Ultra-high dose-rate ‘FLASH’ radiotherapy may be a pivotal step forward for cancer treatment, widening the therapeutic window between radiation tumour killing and damage to neighbouring normal tissues. The extent of normal tissue sparing reported in pre-clinical FLASH studies typically corresponds to an increase in isotoxic dose-levels of 5–20%, though gains are larger at higher doses. Conditions currently thought necessary for FLASH normal tissue sparing are a dose-rate ≥40 Gy s-1, dose-per-fraction ≥5–10 Gy and irradiation duration ≤0.2–0.5 s. Cyclotron proton accelerators are the first clinical systems to be adapted to irradiate deep-seated tumours at FLASH dose-rates, but even using these machines it is challenging to meet the FLASH conditions. In this review we describe the challenges for delivering FLASH proton beam therapy, the compromises that ensue if these challenges are not addressed, and resulting dosimetric losses. Some of these losses are on the same scale as the gains from FLASH found pre-clinically. We therefore conclude that for FLASH to succeed clinically the challenges must be systematically overcome rather than accommodated, and we survey physical and pre-clinical routes for achieving this.

Objectives.To derive a collection efficiency formula,fGauss, for cylindrical ionization chambers in pulsed radiation beams from a volume recombination model of Boaget al(1996Phys. Med. Biol.41885-97) including free electrons. To validatefGaussand a parallel plate chamber formulafexpusing an ion transport code and calculate changes in collection efficiencies caused by electric field charge screening at 0.1-100 mGy doses-per-pulse. And to determine collection efficienciesCE∞predicted at infinite voltage in the absence of avalanche effects by fitting scaled formulae to efficiencies computed for 100-400 V chamber voltages and 10 and 100 mGy doses-per-pulse.Approach.Calculations were performed for an idealized parallel plate chamber with 2 mm electrode separationd, and for an idealized cylindrical chamber with 0.5 and 2.333 mm inner and electrode radiirinandrout.Main results.fGaussandfexppredict the same collection efficiencies for cylindrical and parallel plate chambers satisfyingd2=(rout2-rin2)ln(rout/rin)/2, an equivalence condition met by the chambers studied. Without charge screening, efficiencies computed using the code equalledfGaussandfexp. With screening, efficiencies changed by ⩽0.03%, ⩽1.1% and ⩽21.3% at 1, 10 and 100 mGy doses-per-pulse, and differed between the chambers by ⩽0.9% and ⩽19.6% at ⩽10 and 100 mGy dose-per-pulse. For fits offexpandfGauss,CE∞values were ⩽1.2% and ⩽17.6% from unity at 10 and 100 mGy per pulse respectively, closer than for other formulae tested.Significance.Allowing for screening,fGaussandfexpdescribed computed collection efficiencies to within 0.03%, 1.1% and 21.3% at doses-per-pulse ⩽1, 10 and 100 mGy. Equivalence of the two chambers broke down at 100 mGy per pulse. Departures ofCE∞values from unity suggest that collection efficiencies determined experimentally by fittingfGaussorfexpto readings made at multiple voltages will be accurate to within 1.2% and 17.6% at 10 and 100 mGy per pulse respectively.

PurposeAdding immune checkpoint blockade (ICB) to concurrent chemoradiotherapy (cCRT) has improved overall survival (OS) for inoperable locally advanced non-small cell lung cancer (LA-NSCLC). Trials of cCRT-ICB are heterogeneous for factors such as tumor stage and histology, PDL1 status and cCRT-ICB schedules. We therefore aimed to determine the ICB contribution to survival across studies and identify factors associated with survival gain.Methods and materialsData were collated from cCRT-ICB clinical studies published 2018-2022 which treated 2196 NSCLC patients (99% stage III). Associations between 2-year OS and ICB, CRT, patient and tumor factors were investigated using meta-regression. A published model of survival following RT or CRT was extended to include ICB effects. The model was fitted simultaneously to the cCRT-ICB data and data previously compiled for RT/CRT treatments alone. The net ICB contribution (‘OS gain’) and its associations with factors were described by fitted values of ICB terms added to the model. Statistical significance was determined by likelihood-ratio testing.ResultsThe gain in 2-year OS from ICB was 9.9% overall (95% CI: 7.6%, 12.2%; p=0.018). Both OS gain and 2-year OS itself rose with increasing planned ICB duration (p=0.008, 0.002 respectively) and with tumor PDL1 ≥1% (p=0.034, 0.023). Fitted OS gains were also greater for patients with stage IIIB/C disease (p=0.021). OS gain was not associated with tumor histology, patient performance status, RT dose, ICB drug type (anti-PDL1 vs anti-PD1) or whether ICB began concurrently with or after cCRT.ConclusionFitted gains in 2-year OS due to ICB were higher in cohorts with greater fractions of stage IIIB/C patients and patients with tumor PDL1 ≥1%. OS gain was also significantly higher in a single cohort with a planned ICB duration of two years rather than one, but was not associated with whether ICB treatment began during vs after CRT.

