The objectives of this study were to scrutinize surgical feature and analyze local tumor parameters of early cervical cancer to identify patients at-risk for recurrent disease. Three hundred eight patients who underwent radical hysterectomy and pelvic lymphadenectomy between 1984 and 1997 were studied retrospectively. All radical hysterectomies were performed in a referral oncology center, and treatment policies and operating staff were the same during the study period. Operating time gradually decreased significantly during the study period from an average of 270 min to an average of 187 min (P < 0.0001), and blood loss during surgery also decreased continually from 1515 ml to 1071 ml (P < 0.0001). Postoperative radiation treatment was given to 119 patients (40%). The overall five-year survival rate was 83%, 91% for those with negative, and 53% for those with positive pelvic nodes. Univariate analysis showed that lymph node status, parametrial involvement, status of the surgical margins, capillary lymphatic space involvement, tumor size and depth of invasion were all significantly related to the occurrence of recurrent disease. Multivariate analysis revealed that lymph node involvement (hazard ratio 4.4), parametrial involvement, tumor size and depth of invasion were independent factors of prognostic significance for disease-free survival. It was concluded that the local control of cervical tumors infiltrating > 10 mm (hazard ratio 5.1) might be improved by adjuvant radiotherapy, even in the absence of lymph node metastasis, parametrial involvement or affected surgical margins.