Abstract

Background: Prostate cancer (PCa) is the second most common cancer in men, and radiotherapy (RT) is one of the main treatment options. Although effective, RT can cause toxic side effects. The accurate prediction of dosimetric parameters, enhanced by advanced technologies and AI-based predictive models, is crucial to optimize treatments and reduce toxicity risks. This study aims to explore current methodologies for predictive dosimetric parameters associated with RT toxicity in PCa patients, analyzing both traditional techniques and recent innovations. Methods: A systematic review was conducted using the PubMed, Scopus, and Medline databases to identify dosimetric predictive parameters for RT in prostate cancer. Studies published from 1987 to April 2024 were included, focusing on predictive models, dosimetric data, and AI techniques. Data extraction covered study details, methodology, predictive models, and results, with an emphasis on identifying trends and gaps in the research. Results: After removing duplicate manuscripts, 354 articles were identified from three databases, with 49 shortlisted for in-depth analysis. Of these, 27 met the inclusion criteria. Most studies utilized logistic regression models to analyze correlations between dosimetric parameters and toxicity, with the accuracy assessed by the area under the curve (AUC). The dosimetric parameter studies included Vdose, Dmax, and Dmean for the rectum, anal canal, bowel, and bladder. The evaluated toxicities were genitourinary, hematological, and gastrointestinal. Conclusions: Understanding dosimetric parameters, such as DVH, Dmax, and Dmean, is crucial for optimizing RT and predicting toxicity. Enhanced predictive accuracy improves treatment effectiveness and reduces side effects, ultimately improving patients’ quality of life. Emerging artificial intelligence and machine learning technologies offer the potential to further refine RT in PCa by analyzing complex data, and enabling more personalized treatment approaches.

Abstract

In order to limit radiotherapy (RT)-related side effects, effective toxicity prediction and assessment schemes are essential. In recent years, the growing interest toward artificial intelligence and machine learning (ML) within the science community has led to the implementation of innovative tools in RT. Several researchers have demonstrated the high performance of ML-based models in predicting toxicity, but the application of these approaches in clinics is still lagging, partly due to their low interpretability. Therefore, an overview of contemporary research is needed in order to familiarize practitioners with common methods and strategies. Here, we present a review of ML-based models for predicting and classifying RT-induced complications from both a methodological and a clinical standpoint, focusing on the type of features considered, the ML methods used, and the main results achieved. Our work overviews published research in multiple cancer sites, including brain, breast, esophagus, gynecological, head and neck, liver, lung, and prostate cancers. The aim is to define the current state of the art and main achievements within the field for both researchers and clinicians.

Abstract

Summary Background The association between acute and late toxicity following prostate radiotherapy has not been well studied using data from multiple randomised clinical trials and fractionation schedules. We aimed to characterise the relationship between acute and late genitourinary and gastrointestinal toxicity among patients receiving conventionally fractionated or moderately hypofractionated prostate radiotherapy. Methods This was an individual patient data meta-analysis that identified randomised phase 3 trials of conventionally fractionated or moderately hypofractionated prostate radiotherapy in the Meta-Analysis of Randomized trials in Cancer of the Prostate (MARCAP) Consortium that had individual-level acute and late toxicity data available and were available before Dec 1, 2023. Trials without individual patient data were excluded. Data were provided to MARCAP by study investigators. The associations between acute (≤3 months after radiotherapy) and late (>3 months after radiotherapy) grade 2 or greater genitourinary and gastrointestinal toxicities were assessed using adjusted generalised linear mixed models (adjusted for age, androgen deprivation therapy status, type of radiotherapy, radiation dose, and radiation schedule). In the subset of trials that collected Expanded Prostate Cancer Index Composite quality of life (QOL) evaluations, the association between acute genitourinary and gastrointestinal toxicity and decrements at least twice the minimal clinically important difference (MCID) for urinary and bowel QOL were also evaluated. Findings Six of 26 available trials met all the eligibility criteria. 6593 patients were included (conventionally fractionated: n=4248; moderately hypofractionated: n=2345). Median follow-up was 72 months (IQR 61–94). Acute grade 2 or greater genitourinary toxicity was associated with both late grade 2 or greater genitourinary toxicity (odds ratio 2·20 [95% CI 1·88–2·57], p<0·0001) and decrement at least twice the MCID in urinary QOL (1·41 [1·17–1·68], p=0·0002). Acute grade 2 or greater gastrointestinal toxicity was associated with both late grade 2 or greater gastrointestinal toxicity (2·53 [2·07–3·08], p<0·0001) and decrement at least twice the MCID in bowel QOL (1·52 [1·26–1·83], p<0·0001). Interpretation Acute toxicity following prostate radiotherapy was statistically significantly associated with late toxicity and with decrement in patient-reported QOL metrics. These data support efforts to evaluate whether interventions that reduce acute toxicity ultimately reduce the risk of late toxicity.

