Predicting 1p/19q Codeletion Status in Glioma Using MRI-Derived Radiomics; A Systematic Review and Meta-Analysis of Diagnostic Accuracy.

Abstract

Background: The 1p/19q codeletion is a key genetic marker in gliomas and plays a crucial role in prognosis and treatment decisions. Traditional methods for detecting this genetic alteration rely on invasive tissue biopsies.

Purpose: This systematic review and meta-analysis aimed to evaluate the performance of magnetic resonance imaging (MRI)-derived radiomics-based models to predict glioma 1p/19q codeletion status.

Data sources: A literature search was conducted in four databases-PubMed, Web of Science, Embase, and Scopus.

Study selection: We selected the studies that assessed the performance of radiomics-based models in determining 1p/19q codeletion status.

Data analysis: The METhodological RadiomICs Score (METRICS) was used to evaluate study quality. Pooled diagnostic estimates were calculated, and heterogeneity was assessed using the I² statistic. Subgroup and sensitivity analyses were performed to investigate potential sources of heterogeneity. Deeks' funnel plot was used to assess publication bias.

Data synthesis: Twenty-eight studies met the inclusion criteria for the systematic review. A meta-analysis of 10 studies yielded a pooled sensitivity of 0.82 (95% CI: 0.67-0.91), specificity of 0.80 (95% CI: 0.70-0.88), positive diagnostic likelihood (DLR) of 4.14 (95%CI: 2.62-6.52), negative DLR of 0.23 (95% CI: 0.12-0.43), diagnostic odds ratio of 18.37 (95% CI: 7.36-45.85), and area under the curve of 0.87 (95% CI: 0.84-0.90). Subgroup analysis revealed significant differences based on the country and segmentation method.

Limitations: Our meta-analysis is limited by small number of studies with external validation cohorts.

Conclusions: MRI-derived radiomics-based models demonstrated good predictive performance for glioma 1p/19q codeletion status, highlighting their potential as a non-invasive tool for glioma characterization and for aiding in treatment decision-making.

Abbreviations: DLR: diagnostic likelihood ratio, DOR: diagnostic odds ratio, AUC =area under the curve; HOIV: holdout internal validation, EV = external validation.