Importance of trends in the interpretation of an overall odds ratio in the meta-analysis of clinical trials

OBJECTIVE: To investigate the predictive value of fasting insulin concentrations for subsequent fatal or non-fatal ischaemic heart disease at five year follow up and to examine the associations between insulin and other indicators of risk. DESIGN: A prospective population study among 2512 men aged 45 to 59 at recruitment. SETTING: A whole population sample of men resident in Caerphilly, South Wales. MEASUREMENTS: At recruitment fasting blood samples were taken for measurement of plasma lipids and serum insulin. Men were re-examined at a five year follow up and ischaemic heart disease events during this period were assessed from hospital notes, death certificates, and electrocardiograms. MAIN RESULTS: Diabetic men and those men with a fasting blood glucose of > or = 8 mmol/l were excluded from all analyses. In a univariate analysis the incidence of ischaemic heart disease increased with increasing concentration of fasting insulin, such that for men in the top 20% of the insulin distribution the odds of developing ischaemic heart disease were 1.87 relative to men in the bottom 20%. On multivariate analysis this relation disappeared on adjusting for plasma triglycerides, body mass index, and evidence of ischaemic heart disease at recruitment. CONCLUSION: In this population in South Wales there was no evidence that the fasting insulin concentration is an independent risk factor for ischaemic heart disease. The univariate association between insulin and incident disease was almost entirely explained by the association of both with triglycerides and body mass index.