The results of our simulations and EPIC-CVD analyses are directly generalisable to diseases with cumulative incidence between 5 and 15% over a 20-year period of follow-up. However, our results could probably be safely extrapolated to other disease incidences over this period of follow-up (i.e. the Cox and logistic model estimates will become increasingly similar as the disease becomes rarer (cumulative disease incidence<5%) and will diverge more as the disease becomes more common (cumulative disease incidence>15%)). Additional avenues of research could include comparing further time-to-event models (e.g. parametric survival models), examining the effect of violating the proportional hazards assumption, meta-analysing case–control studies alongside prospective studies and the inclusion of prevalent cases in the analysis.
