Self-administered questionnaires are commonly used to measure exposures and outcomes in epidemiological research and thus need good validity. With increasing numbers of cancer survivors, there is interest in the ongoing assessment of therapy-related complications. A medical record validation of patient-reported complications following bone marrow transplantation (BMT) was performed using a self-administered questionnaire. The study population consisted of 100 patients who had undergone BMT at the City of Hope. The following self-reported complications were validated using medical records: ocular, endocrine, cardiovascular, musculoskeletal, pulmonary, gastrointestinal, neurological, graft-versus-host disease, and subsequent cancers. Using information from medical records as the standard, sensitivities ranged from 52.9% for subsequent cancers to 100% for avascular necrosis and hypothyroidism, Specificities ranged from 75.4% for ocular complications to 100% for avascular necrosis. There was intermediate to excellent agreement (kappa = 0.4-1.0) for all complications evaluated. Thus, the agreement between self-reporting and medical records was good for complications with clear diagnostic criteria that are easily communicated to the patient, but was diminished for complications with non-established diagnostic criteria (xerophthalmia) or a fluctuating course (peripheral neuropathies and hypertension). Overall these results suggest that cancer survivors can self-report serious complications with an acceptable level of accuracy in epidemiological research.