Purpose: To investigate whether a novel signal derived from tumor motion allows more precise sorting of 4D-magnetic resonance (4D-MR) image data than do signals based on normal anatomy, reducing levels of stitching artifacts within sorted lung tumor volumes.Methods: (4D-MRI) scans were collected for 10 lung cancer patients using a 2D T2-weighted single-shot turbo spin echo sequence, obtaining 25 repeat frames per image slice. For each slice, a tumor-motion signal was generated using the first principal component of movement in the tumor neighborhood (TumorPC1). Signals were also generated from displacements of the diaphragm (DIA) and upper and lower chest wall (UCW/LCW) and from slice body area changes (BA). Pearson r coefficients of correlations between observed tumor movement and respiratory signals were determined. TumorPC1, DIA, and UCW signals were used to compile image stacks showing each patient's tumor volume in a respiratory phase. Unsorted image stacks were also built for comparison. For each image stack, the presence of stitching artifacts was assessed by measuring the roughness of the compiled tumor surface according to a roughness metric (Rg). Statistical differences in weighted means of Rg between any two signals were determined using an exact permutation test.Results: The TumorPC1 signal was most strongly correlated with superior inferior tumor motion, and had significantly higher Pearson r values (median 0.86) than those determined for correlations of UCW, LCW, and BA with superior-inferior tumor motion (p < 0.05).Weighted means of ratios of Rg values in TumorPC1 image stacks to those in unsorted, UCW, and DIA stacks were 0.67, 0.69, and 0.71, all significantly favoring TumorPC1 (p = 0.02-0.05). For other pairs of signals, weighted mean ratios did not differ significantly from one.Conclusion: Tumor volumes were smoother in 3D image stacks compiled using the first principal component of tumor motion than in stacks compiled with signals based on normal anatomy.

This Roadmap paper covers the field of precision preclinical x-ray radiation studies in animal models. It is mostly focused on models for cancer and normal tissue response to radiation, but also discusses other disease models. The recent technological evolution in imaging, irradiation, dosimetry and monitoring that have empowered these kinds of studies is discussed, and many developments in the near future are outlined. Finally, clinical translation and reverse translation are discussed.

Objective. Boag et al (1996) formulated a key model of collection efficiency for ionization chambers in pulsed radiation beams, in which some free electrons form negatively charged ions with a density that initially varies exponentially across the chamber. This non-uniform density complicates ion recombination calculations, in comparison with Boag’s 1950 work in which a collection efficiency formula, f , was straightforwardly obtained assuming a uniform negative ion cloud. Boag et al (1996) therefore derived collection efficiency formulae f ′, f ″ and f ′″ based on three approximate descriptions of the exponentially-varying negative ion cloud, each uniform within a region. Collection efficiencies calculated by Boag et al (1996) using these formulae differed by a maximum of 5.1% relative (at 144 mGy dose-per-pulse with 212 V applied over a 1 mm electrode separation) and all three formulae are often used together. Here an exact solution of the exponentially-varying model is obtained. Approach. The exact solution was derived from a differential equation relating the number of negative ions collected from within some distance of the anode to numbers of ions initially located within that region. Using the resulting formula, f exp , collection efficiencies were calculated for a range of ionization chamber properties and doses-per-pulse, and compared with f , f ′, f ″ and f ″′ values and results from an ion transport code. Main results . f exp values agreed to 5 decimal places with ion transport code results. The maximum relative difference between f exp and f ″′, which was often closest to f exp , was 0.78% for the chamber properties and doses-per-pulse studied by Boag et al (1996), rising to 6.1% at 1 Gy dose-per-pulse and 2 mm electrode separation. Significance. Use of f exp should reduce ambiguities in collection efficiencies calculated using the approximate formulae, although like them f exp does not account for electric field distortion, which becomes substantial at doses-per-pulse ≥100 mGy.

