Session Item

A quantitative understanding of magnitudes of normal tissue protection by the FLASH effect and its basic dependencies on dose delivery parameters is an essential requirement for a successful and optimized clinical translation of FLASH RT. However, to date, the mechanism for the FLASH effect is still under investigation. We gathered therefore available in vivo data of normal tissue sparing of ultra-high dose rate (UHDR) beams and followed a phenomenological data-driven approach that is not related to a mechanistic modelling of the FLASH effect to parameterize the data as a function of dose.

We gathered available experimental data of normal tissue reactions for UHDR versus conventional dose rates (CONV) and expressed them in terms of dose modifying factor (DMF) on a common scale. We then performed linear and non-linear regressions using different functions including logistic and log-logistic functions. An augmented covariance matrix and Monte Carlo sampling was used to evaluate confidence and prediction intervals of the fits.

The gathered data allows to evaluate general trends and magnitudes of DMF of UHDR radiation as a function of dose. It was found that individual data series can be mostly well described by a bi-linear function when expressed as D/DMF versus D. This is corroborated by R² > 0.9 for linear fits to data at doses exhibiting a FLASH effect for most individual data series. Furthermore, the analysis highlights that the DMF of pooled skin reaction data follow a consistent general trend as a function of dose that can be parametrized by relatively simple functions (see Figure 1). Skin reactions data show an onset of the FLASH effect for doses of about 17 Gy and reach DMF of about 1.4 only for high doses >30 Gy, where the DMF starts to saturate. Other normal tissues and endpoints show an earlier onset of the FLASH effect (~7 Gy) and significant sparing factors are already achieved at lower doses.

Figure 1: One over DMF versus dose for the skin reactions of different species (mouse, rat, mini pig, human). A log-logistic function (solid line) fitted to data together with the corresponding 95% confidence (dark gray) and 95% prediction (light gray) intervals visualizes the general trend and indicates data spread.

We analyzed and parameterized DMF of available in vivo data as a function of dose. Such parameterizations of combined data provide a means to assess expected toxicities for doses delivered with FLASH RT compared to doses delivered with CONV RT while providing uncertainty margins based on data. The approach was not related to a mechanism that explains the FLASH effect. This may help to guide experiment and clinical trial design and enables explorative treatment planning studies factoring in the FLASH effect. At the same time, the analysis highlights scarcity, spread and uncertainties in the available experimental data for the clinical translation of the FLASH effect.