BACKGROUND: The severity of myelotoxicity after radioimmunotherapy has been predicted from body and blood radiation doses to marrow. However, marrow radiation can be increased substantially if the marrow or skeleton contains the malignancy targeted by the radiolabeled monoclonal antibodies. A study of 29 patients treated with iodine-131 (131I)-Lym-1 showed that radiation doses to marrow from body and blood had little correlation with myelotoxicity. The purpose of the present study was to assess the significance of marrow targeting and other factors for prediction of myelotoxicity. METHODS: Injected radioactivity and nontargeted radiation doses to marrow were compared with peripheral blood cell counts after the first therapy dose of 131I-Lym-1 in 16 heavily pretreated patients with non-Hodgkin's lymphoma (NHL). Bone marrow biopsy, targeted marrow radiation doses, marrow image uptake scores, age, Karnofsky performance score (KPS), previous chemotherapy, and tumor burden were also compared with blood counts. RESULTS: Myelotoxicity was not predicted well by injected radioactivity, total body radiation, or body and blood radiation doses contributed to marrow (P > 0.1). Biopsy-proven bone marrow lymphoma also failed to predict myelotoxicity (P > 0.1). Thrombocytopenia and leukopenia were predicted well by targeted radiation dose to marrow (P < 0.05) obtained by 131I imaging. Similarly, marrow image scores predicted decreases in platelets and white blood cells (WBCs; P < 0.05). Prediction of myelotoxicity using marrow radiation dose methods was slightly improved when serum lactic dehydrogenase (LDH), age, KPS, and prior chemotherapy were included in the analysis (P < or = 0.01). CONCLUSIONS: Prediction of myelotoxicity was improved in this group of patients by assessment of the targeting component of marrow radiation and was better predicted and obtained more easily by semiquantitative marrow image scores. Further improvement in prediction was slight when other factors were considered.