Risk factors and coronary events: attenuation of the association in secondary prevention.

Abstract

Risk factor associations for subsequent events among coronary heart disease (CHD) survivors often appear weakened or paradoxical compared with first CHD event associations, confounding clinical interpretation and risk prediction. This study sought to test whether these observations are a systemic feature of studying disease survivors or unique to specific risk factors requiring biological explanation. A retrospective longitudinal cohort study of 3 275 736 cardiovascular disease (CVD)-free adults and 57 165 CHD survivors was conducted using linked electronic health records from England (Clinical Practice Research Datalink, CPRD). The association of 15 CHD risk factors, documented at cohort entry and again at the time of the first CHD, was compared with both first and subsequent CHD, respectively. The performance of a 10-year primary prevention risk score was also assessed in both settings. Absolute risk over 5 years was 32.4% in those with CHD compared with 1.33% in the CVD-free cohort. There was attenuation of all subsequent CHD event risk factor associations, proportional to first CHD event effect size. Nine risk factors retained a concordant association, while others showed neutral or paradoxical association with subsequent CHD events compared with first CHD events. A 10-year risk model for primary prevention performed well for first events [area under the curve (AUC) .84-.85] but poorly for subsequent events (AUC ∼.55). Risk factor 'weakness' after CHD appears to be a systemic phenomenon and consistent with the operation of index event bias. Association findings in this context should not be used to infer or dismiss causality nor to deprioritize proven therapies such as low-density lipoprotein cholesterol lowering. Broader progress in understanding subsequent event risk associations will require new methods that address index event bias.