Normal Tissue Complication Probability Research Articles

Observed and Predicted Toxicity in Daily Cone-Beam CT-Based Online Adaptive Stereotactic Partial Breast Irradiation with Decreased PTV Margins

Accelerated partial breast irradiation (APBI) delivers smaller radiation volumes over less time compared to whole breast irradiation (WBI), but the organ-at-risk (OAR) sparing allowed by its large (up to 1 cm) planning target volume (PTV) can be improved. PTV can be decreased with daily online adaptive planning, which we hypothesized yields low rates of adverse events observed and predicted by normal tissue complication probability (NTCP) models. Intensity-modulated (IMRT) cone-beam CT (CBCT)-based daily online adaptive stereotactic PBI (A-SPBI) plans with 3-mm PTV from 8 patients were recreated with 1-cm PTV per the Florence APBI IMRT trial planning guideline. Dose statistics with evidence for association with toxicity were compared. Documented toxicities were collected for patients treated with A-SPBI with a minimum follow-up of 3.5 months and Common Terminology Criteria for Adverse Events (CTCAE) v.5.0 grade was assigned. Using α/β = 3 for breast and lung, dose statistics were converted to equivalent doses in 2-Gy fractions (EQD2) for use in NTCP models and for comparison using paired t tests, with differences considered significant if p≤0.05. The table details EQD2 dose statistics for breast, lung, and cosmetic toxicity for A-SPBI plans with 3-mm PTV and their 1-cm PTV re-plans in 8 patients. PTV volume, mean lung dose (MLD), and lung V5, V20, and V30 were significantly lower in 1-cm plans. Acute, subacute (3-6 months), and late toxicities were collected for 30 patients followed for a median of 8 months (range 4-13 months). Radiation dermatitis was the most common acute toxicity (n = 16, 53%), followed by hyperpigmentation (n = 12, 40%), fibrosis (n = 9, 30%), and fatigue (n = 9, 30%). One grade 3 radiation dermatitis was the only grade ≥3 toxicity. Six patients (20%) acutely developed breast or axillary edema: 4 (13.3%) resolved, and 2 (6.7%) developed acutely and persist at last follow-up, >6 months after RT. No patient had a lung V20, V30, or MLD meeting thresholds for radiation-induced lung injury, radiation pneumonitis, or symptomatic or imaging-based pneumonitis models, respectively. The breast V55 model predicted a median risk of unfavorable cosmesis of 33% (range 26-44%) for A-SBPI plans and 35% (range 28-51) for 1-cm PTV plans (p = 0.28). Observed acute toxicities are tolerable and rarely persist in patients treated with A-SPBI with 3-mm PTV margins with daily CBCT-based online adaptation. NTCP modeling predicts similar cosmetic outcome to 1-cm margins. The significant reduction in ipsilateral lung dose with a 3-mm PTV in our first 8 patients especially supports daily adaptation in low-risk breast cancer patients with smoking history and/or lung comorbidity.

A Risk Prediction Model for Severe Radiation Induced Lymphopenia in Patients with Pancreatic Cancer Treated with Concurrent Chemoradiotherapy

In pancreatic cancer, radiation induced lymphopenia (RIL) is associated with a poor prognosis. However, normal tissue complication probability (NTCP) models predicting RIL in pancreatic cancer treated with concurrent chemoradiotherapy (CCRT) have yet to be developed. This study aims to develop a least absolute shrinkage and selection operator (LASSO)-based multivariate NTCP model to predict severe RIL in patients with pancreatic cancer during CCRT and to validate the model internally. We retrospectively reviewed patients with localized pancreatic cancer who underwent CCRT using three-dimensional conformal radiation therapy from 2013 to 2021. The exclusion criteria were patients with distant metastasis; patients who did not complete RT due to tumor progression; patients who did not have absolute lymphocyte count (ALC) data available before or during RT. An ALC of < 0.5 K/μL during CCRT was defined as severe RIL. A NTCP model of severe RIL was developed by LASSO-based multivariate analysis. We used age, sex, Karnofsky performance status, maximum tumor size, carbohydrate antigen 19-9 level before RT, ALC before RT, volume of planning target volume (PTV), and dosimetric parameters for surrounding organs (including spleen, vertebrae, liver, bilateral kidneys, gastrointestinal tracts) as variables for LASSO. In addition, internal validation was performed by the bootstrap method. The predictive performance of the model was evaluated by the area under the curve (AUC) of the receiver operating characteristic curve and scaled Brier score. Of the 131 patients included in the study, the median age was 68 years (range, 42-84), and 55% were male. The median ALC before RT was 1.37 K/µL (0.52-3.50). The median PTV volume was 315.4 ml (86.3-1079.3). The median dose of radiotherapy was 50.4 Gy (16.2-50.4), with 1.8 Gy per fraction. Combination chemotherapy was S-1 in 99 cases (75.6%) and gemcitabine in 32 cases (24.4%). Induction chemotherapy before CCRT was performed in 39 patients (29.8%). Severe RIL was observed in 84 (63.6%) patients. The LASSO showed that low baseline ALC (p = 0.0002), large PTV volume (p < 0.0001), and a large kidney V5 defined as the percentage of bilateral kidneys receiving 5 Gy or more (p = 0.0338) were selected as parameters of the prediction model for severe RIL (AUC = 0.917) and scaled Brier score was 0.511. As a result of internal validation by the bootstrap method, the average AUC was 0.918 (95% confidence interval, 0.849-0.954). Severe RIL occurred frequently during CCRT for pancreatic cancer, and a NTCP model for severe RIL developed and validated internally in this study showed good predictive performance. External validation is needed before this NTCP model can be used as a benchmark for treatment planning to reduce the risk of severe RIL and for considering future treatment approaches.

