Three surgeons performed 180 atherectomy procedures in 161 patients using the Transluminal Extraction Catheter in 144 and the Auth Rotablator in 36. The primary patency rate was 55% at 1 year and 46% at 2 years, and failure was caused by stenosis in 28 (15.6%) and occlusion in 61 (33.7%) limbs. Multivariate Cox regression analysis showed significantly better outcome if the indication was claudication, the lesion was short or there was associated stenting. Vascular laboratory surveillance was performed in 93 limbs in 83 patients. Cox regression analysis in this subgroup also showed a significant relationship between outcome and the maximum peak systolic velocity from a duplex scan at the last study performed. Receiver operating characteristics curves showed that a raised maximum peak systolic velocity best predicted late failure (sensitivity 84%, specificity 66% for > 200 cm/s; sensitivity 72%, specificity 84% for > 250 cm/s); the velocity ratio at the stenosis to that in the segment above or the resting ankle/brachial pressure index were less predictive. For 50 procedures studied in the vascular laboratory which remained successful to the end of the study, maximum peak systolic velocities were > 250 cm/s from the first postoperative study, suggesting residual stenosis in 6%, or increased to become > 250 cm/s by the last study, suggesting recurrent stenoses in 12%. For 43 procedures which were studied and later failed, velocities were > 250 cm/s from the first test in 26% or increased to > 250 cm/s by the last test before failure in 40%. Vascular laboratory surveillance helps to predict outcome after atherectomy. Failure may be a result of residual disease from the time of the procedure or from restenosis. The apparent high incidence of clinically manifest or developing stenoses raises doubts as to the benefit of atherectomy over balloon dilatation alone.