
One point stands out: does rosiglitazone improve outcome in diabetes,
or not? A quick perusal of effectiveness papers on the web shows that it
reduces HbA1c level. This is all well and good, but one must remember that
improving glycaemic control is only a surrogate outcome measure - the
important outcome is reducing the rate of complications of diabetes, like
myocardial infaction, heart failure and stroke. Papers demonstrating such
concrete benefits seem to be somewhat scarcer.
It seems to me that here we have a drug which improves a surrogate
outcome, but (at best) has no benefit on concrete outcomes and may even
make them worse. We have been here before. HRT improves lipid profiles,
but not cardiovascular events. Fluoride improves bone mineral density, but
increases fracture risk.
What cost effectiveness analyses I can find seem to calculate
cost/QUALY based on an assumed reduction in expensive events for a given
change in HbA1c: that assumption is clearly suspect when the literature
indicates at best no change and at worse an increase in events.
Whether or not rosiglitazone is dangerous, surely it should be
withdrawn because it isn't cost effective in terms of concrete outcomes?
Competing interests: No competing interests