Main Session
Sep 29
PQA 06 - Genitourinary Cancer, Gynecological Cancer, and Health Care Access and Engagement

3742 - Integrating HPV ctDNA Dynamics and Tumor Geometry to Guide Selection of 3D-Printed Applicators in Cervical Cancer Brachytherapy: Development and External Validation of a Prognostic Model

02:15pm - 03:30pm ET
Poster Hall - Exhibit Hall A
Screen: 24
POSTER

Presenter(s)

Wiwatchai Sittiwong, MD - Siriraj Hospital Mahidol University, Bangkok, Bangkok

W. Sittiwong1, L. Colbert2, Y. Kim3, T. Prasartseree1, S. Su3, K. Court4, T. Cisneros Napravnik5, Z. Yu3, C. R. Weil2, P. Dankulchai1, C. Vatanashevanopakorn1, C. Metheetrairut6, K. Trakarnsanga6, P. Jintanapunya6, P. Tuntapakul1, W. Thaweerat1, and A. H. Klopp2; 1Division of Radiation Oncology, Department of Radiology, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand, 2Division of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 3Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, 4Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 5MD Anderson Cancer Center, Houston, TX, 6Department of Biochemistry, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand

Purpose/Objective(s):

To develop and externally validate a prognostic model to guide selection of patients with locally advanced cervical cancer most likely to benefit from 3D-printed applicator templates in brachytherapy.

Materials/Methods:

The model was developed from 49 patients treated with HDR or PDR brachytherapy. For each case, 3D-printed plans were generated using two strategies: planTarget (OAR preservation with target dose escalation) and planOAR (target preservation with OAR dose reduction). Benefit was scored using predefined criteria. Candidate predictors included HPV ctDNA changes, clinical stage, tumor characteristics, brachytherapy modality, GTV volumes, tumor response, and geometric parameters (RL/AP ratio, cross-diameter ratio, radial max/min ratio). Multivariable logistic regression was used to construct nomograms. Internal validation was performed using bootstrap resampling to obtain optimism-corrected R² and RMSE. Calibration was assessed by regression of observed on predicted values. External validation was conducted in 11 independent patients. Decision curve analysis (DCA) evaluated clinical utility for benefit defined as observed score =4.5.

Results:

In the training cohort, multivariable analysis identified HDR/PDR, tumor response, and radial max/min ratio as independent predictors of benefit for the PlanTarget model, while HDR/PDR, time-adjusted ctDNA change, and radial max/min ratio were independently associated with benefit in the PlanOAR model. Internal validation demonstrated good discrimination for PlanTarget (optimism-corrected R² 0.627) and moderate discrimination for PlanOAR (0.496). Predictive accuracy was acceptable, with RMSE of 1.19 for PlanTarget and 1.59 for PlanOAR. Calibration plots showed good overall agreement, with minor deviation at the extremes. In external validation, PlanTarget maintained modest discrimination (R² = 0.15) and moderate accuracy (RMSE 2.08). Calibration showed a slope of 1.28 and intercept of -2.76, suggesting mild deviation at higher predicted values. Decision curve analysis demonstrated net benefit across threshold probabilities of 10–45%. In contrast, PlanOAR showed weak discrimination (R² = 0.10) limited accuracy (RMSE 3.32), substantial calibration uncertainty (slope 0.34; intercept 0.73), and net benefit mainly at lower thresholds (10–30%).

Conclusion:

A prognostic model integrating clinical factors, HPV ctDNA dynamics, and tumor geometric parameters may help identify patients with LACC most likely to benefit from 3D-printed brachytherapy applicator templates. The PlanTarget model demonstrated robust internal performance and maintained discrimination and clinical utility on external validation, whereas the PlanOAR model showed limited generalizability. Further validation in larger, independent cohorts is warranted.