There is evidence of synergy between radiotherapy and immunotherapy. Radiotherapy can increase liberation of tumor antigens, causing activation of antitumor T-cells. This effect can be boosted with immunotherapy. Radioimmunotherapy has potential to increase tumor control rates. Biomathematical models of response to radioimmunotherapy may help on understanding of the mechanisms affecting response, and assist clinicians on the design of optimal treatment strategies. In this work we present a biomathematical model of tumor response to radioimmunotherapy. The model uses the linear-quadratic response of tumor cells to radiation (or variation of it), and builds on previous developments to include the radiation-induced immune effect. We have focused this study on the combined effect of radiotherapy and αPDL1/ αCTLA4 therapies. The model can fit preclinical data of volume dynamics and control obtained with different dose fractionations and αPDL1/ αCTLA4. A biomathematical study of optimal combination strategies suggests that a good understanding of the involved biological delays, the biokinetics of the immunotherapy drug, and the interplay between them, may be of paramount importance to design optimal radioimmunotherapy schedules. Biomathematical models like the one we present can help to interpret experimental data on the synergy between radiotherapy and immunotherapy, and to assist in the design of more effective treatments.

Purpose/Objective(s)Immune checkpoint inhibitors (ICI) have shown significant improvement in overall survival in locally advanced non-small cell lung cancer (LA-NSCLC), when combined with concurrent chemoradiotherapy (cCRT). The best schedule of cCRT ICI combination is still uncertain. We aim to develop a novel prediction survival model and investigate dose-relationship following cCRT and ICI combinations for LA-NSCLC.Materials/MethodsICI and cCRT studies between 2010 and 2021 were collated. ICI were intended to be delivered to 12 months. We modelled the 2-year overall survival (OS2-yr) through sigmoidal TCP and NTCP, where EQD2 of tumor control were modified to quantify immune-related survival gain as dose effects via adding an immune related term "I" – potentially dependent on the ICI doses and duration. Models were fitted using maximum likelihood estimation with 100 times bootstrapping, comparing fits via Akaike Information Criteria and likelihood ratio test. Dose-responses of OS2-yr were plotted for 60 – 74Gy cCRT given in 6 weeks and standard 2Gy-per-fraction (F).ResultsThe data comprised 1001 NSCLC patients from 6 trials. All treated with 2Gy F cCRT, 60 – 66Gy in 30 – 33 fractions. ICI were PD-1/PD-L1 inhibitors 17 doses (range: 13 – 22). 301 patients had cCRT only, 592 patients had ICI sequentially, 108 patients concurrently with cCRT. OS2-yr was 48% (range: 41% – 55%) using cCRT, 67% (range: 44% – 87%) using cCRT ICI. In the generalized model, I was 0.37Gy/ICI-dose (95% CI: 0.12 – infinite), indicating that ICI augmented equivalent 5.6Gy or 7.4Gy EQD2 dose escalation per 15 or 20 ICI doses respectively. The fitting efficacy improved significantly after incorporating ICI (p = 1.5*10−5). The best modelling OS2-yr for cCRT was 63% (95% CI: 53% – 75%) for stage IIIA and 54% (33% – 69%) for stage IIIB/C; while it raised to 75% (66% – 86%) for stage IIIA and 67% (51% – 79%) for stage IIIB/C using cCRT ICI. Once ICI were added, best OS2-yr occurred using radiation dose of 60Gy, and dose-escalation did not improve further outcomes. (table)ConclusionThe model suggests that ICI increase the OS2-yr by ∼12% for stage IIIA and ∼13% for stage IIIB/C following the cCRT treatment of LA-NSCLC at 60Gy biological dose regardless of fractionation. Immune checkpoint inhibitors (ICI) have shown significant improvement in overall survival in locally advanced non-small cell lung cancer (LA-NSCLC), when combined with concurrent chemoradiotherapy (cCRT). The best schedule of cCRT ICI combination is still uncertain. We aim to develop a novel prediction survival model and investigate dose-relationship following cCRT and ICI combinations for LA-NSCLC. ICI and cCRT studies between 2010 and 2021 were collated. ICI were intended to be delivered to 12 months. We modelled the 2-year overall survival (OS2-yr) through sigmoidal TCP and NTCP, where EQD2 of tumor control were modified to quantify immune-related survival gain as dose effects via adding an immune related term "I" – potentially dependent on the ICI doses and duration. Models were fitted using maximum likelihood estimation with 100 times bootstrapping, comparing fits via Akaike Information Criteria and likelihood ratio test. Dose-responses of OS2-yr were plotted for 60 – 74Gy cCRT given in 6 weeks and standard 2Gy-per-fraction (F). The data comprised 1001 NSCLC patients from 6 trials. All treated with 2Gy F cCRT, 60 – 66Gy in 30 – 33 fractions. ICI were PD-1/PD-L1 inhibitors 17 doses (range: 13 – 22). 301 patients had cCRT only, 592 patients had ICI sequentially, 108 patients concurrently with cCRT. OS2-yr was 48% (range: 41% – 55%) using cCRT, 67% (range: 44% – 87%) using cCRT ICI. In the generalized model, I was 0.37Gy/ICI-dose (95% CI: 0.12 – infinite), indicating that ICI augmented equivalent 5.6Gy or 7.4Gy EQD2 dose escalation per 15 or 20 ICI doses respectively. The fitting efficacy improved significantly after incorporating ICI (p = 1.5*10−5). The best modelling OS2-yr for cCRT was 63% (95% CI: 53% – 75%) for stage IIIA and 54% (33% – 69%) for stage IIIB/C; while it raised to 75% (66% – 86%) for stage IIIA and 67% (51% – 79%) for stage IIIB/C using cCRT ICI. Once ICI were added, best OS2-yr occurred using radiation dose of 60Gy, and dose-escalation did not improve further outcomes. (table) The model suggests that ICI increase the OS2-yr by ∼12% for stage IIIA and ∼13% for stage IIIB/C following the cCRT treatment of LA-NSCLC at 60Gy biological dose regardless of fractionation.

