In up to 21 years of follow-up in a study to periodically screen pharmaceuticals for possible carcinogenic eects, 1146 persons who received lindane, an established carcinogen in animal experiments, showed for the ยฎrst time a statistically signiยฎcant increase in incidence of cancer (43 cases observed, 30.2 expected, standardized morbidity ratio 1.42, 95% conยฎdence interval 1.03ยฑ1.92). The computer-stored data on drug dispensing and cancer occurrence also showed some evidence of a doseยฑresponse relationship. Review of the complete medical records suggested that at least part of the association was due to confounding by indication. Four of the cancer patients had Kaposi's sarcoma due to AIDS and the acquisition of pediculosis pubis or scabies, commonly treated with lindane, is associated with behaviour that also predisposes to AIDS. Four other patients also had strong risk factors that could readily be blamed for their cancer and one had not used the drug. Also, the location of the skin lesions or infestations treated with lindane seemed to bear no relation to the cancer sites. It is concluded that this study provides no convincing evidence of carcinogenicity of lindane; nor does it rule it out. In pharmacoepidemiology, it is advisable to supplement analyses of routinely collected data on drugs and clinical events with review of medical records when interesting associations are noted.