Abstract

AIMS Patient factors affect the risk of radiotherapy toxicity, but many are poorly defined. Studies have shown that race affects cancer incidence, survival, drug response, molecular pathways and epigenetics. Effects on radiosensitivity and radiotherapy toxicity are not well studied. The aim of the present study was to identify the effects of race and ethnicity on the risk of radiotherapy toxicity. MATERIALS AND METHODS A systematic review was carried out of PubMed, Ovid Medline and Ovid Embase with no year limit. PRISMA 2020 guidelines were followed. Two independent assessors reviewed papers. RESULTS Of 607 papers screened, 46 fulfilled the inclusion criteria. Papers were published between 1996 and 2021 and involved 30-28,354 individuals (median 433). Most involved patients with prostate (33%), breast (26%) and lung (9%) cancer. Both early and late toxicities were studied. Some studies reported a higher risk of toxicity in White men with prostate cancer compared with other races and ethnicities. For breast cancer patients, some reported an increased risk of toxicity in White women compared with other race and ethnic groups. In general, it was difficult to draw conclusions due to insufficient reporting and analysis of race and ethnicity in published literature. CONCLUSIONS Reporting of race and ethnicity in radiotherapy studies must be harmonised and improved and frameworks are needed to improve the quality of reporting. Further research is needed to understand how ancestral heritage might affect radiosensitivity and risk of radiotherapy toxicity.

Abstract

Machine learning technology has a growing impact on radiation oncology with an increasing presence in research and industry. The prevalence of diverse data including 3D imaging and the 3D radiation dose delivery presents potential for future automation and scope for treatment improvements for cancer patients. Harnessing this potential requires standardization of tools and data, and focused collaboration between fields of expertise. The rapid advancement of radiation oncology treatment technologies presents opportunities for machine learning integration with investments targeted towards data quality, data extraction, software, and engagement with clinical expertise. In this review, we provide an overview of machine learning concepts before reviewing advances in applying machine learning to radiation oncology and integrating these techniques into the radiation oncology workflows. Several key areas are outlined in the radiation oncology workflow where machine learning has been applied and where it can have a significant impact in terms of efficiency, consistency in treatment and overall treatment outcomes. This review highlights that machine learning has key early applications in radiation oncology due to the repetitive nature of many tasks that also currently have human review. Standardized data management of routinely collected imaging and radiation dose data are also highlighted as enabling engagement in research utilizing machine learning and the ability integrate these technologies into clinical workflow to benefit patients. Physicists need to be part of the conversation to facilitate this technical integration.

Abstract

Purpose To evaluate the efficacy of prominent machine learning algorithms in predicting normal tissue complication probability using clinical data obtained from 2 distinct disease sites and to create a software tool that facilitates the automatic determination of the optimal algorithm to model any given labeled data set. Methods and Materials We obtained 3 sets of radiation toxicity data (478 patients) from our clinic: gastrointestinal toxicity, radiation pneumonitis, and radiation esophagitis. These data comprised clinicopathological and dosimetric information for patients diagnosed with non-small cell lung cancer and anal squamous cell carcinoma. Each data set was modeled using 11 commonly employed machine learning algorithms (elastic net, least absolute shrinkage and selection operator [LASSO], random forest, random forest regression, support vector machine, extreme gradient boosting, light gradient boosting machine, k-nearest neighbors, neural network, Bayesian-LASSO, and Bayesian neural network) by randomly dividing the data set into a training and test set. The training set was used to create and tune the model, and the test set served to assess it by calculating performance metrics. This process was repeated 100 times by each algorithm for each data set. Figures were generated to visually compare the performance of the algorithms. A graphical user interface was developed to automate this whole process. Results LASSO achieved the highest area under the precision-recall curve (0.807 ± 0.067) for radiation esophagitis, random forest for gastrointestinal toxicity (0.726 ± 0.096), and the neural network for radiation pneumonitis (0.878 ± 0.060). The area under the curve was 0.754 ± 0.069, 0.889 ± 0.043, and 0.905 ± 0.045, respectively. The graphical user interface was used to compare all algorithms for each data set automatically. When averaging the area under the precision-recall curve across all toxicities, Bayesian-LASSO was the best model. Conclusions Our results show that there is no best algorithm for all data sets. Therefore, it is important to compare multiple algorithms when training an outcome prediction model on a new data set. The graphical user interface created for this study automatically compares the performance of these 11 algorithms for any data set.