Natural History and Clinical/Dosimetric Determinants of Osteoradionecrosis in a Large Cohort of Head and Neck Cancer Following Curative Radiotherapy: Debunking the Myth of Decreased Rates of Osteoradionecrosis in the Modern Radiotherapy Era

Osteoradionecrosis (ORN) of the mandible is the most devastating toxicity following head and neck cancer (HNC) radiotherapy (RT). The rate of ORN occurrence has been debatable in the modern era of advanced RT. We aim to determine the natural history and time to ORN development in a large cohort of HNC. After IRB approval, we identified HNC patients treated with curative-intent RT between 2005 and 2020 at MD Anderson Cancer Center. Dental oncology notes were reviewed and all dental procedures were recorded. Available dose volume histograms (DVHs) for the segmented mandibular volumes were extracted. Based on our previously published ORN normal tissue complication probability (NTCP) model, any dental procedure pre-RT is a clinical high-risk factor and, therefore, data were coded as (high vs. low clinical risk) accordingly. We also coded the dosimetric risk factors to (high vs. low dosimetric risk) according to our published DVH thresholds (high dosimetric risk if any applies: V45≥40%, V55≥25%, and/or D30 ≥40 Gy). Finally, patients were classified to four groups according to combined clinical and dosimetric risk factors (No, clinical, dosimetric, and both). We used the Kaplan-Meier method to calculate time to ORN development and ORN-free survival. For time to ORN development, any grade ORN occurrence was coded as event and all others were censored. For ORN-free survival, death and ORN were coded as events and all others were censored. Log-rank test was used to compared curves of different risk groups. One thousand eight hundred sixty-six patients were included. Median follow-up was 38 months (range 4-162). ORN was reported in 252 patients (13.5%). The median time to ORN development was 18.5 months (range 4-145). 95 patients (37.7%) developed ORN after 2-years post-RT. The 1-, 3-, and 5-year ORN rates were 4.7%, 12.7%, and 17.8%, respectively. The 1-, 3-, and 5-year ORN-free survival were 94.3%, 85.5%, and 80.3%, respectively. There were statistically significant differences (P < 0.0001) between ORN-free survival in different clinical/dosimetric risk factors. The patients with no, clinical-only, dosimetric-only, and both clinical and dosimetric risk factors were 35%, 19%, 22%, and 24%, respectively. The 5-year ORN-free survival was 94.3%, 89.8%, 76.3%, and 69.6% for patients with no, clinical-only, dosimetric-only, and both clinical and dosimetric risk factors, respectively. The hazard-ratio (HR) of ORN development in clinical-only, dosimetric-only, and both clinical and dosimetric risk groups was 2.1, 5.4, and 7.5 compared to the no risk group (P<0.05 for all). Our findings indicate that ORN remains a remarkable toxicity hazard for HNC survivors. A prolonged surveillance time is required for the majority of HNC survivor since more than one-third of the ORN events occurred after 2-year follow-up. Patients with combined clinical and dosimetric risk factors have a staggering ORN risk profile and are proper candidates for future prophylactic pharmacotherapy clinical studies.

Radiation-Induced Brachial Plexopathy (RIBP) after Stereotactic Body Radiotherapy (SBRT): Pooled Analyses of Risks

RIBP, with symptomatic upper extremity motor or sensory deficits, is a risk after SBRT. We herein model dosimetric factors associated with risks of RIBP of the inferior aspect of the brachial plexus following SBRT for apical lung tumors. Literature searches (PubMed & Embase databases) were performed to identify reports published from 2000-2021, using search criteria of ["brachial plex*" and stereotactic]. From a PRISMA systematic review, studies were identified that included individual patient data on: (1) RIBP endpoints after SBRT for apical lung tumors; and (2) brachial plexus Dmax, or maximum point doses (i.e., D0.035cc or D0.03cc). These data were amenable to normal tissue complication probability (NTCP) modelling. Doses were converted using the linear-quadratic model with an alpha-beta ratio of 3 Gy. For the probit models, the parameter values were determined using the maximum likelihood method and the 95% confidence intervals (CI) were determined using the profile likelihood method. The ability of the NTCP models to distinguish patients with and without RIBP was evaluated using the area under the curve (AUC). Two probit NTCP models were derived: one from 3 studies (185 patients with 192 targets and 11 events) and another from 4 studies (221 patients with 229 targets and 18 events). NTCP models (summarized in table) suggest ≈10% risks associated with brachial plexus maximum dose (Dmax) of ∼32-34 Gy in 3 fractions and ∼40-43 Gy in 5 fractions, with a clear dose response. These dose-responses with SBRT (with steep dose gradients beyond the target volume and thus only partial-irradiation of the brachial plexus) are far less steep than those reported following conventionally-fractionated or moderately-hypofractionated radiotherapy used for breast, lung and head and neck cancers (that tend to use radiotherapy fields that circumferentially irradiate the brachial plexus). A dose-response for risk of RIBP after SBRT is observed relative to brachial plexus Dmax. The less-steep dose-response compared to that seen from conventionally-fractionated or moderately-hypofractionated radiotherapy (with large irradiated plexus volumes) suggest a possible volume dependence of RIBP risks. Future work should focus on understanding possible volume effects.