Introduction: We analyzed a comprehensive national radiotherapy data set to compare outcomes of the most frequently used moderate hypofractionation regimen (55 Gy in 20 fractions) and conventional fractionation regimen (60-66 Gy in 30-33 fractions). Methods: A total of 169,863 cases of NSCLC registered in England from January 2012 to December 2016 obtained from the Public Health England were divided into cohort 1 (training set) diagnosed in 2012 to 2013 and cohort 2 (validation set) diagnosed in 2014 to 2016. Radiotherapy data were obtained from theNational Radiotherapy Dataset and linked by National Health Service number to survival data from the Office of National Statistics and Hospital Episode Statistics, from which surgical data and Charlson comorbidity index were obtained. Of 73,186 patients with stages I to III NSCLC, 12,898 received radical fractionated radiotherapy (cohort 1-4894; cohort 2-8004). The proportional hazards model was used to investigate overall survival from time of diagnosis. Survival was adjusted for the prognostic factors of age, sex, stage of disease, comorbidity, other radical treatments, and adjuvant chemotherapy, and the difference between the treatment schedules was summarized by hazard ratio (HR) and 95% confidence interval. The significance of any difference was evaluated by the log likelihood test. Results: Of patients with stages I to III NSCLC, 17% to 18% received radical fractionated radiotherapy. After adjustment for independent prognostic factors of age, stage, comorbidity, and other radical and adjuvant treatments, patients in cohort 1 treated with the 2.75 Gy per fraction regimen had a median survival of 25 months compared with 29 months for patients treated with the 2 Gy per fraction regimen(HR = 1.16, p = 0.001). Similarly, in cohort 2, the respective median survival values were 25 and 28 months (HR = 1.10, p = 0.02). Conclusions: Big data analysis of a comprehensive national cohort of patients with NSCLC treated in England suggests that compared with a 4-week regimen of 55 Gy in 20 fractions, a 6-week regimen of conventional daily fractionation to a dose of 60 to 66 Gy at 2 Gy per fraction is associated with a survival benefit. Within the limitations of the retrospective big data analysis with potential selection bias and in the absence of randomized trials, the results suggest that conventional fractionation regimens should remain the standard of care. (C) 2022 International Association for the Study of Lung Cancer. Published by Elsevier Inc.

