Working through the lens of the Bradford Hill criteria for deriving causal inference from observational evidence, Uffe Ravnskov et al. argue that recent literature suggesting statins reduce heart-disease risk by lowering total (TC) and LDL cholesterol (LDL-C) is flawed, as is the broader belief in the inverse relationship between TC and LDL-C and heart disease.
The evidence presented includes the following:
- Across multiple populations, there is not a correlation between rates of atherosclerosis and either TC or LDL-C; in particular, those admitted to hospitals for acute myocardial infarction (heart attack) do not show higher TC or LDL-C levels on average than population averages.
- Studies that have attempted to lower TC and LDL-C over time are inconsistent at best, with some studies showing decreasing cholesterol levels to be associated with increased mortality (total and/or cardiovascular) and others showing decreased cardiovascular mortality only with very low absolute risk reductions (less than 1 percent).
- There is no evidence of an exposure-response relationship in the literature; i.e., greater lowering of cholesterol levels does not associate with greater lowering of risk of mortality.
- The majority of statin trials were completed prior to 2005, and the underlying data are unavailable for public analysis (and so have been questioned).
- The side effects of statins might be underestimated, particularly given the size of the prescribed population and the duration of treatment.
- While cardiovascular-disease (CVD) mortality has decreased substantially since statins were introduced, these decreases have not been correlated with degree of statin use and so may be primarily due to other factors.
Summarizing key points, the authors note: “The usual argument in support of the lipid hypothesis is that numerous studies of young and middle-aged people have shown that high TC or LDL-C predict future CVD. This is correct, but association is not the same as causation.” In other words, TC or LDL-C may be effective markers for atherosclerotic risk within an untreated (or minimally treated) population, but this does not mean these markers are causal in the disease state.