UNDERSTANDING THE CLINICAL LIMITS OF LAPAROSCOPIC MYOMECTOMY: RESULTS FROM AN ARTIFICIAL NEURAL NETWORK ANALYSIS

Objective to evaluate whether the pre-operative patient’s and myoma’s characteristics may predict the overall peri-operative outcomes and the overall clinical benefit of the surgical approach. Design Observational cohort study Setting Single-center University Hospital Population or Sample Consecutive series collected over a 10-year period Methods and Main Outcome Measures Patients submitted to laparoscopic (LM) and open myomectomy (OM). Multivariable (MLR) and Artificial Neural Network (ANN) analyses were performed to identify the predictors of conversion to laparotomy or impaired outcomes. Results 399 (78.1%) and 112 (21.9%) patients had LM and OM; 27 (6.8%) case were converted to OM. Seventy-one (17.8%) recorded overall outcomes at least comparable to OM (Operative time ≥ 120 min and Blood Loss ≥ 300 mL and hospitalization ≥3 days combined) or received conversion to OM (OOCO group). Dominant myoma size ≥8cm [OR 5.2;p<0.0001], anterior location [OR 3.0;p=0.012], intramural type [OR 2.5; p=0.039] and age≥40years [OR 2.1;p=0.08] were all independent predictors of conversion to laparotomy at MLR and ANN analyses. Myoma size ≥8cm [OR 4.4;p<0.0001], number of myomas ≥3 [OR 2.4;p=0.004], age≥40years [OR 2.3;p=0.003], intramural type [OR 2.2;p=0.008] and vaginal delivery [OR 1.6;p=0.045] were all independent predictors of OOCO at MLR and ANN analyses. Conclusions in case of LM, both patient’s and myoma’s characteristics may predict the risk of conversion or of overall outcomes similar to laparotomy. This information could be taken into account to better address the preoperative counseling and the surgical planning based not only on the technical feasibility, but also on the overall clinical benefit.