The authors evaluated clinical and pathologic factors that predicted for recurrence after patients underwent radical surgery for International Federation of Gynecology and Obstetrics (FIGO) Stage IA2–IB1–2 cervical carcinoma and developed a simple method of scoring those predictive factors to quantify outcome.
An analysis was conducted of a prospective radical surgery cervical carcinoma data base. A Cox proportional hazards regression analysis was done for each of the individual factors to estimate individual risk ratios using all available data for each factor. Stepwise and best-model options were used to identify the best combinations as predictors and to calculate adjusted risk ratios. Based on the information obtained, each patient was assigned a categorical score to predict recurrence. The variables used for the score were dichotomized. The differences between the scores in time to recurrence were evaluated using the log-rank test to compare the time to recurrence curves that were generated with the Kaplan–Meier method.
Eight hundred seventy-one patients were included in the study, and 66 patients who developed recurrent disease after a median follow-up of 49 months. Tumor size, maximum depth of invasion, pelvic lymph node status, tumor grade, and capillary lymphatic space (CLS) were single predictors for recurrence, and the score, which was based on combinations of these factors, predicted the disease free survival. Maximum depth of invasion, pelvic lymph node status, and CLS were the best combined predictors for recurrence, and they were used to form a second, precise scoring system to predict disease free survival (P < 0.0001; log-rank test).
Radical hysterectomy has been demonstrated as a viable method of treatment for patients with Stage IA–IB carcinoma of the uterine cervix. Despite advances over the past 3 decades in decreasing the morbidity of treatment, the cure rate associated with radical surgery (approximately 90%) has not changed appreciably.1 Traditionally, patients were chosen for adjuvant radiotherapy on the basis of lymph node metastasis as a single variable. However, because such an approach failed to identify approximately 40–50% of patients who were destined to develop recurrent disease,2, 3 it was found that the use of a combination of factors was more reliable.4–6 These factors include tumor size, cell type, tumor grade, the presence of capillary lymphatic space (CLS), the depth of invasion, and lymph node status.1, 4–10
We analyzed our prospective data base of patients with cervical carcinoma who underwent radical surgery to identify prognostic predictors of disease recurrence after surgery. This served as a basis for a proposed scoring system to predict recurrence.
MATERIALS AND METHODS
Since July 1, 1984, all patients who underwent radical surgery for International Federation of Gynecology and Obstetrics (FIGO) Stage IA2/IB1–2 cervical carcinoma by members of the Division of Gynecologic Oncology at the University of Toronto have been recorded prospectively (all pertinent data, including follow-up) and entered into a computerized data base. The management philosophy and technical details of surgery at our institution during this period for patients with FIGO Stage IA2–IB1–2 cervical carcinoma have been reported previously.11–13 Radical vaginal trachelectomy and laparoscopic pelvic lymphadenectomy were offered to patients with tumors measuring ≤ 2 cm who wanted to preserve fertility.12 Adjuvant postoperative therapy was prescribed as described previously published.14
Prognostic pathologic parameters were identified as part of the data base and were recorded prospectively by gynecologic pathologists. Specimens from cervical cones, specimens from large-loop electrosurgical removal of transformation zone, and radical hysterectomy or radical trachelectomy specimens were used for pathologic evaluation. Definitions of maximum tumor size, CLS, and maximum depth of invasion also have been published previously.12
Data for this study came from the University of Toronto Cervical Cancer data base. This data base is updated regularly in Medlog (Information Analysis Systems, Crystal Bay, NV). For the purposes of analysis, the data was exported into SAS software (SAS Institute, Cary, NC). An inverse Kaplan–Meier analysis was used to calculate the median follow-up (censoring death patients).
Predictors for disease recurrence included patient age, tumor grade, presence of CLS, depth of invasion, margins, and lymph node metastases. A Cox proportional hazards regression analysis was performed on each of the individual factors to estimate individual risk ratios. Cox proportional hazards regression using stepwise and best-model options identified the best combinations as predictors and obtained adjusted risk ratios. Based on the information obtained, each patient was assigned a histopathology (categorical) score to predict recurrence. Several ways of combining the information obtained from the Cox proportional hazards regression analysis were used to identify the scoring system that would best differentiate between the groups. For example, the hazards ratios of the variables identified were used as weights in one of the scoring systems. The variables used for the histopathology scores were dichotomized. The difference between the histopathology scores in time to recurrence was evaluated using the log-rank test to compare the time to recurrence curves generated with the Kaplan–Meier method.
