To determine the accuracy with which Medicare claims data measure chemotherapy use in elderly Medicare beneficiaries with cancer, we performed a criterion validation study. We compared gold-standard clinical trial data for 175 elderly cancer patients treated in two Cancer and Leukemia Group B (CALGB) breast and lung cancer trials (i.e., 45 from trial 9344 and 130 from trial 9730) with contemporaneous ambulatory and in-patient Medicare health insurance claims data from Centers for Medicare and Medicaid Services (CMS). The breast trial participants studied were those elderly enrolled between 1995 and 1997 and treated with doxorubicin and cyclophosphamide or this combination with paclitaxel. The lung trial participants studied were those elderly enrolled between 1998 and 2000 and treated with paclitaxel and carboplatin or paclitaxel alone. Comparing CALGB data with Medicare claims, we found the crude sensitivity for chemotherapy administration was 93% (95% confidence interval [CI] = 88% to 96%). Individual chemotherapy agents had similarly high sensitivities, ranging from 81% (95% CI = 70% to 89%) for carboplatin to 91% (95% CI = 79% to 98%) for cyclophosphamide. Agent-specific specificities were 100%. CMS data reliably captured repeat administration of chemotherapy to within one cycle. Administrative Medicare claims data appear to be a valid source of information for chemotherapy administered to elderly Medicare beneficiaries with cancer.