OBJECTIVE To compare the effect on the recovery of incontinence after retropubic radical prostatectomy (RRP) of intensive physiotherapist-guided pelvic floor muscle exercises (PG-PFME) in addition to an information folder, with PFME explained to patients by an information folder only (F-PFME), and to determine independent predictors of failure to regain continence after RRP. PATIENTS AND METHODS We postulated that a 10% increase in the proportion of men who regained continence at 6 months with PG-PFME compared with men treated with F-PFME only would constitute a clinically relevant effect. To show statistical significance of this difference with a power of 80%, 96 men should be randomized to each of the two arms. One day before operation, all patients received verbal instruction and an information folder on PFME. Patients randomized to the F-PFME arm received no further physiotherapist guidance, whereas those in the PG-PFME arm received a maximum of nine sessions with the physiotherapist. The men underwent a 1-h pad-test at 1, 12 and 26 weeks, and a 24-h pad-test at 1, 4, 8, 12 and 26 weeks after catheter removal. We defined 'continence' as urine loss of < 1 g at the 1-h and < 4 g at the 24-h pad-test. RESULTS During the 2-year recruitment period, the number of patients randomized fell short of the target determined by the sample size calculation, because of limitations of resources and unexpected changes in treatment preferences. Despite this, we analysed the data. Of the 82 randomized patients, 70 completed the study. Of these, 34 and 36 men had been assigned to the PG-PFME and the F-PFME group, respectively. At 6 months after RRP, 10 (30%) and nine (27%) men were completely dry on both the 1-h and 24-h pad-test in the PG-PFME and the F-PFME group, respectively (difference not significant). In a multivariate analysis the amount of urine loss at 1 week after catheter removal seemed to be an independent prognostic factor for failure to regain continence. CONCLUSION PG-PFME seems to have no beneficial effect on the recovery of continence within the first 6 months after RRP, over an instruction folder-guided approach. However, due to under-powering there is a high risk of type II error. Nevertheless, these findings add to the knowledge base for availability in meta-analyses and can serve as a starting point for the design of new randomized studies.