Evidence-Based Multivariate Normal Tissue Complication Probability (NTCP) Study of Domain-Specific Cognitive Decline after Partial Brain RT

Beyond the hippocampus, there are no evidence-based dose constraints for eloquent brain structures which subserve memory and attention/processing speed. We performed a multivariate normal tissue complication probability analysis of post-RT neurocognitive decline, examining dosimetric predictors of eloquent brain regions. Data were analyzed from a prospective longitudinal clinical trial. Patients (n = 78) with primary brain tumors receiving fractionated RT complete a comprehensive neurocognitive evaluation and high-resolution volumetric and diffusion MRI at baseline and 6 months post-RT. Image processing using robust, validated automated segmentation parcellated individual WM tracts, cortical regions, and hippocampi. Well-validated neurocognitive tests including Delis-Kaplan Executive Function System and Wechsler Adult Intelligence Scale-IV coding (attention/processing), Boston Naming Test (language) and Hopkins Verbal Learning Test and Brief Visuospatial Memory Test (verbal/visuospatial memory) were assessed. Reliable change indices adjusted for practice effects (RCI-PE) were calculated for each patient between baseline and 6 months; a negative RCI-PE was scored as decline. Univariate logistic regression was performed with mean and max dose to structures of interest as well as clinical variables. Multivariate model building was performed using automated bootstrapped logistic regression, LASSO and random forest modeling. On univariate analysis mean and max dose to multiple regions of the corpus callosum (CC) were correlated with attention/processing speed decline; most significantly in WAIS coding, including Dmax to the anterior CC (p = 0.011) and central CC (p = 0.010), and Dmax and Dmean to the mid anterior CC (p = 0.006 and 0.010). Mean dose to the left fornix was associated with decline in memory (p = 0.023, cutoff 12.9 Gy), as were increasing age and both concurrent and adjuvant chemotherapy. On multivariate analysis for attention, automated bootstrapped logistic regression showed the most frequently selected variable was mean dose to the mid anterior CC. Performance at nested cross-validation by AUC was 0.80 (0.75-0.84); LASSO model performance by AUC was 0.76 (0.72-0.81) with Dmean to the mid anterior CC being the most frequent variable. The top five most important variables in the Random Forest as ranked by mean decrease in Gini coefficient were mean dose to mid anterior CC, all white matter, combined CC and max dose to CC and posterior CC. Model performance by AUC was 0.66 (0.60-0.71). Here, we present the first, to our knowledge, NTCP model for decline in attention/processing speed, along with dosimetric predictors of memory decline beyond the hippocampus. We found that after partial brain RT, dose to several ROIs significantly correlated with post-RT impairment. These data can guide future cognitive-sparing strategies for brain RT.

Dosimetric Parameters Correlate with Taste Alterations in Head and Neck Radiotherapy

Dysgeusia is an acute and chronic side effect of head and neck chemoradiation, with persistent taste dysfunction 1-2 years post radiotherapy occurring in 23-50% of patients. Several head and neck clinical trials set oral cavity mean doses between 30 to 40 Gy, however, there are no set guidelines accurately defining the composite structure and anatomic boundaries for the oral cavity or separating the tongue into specific quadrants. In this single institution cross sectional study, we sought to determine the effects of radiation doses to specific regions of the oral cavity and tongue on patient-reported long term dysgeusia. We hypothesize that radiation Dose to specific structures in the oral cavity will correlate with long-term dysgeusia in patients who have received head and neck radiotherapy. Patients with primary head and neck cancer receiving definitive intensity modulated radiation therapy (RT) completed quality of life assessments that included patient-reported gustatory function, 1 year post RT. Dosimetric data from RT plans were collected for specific regions and organs at risk within the oral cavity. These structures included the anterior, posterior, superior and inferior tongue (OT), pharyngeal constrictors (as a surrogate for the corda-tympani) and the oral cavity. A forward binary logistic regression model was performed using these data points with scoring from the QLQ-H&N43 questionnaire subsection on taste (Q45) to determine the most predictive values for dysgeusia and the OR for moderate/severe alterations in taste. A total of 78 patients were included in this study, of these patients, 63 had dosimetric data for the pharyngeal constrictors. The Oral Cavity Volume percentage receiving 50Gy (OC V50) and Pharyngeal Constrictors Volume percentage receiving 55Gy (PC V55) were the most predictive constraints for dysgeusia. (0.042, p = 0.011 and 0.041, p = 0.033). Holding all other variables constant in a forward binary logistic regression including 78 patients, the odds of having moderate to severe taste alterations increased by 56% (OR 1.045, 95% CI 1.012-1.079) for a 1% increase in OCV50. When analyzing the 68 patients with available pharyngeal constrictors data, taste alteration increased 20% for every 1% increase in PCV55 (OR 1.042, 95% CI 1.003-1.082). There was no correlation with taste alterations in any of the dose parameters tested (Dmax, mean and V50 up to V70) for superior, inferior, anterior or posterior regions of the tongue. Dosimetric parameters for the oral cavity (OC V50) and pharyngeal constrictors (PC V55) were positively correlated with taste alterations, however, no other specific regions within the oral tongue predicted taste disfunction. Normal tissue complication probability modeling curves to identify OC V50 and PC V55 constraints for dysgeusia are needed.

