To evaluate the efficacy and safety of high-dose tacrine hydrochloride over 30 weeks in patients with probable Alzheimer's disease. A 30-week randomized, double-blind, placebo-controlled, parallel-group trial. Outpatients at 33 US centers. Men and women at least 50 years of age with mild to moderate Alzheimer's disease and otherwise in good health. Group 1 received placebo; group 2 received 40 mg/d of tacrine for 6 weeks, then 80 mg/d for 24 weeks; groups 3 and 4 received 40 mg/d of tacrine for 6 weeks, 80 mg/d for 6 weeks, and 120 mg/d for 6 weeks. Group 3 remained on a dosage of 120 mg/d for a total of 18 weeks; after 6 weeks at 120 mg/d, group 4 titrated to 160 mg/d for the last 12 weeks. Clinician Interview-Based Impression (CIBI), Alzheimer's Disease Assessment Scale--Cognitive subscale (ADAS-Cog), and Final Comprehensive Consensus Assessment (FCCA). A total of 663 patients entered the study; 653 patients were included in an intent-to-treat (ITT) analysis; 263 had evaluable data at 30 weeks. The results of the ITT analysis revealed significant (P < or = .05) dose-response trends and between-group comparisons on CIBI and ADAS-Cog. In evaluable patients, significant dose-response trends were observed for all three primary measures (P < or = .001). Significant differences in favor of 160 mg/d of tacrine vs placebo were observed on the CIBI (P < or = .002) and ADAS-Cog and FCCA (P < or = .001), as well as caregiver-global and quality-of-life assessments (P < or = .05). On the CIBI, 23% and 42% of tacrine-treated patients in the ITT and evaluable-patient populations, respectively, were rated improved compared with 17% and 18% of placebo patients, respectively. The primary reasons for withdrawal of tacrine-treated patients were asymptomatic liver transaminase elevations (28%) and gastrointestinal complaints (16%). These adverse events were reversible on discontinuation of treatment, and many patients were able to restart tacrine. Tacrine produced statistically significant, dose-related improvements on objective performance-based tests, clinician- and caregiver-rated global evaluations, and measures of quality of life. There was no evidence that the large number of patient withdrawals biased the overall conclusions of the study.