We present a novel classification system of the parenchymal features of radiation-induced lung damage (RILD). We developed a deep learning network to automate the delineation of five classes of parenchymal textures. We quantify the volumetric change in classes after radiotherapy in order to allow detailed, quantitative descriptions of the evolution of lung parenchyma up to 24 months after RT, and correlate these with radiotherapy dose and respiratory outcomes. Diagnostic CTs were available pre-RT, and at 3, 6, 12 and 24 months post-RT, for 46 subjects enrolled in a clinical trial of chemoradiotherapy for non-small cell lung cancer. All 230 CT scans were segmented using our network. The five parenchymal classes showed distinct temporal patterns. Moderate correlation was seen between change in tissue class volume and clinical and dosimetric parameters, e.g., the Pearson correlation coefficient was ≤0.49 between V30 and change in Class 2, and was 0.39 between change in Class 1 and decline in FVC. The effect of the local dose on tissue class revealed a strong dose-dependent relationship. Respiratory function measured by spirometry and MRC dyspnoea scores after radiotherapy correlated with the measured radiological RILD. We demonstrate the potential of using our approach to analyse and understand the morphological and functional evolution of RILD in greater detail than previously possible.

(1) Purpose: We analysed overall survival (OS) rates following radiotherapy (RT) and chemo-RT of locally-advanced non-small cell lung cancer (LA-NSCLC) to investigate whether tumour repopulation varies with treatment-type, and to further characterise the low α/β ratio found in a previous study. (2) Materials and methods: Our dataset comprised 2-year OS rates for 4866 NSCLC patients (90.5% stage IIIA/B) belonging to 51 cohorts treated with definitive RT, sequential chemo-RT (sCRT) or concurrent chemo-RT (cCRT) given in doses-per-fraction ≤3 Gy over 16–60 days. Progressively more detailed dose-response models were fitted, beginning with a probit model, adding chemotherapy effects and survival-limiting toxicity, and allowing tumour repopulation and α/β to vary with treatment-type and stage. Models were fitted using the maximum-likelihood technique, then assessed via the Akaike information criterion and cross-validation. (3) Results: The most detailed model performed best, with repopulation offsetting 1.47 Gy/day (95% confidence interval, CI: 0.36, 2.57 Gy/day) for cCRT but only 0.30 Gy/day (95% CI: 0.18, 0.47 Gy/day) for RT/sCRT. The overall fitted tumour α/β ratio was 3.0 Gy (95% CI: 1.6, 5.6 Gy). (4) Conclusion: The fitted repopulation rates indicate that cCRT schedule durations should be shortened to the minimum in which prescribed doses can be tolerated. The low α/β ratio suggests hypofractionation should be efficacious.

