In the applied example of the paper, there may also be deterministic dependencies between the risk factors. LDL‐C is rarely measured directly but is estimated from measurements of total cholesterol, triglycerides, and HDL‐C via the Friedewald equation as total cholesterol minus HDL‐C minus 0.2 times triglycerides (assuming all measurements in mg/dL).26 It has previously been shown that the coefficient for LDL‐C is the same as the coefficient for non‐HDL‐C (calculated as total cholesterol minus HDL‐C) in a regression model including HDL‐C and triglycerides (see Appendix 2 in the paper by Di Angelantonio et al).27 However, the coefficient for triglycerides will change, as the non‐HDL‐C measure contains more triglycerides than the LDL‐C measure. Hence, in the case that there are deterministic relationships between the risk factors, effect estimates may change as the choice of risk factors varies due to their interpretation as direct effects conditional on other risk factors in the regression model.
