Adaptive workflow with EQD2 conversion for side effect prediction in prostate re-irradiation

Re-irradiation is increasingly applied in prostate cancer management, supported by improved techniques and radiobiological modeling. Accurate cumulative dose estimation is crucial for predicting late OAR side effects. This study introduces an adaptive workflow that integrates session-based dose accumulation with EQD2/BED conversion to enhance side effect prediction.

A retrospective analysis of 15 patients with recurrent prostate cancer was performed. The first course used conventional fractionation followed by hypofractionated re-irradiation. Two cumulative dose evaluation strategies were compared: (i) plan-sum method with rigid CT registration and EQD2/BED conversion, and (ii) adaptive workflow using resampled CTs from available CBCTs, deformable registration, and session-specific EQD2/BED. Cumulative bladder and rectum doses were assessed against different proposed constraints. Side effect prediction was evaluated using true positive rate (TPR), true negative rate (TNR) and positive predictive value (PPV), metrics with confusion matrix analysis.

The plan-sum method yielded limited bladder prediction accuracy, with TPR ranging from 0.3 to 0.9 and PPV up to 1.0, while rectum prediction showed high TPR (1.0) but low PPV (0.2-0.3). The adaptive workflow showed a different balance for bladder prediction, particularly under Delphi constraints and α/β = 3, achieving TPR 0.5, PPV 0.9, and TNR 0.5. BED-based models showed comparable findings at α/β = 3. Significant differences between methods were observed for bladder dose indices (p < 0.05).

The adaptive EQD2 workflow provided more favorable prediction of bladder side-effect compared with plan-sum methods, while rectal prediction remained unreliable. Despite being time-intensive, this approach better reflects delivered dose distributions and may improve clinical decision-making in prostate re-irradiation.