As reported,1Brada M. Forbes H. Ashley S. Fenwick J. Improving outcomes in NSCLC: optimum dose fractionation in radical radiotherapy matters.J Thorac Oncol. 2022; 17: 532-543Abstract Full Text Full Text PDF PubMed Scopus (10) Google Scholar we agree that some uncertainty remains around patient selection as one of the potential determinants of outcome. Nevertheless, the marked geographic variation in the relative use of the two main dose fractionation regimens suggests that the selection is not principally patient specific. In the cohorts reported (patients receiving radiotherapy with radical intent), deprivation index was not an independent prognostic factor and does not explain the regional differences. The authors suggest, without providing evidence, that data completeness and data quality may have compromised the analysis. It is correct that the data collection was transferred from NATCANSAT to the National Cancer Registration and Analysis Service within the Public Health England in April 2016 and therefore approximately 22% of the data in the validation cohort was collected by the Public Health England. There is no suggestion that either the amount of data captured or the data quality was compromised in the period between April 2016 and December 2016 and no reason to believe that, if even that had been the case, it would have resulted in uneven distribution between the two principal fractionation cohorts. In the list of possible confounders suggested by Salem et al., all, except for performance status and volume of disease, have been corrected for. We believe that comorbidity, as obtained from the Hospital Episode Statistics, is a reasonable but admittedly not a fool-proof surrogate for performance status. The authors appropriately point out the potential higher normal tissue toxicity associated with the hypofractionation regimen. Patients with larger volume tumors would therefore be more likely to be offered conventional fractionation, and this is particularly likely to be the case when a substantial portion of the heart and lung were to receive high radiation doses. This selection would compromise the outcome in the conventional fractionation group and hence is likely to strengthen the conclusion. In our previous publication,2Brada M. Ball C. Mitchell S. Forbes H. Ashley S. Improving outcomes in non-small cell lung cancer; population analysis of radical radiotherapy.Radiother Oncol. 2019; 132: 204-210Abstract Full Text Full Text PDF PubMed Scopus (8) Google Scholar we reported that NSCLC population outcome (i.e., all patients potentially eligible for radical radiotherapy, of which 17.6% received it) was related to the regional variation in radiotherapy utilization, and the independent prognostic factors included among others the deprivation index. The study revealed that the best outcome was not associated with the highest utilization such as would be enabled by the increasing use of intensity-modulated radiation therapy, but that treating beyond the “optimum” proportion of eligible patients results in decline in the NSCLC population survival. Hypofractionated radiotherapy for NSCLC has been introduced in the United Kingdom on empirical grounds without high-level evidence to support its widespread use. This has not been mirrored by general acceptance in other countries. We welcome the suggestion that the challenging and unexpected data warrant a future randomized controlled trial. Although some of the differences in survival may be a result of as-yet unaccounted for prognostic factors, the magnitude of difference, well beyond that found with adjuvant and concomitant chemotherapy, would indeed argue in favor of revisiting dose fractionation to offer the best treatment to patients with NSCLC requiring radical radiotherapy. Michael Brada: Conceptualization, Methodology, Formal analysis, Writing - original draft, Writing - review & editing. Helen Forbes: Resources, Data curation; Investigation, Methodology, Software, Writing - review & editing. Susan Ashley: Data curation, Statistical analysis, Writing - original draft, Writing - review & editing. John Fenwick: Methodology, Writing - review & editing. Unaccounted Confounders Limit the Ability to Draw Conclusions From Big Data Analysis Comparing Radiotherapy Fractionation Regimens in NSCLCJournal of Thoracic OncologyVol. 17Issue 6PreviewWe read with interest the study analyzing England’s National Radiotherapy Data set and outcomes by radiotherapy fractionation in 169,863 patients with stage I to III NSCLC.1 The authors reported that moderate hypofractionation (55 Gy in 20 fractions]) is associated with worse overall survival (OS) compared with conventional fractionation (60–66 Gy in 30–33 fractions). Full-Text PDF Open Archive

Background PET imaging of 18F-fluorodeoxygucose (FDG) is used widely for tumour staging and assessment of treatment response, but the biology associated with FDG uptake is still not fully elucidated. We therefore carried out gene set enrichment analyses (GSEA) of RNA sequencing data to find KEGG pathways associated with FDG uptake in primary breast cancers. Methods Pre-treatment data were analysed from a window-of-opportunity study in which 30 patients underwent static and dynamic FDG-PET and tumour biopsy. Kinetic models were fitted to dynamic images, and GSEA was performed for enrichment scores reflecting Pearson and Spearman coefficients of correlations between gene expression and imaging. Results A total of 38 pathways were associated with kinetic model flux-constants or static measures of FDG uptake, all positively. The associated pathways included glycolysis/gluconeogenesis (‘GLYC-GLUC’) which mediates FDG uptake and was associated with model flux-constants but not with static uptake measures, and 28 pathways related to immune-response or inflammation. More pathways, 32, were associated with the flux-constant K of the simple Patlak model than with any other imaging index. Numbers of pathways categorised as being associated with individual micro-parameters of the kinetic models were substantially fewer than numbers associated with flux-constants, and lay around levels expected by chance. Conclusions In pre-treatment images GLYC-GLUC was associated with FDG kinetic flux-constants including Patlak K , but not with static uptake measures. Immune-related pathways were associated with flux-constants and static uptake. Patlak K was associated with more pathways than were the flux-constants of more complex kinetic models. On the basis of these results Patlak analysis of dynamic FDG-PET scans is advantageous, compared to other kinetic analyses or static imaging, in studies seeking to infer tumour-to-tumour differences in biology from differences in imaging. Trial registration NCT01266486, December 24th 2010.

