To investigate whether molecular classification could support individualized selection of patients with grade II endometrioid endometrial cancer for fertility-sparing approaches. This is a retrospective multi-institutional study. Data of patients undergoing fertility-sparing treatment with a levonorgestrel intrauterine device (with or without oral hormonal therapy) were retrieved. Surrogate molecular classification was used to categorize patients into 4 classes: (1) POLE-mutated, (2) mismatch repair deficient/microsatellite instability high, (3) p53 abnormal, and (4) no specific molecular profile. Overall, data from 23 patients with grade II endometrioid endometrial cancer starting a fertility-sparing attempt were retrieved. The median patient age was 36 (range; 30-41) years. All patients underwent hysteroscopic-guided endometrial biopsies. Hysteroscopic resection of the tumor was performed in 9 (39.1%) patients. According to surrogate molecular classification, 1 (4.3%), 2 (8.7%), 3 (13.1%), and 17 (73.9%) patients were classified as POLE-mutated, p53 abnormal, mismatch repair deficient/microsatellite instability high, and no specific molecular profile, respectively. Molecular classification was not associated with 6-month response rates (p = .080) nor with best response rates (p = .366). Overall, 7 women attempted to achieve a pregnancy; 3 underwent in vitro fertilization. Three patients achieved a pregnancy (1 first-trimester miscarriage and 2 term live births). A total of 10 (43.4%) patients were diagnosed with progressive disease during hysteroscopic surveillance. Overall, 19 (82.6%) patients required hysterectomy. Three (13%) patients required adjuvant therapy for the presence of locally advanced disease (1 stage II, and 2 stage III). Over a median (range) follow-up of 28.7(8.2-91.1) months, no recurrence or disease-related death occurred. Less than 10% of women with grade II endometrioid endometrial cancer starting a fertility-sparing attempt achieve a term pregnancy. In our series, molecular classification did not influence response rate. Further collaborative registers are needed.
Bogani et al. (Tue,) studied this question.