eta-analysis, which is achieved by pooling data from several studies that meet certain quality (and size) criteria, tends to be done in arrears. A research group decides that more robust conclusions might be drawn from a summation of published numbers than from a conventional literature review, and then proceeds to seek cooperation from those who hold the individual patient data. The Cholesterol Treatment Trialists’ Collaboration, which is just one of many projects handled by the University of Oxford’s Clinical Trial Service Unit and Epidemiological Studies Unit (in collaboration here with the University of Sydney, Australia), is a meta-analysis planned in advance, in 1994.¹ The first findings appeared in 2005 (#7) and concern just one type of cholesterol-lowering drug, the statins, of which four were the focus of the 14 randomized trials (in a total of just over 90,000 patients) that contribute to this paper. These individual trials were published several years ago but one of them is fresh enough to be listed, albeit just outside the Top Ten (#12, J.C. LaRosa, et al., New Engl. J. Med., 352[14], 1425-35, 2005; total cites 253, latest count 36).

Millions of people now take statins long-term. These drugs lower levels of the "bad" or low-density lipoprotein (LDL) form of cholesterol and do so in people with a raised cholesterol, whether or not they have any personal history of heart disease. This biochemical benefit converts to a clinical one in the form of reduced cardiovascular risks (e.g., a heart attack or further such attack). So what can be added by further number crunching? There are issues with statins. More accurate estimates of the benefit can be had from a meta-analysis. Also, safety can be explored in more detail, and this is perhaps a particular concern in countries such as the U.K., where a statin can be purchased at a pharmacy without a doctor’s prescription. The U.S.A. has yet to go down this over-the-counter route. Furthermore, the bigger datasets that a meta-analysis provides may help clarify effects in various patient subgroups.

With LDL-cholesterol lowering, size matters. This analysis permitted the reduction in clinical endpoints such as major coronary events to be plotted against the reduction in LDL-cholesterol, which overall was about 1 mmol/L at one year in these trials. The result was a straight line. The fall in major vascular events was about one-fifth but the data can be extrapolated to predict a reduction by about one-third if the LDL-cholesterol were to be lowered by 50% more (i.e., by 1.5 mmol/L). The Collaboration concludes that "major clinical and public-health benefits" would result if substantial reductions in LDL-cholesterol are achieved in high-risk patients.

On the safety side, the main concerns with statins have been over muscle problems (especially rhabdomyolysis) and cancer; the latter has been for many years a long-running worry for cholesterol lowering more generally. On both points this meta-analysis is reassuring. The five-year excess risk of rhabdomyolysis was just 0.01% and the relative risk of cancer was a non-significant 1.01. (Nor did statins reduce the risk of particular cancers, as some non-randomized studies had hinted.) This 2005 meta-analysis restricted itself to large trials having at least 1,000 participants. Two analyses from last year looked separately at the same two aspects of statin safety but let in trials of 100 patients and more. The conclusions were the same as in #7 (on cancer risk, K.M. Dale, et al., JAMA, 295[1], 74-80, 2006; on musculoskeletal and other complications, A. Kashani, et al., Circulation, 114[25], 2788-97, 2006). One way and another, both the efficacy and the safety evidence on statins are solidifying in favor of this drug class being an important contributor to public health by way of secondary prevention. A statin’s effect on total mortality in people with no occlusive vascular disease at all (true primary prevention) remains under discussion (J. Abramson, J.M. Wright, Lancet, 369[9557]: 168-9, 2007).