Purpose The purpose of this study is to present a biomathematical model based on the dynamics of cell populations to predict the tolerability/intolerability of mucosal toxicity in head‐and‐neck radiotherapy. Methods and Materials Our model is based on the dynamics of proliferative and functional cell populations in irradiated mucosa, and incorporates the three As: Accelerated proliferation, loss of Asymmetric proliferation, and Abortive divisions. The model consists of a set of delay differential equations, and tolerability is based on the depletion of functional cells during treatment. We calculate the sensitivity (sen) and specificity (spe) of the model in a dataset of 108 radiotherapy schedules, and compare the results with those obtained with three phenomenological classification models, two based on a biologically effective dose (BED) function describing the tolerability boundary (Fowler and Fenwick) and one based on an equivalent dose in 2 Gy fractions (EQD2) boundary (Strigari). We also perform a machine learning‐like cross‐validation of all the models, splitting the database in two, one for training and one for validation. Results When fitting our model to the whole dataset, we obtain predictive values (sen + spe) up to 1.824. The predictive value of our model is very similar to that of the phenomenological models of Fowler (1.785), Fenwick (1.806), and Strigari (1.774). When performing a k = 2 cross‐validation, the specificity and sensitivity in the validation dataset decrease for all models, from ˜1.82 to ˜1.55–1.63. For Fowler, the worsening is higher, down to 1.49. Conclusions Our model has proved useful to predict the tolerability/intolerability of a dataset of 108 schedules. As the model is more mechanistic than other available models, it could prove helpful when designing unconventional dose fractionations, schedules not covered by datasets to which phenomenological models of toxicity have been fitted.

Introduction: In 'IDEAL-60 patients (N = 78) treated for locally-advanced non-small-cell lung cancer using isotoxically dose-escalated radiotherapy, overall survival (OS) was associated more strongly with VLAwall-64-73-EQD2, the left atrial (LA) wall volume receiving 64-73 Gy equivalent dose in 2 Gy fractions (EQD2), than with whole-heart irradiation measures. Here we test this in an independent cohort 'OX-RT' (N = 64) treated routinely. Methods: Using Cox regression analysis we assessed how strongly OS was associated with VLAwall-64-73-EQD2, with whole-heart volumes receiving 64-73 Gy EQD2 or doses above 10-to-70 Gy thresholds, and with principal components of whole-heart dose-distributions. Additionally, we tested associations between OS and volumes of cardiac substructures receiving dose-ranges described by whole-heart principal components significantly associated with OS. Results: In univariable analyses of OX-RT, OS was associated more strongly with VLAwall-64-73-EQD2 than with whole-heart irradiation measures, but more strongly still with VAortV-29-38-EQD2, the volume of the aortic valve region receiving 29-38 Gy EQD2. The best multivariable OS model included LA wall and aortic valve region mean doses, and the aortic valve volume receiving >38 Gy EQD2, VAortV-38-EQD2. In a subsidiary analysis of IDEAL-6, the best multivariable model included VLAwall-64-73-EQD2, VAortV-29-38-EQD2, VAortV-38-EQD2 and mean aortic valve dose. Conclusion: We propose reducing heart mean doses to the lowest levels possible while meeting protocol dose-limits for lung, oesophagus, proximal bronchial tree, cord and brachial plexus. This in turn achieves large reductions in VAortV-29-38-EQD2 and VLAwall-64-73-EQD2, and we plan to closely monitor patients with values of these measures still >0% (their median value in OX-RT) following reduction. (C) 2021 Elsevier B.V. All rights reserved.