Organiser: Fysikum, Stockholm University
Zoom: Shall not be streamed online
Contact: Fredrik Kalholm
No registration required

Abstract

In clinical proton radiotherapy, a fixed relative biological effectiveness (RBE) of 1.1 is commonly utilized, assuming a consistent 10% increase in cell inactivation effectiveness compared to photons, regardless of proton energy and cell type. While this fixed RBE assumption has generally led to satisfactory clinical outcomes, various studies based on both in vitro data and patient outcomes suggest that the RBE may actually vary with proton energy. However, there is no widely accepted method to quantify this variability using a specific RBE model. In clinical practice, concerns regarding RBE variability are typically addressed by reducing the dose near the end of the proton range, where the variable RBE is presumed to be highest, particularly near distal risk organs.
    Many proton variable RBE models have been proposed historically, mainly based on in vitro data. Most commonly, this involves describing the α and β parameters of the linear quadratic (LQ) model as a function of a chosen radiation quality metric. Typically, this metric involves an averaged value of linear energy transfer (LET) at a given location. By comparing the obtained α and β values against corresponding parameters obtained under reference conditions (typically megavolt photon irradiation), the RBE value can be modelled.
   While the radiation quality metric used for proton variable RBE models typically is dose-averaged LET (LETd), the details on how the averaged LET value has been calculated or determined are often not fully provided, possibly introducing a source of error in the estimated RBE value. This can vary with respect to the averaging method (typical dose- or track averaging), included particles (only primary, or also including secondary protons and/or ions) and other aspects. Furthermore, while LET is the most commonly used beam quality descriptor, other quantities exist such as Q and z*2/β2, renamed in this work as Qeff. These alternative metrics have been shown to better correlate with RBE across different particle species compared to LET, and can potentially perform better for a single particle species as well. However, these has so far not been systematically tested or verified.
   Paper I investigates which kind of averaged LET is provided in the scientific literature for the purpose of RBE determination, for both protons and other hadronic particles. It also attempts to quantify the corresponding impact to the calculated RBE values. Paper II investigates which beam quality descriptor is most suitable for predicting RBE by simulating the experimental setup of recently published high throughput in vitro cell survival studies for RBE determination by a Monte Carlo particle transport code, and fitting parameters to a phenomenological LQ-based model based on the cell survival data. Different variants of LET, Q and Qeff are included, to generate both linear and non linear variable RBE models. Paper III explores if novel averaging techniques, deviating from conventional linear weighing when compiling the LET spectrum, can improve RBE predictions, while paper IV finally investigates if a binary weighing technique (dirty dose) can be utilized as a radiation quality metric.
   In paper I, it is shown that averaged LET for the purpose of RBE determination is, typically, not entirely well defined with a significant minority not mentioning which averaging method is used, and a majority not mentioning what particles are included when averaging. The impact of using different definitions for proton variable RBE models is, in most cases, small, unless heavier secondary particles are included. In paper II it is shown that Q and especially Qeff are expected to better predict RBE compared to LET by a statistically significant margin, for both linear and non-linear models, suggesting they are likely to be more suitable beam quality descriptors to use in a LQ based phenomenological variable RBE model. The results from paper III suggest that performing a non-linear weighting when compiling the LET spectrum into a radiation quality metric improves the performance of the variable RBE model, suggesting a non-linear underlying RBE(LET) relationship of individual protons. Paper IV finally shows that variable RBE models based on the pragmatic dirty dose approach performs on par with conventional radiation quality metrics, while offering improvements by enabling both simplified calculations and efficient measurement techniques.