One hundred and eighteen consecutive adults with newly diagnosed intermediate and high-grade non-Hodgkin's lymphoma were treated with chemotherapy comprising Doxorubicin, Cyclophosphamide, Vincristine and Prednisolone with mid-cycle Methotrexate (MTX) and leucovorin rescue ('CHOP-M'). Intrathecal MTX was given with each treatment cycle as central nervous system (CNS) prophylaxis. 'Clinical remission' was achieved in 70/110 evaluable patients (64%), complete remission: 45/110, (41%), good partial remission: 25/110 (23%). Twenty two patients (19%) died prior to completion of therapy, 18 patients had persistent disease. Hyponatremia (< 137 mmol l-1), advanced age and hypoalbuminaemia (< 32 g l-1) correlated adversely with achievement of CR (P = 0.0007, 0.0005 and 0.04 respectively). With a minimum follow up of 41 years, 47 of the seventy patients (67%) in whom clinical remission was achieved remain well, 19 have developed recurrent disease, resulting in an actuarial projected remission duration of 70% at 8 years. Four died in CR. There has been only one isolated CNS recurrence. On univariate analysis, hypoalbuminaemia, hyponatremia and beta 2 microglobulin (> 3) correlated with unfavourable outcome in terms of duration of remission (P = 0.0009, 0.007 and 0.04 respectively). On multivariate analysis, only the serum sodium (0.002) and advanced age (0.01) were predictive for remission duration. Fifty patients (45%) are alive, the overall actuarial projected survival is thus 42% at 8 years. On univariate analysis, age, hypoalbuminaemia, hyponatraemia, liver involvement and the presence of B symptoms correlated unfavourably with survival. On multivariate analysis, hypoalbuminaemia, advanced age, hyponatraemia, male gender (aged > 50) and diffuse large cell or large cell, immunoblastic histology (Working Formulation) had an adverse effect (P = 0.003, < 0.0001, < 0.0001, 0.002, and 0.03). It was further possible, using cut-off points of 32 g l-1 and 136 mmol l-1 for albumin and sodium respectively to define prognostic categories for patients who differed significantly in terms of survival.