Radiomics and Clinical Biomarkers Improve Prediction of Vertebral Compression Fracture in Multiple Myeloma

Djennifer K. Madzia-Madzou, Margot Jak, Bart de Keizer, Jorrit-Jan Verlaan, Nicole Erler, Monique Minnema, Kenneth Gilhuijs

Keywords: Multiple myeloma, vertebral compression fracture, radiomics, SINS, survival analysis, multimodal, risk prediction

TL;DR: CT radiomics and clinical biomarkers predict vertebral compression fractures in multiple myeloma better than the current clinical standard (SINS).

Registration Requirement: Yes

Abstract: Vertebral compression fractures (VCFs) are a debilitating complication of multiple myeloma (MM), affecting up to 60\% of patients. The Spine Instability Neoplastic Score (SINS), developed for solid tumour metastases, is commonly used but unvalidated for MM. We evaluated whether CT radiomics features improve VCF prediction compared to SINS. In a retrospective cohort of 271 MM patients (1,331 CT scans), 1,271 radiomics features were extracted and reduced to 121 robust features. Five Cox models were evaluated using internal 5-fold cross-validation with patient-level bootstraps (R=1,000). The full combined model (SINS + clinical biomarkers + radiomics) significantly outperformed SINS (C-index: 0.70 vs 0.57, $p = 0.003$; 2-year AUC: 0.71 vs 0.62, $p = 0.007$). Radiomics alone also showed significant improvement (C-index: 0.66, $p = 0.037$). CT radiomics and clinical biomarkers substantially improved VCF prediction over SINS, enabling personalized risk assessment for MM patients.

Visa & Travel: No

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 96