A prospective multicenter study across 16 hospitals in Spain evaluated a plasma circulating cell-free RNA (cfRNA) machine learning classifier to differentiate uterine leiomyosarcoma (UMS) from uterine leiomyoma (UM) preoperatively in 102 women (78 UM, 24 UMS) undergoing surgery for suspected myometrial tumors.
A 100-gene regularized logistic regression cfRNA classifier achieved an AUROC of 0.868, sensitivity of 0.732, and specificity of 0.813 in Monte Carlo cross-validation; performance was consistent across age groups (AUROC 0.879 for <55 years, 0.882 for >55 years) and reached AUROC 0.892 in an independent external tissue cohort.
- Small sample size (only 24 UMS cases), limiting statistical power and generalizability. - Internal validation only via Monte Carlo cross-validation — no independent prospective test cohort. - Positive predictive value will vary substantially with UMS prevalence in applied populations, restricting direct clinical use.
This cfRNA classifier is not yet ready for clinical deployment, but it shows promise as a noninvasive preoperative risk-stratification tool that could help guide surgical planning for suspected myometrial tumors. Clinicians should await prospective validation before incorporating cfRNA profiling into practice.
Explore related topics