A systematic review of the normal tissue complication probability models and parameters: Head and neck cancers treated with conformal radiotherapy.

This systematic review study aims to provide comprehensive data on different radiobiological models, parameters, and endpoints used for calculating the normal tissue complication probability (NTCP) based on clinical data from head and neck cancer patients treated with conformal radiotherapy. A systematic literature search was carried out according to the PRISMA guideline for the identification of relevant publications in six electronic databases of Embase, PubMed, Scopus, and Google Scholar to July 2022 using specific keywords in the paper's title and abstract. The initial search resulted in 1368 articles for all organs for the review article about the NTCP parameters. One hundred and seventy-eight articles were accepted for all organs with complete parameters for the mentioned models and finally, 20 head and neck cancer articles were accepted for review. Analysis of the studies shows that the Lyman-Kutcher-Burman (LKB) model properly links the NTCP curve parameters to the postradiotherapy endpoints. In the LKB model for esophagus, the minimum, and maximum corresponding parameters were reported as TD50 = 2.61 Gy with grade ≥3 radiation-induced esophagitis endpoints as the minimum TD50 and TD50 = 68 Gy as the maximum ones. nmin = 0.06, nmax = 1.04, mmin = 0.1, and mmax = 0.65, respectively. Unfortunately, there was not a wide range of published articles on other organs at risk like ear or cauda equina except Burman et al. (Fitting of normal tissue tolerance data to an analytic function. Int J Radiat Oncol Biol Phys Ther. 1991;21:123-135). Findings suggest that the validation of different radiobiological models and their corresponding parameters need to be investigated in vivo and in vitro for developing a more accurate NTCP model to be used for radiotherapy treatment planning optimization.

Dosimetric and radiobiological comparison between conventional and hypofractionated breast treatment plans using the Halcyon system.

The objective of this research is to compare the efficacy of conventional and hypofractionated radiotherapy treatment plans for breast cancer patients, with a specific focus on the unique features of the Halcyon system. The study collected and analyzed dose volume histogram (DVH) data for two groups of treatment plans implemented using the Halcyon system. The first group consisted of 19 patients who received conventional fractionated (CF) treatment with a total dose of 50 Gy in 25 fractions, while the second group comprised 9 patients who received hypofractionated (HF) treatment with a total dose of 42.56 Gy in 16 fractions. The DVH data was used to calculate various parameters, including tumor control probability (TCP), normal tissue complication probability (NTCP), and equivalent uniform dose (EUD), using radiobiological models. The results indicated that the CF plan resulted in higher TCP but lower NTCP for the lungs compared to the HF plan. The EUD for the HF plan was approximately 49 Gy (114% of its total dose) while that for the CF plan was around 53 Gy (107% of its total dose). The analysis suggests that while the CF plan is better at controlling tumors, it is not as effective as the HF plan in minimizing side effects. Additionally, it is suggested that there may be an optimal configuration for the HF plan that can provide the same or higher EUD than the CF plan.

Quality assessment of automatically planned O-Ring linac SBRT plans for pelvic lymph node metastases, finding the optimal minimum target size by comparison with robotic SBRT.

