Added-value of dynamic contrast-enhanced MRI on prediction of tumor recurrence in locally advanced cervical cancer treated with chemoradiotherapy

To evaluate whether the DCE-MRI derived parameters integrated into clinical and conventional imaging variables may improve the prediction of tumor recurrence for locally advanced cervical cancer (LACC) patients following concurrent chemoradiotherapy (CCRT).Between March 2014 and November 2019, 79 consecutive LACC patients who underwent pelvic MRI examinations with DCE-MRI sequence before treatment were prospectively enrolled. The primary outcome was disease-free survival (DFS). DCE-MRI derived parameters, conventional imaging, and clinical factors were collected. Univariate and multivariate Cox hazard regression analyses were performed to evaluate these parameters in the prediction of DFS. The independent and prognostic interested variables were combined to build a prediction model compared with the clinical International Federation of Gynecological (FIGO) staging system.Lymph node metastasis (LNM) and the mean value of ve (ve_mean) were independently associated with tumor recurrence (all p < 0.05). The prediction model based on T stage, LNM, and ve_mean demonstrated a moderate predictive capability in identifying LACC patients with a high risk of tumor recurrence; the model was more accurate than the FIGO staging system alone (c-index: 0.735 vs. 0.661) and the combination of ve_mean and the FIGO staging system (c-index: 0.735 vs. 0.688). Moreover, patients were grouped into low-, medial-, and high-risk levels based on the advanced T stage, positive LNM, and ve_mean < 0.361, with which the 2-year DFS was significantly stratified (p < 0.001).The ve_mean from DCE-MRI could be used as a useful biomarker to predict DFS in LACC patients treated with CCRT as an assistant of LNM and T stage.Lower ve_mean is an independent predictor of poor prognosis for disease-free survival in locally advanced cervical cancer patients treated with concurrent chemoradiotherapy (hazard ratio [HR]: 0.016, p<0.023). A combined prediction model based on advanced T stage, LNM, and ve_mean performed better than the FIGO staging system alone.