Blood glucose and subsequent cardiovascular disease: update of a meta-analysis

BACKGROUND/PURPOSE: A 1999 meta-analysis described the relationship between blood glucose and subsequent cardiovascular events (MI, stroke, cardiovascular mortality). More studies have been published; therefore, we updated and refined estimates of this relationship in people without diabetes. ACCEPTABILITY CRITERIA: We accepted prospective studies that reported screening results for blood glucose levels (either fasting, 2-hour postprandial, 1-hour postprandial, or casual) divided into ≥3 quantiles. Required data within each quantile were numbers exposed plus study duration or person-years at risk, and numbers of cardiovascular outcomes (myocardial infarction, stroke, death). No restrictions were placed on language or publication date. DATA SOURCES: Two reviewers searched Medline, Embase, Scopus, and Cochrane databases from inception until December 2009. Consensus settled discrepancies. DATA SYNTHESIS: Poisson regression quantified the relationship between glucose quantile and outcomes. Beta values were combined with inverse variance weightings using a random effects meta-analytic model. RESULTS: We found 36 articles with 141 datasets examining the relationship between blood glucose (32 fasting, 52 2-hour postprandial, 37 1-hour postprandial, 20 casual) in 191,249 patients without diabetes (73% male) for 3 million person-years. There were 12,537 (6.6%) cardiovascular deaths, 14,445 (7.6%) cardiovascular events, 6862 (3.6%) cardiac and 3412 (1.7%) stroke deaths. Relative risks/unit increase in blood glucose were all significant for total cardiovascular events (RRs ranged from 1.09-1.51, all p-values < 0.005) and cardiovascular deaths (RR = 1.05-1.24, p < 0.007), and all for cardiac deaths (p < 0.05) except casual glucose; stroke mortality was less clear. The two strongest relationships were found between fasting levels and all events (RR = 1.51, CI: 1.20-1.89) and with cardiovascular mortality (RR = 1.40, CI: 1.18-1.60). With 2-hour postprandial levels, the respective RRs were 1.22 (1.17-1.28) and 1.24 (1.19-1.30). A limitation is the assumption of a continuous relationship between variables. CONCLUSIONS: We have provided refined estimates confirming the association between elevated blood glucose and subsequent cardiovascular events.