We have studied determinants of outcome of 7 day courses of treatment in 77 middle aged and elderly patients, in whom the general practitioner's diagnosis of urinary tract infections had been confirmed microbiologically. Bacteria were sensitive to cephalexin or trimethoprim. Where there was no preference, treatments were allocated randomly. Compliance was monitored using a pill box with a concealed electronic device which recorded openings of the box. 2. Prescribing trimethoprim, 200 mg twice daily, was more effective than cephalexin, 250 mg four times daily (cure rates 93 and 67%) (P less than 0.006). Those cured and not cured were not distinguished by age, gender, genitourinary history, or infecting organism. 3. Compliance as measured by box openings was worse for cephalexin than for trimethopim (P = 0.01). However, both totality and pattern of compliance were similar in patients cured and not cured by cephalexin. Thus rigid adherence to a conventional course did not promote cure: fewer doses could have been prescribed. 4. Estimating compliance is essential to clinical trials where medication is self-administered. Poor compliance may establish over exacting regimens. Counting box openings did overestimate compliance, but counting residual tablets overestimated it grossly: given the number of openings less than the ideal, there should have been 171 residual tablets, only 55 were found.
