We present 8 years' experience of primary retroperitoneal lymph node dissection (RLND) in 190 patients with low stage non-seminoma; 154 patients had clinical stage I (CSI) and 36 had clinical stage IIa (CSIIa) disease. Of the 154 patients with CSI tumours, 33 had increased serum AFP and/or HCG before RLND (CSIM+) and 121 had normal tumour markers (CSIM-). Retroperitoneal lymph node metastases (pathological stage II) (PSII) were found in 38 of 121 patients with CSIM-, in 19 of 33 patients with CSIIM+ and in 26 of 36 patients with CSIIa. In a multivariate analysis, the presence of small vessel infiltration (demonstrated in histological sections of the primary tumour) and a prolonged tumour marker half-life were predictive factors for PSII. These 2 factors enabled a group of non-seminoma patients with CSI disease to be identified who had a 15% risk of retroperitoneal tumour growth (low risk group) as compared with a high risk group where 60 to 70% of patients had retroperitoneal lymph node metastases. Relapses occurred in 7 of 107 patients with PSI and in 6 of 83 patients with PSII disease; in the latter group, 5 relapses developed before the start of routine adjuvant chemotherapy; 6% of patients developed major post-operative complications. In addition, "dry ejaculation" was the principal side effect following RLND (unilateral RLND: 20/132 patients; bilateral RLND: 50/54 patients). The comparative cost to the health service during the first year of follow-up was estimated for low risk non-seminoma patients with CSI subjected to RLND and for those in whom a surveillance policy was adopted. The latter approach was preferable. It was concluded that a surveillance policy should be followed in low risk non-seminoma CSI patients provided that frequent follow-up is possible. A more active policy is recommended in high risk patients (e.g. adjuvant chemotherapy without RLND). Nerve-sparing RLND may be considered in patients with CSIIa disease and negative tumour markers.