Purpose: Endometrial carcinoma is the most common malignancy of the female genital tract. Recognized prognostic factors include FIGO stage, histologic grade, depth of myometrial invasion, and age. Although determination of these factors may seem clear and reproducible, the histologic grade has recently been the subject of debate. A retrospective analysis of long-term outcome and predictive factors in endometrial carcinoma was conducted, focusing on the prognostic value of tumor grade.
Materials and methods: The study included 253 patients with endometrial carcinoma Stages I to III, who were treated between 1984 and 1993. The histologic slides were reviewed and the prognostic value of stage, age, myometrial invasion (depth and pattern), tumor grade, and histologic subtype was analyzed. The end point was cancer-specific death; the median follow-up time was 11.7 years.
Results: The actuarial 5- and 10-year cancer-specific survival rates (CSS) were 85% and 82%, respectively. Five-year vaginal and/or pelvic recurrence and distant relapse rates were 7% and 15%. In multivariate analysis, stage, pattern of myometrial invasion, tumor grade, and age were independent prognostic factors. At pathology review, a shift from Grade 2 to Grade 1 was seen in 112 of the original 144 Grade 2 (78%). There was no difference in CSS between Grade 1 and Grade 2 (94 vs. 90% for original grade and 92 vs. 95% for grade after review), whereas Grade 3 was found to be a significant adverse prognostic factor (p < 0.001).
Conclusions: The independent prognostic factors for patients with endometrial cancer were stage, pattern of myometrial invasion, tumor grade, and age. Systematic grading led to a considerable shift from Grade 2 to Grade 1. However, there was no difference in prognostic significance between Grade 1 and 2, whereas Grade 3 was a major adverse prognostic factor. A two-tiered grading system, instead of the currently used three-tiered system seems preferable, because it has a better correlation with clinical outcome and is expected to have less interobserver variability.