Eight hundred seventy-one patients with Stage IA2–IB1–2 cervical carcinoma underwent bilateral pelvic lymphadenectomy in conjunction with radical abdominal hysterectomy, radical vaginal hysterectomy, or radical trachelectomy as the primary treatment between July 1984 and January 2000. After a median follow-up of 49 months, (range, 0.1–166.2 months), 66 patients developed recurrent disease in the following sites: pelvis, 40 patients; distant sites, 13 patients; and pelvis plus distant sites, 13 patients. The median time to recurrence was 14 months, and 67% of patients developed recurrent disease within 2 years of surgery. The demographic and pathologic data of the patients are summarized in Table 1.
Tumor size, maximum depth of invasion, pelvic lymph node status, tumor grade, and CLS were single predictors for recurrence both in adjusted and unadjusted Cox proportional hazards regression analyses (α ≤ 0.05) (Table 2). To calculate the scores, variables were categorized as follows: maximum depth of invasion, < 10 mm versus ≥ 10 mm; tumor grade, Grade 1 versus Grade 2–3; pelvic lymph nodes, positive versus negative; and CLS, positive versus negative. The first scoring system used all of the single predictors identified above (Fig. 1). Because maximum depth of invasion, pelvic lymph node status, and CLS were the best combined predictors for recurrence (Table 3), a second scoring system used these three predictors with equal weight (Fig. 2). A third scoring system used hazard ratios as weights (data not shown; the results were very similar to the second scoring system, possibly due to the large confidence intervals for the hazard ratios). The 5-year recurrence free survival for each score (0–3, respectively) was 0–97%, 1–86%, 2–80%, and 3–65%. The proportion of patients who received postoperative adjuvant radiation therapy with or without chemotherapy (a confounder) was 3.2% (score = 0), 15.6% (score = 1), 33.6% (score = 2), and 66.7% (score = 3).
Forty-one of 553 patients developed recurrent disease.
Positive lymph nodes
Maximum depth of invasion
Primary surgery is successful as the sole therapeutic modality for the majority of patients with early-stage cervical carcinoma. Historic attempts to improve patient outcome have focused mainly on the use of adjuvant pelvic radiation (with or without chemotherapy), with modest success in decreasing locoregional recurrence and without a significant effect on survival.2, 15 The reliable identification of various pathologic risk factors that portend a higher rate of recurrence and decreased survival after primary surgical management of patients with early-stage cervical carcinoma–3, 16 is an important basis for the investigation of postoperative adjuvant therapies to improve outcome.
Pathologic findings that previously have been identified with poor clinical outcome include tumor size, depth of cervical stromal penetration, histologic cell type, tumor grade, vascular space invasion, positive surgical margins, and lymph node metastases.1–3, 16 The risk of tumor recurrence in a given patient population exists as a continuous spectrum influenced by multiple, interrelated elements. This may explain why individual pathologic features or varying combinations of such features had diverse predictive values among different investigators.
Failure analysis of patients with Stage IA2–IB1–2 cervical carcinoma after radical surgery is important to help identify adjuvant treatment strategies. These patients generally are young and healthy. The appropriate selection of patients for primary surgical treatment usually results in a favorable outcome. To date, several studies have attempted to identify patients who may require adjuvant therapy after primary surgical intervention. However, small numbers and/or the retrospective nature of data collection limit those results.1, 4, 6, 7, 9, 10, 17, 18 Therefore, the current study is novel in three ways: the comprehensive prospective nature of data collection, the large number of patients treated in a single institution, and the quantification of the risk of recurrence based on a simple scoring method. The only prospective data published to date were from a multicenter study,1 which was subject to differences in surgical technique at different centers. Our center is the only gynecologic oncology referral center for the Toronto area. This also allowed uniform management of patients with respect to decision making, surgical procedure, pathologic evaluation, and postoperative radiotherapy.
In the current study, we used these results to create a scoring system for predicting the outcome of patients with early-stage cervical carcinoma who undergo surgery. We were able to show that each adverse prognostic factor indicated a slight decrease in patient outcome. However, we found that a combination of three factors was especially deleterious to recurrence free survival. The proposed prognostic scoring system is easy to apply and requires only three factors (depth of invasion, CLS, and lymph node status).
A confounding factor in our analysis was the addition of postoperative adjuvant radiation therapy with or without chemotherapy in patients with poor prognostic factors. Several studies have shown that the administration of postoperative adjuvant chemoradiation therapy in patients with poor prognostic factors—like positive or close surgical margins, positive pelvic lymph nodes, the presence of CLS, poorly differentiated tumor, and deep stromal invasion—is beneficial, demonstrating a lower pelvic failure rate.15, 19–22 New adjuvant therapeutic approaches are needed for these patients to improve their outcome. The proposed scoring system in this study may serve as a base for such studies; however, validation by others with similar data will be important.