ABSTRACT
PURPOSE: To study a male Italian cohort (initially aged 40-59, n = 1712) during 61 years and the natural history of major CVD mortality categories including coronary heart disease (CHD), stroke and other heart diseases of uncertain etiology (HDUE), including congestive heart failure) along with their risk factor relationships. METHODS AND RESULTS: Cox models were run with 12 covariates as possible predictors measured at entry to the study. About 93% of all CVD deaths were covered by the three major groups selected here (N = 751): 37.4% of them were diagnosed as CHD, 30.6% as stroke and 28.5% as HDUE. CHD declined in the last 20 years of follow-up, while a sharp increase in HDUE mortality was seen. Baseline mean levels of serum cholesterol were 209.6, 204.2 and 198.0 mg/dL, respectively, for CHD, stroke and HDUE deaths: the multivariable coefficients of serum cholesterol were positive and significant for CHD (p < 0.0001), and stroke (p = 0.0203) and not significant for HDUE (p = 0.3467). In Fine-Gray models, the algebraic signs of cholesterol coefficients were opposite for CHD versus the other mortality categories (t = 3.13). The predictive performances of remaining risk factors were varied whereas that of Cox models was not very good, probably due to the attrition phenomenon and possible competing risks. CONCLUSION: Large differences in natural history and risk factors were found comparing the three CVD conditions, potentially indicating different etiologies and pointing to the need of not mixing them up in a grouped CVD category.
ABSTRACT
OBJECTIVES: To assess whether competing risks help explain why regions with initially high serum cholesterol have higher mortality from coronary heart disease (CHD) and lower mortality from stroke and other major heart diseases, while the reverse is found for those with initially lower serum cholesterol. MATERIAL AND METHODS: Ten cohorts of men (N = 9063) initially aged 40-59 in six countries were examined and followed for fatal outcomes for 60 years. Major cardiovascular disease (CVD) groups were CHD, stroke, and other Heart Diseases of Uncertain Etiology (HDUE), or the combination of stroke and HDUE (STHD), along with all other causes of death. Fine-Gray competing risk analysis was applied with CHD versus all other causes of death or STHD (direct mode) and all other causes of death or STHD versus CHD (inverse mode), and the effects of 19 covariates (of which 3 references) on the cause-specific hazard of the outcomes were assessed, thus investigating potential etiologic roles. A systematic comparison with results obtained by running the Cox model in direct and inverse modes with the same end-point results was also performed and illustrated graphically. RESULTS: CHD mortality is bound to different risk factor relationships when compared with all other causes of death and with STHD. The role of serum cholesterol is crucial since, in both comparisons, by Fine-Gray, its coefficients are positive and significant for CHD and negative and significant for all other causes of death and STHD. Risk factor capabilities in specific outcome types of the CVD domain (CHD versus STHD) are different depending on the outcome types considered. Risk factor coefficients are smaller in Fine-Gray modelling and larger in the Cox model. Fine-Gray detects different risk factors whose coefficients may have opposite algebraic signs. CONCLUSIONS: This is the first report whereby a large group of risk factors are investigated in connection with life-long CVD outcomes by Fine-Gray competing risk analysis, and a systematic comparison is performed with results obtained by Cox models in both direct and inverse modes. Subtypes of CVD mortality should be summed with full awareness that some risk factors vary by pathology, and they should at least be disentangled into CHD and STHD.