The purpose of this study is to assess the quality of automatic planned O-Ring Halcyon linac SBRT plans for pelvic lymph node metastases and to establish an absolute PTV volume threshold as a plan quality prediction criterion. Compliance of the plans to institutional SBRT plan evaluation criteria and differences in plan quality and treatment delivery times between Halcyon Linac and CyberKnife robotic SBRT were evaluated. Twenty-one CyberKnife treatment plans were replanned for Halcyon. Prescription doses range was 26-40Gy in mean three fractions. The mean/median planning target volume was 4.0/3.6 cm3 . Institutional criteria for the plan evaluation were: New Conformity Index (NCI), Conformity Index (CI), Modified Gradient Index (MGI), selectivity index reciprocal (PIV/TVPIV ), and the target coverage by prescription isodose (%PIV). Statistical analysis based on the receiver operating characteristic (ROC) curve was used to determine a plan quality predictor threshold of the PTV volume. Comparative analysis of normal tissue complication probabilities (NTCP) was used to assess the risk of toxicity in healthy tissues. Seventy-one percent (n=15)/95% (n=20) of Halcyon and 81% (n=17)/100% (n=21) of CK plans fulfilled all ideal/tolerance criteria. For PTVs above a found optimal threshold of 2.6 cm3 (71%, n=15), no statistically significant difference was observed between the CI, NCI, PIV/TVPIV , and MGI indexes of both groups, while the coverage (%PIV) was statistically but not clinically significantly different between cohorts. Significantly shorter delivery times are expected with Halcyon. No significant differences in NTCP were observed. All but one automatically optimized Halcyon treatment plans demonstrated ideal or acceptable performance. PTV threshold of 2.6 cm3 can be used as decision criteria in clinical settings. The results of our study demonstrated the promising performance of the Halcyon for pelvic SBRT, although plan-specific QA is required to verify machine performance during plan delivery.

Comparing supervised and semi-supervised machine learning approaches in NTCP modeling to predict complications in head and neck cancer patients

Background and purposeHead and neck cancer (HNC) patients treated with radiotherapy often suffer from radiation-induced toxicities. Normal Tissue Complication Probability (NTCP) modeling can be used to determine the probability to develop these toxicities based on patient, tumor, treatment and dose characteristics. Since the currently used NTCP models are developed using supervised methods that discard unlabeled patient data, we assessed whether the addition of unlabeled patient data by using semi-supervised modeling would gain predictive performance. Materials and methodsThe semi-supervised method of self-training was compared to supervised regression methods with and without prior multiple imputation by chained equation (MICE). The models were developed for the most common toxicity outcomes in HNC patients, xerostomia (dry mouth) and dysphagia (difficulty swallowing), measured at six months after treatment, in a development cohort of 750 HNC patients. The models were externally validated in a validation cohort of 395 HNC patients. Model performance was assessed by discrimination and calibration. ResultsMICE and self-training did not improve performance in terms of discrimination or calibration at external validation compared to current regression models. In addition, the relative performance of the different models did not change upon a decrease in the amount of (labeled) data available for model development. Models using ridge regression outperformed the logistic models for the dysphagia outcome. ConclusionSince there was no apparent gain in the addition of unlabeled patient data by using the semi-supervised method of self-training or MICE, the supervised regression models would still be preferred in current NTCP modeling for HNC patients.

Comparing Predicted Toxicities between Hypofractionated Proton and Photon Radiotherapy of Liver Cancer Patients with Different Adaptive Schemes.

With the availability of MRI linacs, online adaptive intensity modulated radiotherapy (IMRT) has become a treatment option for liver cancer patients, often combined with hypofractionation. Intensity modulated proton therapy (IMPT) has the potential to reduce the dose to healthy tissue, but it is particularly sensitive to changes in the beam path and might therefore benefit from online adaptation. This study compares the normal tissue complication probabilities (NTCPs) for liver and duodenal toxicity for adaptive and non-adaptive IMRT and IMPT treatments of liver cancer patients. Adaptive and non-adaptive IMRT and IMPT plans were optimized to 50 Gy (RBE = 1.1 for IMPT) in five fractions for 10 liver cancer patients, using the original MRI linac images and physician-drawn structures. Three liver NTCP models were used to predict radiation-induced liver disease, an increase in albumin-bilirubin level, and a Child-Pugh score increase of more than 2. Additionally, three duodenal NTCP models were used to predict gastric bleeding, gastrointestinal (GI) toxicity with grades >3, and duodenal toxicity grades 2-4. NTCPs were calculated for adaptive and non-adaptive IMRT and IMPT treatments. In general, IMRT showed higher NTCP values than IMPT and the differences were often significant. However, the differences between adaptive and non-adaptive treatment schemes were not significant, indicating that the NTCP benefit of adaptive treatment regimens is expected to be smaller than the expected difference between IMRT and IMPT.

Consistency in contouring of organs at risk by artificial intelligence vs oncologists in head and neck cancer patients

