Quantitative liver function tests such as the determination of galactose elimination capacity (GEC) or the aminopyrine breath test (ABT) may have the potential to serve as refined entry criteria and surrogate markers for end-points in controlled clinical trials. The magnitude of a statistically detectable difference in test results and the period of observation required to document such a difference must be known to properly design such trials. Therefore, we explored retrospectively the time course of changes in GEC and ABT and their reproducibility from a cohort of patients with alcoholic cirrhosis followed for 12 to 42 months, with a median of 34 months. In 15 patients who stopped drinking, GEC improved significantly by 0.64 mg/min/kg within 1 year (mean; 95% confidence interval [CI]: 0.42; 0.86). In contrast, it deteriorated by 0.53 mg/min/kg within 1 year (95% CI: 0.32; 0.74) in another 17 patients who continued to drink (P < .01). The residual standard deviation of the changes in GEC with respect to the patients' initial values was 0.43 mg/min/kg (95% CI: 0.32; 0.52). In addition, ABT improved significantly by 0.14% dose · kg/mmol CO2 (95% CI: 0.09; 0.18) in the abstinent group, and deteriorated by 0.09% dose · kg/mmol CO2 (95% CI: 0.06; 0.13) in the nonabstinent group (P < .01). The residual standard deviation in the above sense for ABT was 0.08% dose · kg/mmol CO2 (95% CI: 0.06; 0.10). These data indicate that clinical trials with a sample size of n = 20 in each group must achieve absolute differences (ADs) in GEC of 0.6 mg/min/kg and of 0.7 mg/min/kg to reach statistical significance at the 5% and 1% level, respectively. In the present study, a period of 11 and 12 months was necessary to observe such differences. The corresponding results for the ABT are 0.11% dose · kg/mmol CO2 (9 months of follow- up; 5% level) and 0.13% dose · kg/mmol CO2 (11 months of observation; 1% level), respectively. Provided that patients with liver diseases treated with drugs are similar to the abstinent and nonabstinent patients with alcoholic liver disease investigated in this study, such numbers could serve for the planning of controlled clinical trials, in which the control group is likely to deteriorate and the treated group is expected to improve. Trials based on GEC or ABT would require only 37 or 30 patient years of observation compared with a median of 444 patient years (range, 50-2,100 patient years) reported for various published controlled clinical trials using survival analysis.