Predictors of ≥50% volume reduction after radiofrequency ablation of uterine fibroids: a single-center retrospective case series in Romania

TL;DR

This study explores how pre-procedure data can predict successful radiofrequency ablation outcomes for uterine fibroids, suggesting potential for future predictive models.

Contribution

The paper introduces a proof-of-concept for using machine learning with clinical data to predict RFA outcomes for fibroids.

Findings

Fibroid volume decreased by an average of 54% after RFA treatment.

Doppler score decreased by 59.8% post-procedure.

No significant association was found between treatment outcome and age or FIGO classification.

Abstract

The study aimed to explore the potential of routinely collected pre-procedural clinical and ultrasonographic parameters (age, classification, baseline fibroid volume, and baseline Doppler score) in predicting treatment response following transvaginal radiofrequency ablation (RFA). We employed a combined machine learning-based feature ranking and regression interpretability approach to provide a proof-of-concept for future data-driven predictive models.

This study was designed as a single-center retrospective case series.

The study was conducted at a public hospital in Romania, where RFA was introduced as a minimally invasive alternative for uterine fibroid treatment.

Twenty-two fibroids were treated in women aged 28–52 years (mean age 38), who were selected based on the presence of symptoms or documented fibroid growth.

Transvaginal ultrasound-guided radiofrequency ablation was performed using the VIVA RF system. Baseline and post-procedural measurements included fibroid size and volume, vascularity as measured by Doppler score, and anatomical classification according to FIGO criteria.

Fibroid volume decreased from a mean of 60.82 cm3 to 28.3 cm3 (−54.0%), while the Doppler score decreased from 3.59 to 1.41 (−59.8%). An exploratory feature-based analysis using four pre-procedural variables (age, FIGO classification, baseline fibroid volume, and baseline Doppler score) was conducted as a proof-of-concept, highlighting the challenges of robust modeling in small cohorts. In the multivariable logistic regression analysis, no statistically significant association was observed between treatment outcome and patient age or FIGO classification.

This proof-of-concept study demonstrates the potential of using routinely collected pre-procedural data for predictive analytics in RFA of uterine fibroids. The successful outcomes observed in two submucosal fibroids (G0 and G1) suggest that RFA followed by hysteroscopic myomectomy may serve as a viable two-step fertility-preserving treatment approach. However, the small sample size and class imbalance highlight the critical need for larger, prospective, multicentric studies to develop clinically valid predictive models for minimally invasive fibroid treatment.

Figures