Dynamic Radio-Immuno-Dosimetric Profiling for Outcome Prediction in Definitive Chemoradiation of Locally Advanced Carcinoma of the Cervix

Abstract

For locally advanced carcinoma of the cervix (LACC), the standard, most definitive therapy is chemo-radiation followed by image-guided adaptive brachytherapy (IGABT). Current tools are still mostly based on baseline clinical stage and blood markers pretreatment, and the tools used for brachytherapy are rarely adapted for centers where the MRI-based brachytherapy is not routinely performed. The disease courses of systemic inflammation during therapy, the dose delivered to the pelvic hematopoietic marrow, the duration of overall therapy and the dose of voluminous brachytherapy are all individually predictive, but have never been linked into one deployable index. We developed the design of a single-institution observational cohort of FIGO stage IB3–IVA patients treated with chemo radiation with platinum-based drugs and high-dose-rate brachytherapy. Four candidate domains (inflammatory kinetics: ΔNLR, ΔSII, absolute-lymphocyte nadir and recovery, pelvic bone-marrow dose-volume metrics: V10, V20, V40, overall treatment time (OTT) and HR-CTV D90 (EQD2)) are combined into a weighted composite index, the Dynamic Radio-Immuno-Dosimetric (DRID) index. Harrell C-index and time dependent AUC against FIGO stage and pretreatment NLR are used to assess discrimination with bootstrap internal validation and decision curve analysis. In a worked exemplar cohort (n = 214; median duration of follow up 38 months), the index discriminated between low-, intermediate- and high-risk groups with 3-year PFS rates of 86%, 64% and 38%, respectively. The composite (C-index 0.78) performed better than FIGO stage (0.62) and pretreatment NLR (0.58) and showed a positive net clinical benefit at many thresholds. These are estimates, and not based on actual data yet. The new dynamic, multi-domain index combining inflammation kinetics to deliverable dosimetric and timing parameters is biologically sensible, and is clearly tailored to resource adapted radiotherapy workflows and is likely to be better than static staging. The framework is hypothesis generating and needs to be validated with a prospective external validation.