*The development of diabetes is one such unanticipated harm found in a recent large study and it seems likely therefore that this applies to the data above, although this is a best guess.
Source: Thavendiranathan P. Primary prevention of cardiovascular disease with statin therapy. Arch Int Med. 2006; 166: 2307-13.
CTT Collaborators. Efﬁcacy and safety of cholesterol-lowering treatment: prospective meta-analysis of data from 90 056 participants in 14 randomised trials of statins. Lancet. 2005; 366: 1267-1278.
Ridker et al. Rosuvastatin to prevent vascular events in men and women with elevated c-reactive protein. NEJM. 2008; 359(21): 2195-2207.
Efficacy Endpoints: Death, heart attacks, strokes
Harm Endpoints: Rhabdomyolysis (muscle breakdown), cancer, diabetes
Narrative: Statin medications are aimed at an enzyme in the cholesterol production pathway, and are therefore intended to reduce cholesterol. In some patients elevated cholesterol plays a role in the development of coronary artery disease and heart attacks, so the drugs are intended to reduce the chance of death by reducing heart attacks and strokes.
These data examine the effect of statins for people who have known heart disease or a history of stroke. The effectiveness of the statins is fairly consistent across studies in this group—they lower cholesterol in most people who took them. A few people will also avoid a heart attack or stroke by virtue of this change. After 5 years of daily statin therapy study subjects achieved a 1.2% lower chance of death, a 2.6% lower chance of heart attack, and a 0.8% lower chance of stroke. As a public health measure, this suggests that statins may have an identifiable effect, because while the chances of any one individual being affected are small (19 out of 20 people who took the drugs for five years saw no effect), when one million people take them roughly 45,000 people saw some benefit, while another 6,000 may see a harm.
Caveats: The raw numbers of 270 and 216 new onset diabetes cases from 24 months of exposure to a statin and a placebo (respectively) can be extrapolated, assuming that increased diabetes risk is likely to continue linearly with exposure. This yields 675 and 540 cases at 5 years. Combining these numbers with the AFCAPS numbers yields a relative risk of 1.22 for development of diabetes (not 1.18 as reported by Cochrane, and NOT 1.81 as we had erroneously reported in an earlier version of this review, which also led to an erroneously low NNH of 50). This updated relative risk increase of 1.22% translates to an absolute risk of 1.12%, or 1 in 89.
This may, however, be flawed in its assumption of a linear increase. Perhaps the increase in diabetes risk is frontloaded, and does not rise above that found at two years. In this case the risk increase would be 1.18 as suggested by the Cochrane group, and the absolute risk increase in this case would be 0.43%, or 1 in 233. The answer likely lies between 0.4% and 1.5%, and we have chosen what we believe to be a conservative estimate of 1% as a midway point in this credible interval.
Author: David Newman, MD
Published/Updated: July 17, 2015
The title bar is color-coded with our overall recommendation.
If you have suggestions, requests, or questions about a particular NNT review, please send us a message and we’ll try to address it as soon as possible.