In studying patients with mild hypertension, we used a double-blind, randomly allocated crossover, 2 x 2 factorial design to obtain more precise estimates of the effectiveness of individual hypertensive drugs and of the interactions when used in combination. In the present study, we reanalysed data from our own previously completed crossover factorial studies in order to determine the optimal length of phases and to try to exploit patient heterogeneity by comparing blood pressure responses within each patient to the two different drugs in each study. An analysis of the 'weeks within phases' variation (analysis of variance) applied to all phases, to the placebo phases alone and to the active treatment phases alone, of the individual factorial studies and of pooled data from studies of similar design, only revealed differences in the first week of our 4-week phases. Accordingly, we suggest that 4 weeks is an optimal phase length, since only the values from the third and fourth weeks showed the full expression of a drug's antihypertensive effect, avoiding the problem of a carryover effect and the need for a more complex intervening washout placebo phase. Comparing the blood pressure reductions shown by individual patients in response to different antihypertensive drugs using arbitrary criteria of greater than or equal to 5, 10 and 15 mmHg for diastolic and greater than or equal to 5, 10 and 20 mmHg for systolic pressure, we were unable to distinguish between the two separate drugs in each factorial study, due to the magnitude of the background variation which had a coefficient of variation of 5-7%.(ABSTRACT TRUNCATED AT 250 WORDS)