Seventy-nine patients with probable Alzheimer's disease were enrolled into a double-blind, placebo-controlled cross-over study to assess the therapeutic effect and safety of THA (tetrahydroaminoacridine; tacrine) without concomitant lecithin administration. Forty-one patients completed the trial which consisted of two 12-week treatment phases separated by a 4-week wash-out period. Twenty-six subjects were withdrawn during the active treatment phase, mostly because of elevated transaminases or cholinergic side-effects, and ten during treatment with placebo. Statistical analyses were conducted on two groups of patients; those completing the cross-over and those with at least one evaluation in the first treatment period. This latter analysis, using the last observation carried forward was used to approximate an intention-to-treat analysis. THA was favoured over placebo in all three primary outcome measures (MMSE, ADAS Non-cognitive Scale, and the Functional Life Scale), but the results did not reach statistical significance. THA was favoured over placebo in five of the seven secondary outcome measures, but for only two of these was statistical significance attained. In terms of a three-point or greater increase in MMSE score, three to four times as many subjects improved on THA as on placebo.