Background In the Danish Head and Neck Cancer Group (DAHANCA) 35 trial, patients are selected for proton treatment based on simulated reductions of Normal Tissue Complication Probability (NTCP) for proton compared to photon treatment at the referring departments. After inclusion in the trial, immobilization, scanning, contouring and planning are repeated at the national proton centre. The new contours could result in reduced expected NTCP gain of the proton plan, resulting in a loss of validity in the selection process. The present study evaluates if contour consistency can be improved by having access to AI (Artificial Intelligence) based contours. Materials and Methods The 63 patients in the DAHANCA 35 pilot trial had a CT from the local DAHANCA centre and one from the proton centre. A nationally validated convolutional neural network, based on nnU-Net, was used to contour OARs on both scans for each patient. Using deformable image registration, local AI and oncologist contours were transferred to the proton centre scans for comparison. Consistency was calculated with the Dice Similarity Coefficient (DSC) and Mean Surface Distance (MSD), comparing contours from AI to AI and oncologist to oncologist, respectively. Two NTCP models were applied to calculate NTCP for xerostomia and dysphagia. Results The AI contours showed significantly better consistency than the contours by oncologists. The median and interquartile range of DSC was 0.85 [0.78 − 0.90] and 0.68 [0.51 − 0.80] for AI and oncologist contours, respectively. The median and interquartile range of MSD was 0.9 mm [0.7 − 1.1] mm and 1.9 mm [1.5 − 2.6] mm for AI and oncologist contours, respectively. There was no significant difference in NTCP. Conclusions The study showed that OAR contours made by the AI algorithm were more consistent than those made by oncologists. No significant impact on the NTCP calculations could be discerned.

Development and validation of a machine learning model of radiation-induced hypothyroidism with clinical and dose–volume features

Background and purposeRadiation-induced hypothyroidism (RIHT) is a common but underestimated late effect in head and neck cancers. However, no consensus exists regarding risk prediction or dose constraints in RIHT. We aimed to develop a machine learning model for the accurate risk prediction of RIHT based on clinical and dose–volume features and to evaluate its performance internally and externally. Materials and methodsWe retrospectively searched two institutions for patients aged >20 years treated with definitive radiotherapy for nasopharyngeal or oropharyngeal cancer, and extracted their clinical information and dose–volume features. One was designated the developmental cohort, the other as the external validation cohort. We compared the performances of machine learning models with those of published normal tissue complication probability (NTCP) models. ResultsThe developmental and external validation cohorts consisted of 378 and 49 patients, respectively. The estimated cumulative incidence rates of grade ≥1 hypothyroidism were 53.5% and 61.3% in the developmental and external validation cohorts, respectively. Machine learning models outperformed traditional NTCP models by having lower Brier scores at every time point and a lower integrated Brier score, while demonstrating a comparable calibration index and mean area under the curve. Even simplified machine learning models using only thyroid features performed better than did traditional NTCP algorithms. The machine learning models showed consistent performance between folds. The performance in a previously unseen external validation cohort was comparable to that of the cross-validation. ConclusionsOur model outperformed traditional NTCP models, with additional capabilities of predicting the RIHT risk at individual time points. A simplified model using only thyroid dose–volume features still outperforms traditional NTCP models and can be incorporated into future treatment planning systems for biological optimization.

Contouring variation affects estimates of normal tissue complication probability for breast fibrosis after radiotherapy

BackgroundNormal tissue complication probability (NTCP) models can be useful to estimate the risk of fibrosis after breast-conserving surgery (BCS) and radiotherapy (RT) to the breast. However, they are subject to uncertainties. We present the impact of contouring variation on the prediction of fibrosis. Materials and methods280 breast cancer patients treated BCS-RT were included. Nine Clinical Target Volume (CTV) contours were created for each patient: i) CTV_crop (reference), cropped 5 mm from the skin and ii) CTV_skin, uncropped and including the skin, iii) segmenting the 95% isodose (Iso95%) and iv) 3 different auto-contouring atlases generating uncropped and cropped contours (Atlas_skin/Atlas_crop). To illustrate the impact of contour variation on NTCP estimates, we applied two equations predicting fibrosis grade ≥ 2 at 5 years, based on Lyman-Kutcher-Burman (LKB) and Relative Seriality (RS) models, respectively, to each contour. Differences were evaluated using repeated-measures ANOVA. For completeness, the association between observed fibrosis events and NTCP estimates was also evaluated using logistic regression. ResultsThere were minimal differences between contours when the same contouring approach was followed (cropped and uncropped). CTV_skin and Atlas_skin contours had lower NTCP estimates (−3.92%, IQR 4.00, p < 0.05) compared to CTV_crop. No significant difference was observed for Atlas_crop and Iso95% contours compared to CTV_crop. For the whole cohort, NTCP estimates varied between 5.3% and 49.5% (LKB) or 2.2% and 49.6% (RS) depending on the choice of contours. NTCP estimates for individual patients varied by up to a factor of 4. Estimates from “skin” contours showed higher agreement with observed events. ConclusionContour variations can lead to significantly different NTCP estimates for breast fibrosis, highlighting the importance of standardising breast contours before developing and/or applying NTCP models.

