A double-blind, placebo-controlled, parallel-group, multicentre study was conducted to evaluate the efficacy of pentoxifylline (Trental) in patients with multi-infarct dementia (MID) according to DSM-III-R criteria. Men and women aged 45 years or older, with a Hachinski Ischemia Scale score > or = 7 and a Mini Mental State Examination (MMSE) score of 10-25 at entry, and computed tomographic evidence of vascular disease were enrolled. A total of 289 patients were randomised to receive either oral pentoxifylline 400 mg t.i.d. or placebo for 9 months, and efficacy was assessed every 3 months. The primary outcome variable was the difference in scores between the two treatment groups, as measured on the Gottfries, Bråne, Steen (GBS) scale. Secondary outcome variables included the scores achieved on the Sandoz Clinical Assessment Geriatric (SCAG) scale and MMSE, and a battery of psychological and other tests. The intention-to-treat analysis for patients completing the study (n = 239) showed a statistically significant difference in the total GBS score in favour of pentoxifylline (improvement of 3.5 points, p = 0.028). A significant difference in the total GBS score in favour of pentoxifylline was even almost achieved in the intention-to-treat analysis for all evaluable patients (n = 269, improvement of 2.1 points, p = 0.065). It is concluded that treatment with pentoxifylline is beneficial for patients with MID, the global results of the GBS and SCAG scales being reinforced by significant improvements in those subscales specific for intellectual and cognitive function.