Calculating the relative biological effectiveness of carbon ion radiotherapy by common models using measurable microdosimetric inputs

Inconsistencies in dose delivered during carbon radiotherapy is largely due to uncertainties in RBE, which can be calculated by several models. While RBE by the modified Microdosimetric Kinetic Model (MKM) can theoretically be calculated using measured microdosimetric values, clinical implementation relies on lookup tables. Therefore, there exist no means of measuring RBE calculated by MKM, or other common models, including Repair Misrepair Fixation (RMF) and Local Effect Model I (LEM). This study investigates whether a single, uniform microdosimetric measurement can be used to estimate RBE across models, thereby enabling measurement-based validation and inter-model comparison. Monte Carlo simulations were used to generate 260 data points across nine modeled carbon ion beams. All model-specific input parameters required for RBE calculation, such as microdosimetric spectra and DNA double-strand break yields, were scored. The resulting α and β parameters, derived from each model's definitions, were then fit as functions of saturation-corrected dose-mean lineal energy (ȳD*), which was calculated in a simulated tissue-equivalent proportional counter (TEPC) for two representative test beams. Polynomial fits were generated from this relationship and subsequently used to estimate α and β values for the remaining datasets. For experimental validation, TEPC measurements were performed in the clinical carbon beamline at the National Center for Oncological Hadrontherapy (CNAO). From these measured spectra, ȳD* values were derived and used to estimate RBE, which was then compared to model-based RBE calculated directly from Monte Carlo data. RBE calculated using true model input parameters versus parameters estimated with the measurement-based fit were compared to quantify estimation error. RBE was estimated using microdosimetric parameters within ± 10% accuracy in 96% of points for MKM, and 100% for LEM and RMF. Average uncertainty due to estimation was below 5% for all models. When using measured TEPC data as input, estimated RBE values agreed with Monte Carlo-derived RBE within 5% on average across all models. The largest deviations occurred near the Bragg peak, where precise detector positioning is critical for accurate LET and microdosimetric measurement. Although true RBE carries inherent uncertainties, this study demonstrates that modeled RBE can be estimated with reasonable accuracy using a measurement-informed framework. These results support the potential for microdosimetry to serve as a practical quality assurance tool for evaluating complex mixed fields in carbon ion therapy and verifying correct implementation of RBE models in clinical settings.