Trends in Cardiovascular Complications of Diabetes

Abstract

To the Editor: The study by Dr Fox and colleagues compared cardiovascular disease (CVD) outcomes for a Framingham cohort examined in 1950-1966 with one examined in 1970-1995 and found that CVD incidence/mortality declined 49% for adults with diabetes and 35% for adults without diabetes. In contrast, studies using National Health and Nutrition Examination Survey (NHANES) data and Rochester Epidemiology Project (REP) resources found that declines in CVD mortality rates between 1970 and 1995 were significantly less for adults with diabetes compared with adults without diabetes. Fox et al attributed this difference to shorter follow-up in NHANES and REP studies and to limitations in the determination of CVD and diabetes in these studies. However, the criticismregardingCVDdeterminationdoesnotaddress those studies’ findings of significant temporal increases in the diabetesassociated risk of all-cause-mortality. While the NHANES and REP studies are more vulnerable to diabetes detection bias than is the Framingham study, earlier detection of milder diabetes in recent periods should contribute to artifactual improvement in relative survival for persons with diabetes, while both NHANES and REP studies found the opposite. Because diabetes duration is a strong predictor of CVD, the risk of CVD associated with diabetes would be artifactually reduced by any bias toward shorter duration. In the Framingham analysis, some individuals belonged to both cohorts; thus, diabetes duration at baseline was likely shorter in the later than earlier cohort. Temporal changes in the Framingham definition of diabetes also favor shorter duration in the later cohort. While follow-up in the NHANES and REP studies was intentionally truncated at 10 years, this was not the case in the Framingham study, so that relatively shorter allowable follow-up for the later Framingham cohort could also contribute to the observed reductions in the diabetes-associated risk of CVD for that cohort. The Framingham study followed up patients to 2000 and is thus more informative than the REP and NHANES studies regarding recent improvements in diabetes treatment and care. However, the latest entry in the earlier Framingham cohort (1966) preceded the earliest entry date for the NHANES and REP studies (1970). Relative survival associated with diabetes improved between 1945-1958 and 19551969. The extent to which the Framingham trends were driven by these earlier years is unknown.