Abstract

Genetic variants associated with acute side effects of radiotherapy in nasopharyngeal carcinoma (NPC) remain largely unknown. We performed a two-stage genome-wide association analysis including a total of 1084 patients, where 319 individuals in the discovery stage were genotyped for 688,783 SNPs using whole genome-wide screening microarray. Significant variants were then validated in an independent cohort of 765 patients using the MassARRAY system. Gene mapping, linkage disequilibrium, genome-wide association analysis, and polygenic risk score were conducted or calculated using FUMA, LDBlockShow, PLINK, and PRSice software programs, respectively. Five SNPs (rs6711678, rs4848597, rs4848598, rs2091255, and rs584547) showed statistical significance after validation. Radiotherapy toxicity was more serious in mutant minor allele carriers of all five SNPs. Stratified analysis further indicated that rs6711678, rs4848597, rs4848598, and rs2091255 correlated with skin toxicity in patients of EBV positive, late stage (III and IV), receiving both concurrent chemoradiotherapy and induction/adjuvant chemotherapy, and with OR values ranging from 1.92 to 2.66. For rs584547, high occurrence of dysphagia was found in A allele carriers in both the discovery (P = 1.27 × 10− 6, OR = 1.55) and validation (P = 0.002, OR = 4.20) cohorts. Furthermore, prediction models integrating both genetic and clinical factors for skin reaction and dysphagia were established. The area under curve (AUC) value of receiver operating characteristic (ROC) curves were 0.657 (skin reaction) and 0.788 (dysphagia). Rs6711678, rs4848597, rs4848598, and rs2091255 on chromosome 2q14.2 and rs584547 were found to be novel risk loci for skin toxicity and dysphagia in NPC patients receiving radiotherapy. Chinese Clinical Trial Register (registration number: ChiCTR-OPC-14005257 and CTXY-140007-2).

Abstract

PURPOSE The Common Terminology Criteria for Adverse Events (CTCAE) is the established toxicity scoring system that assigns severity grades (G1 = mild to G5 = death) to Adverse Events (AEs). Compared to CTCAEv3.0 (2006), updated versions introduced changes in severity grade definitions. This study evaluated changes between v3.0 and v5.0 (2017) for AEs in gynaecological radiotherapy. MATERIAL AND METHODS After selecting AEs relevant for gynaecological radiotherapy in v3.0, changes in severity grades were identified using CTCAE v3.0-to-v5.0 mapping tables. Six radiation oncologists (ROs) evaluated severity grade definitions for changes in: clinical interpretation, subjective (patient-reported symptoms) and objective (details on medication/intervention) information, and expected severe (≥G3) events. Agreement was based on at least five (≥5)ROs. RESULTS Gastrointestinal, urinary, reproductive, general and injury/musculoskeletal AEs were selected (n = 118). G4 definitions in v5.0 were removed in 22 % of AEs. ≥5ROs agreed on changes affecting clinical interpretation especially for G2 (31 %) and G3 (30 %). For subjective information, 18 % of G2 and 15 % of G3 were judged relying more on patient-reported symptoms. Less objective information was found in 51 % of G3 definitions. Variability in agreement was observed especially for subjective information in G3 and expected ≥G3 events. CONCLUSION This analysis revealed that severity grade definitions in v3.0 and v5.0 for AEs in gynaecological radiotherapy present changes with potential impact on scoring in clinical studies. Notably, 22 % of AEs in v5.0 no longer have G4 defined, and G3 definitions often include fewer details on medication/intervention. Variability in ROs' interpretations is frequently observed, highlighting the need for education to standardise toxicity scoring.

Abstract

PURPOSE Translating FLASH radiotherapy into clinical practice is limited by technological challenges in achieving highly conformal dose delivery. A hybrid approach, combining ultra-high dose rate (UHDR) and conventional dose rate (CDR) irradiation, could potentially improve dose conformity, but biological validation is required. This study presents the first investigation into the biological response to hybrid UHDR/CDR regimens, evaluating normal tissue sparing in murine models and assessing its dependence on the UHDR dose. METHODS AND MATERIALS Two murine models were utilized: BALB/c mice evaluated for radiation dermatitis following localized hindlimb irradiation, and C57BL/6 mice evaluated for gastrointestinal (GI) toxicity following whole-abdomen irradiation. The hybrid irradiation protocol consisted of an initial UHDR boost (10 or 20 Gy for skin; 10 or 14 Gy for abdomen), a one-minute beam-off interval, and escalating CDR doses. Irradiation was delivered using prototype electron FLASH irradiators. Normal tissue complication probability (NTCP) curves were generated to determine the median toxic dose (TD50) and dose modifying factors (DMF), comparing hybrid treatments against reference split-dose regimens delivered entirely at CDR. RESULTS Hybrid irradiation effectively preserved the normal tissue sparing characteristic of the FLASH effect, though the magnitude of protection was highly dependent on the proportion of the UHDR boost. In the skin model, the 20 Gy UHDR boost yielded a DMF of 1.26, which decreased to 1.14 when the boost was reduced to 10 Gy. For GI toxicity, the 14 Gy UHDR boost demonstrated a tissue-sparing effect with a DMF of 1.09. However, reducing the UHDR boost to 10 Gy in the abdominal model resulted in complete overlap with the CDR reference curve, eliminating the sparing effect (DMF = 1.00). CONCLUSIONS The FLASH normal tissue sparing effect is partially maintained when UHDR radiation is combined with CDR radiation in a hybrid protocol, with the degree of sparing relying heavily on the overall UHDR dose contribution. While these biological findings support hybrid UHDR/CDR regimens as a promising approach for the clinical implementation of FLASH-RT, the achievable dose conformity has yet to be practically validated.