Induction of protective hemagglutination-inhibition (HI) antibodies in response to influenza virus vaccine and the effectiveness of two doses versus a single dose of vaccine were studied in 48 BMT recipients. The patients were 1-50 years old (median 21 years), 33 with malignant and 15 with non-malignant disease. Thirty-five of the patients underwent allogeneic, T lymphocyte-depleted, BMT and 13, autologous BMT. Nine patients had GVHD at initial immunization. The time interval from BMT to influenza vaccination ranged from 2 to 82 months (median 14.5 months). Two doses of vaccine, administered 1 month apart, consisted of trivalent influenza subunit inactivated vaccine with the following strains: A/Singapore/6/86 (H1N1), A/Sichuan/2/87 (H3N2), and B/Beijing/1/87. There was a statistically significant association between development of protective antibody level (> or = 1:40) and the time interval between BMT and initial vaccination (p < or = 0.001). Regression analysis revealed that longer time interval between the BMT and immunization was positively correlated with seroconversion (a fourfold or greater rise in titers). In the presence of GVHD, there was reduced seroconversion to H1N1, but not to H3N2 or B strains. Influenza vaccination within the first 6 months following BMT was totally ineffective. The efficacy of the vaccine was similar to that described in non-immunocompromised hosts initiated 2 years following BMT. As, overall, specific response was only marginally enhanced by the second dose of vaccine, its indication is questionable.