During a 5-year period (1981-86) 588 consecutive patients with nonseminatous germ cell tumors of the testis were included into a prospective Swedish-Norwegian multicenter study (SWENOTECA) and clinically staged according to the Royal Marsden system. A total of 370 patients (63%) had early clinical stages (CS) of disease; 295 (50%) had CS1, 32 (5%) had CS1Mk+ (CS1 with pathological serum tumor marker patterns after orchiectomy) and 43 (7%) had CS2A disease. Pathological staging with retroperitoneal lymph node dissection (RPLND) of the retroperitoneum was performed in 345 (93%) of the early CS patients and 128 (37%) had pathological stage 2 (PS2) disease; 27% of the CS1, 100% of the CS1Mk+ and 66% of the CS2A patients. The overall clinical staging accuracy was 75%. All the 40 patients with pathological serum AFP and/or HCG patterns before RPLND had PS2 disease, compared to 81/282 (29%) of patients with normal marker patterns. The PS2 patients with pathological marker patterns had significantly more and larger retroperitoneal metastases than those with normal AFP and HCG values. Elevated pre-orchiectomy AFP level indicated significantly reduced risk of PS2 disease in CS1 patients, but this effect became non-significant if the CS1Mk+ and CS2A cases were included into univariate or multivariate analyses. We suggest that the 'good risk' effect of pre-orchiectomy AFP elevation for CS1 cases may be caused by a selection mechanism during the clinical staging process.