When "flawed" translates into "flood": the unproven association between cancer incidence and glargine insulin therapy

Abstract

A few months ago, a hot spot emerged in the area of diabetes and cancer epidemiology. Hemkens et al., from the Institute for Quality and Efficiency in Health Care, published in Diabetologia a report of an observational cohort study based on a large health insurance database representing almost 18 million people from Germany. Their hypothesis was to test whether glargine, an insulin analog (Lantus®; sanofi-aventis, Bridgewater, NJ), was associated with a higher incidence of cancer than human insulin. Initial results not only rejected this hypothesis, but found a protective effect for glargine in the age-sex-adjusted analysis (for glargine and the risk for malignant neoplasm: hazard ratio [HR], 0.86; 95% confidence interval [CI], 0.79–0.94), yet after a bias-introducing adjustment for the glargine insulin dose, the insulin analog seemed to confer a higher risk for neoplasm incidence (HR, 1.14; 95% CI, 1.05–1.24). Three other papers in the same issue did not find a convincing association [2–4].