Dose-Volume Tolerance of the Brain and Predictors of Radiation Necrosis After 3-Fraction Radiosurgery for Brain Metastases: A Large Single-Institutional Analysis

Stereotactic radiosurgery (SRS) is the current standard of care in patients with brain metastases and controlled extracranial disease. Radiation necrosis (RN) is the dose-limiting side effect of SRS, but the dose constraints especially for fractionated SRS remain poorly defined. We assessed the risk of RN after 3-fraction SRS with a goal to identify specific dose-volume constraints associated with grade 3 or higher RN (G3RN). A single-institutional retrospective review of patients treated with 3-fraction SRS was performed. The primary endpoint was G3RN, which was defined as severe symptoms with evidence of necrosis on magnetic resonance imaging with perfusion and/or biopsy confirmation. Tissue volume around each target lesion was contoured, and volumetric doses per lesion were recorded. Logistic regression models were used to estimate the relationship between RN and each volumetric dose, and normal tissue complication probability modeling was performed using a modified Lyman-Kutcher-Burman model. From 2015 to 2021, 434 patients underwent 539 courses of linear accelerator-based SRS; 2518 lesions were treated. Median SRS dose was 24 Gy. Median follow-up after SRS was 7.9 months, and the median overall survival was 9 months. A total of 93 patients (17.2%) and 123 lesions (4.9%) developed any RN. Forty-two patients (7.8%) and 57 lesions (2.3%) developed G3RN. On logistic regression, V20 and V23 were best predictors of any grade RN and G3RN, respectively, with cutoff values of 4 cc, 10 cc, and 20 cc associated with <5%, <7.5%, and <10% risk of any RN, respectively, and V23 < 15 cc associated with <5% risk of G3RN. With constrained optimization of the normal tissue complication probability Lyman-Kutcher-Burman model for G3RN, we obtained a TD50 (uniform dose resulting in a 50% complication risk) of 31.4 Gy (95% CI, 27.8-35.1 Gy). In patients receiving 3-fraction SRS, G3RN was seen in 7.8% of patients, and 2.3% of the lesions were treated. V20 and V23 were the most robust dosimetric parameters associated with RN. Further studies evaluating the outcomes and RN in patients treated with fractionated SRS compared with single-fraction SRS are warranted.

The institutional experience of the implementing 4DCT in NSCLC radiotherapy planning.

The study was to evaluate the effectiveness of dose distribution of four-dimensional computed tomography (4DCT) simulation. The gross tumor volume (GTV) and clinical target volume (CTV) were contoured in all 10 respiratory phases of 4DCT in 30 patients with non-small cell lung cancer (NSCLC). Both 3D and 4D treatment plans were made individually for each patient using the planning volume (PTV). The PTV3D was taken from a single CTV plus the recommended margin, and the PTV4D was taken from the 4D internal target volume, including all 10 CTVs plus the setup margins. The mean PTV was 460 ± 179 (69-820) cm3 for 3DCT and 401 ± 167 (127-854) cm3 for 4DCT (p = 0.0018). The dose distribution (DD) of organs at risk, especially the lungs, was lower for the 4DCT simulation. The V5%, V10%, and V20% of the total lung dose for 4DCT were significantly lower for the 3DCT. However, lung V30% the heart, esophagus, and spinal cord were not significantly different. In addition, the conformity index and the dose heterogeneity index of the PTV were not significantly different. The normal tissue complication probability (NTCP) of the lung and heart was significantly lower for 4DCT than for 3DCT. The 4DCT simulation gives better results on the NTCP. The organs at risk, especially the lungs, receive a significantly lower DD compared with the 3DCT. The conformity index (CI), heterogeneity index (HI) and the DD to the heart, spinal cord, and esophagus were not significantly different between the two techniques.

Carbon-ion radiotherapy (CIRT) as treatment of pancreatic cancer at HIT: initial radiation plan analysis of the prospective phase II PACK-study

PurposeTo analyze the dose objectives and constraints applied at the prospective phase II PACK-study at Heidelberg ion therapy center (HIT) for different radiobiological models. MethodsTreatment plans of 14 patients from the PACK-study were analyzed and recomputed in terms of physical, biological dose and dose-averaged linear energy transfer (LETd). Both LEM-I (local effect model 1) and the adapted NIRS-MKM (microdosimetric kinetic model), were used for relative biological effectiveness (RBE)-weighted dose calculations (DBio|HIT and DBio|NIRS). A new constraint to the gastrointestinal (GI) tract was derived from the National Institute of Radiological Science (NIRS) clinical experience and considered for plan reoptimization (DBio|NIRS-const_48Gy and DBio|NIRS-const_50.4Gy). The Lyman-Kutcher-Burman (LKB) model of Normal Tissue Complication Probability (NTCP) for GI toxicity endpoints was computed. Furthermore, the computed LETd distribution was evaluated and correlated with Local Control (LC). ResultsOnly two patients showed a LETd98% in the GTV greater than 44 keV/μm. A HIT-dose constraint to the GI of D2cm3=48.6GyRBEHIT was derived from the NIRS experience, in alternative to the standard at HIT Dmax = 45.6 GyRBEHIT. In comparison with the original DBio|HIT,DBio|NIRS-const_48GyandDBio|NIRS-const_50.4Gy resulted in an increase in the ITV’s D98% of 8.7% and 11.3%. The NTCP calculation resulted in a probability for gastrointestinal bleeding of 4.5%, 12.3% and 13.0%, for DBio|NIRS, DBio|NIRS-const_48Gy and DBio|NIRS-const_50.4Gy, respectively. ConclusionThe results indicate that the current standards applied at HIT for CIRT closely align with the Japanese experience. However, to enhance tumor coverage, a more relaxed constraint on the GI tract may be considered. As the PACK-trial progresses, further analyses of various clinical endpoints are anticipated.

