A simple model for predicting nodal metastasis in patients with clinical stage I nonseminomatous germ cell testicular tumors undergoing retroperitoneal lymph node dissection only.

PURPOSE: In patients with clinical stage I nonseminomatous germ cell testicular tumor the identification of risk categories for nodal metastases and/or distant metastases could permit selective management. We built 2 models for distinguishing these risk categories.
MATERIALS AND METHODS: Data on 322 consecutive patients with clinical stage I nonseminomatous germ cell testicular tumor patients treated with retroperitoneal lymphadenectomy (RPLND) alone between 1985 and 1995 were analyzed. The interval between orchiectomy and RPLND, vascular invasion (VI), pT stage, percent embryonal carcinoma (ECa) and teratoma in the primary tumor were considered clinically relevant for their association with nodal or distant metastases. Two logistic models were constructed. Model 1 was meant to discriminate 2 patient categories, namely those with and without nodal metastases at RPLND. Model 2 was meant to discriminate 3 patient categories, namely those without any metastases, with nodal metastases only and with distant metastases independent of retroperitoneal metastases. The models were based on these above variables, which were inserted as categorical and then processed through a backward selection procedure.
RESULTS: At RPLND nodal metastases were found in 60 patients (18.6%). During followup distant metastases were observed in 43 patients (13.4%) and retroperitoneal recurrences were noted in 6 (1.9%). Of all recurrences 93.8% were within 2 years since RPLND. RPLND had a high curative rate since 73% of all pN+ cases were cured by surgery alone. The final logistic model 1, including percent ECa and VI, was reassessed in 202 patients with available data. Absent VI and ECa 90% or greater identified a category of 110 patients at low risk for nodal metastasis (14%), while VI and/or ECa greater than 90% identified a category of 92 at higher risk (35%). The identified categories were also related to distant metastases, which occurred in 9.3% of low risk and in 23.1% of high risk cases. Model 2 was not clinically suitable because it did not allow us to distinguish patients at risk for nodal metastases only from those at risk for distant metastases.
CONCLUSIONS: Simplicity is the main advantage of model 1 since only 2 well-known prognostic parameters are involved. Although the model must be validated in an independent case series, the identification of a low risk category with few expected nodal metastases could permit us to replace traditional RPLND with a less invasive staging procedure.