The purpose of this study was to investigate whether marrow radiation absorbed dose estimates predict haematotoxicity following radioimmunotherapy with an yttrium-90 labelled anti-CD20 monoclonal antibody in non-Hodgkin's B-cell lymphoma (NHL). Radiopharmaceutical data from 12 NHL radioimmunotherapy patients were analysed retrospectively using three methods of marrow radiation absorbed dose estimation based on serial pretreatment indium-111 labelled anti-CD20 monoclonal antibody activity versus time data (0-144 h): (i) lumbar spine (LS) image counts; (ii) blood clearance (BL); and (iii) whole body (WB) activity. Linear regressions were performed between the methods, and between each method and the 0-6 month post-treatment platelet and white blood cell count nadir and absolute drop in count (ADC). For the range of yttrium-90 activities (740-1547 MBq), absorbed dose estimates (mean +/- sigma) were: LS, 142+/-50 cGy (range 62-233 cGy); BL, 89+/-21 cGy (range 63-140 cGy); and WB, 54+/-10 cGy (range 36-63 cGy). The LS and BL marrow estimates differed significantly (P <0.003), with a correlation coefficient r of 0.36 (P = NS), while WB correlated significantly with both LS (r = 0.50, P < 0.05) and BL (r = 0.58, P < 0.05). The range of r with platelet nadir and ADC was -0.20 < or = r < or = 0.01, except for WB with ADC (r = 0.38) (all P = NS). Values of r for white blood cell nadir were unexpectedly positive, being 0.13 for BL and 0.29 for LS (P = NS), and 0.60 for WB (P < 0.025). Values of r for white blood cell ADC were 0.36 for BL and -0.26 for LS (P = NS), and 0.50 for WB (P < 0.05). These results indicate that different commonly employed methods of estimating marrow radiation absorbed dose may yield significantly differing results, which may not correlate with actual radiation toxicity. Therefore, caution must be exercised in relying on these results to predict haematotoxicity.