OBJECTIVE To test the Gynecologic Oncology Group (GOG) prognostic criteria (based on stromal invasion, tumor size and vascular invasion) for early squamous cervical carcinoma (SCC) in an independent population and to evaluate the prognostic value of a simpler model. METHODS We studied 221 patients who underwent radical hysterectomy and bilateral lymphadenectomy for stage IB SCC between 1987 and 1993. Adjuvant treatment consisting of radiotherapy and/or chemotherapy was given in case of large tumor size, positive lymph nodes or invasion into the parametria. Histological slides from all patients were reviewed by one pathologist. RESULTS The GOG criteria divided the patients from our population in a small low risk group (3-year relapse-free rate (RFR) of 100%), a small high risk group (RFR of 57%) and a bigger intermediate risk group (RFR of 80-90%). These results are in good agreement with those of the original publication. A risk model based on 2 of the 3 factors used by the GOG may perform as well as the 3-factor model, especially when allowing interaction. Tumors measuring >2 cm and invading into the outer third of the cervical wall had a 5-year DFS of 56% compared to 93% for tumors measuring < or =2 cm or invading to less then the outer third of the cervical wall. CONCLUSIONS The GOG criteria could be validated in this independent population. A model based on 2 of the 3 factors may perform as good.