We prospectively followed a group of unimmunized, immunosuppressed children with cancer to determine their relative risk of influenza and the severity of infection compared with those of siblings or matched community controls. The incidence of influenza infection was higher in children with cancer (23/73, 32%) than in control subjects (10/70, 14%, p = 0.02). A preseason hemagglutination inhibition titer greater than or equal to 1:32, generally used as a marker of successful immunization in vaccine trials, was protective for all children in the control groups, but did not prevent influenza infection in 24% of the patients with cancer. Infection rates of patients and community controls with titers greater than or equal to 1:32 differed significantly (p = 0.006). No significant differences were noted in duration of reported symptoms between groups, and clinical complications occurred too infrequently to analyze. However, 2 (11%) of 18 of the cancer patients with positive culture results were hospitalized during the illness and one patient developed a nosocomial infection. None of the control children was hospitalized. These findings suggest the need for further study of the immunologic response of immunosuppressed children to influenza infection and a clinical efficacy trial of the influenza vaccine in these patients.