The changing "natural history" of symptomatic coronary artery disease: basis versus bias.

In contrast to the 20 to 30% reduction in ischemic heart disease mortality that has been demonstrated by national mortality statistics and by several randomized controlled trials, an uncritical reading of recent "natural history" studies suggests far greater improvements in the survival of symptomatic coronary patients. Although the intrinsic accuracy of such natural history studies is not questioned, attempts to compare patients from different studies and different eras may greatly overestimate the true improvement in natural history because of at least 3 biases in the selection of cases from the spectrum of diseased patients. Because of lead-time bias, patients who are diagnosed earlier live longer regardless of whether interventions are efficacious. Because of referral bias, current patients may have symptoms or anatomy that place them at a different stage of severity than patients who were chosen for earlier studies. Because of incidence-prevalence bias, prevalence studies will be over-represented with survivors of previous incidence cohorts who have slower-progressing disease. We suggest that all natural history studies be carefully scrutinized for such biases before they are compared with each other.