Tuesday, March 23, 2010

BBC article's advice: Replace saturated with polyunsaturated fats

The BBC article is here. It is based on meta-analysis of eight previous studies conducted by Harvard researchers, which the article states cover more than 13,000 people. The article also says that: “… [saturated] fats raise the levels of bad cholesterol that block the arteries to the heart.” They are of course referring to LDL cholesterol as the "bad cholesterol".

Sourcing bias is a notorious problem with meta-analyses (i.e., the choice of studies to use in a meta-analysis). Another problem is that you cannot tell what the studies sourced controlled for. Consider a study that compares health markers for smokers and non-smokers, where the smokers eat more saturated fat than the non-smokers. This study may confuse the effect of smoking with that of saturated fat consumption. To be reliable, the study must analyze the effect of saturated fat consumption, controlling for smoking habits.

There are other statistical issues to be considered in meta-analyses. For example, some of the sourced studies may take nonlinear relationships into consideration and others not. In multivariate analysis studies, nonlinearity may lead to significantly different results from those obtained through more conventional linear analyses.

Finally, reaching misleading results with sound statistical analyses is not that hard. As my age went from 1 to 20 years, my weight was strongly correlated with the price of gasoline. Yet, neither my weight caused the price of gasoline, nor the other way around. When you look at an individual study, not a meta-analysis, you can at least try to identify the possible sources of bias and mistakes.

Having said that, a solid refutation of the main argument in the article can be made from many angles. Here is a simple refutation based on what I would call the “HDL cholesterol angle”, with links to posts and various refereed publications:

- Increasing HDL cholesterol levels, especially beyond 60 mg/dl, dramatically decreases the risk of heart disease; and this is an almost universal effect in humans. This reduction in risk occurs even for people who suffer from diabetes and familial hypercholesterolemia. The latter is a genetic condition that is associated with very elevated LDL cholesterol and that is rare, typically afflicting 1 in 500 people in its heterozygous (and most common) form.

- Increasing consumption of saturated fats (present in: lard, fatty meat, coconut oil) and dietary cholesterol (from: fish, organ meats, eggs), while decreasing consumption of refined carbohydrates (e.g., pasta, white bread) and sugars (e.g., table sugar, high fructose corn syrup), significantly increases HDL cholesterol for the vast majority of people. Neither omega-6 nor omega-3 polyunsaturated fats lead to the same results. Omega-3 fats do reduce triglycerides, and increase HDL somewhat, but their effect on HDL pales in comparison with that of saturated fats. Excessive consumption of omega-6 fats is associated with chronic inflammation and related health problems.

- With the exception of cases involving familial hypercholesterolemia, there is no conclusive evidence that LDL cholesterol levels are associated with heart disease. Two widely used online calculators of risk of heart disease, based on the Framingham Heart Study and the Reynold Risk Score, do not even ask for LDL cholesterol levels to estimate risk. And that is not because they calculate LDL cholesterol based on other figures; they do not ask for VLDL cholesterol or triglycerides either.

After reading the BBC article again, it is clear that they are re-stating, in general terms, Rudolph Virchow’s mid-1800s lipid hypothesis. And they do so as if it was big news!