This study demonstrates that evaluation of the variation in cancer risk in relation to treatment duration is of particular importance for enhancing the accuracy of conclusions on the link between exposure to anti-diabetic medication and cancer risk (PLoS ONE)