Lies, damned lies, and statistics 18. April 2009 William Davis (8) In the last Heart Scan Blog post, I discussed the question of whether statin drugs provide incremental benefit when excellent lipid values are already achieved without drugs. But I admit that I was guilty of oversimplification. One peculiar phenomenon is that, when plaque-causing small LDL particles are reduced or eliminated and leave relatively benign large LDL particles in their place, conventional calculated LDL overestimates true LDL. In other words, eliminate wheat from your diet, lose 25 lbs. Small LDL is reduced as a result, leaving large LDL. Now the LDL cholesterol from your doctor's office overestimates the true value. Anne raised this issue in her comment on the discussion:I eliminated wheat - and all grains - from my diet nearly three years ago (I eat low carb Paleo). My fish oils give me a total of 1680 mg EPA and DHA per day, and my vitamin D levels since last year have varied between 50 ng/ml and 80 ng/ml. However, my lipid profile is not like either John's or Sam's:LDL cholesterol 154 mg/dlHDL cholesterol 93 mg/dlTriglycerides 36 mg/dlTotal cholesterol 255 mg/dlMy cardiologist and endocrinologist are happy with my profile because they say the ratios are good, no one is asking me to take a statin. My calcium score is 0. However, if we were to measure LDL, not just calculate it from the miserably inaccurate Friedewald equation, we would likely discover that her true LDL is far lower, certainly <100 mg/dl. (My preferred method is the bull's eye accurate NMR LDL particle number; alternatives include apoprotein B, the main apoprotein on LDL.)So Anne, don't despair. You are yet another victim of the misleading inaccuracy of standard LDL cholesterol determination, a number that I believe should no longer be used at all, but eliminated. Unfortunately, it would further confuse your poor primary care doctor or cardiologist, who--still believe in the sanctity of LDL cholesterol.By the way, the so-called "ratios" (i.e., total cholesterol to HDL and the like) are absurd notions of risk. Take weak statistical predictors, manipulate them, and try to squeeze better predictive value out of them. This is no better than suggesting that, since you've installed new brakes on your car, you no longer are at risk for a car accident. It may reduce risk, but there are too many other variables that have nothing to do with your new brakes. Likewise cholesterol ratios.