In Reply We appreciate the well-considered comments by van den Born and colleagues about our recent study on use of hydrochlorothiazide and risk of malignant melanoma.1 We agree that the possibility of residual confounding should always be entertained in observational studies, particularly when considering risk estimates in the order reported by us (eg, an odds ratio of 1.22 for hydrochlorothiazide use and melanoma risk). Confounding by indication (in this context, hypertension) is often a potential concern in pharmacoepidemiologic studies. However, we find it unlikely that confounding by indication explains the results of the analyses by subtype of melanoma, which we were particularly concerned by, because they yielded higher risk estimates for nodular and lentigo melanoma, with some evidence of a dose-response pattern. Similar associations with melanoma subtypes were not seen for other antihypertensive drugs.