OBJECTIVE To evaluate the efficacy of retrograde balloon dilatation for primary pelvi-ureteric junction (PUJ) stenosis (RBDP) in children. METHODS Sixteen children (12 boys, four girls; mean age 29 months) with primary PUJ stenosis initially underwent RBDP. A double-pigtail catheter was left in the ureter for 6 weeks to provide drainage. The initial diagnosis of PUJ stenosis and the subsequent post-operative assessment were both made by diuretic renography in almost all patients. All children were followed up for between 9 and 38 months (mean 25) after the procedure. RESULTS RBDP was unsuccessful in nine of 17 renal units. In three of these, failure was caused by inability to pass the catheter through the PUJ or vesico-ureteric junction (VUJ). In six cases, failure was caused by recurrent or persistent stenosis. In eight of 17 renal units, RBDP was successful in relieving the stenosis. The morbidity from the procedure was minimal. CONCLUSIONS Long-term follow-up revealed a significant failure rate for RBDP, caused by the inability to pass the catheter through the VUJ or PUJ, or persistent and/or recurrent stenosis. Recurrent stenosis may result from excessive dilatation which traumatizes the PUJ and causes subsequent scarring. Failure to overcome the stenosis is thought to be caused by a large redundant renal pelvis, producing 'kinking' of the upper ureter. In these cases, only reduction pyeloplasty may be successful in relieving the obstruction.