With the use of a multivariate regression model, 21 patients with diffuse lymphoma were identified as having greater than 15% risk of experiencing relapse in the central nervous system (CNS). The efficacy of a combination of sequential intravenous methotrexate (MTX)(l g/m2) and intrathecal MTX in preventing relapse in the CNS and improving survival times was assessed. A comparable historical control group of patients with similar risk of CNS relapse and who did not receive any CNS prophylaxis was used. The CNS relapse-free survival rate (RFS) was improved in patients who received CNS prophylaxis (95% versus 59% at 2 years; P = 0.01). Pretreatment serum lactic dehydrogenase (LDH) levels correlated with the incidence of relapse in the CNS in the control group (P = 0.01). In patients with high pretreatment serum LDH levels (>225 U/L), CNS RFS was improved in those who received CNS prophylaxis (RFS at 2 years: 91% versus 46%; P = 0.02). Both CNS RFS (100% versus 38% at 2 years; P = 0.03) and survival rates (100% versus 38% at 2 years; P = 0.02) were improved in six patients with histologic type other than large cell. In 15 patients with large cell lymphoma, no significant differences in CNS RFS (93% versus 75% at 2 years; P = 0.29) and survival rates (43% versus 44% at 2 years; P = 0.56) were observed. Cerebrospinal fluid MTX levels were above the therapeutic level of 1 X M for at least 20 hours in 90% of courses of combined MTX. The MTX combination used is an effective and non-neurotoxic CNS prophylaxis method. Because the comparison between different methods of CNS prophylaxis is difficult to make without a precise idea of the expected CNS relapse rate, use of the multivariate regression technique is recommended.