NTCP Calculations of Five Different Irradiation Techniques for the Treatment of Thymoma.

This study provided normal tissue complication probability (NTCP) calculations from photon radiotherapy techniques in eleven patients with thymoma. Five plans were created for each participant using three-dimensional conformal radiotherapy (3D-CRT), five-field intensity modulated radiotherapy (5F-IMRT), seven-field IMRT (7F-IMRT), and volumetric modulated arc therapy with full arcs (FA-VMAT) and partial arcs (PA-VMAT). The target coverage, homogeneity index and conformation number for the planning target volume (PTV) and dosimetric parameters for the organs-at-risk (OARs) were taken from the fifty-five generated plans. The patient-specific NTCP of the lungs, heart and esophagus was calculated with an in-house software tool using differential dose-volume histograms and the equivalent uniform dose model. The PTV dose metrics from 3D-CRT were inferior to those from IMRT and VMAT plans. The dose constraints for the OARs were met in all treatment plans. The NTCP range of the lungs, heart and esophagus was 0.34-0.49%, 0.03-0.06% and 0.08-0.10%, respectively. The NTCPs of the heart for the incidence of peridarditis from IMRT and VMAT were significantly smaller than those from conformal treatment (p < 0.05). The 7F-IMRT was significantly superior to FA-VMAT in reducing the NTCP of the lungs and the risk of pneumonitis (p = 0.001). Similar superiority of 5F-IMRT over PA-VMAT for lung protection was found (p = 0.009). The presented results may be employed in the selection of the appropriate irradiation technique for restricting the complications in the adjacent OARs.

Dosimetric and Radiobiological Impact of Flattening Filter‐Free Beam and Dose Calculation Algorithm Using RapidArc Plans for Cervical Cancer Treatment

AbstractPurposeThe aim of this study is to quantify the potential benefits of a flattening filter‐free (FFF) beam and implement a dose‐computation algorithm for cervical radiotherapy through dosimetric and radiobiological analyses using RapidArc.MethodsThirty‐three patients were enrolled, and four RapidArc plans were created for each patient using a dual‐arc flattening filter and 6‐MV FFF photon beams for the two calculation algorithms. Homogeneity index (HI), conformity index (CI), target coverage, monitor units (MUs), and organ‐at‐risk (OAR) dosimetric characteristics were compared between the plans. Radiobiological characteristics and normal tissue complication probability (NTCP) scores were computed for the OAR using different biological models.ResultsNo significant differences were observed in the Dmax, D98%, and CI in the planning target volume (PTV). Both computations estimated a significant difference in V95%, D2%, and HI for the PTV. Furthermore, the FFF beam showed a significant increase in the MUs and a significant reduction in V30% for the femoral heads. The NTCP score showed a significant increase in the late effects on the bladder, rectum, and bowel with FFF beams.ConclusionThe current study recommends FFF beams for better conformity, comparable dose coverage for the target, and OAR sparing invariable to the dose computation algorithm. The difference in the NTCP score for OAR was minimal with the FFF beam.

Improving knowledge-based treatment planning for lung cancer radiotherapy with automatic multi-criteria optimized training plans

Background Knowledge-based planning (KBP) is a method for automated radiotherapy treatment planning where appropriate optimization objectives for new patients are predicted based on a library of training plans. KBP can save time and improve organ at-risk sparing and inter-patient consistency compared to manual planning, but its performance depends on the quality of the training plans. We used another system for automated planning, which generates multi-criteria optimized (MCO) plans based on a wish list, to create training plans for the KBP model, to allow seamless integration of knowledge from a new system into clinical routine. Model performance was compared for KBP models trained with manually created and automatic MCO treatment plans. Material and Methods Two RapidPlan models with the same 30 locally advanced non-small cell lung cancer patients included were created, one containing manually created clinical plans (RP_CLIN) and one containing fully automatic multi-criteria optimized plans (RP_MCO). For 15 validation patients, model performance was compared in terms of dose-volume parameters and normal tissue complication probabilities, and an oncologist performed a blind comparison of the clinical (CLIN), RP_CLIN, and RP_MCO plans. Results The heart and esophagus doses were lower for RP_MCO compared to RP_CLIN, resulting in an average reduction in the risk of 2-year mortality by 0.9 percentage points and the risk of acute esophageal toxicity by 1.6 percentage points with RP_MCO. The oncologist preferred the RP_MCO plan for 8 patients and the CLIN plan for 7 patients, while the RP_CLIN plan was not preferred for any patients. Conclusion RP_MCO improved OAR sparing compared to RP_CLIN and was selected for implementation in the clinic. Training a KBP model with clinical plans may lead to suboptimal output plans, and making an extra effort to optimize the library plans in the KBP model creation phase can improve the plan quality for many future patients.