Interpretable deep learning insights: Unveiling the role of 1 Gy volume on lymphopenia after radiotherapy in breast cancer.

BACKGROUND: Lymphopenia is known for its significance on poor survivals in breast cancer patients. Considering full dosimetric data, this study aimed to develop and validate predictive models for lymphopenia after radiotherapy (RT) in breast cancer. MATERIAL AND METHODS: Patients with breast cancer treated with adjuvant RT were eligible in this multicenter study. The study endpoint was lympopenia, defined as the reduction in absolute lymphocytes and graded lymphopenia after RT. The dose-volume histogram (DVH) data of related critical structures and clinical factors were taken into account for the development of dense neural network (DNN) predictive models. The developed DNN models were validated using external patient cohorts. RESULTS: A total of 918 consecutive patients with invasive breast cancer enrolled. The training, testing, and external validating datasets consisted of 589, 203, and 126 patients, respectively. Treatment volumes at nearly all dose levels of the DVH were significant predictors for lymphopenia following RT, including volumes at very low-dose 1 Gy (V1) of organs at risk (OARs) including lung, heart and body, especially ipsilateral-lung V1. A final DNN model, combining full DVH dosimetric parameters of OARs and three key clinical factors, achieved a predictive accuracy of 75 % or higher. CONCLUSION: This study demonstrated and externally validated the significance of full dosimetric data, particularly the volume of low dose at as low as 1 Gy of critical structures on lymphopenia after radiation in patients with breast cancer. The significance of V1 deserves special attention, as modern VMAT RT technology often has a relatively high value of this parameter. Further study is warranted for RT plan optimization.

Dosimetric Effect of Thymus and Thoracic Duct on Radiation-Induced Lymphopenia in Patients With Primary Lung Cancer Who Received Thoracic Radiation.

Purpose: Radiation-induced lymphopenia is a well-recognized factor for tumor control and survival in patients with cancer. This study aimed to determine the role of radiation dose to the thymus and thoracic duct on radiation-induced lymphopenia. Methods and Materials: Patients with primary lung cancer treated with thoracic radiation therapy between May 2015 and February 2020 with whole blood count data were eligible. Clinical characteristics, including age, gender, histology, stage, chemotherapy regimen, radiation dosimetry, and absolute lymphocyte count (ALC) were collected. The thymus and thoracic duct were contoured by one investigator for consistency and checked by one senior physician. The primary endpoint was radiation-induced decrease in lymphocytes, defined as the difference in ALC (DALC) before and after radiation therapy. Results: The data of a total of 116 consecutive patients were retrospectively retrieved. Significant correlations were found between DALC and several clinical factors. These factors include stage, chemotherapy or concurrent chemoradiation, biologically effective dose (BED), mean lung dose, mean body dose, effective dose to immune cells (EDIC), mean thymus dose (MTD), and mean thoracic duct dose (MTDD) (all P < .05). Ridge regression showed that DALC = 0.0063 × BED + 0.0172 × EDIC + 0.0002 × MTD + 0.0147 × MTDD + 0.2510 (overall P = .00025 and F = 5.85). The combination model has the highest area under the curve of 0.77 (P < .001) when fitting the logistic regression model on DALC categorized as binary endpoint. The sensitivity and specificity of the combined model were 89% and 58%, respectively. Conclusions: This study demonstrated for the first time that radiation doses to the thymus and thoracic duct are strongly associated with radiation-induced lymphopenia patients with lung cancer. Further validation studies are needed to implement thymus and thoracic duct as organs at risk.

Potential Determinants for Radiation-Induced Lymphopenia in Patients With Breast Cancer Using Interpretable Machine Learning Approach.

Radiation-induced lymphopenia is known for its survival significance in patients with breast cancer treated with radiation therapy. This study aimed to evaluate the impact of radiotherapy on lymphocytes by applying machine learning strategies. We used Extreme Gradient Boosting (XGboost) to predict the event of lymphopenia (grade≥1) and conduced an independent validation. Then, we induced feature attribution analysis (Shapley additive explanation, SHAP) in explaining the XGboost models to explore the directional contribution of each feature to lymphopenia. Finally, we implemented the proof-of-concept clinical validation. The results showed that the XGboost models had rigorous generalization performances (accuracies 0.764 and ROC-AUC 0.841, respectively) in the independent cohort. The baseline lymphocyte counts are the most protective feature (SHAP = 5.226, direction of SHAP = -0.964). Baseline platelets and monocytes also played important protective roles. The usage of taxane only chemotherapy was less risk on lymphopenia than the combination of anthracycline and taxane. By the contribution analysis of dose, we identified that firstly lymphocytes were sensitive to a radiation dose less than 4Gy; secondly the irradiation volume was more important in promoting lymphopenia than the irradiation dose; thirdly the irradiation dose promoted the event of lymphopenia when the irradiation volume was fixed. Overall, our findings paved the way to clarifying the radiation dose volume effect. To avoid radiation-induced lymphopenia, irradiation volume should be kept to a minimum during the planning process, as long as the target coverage is not compromised.

Radiation Induced Lymphopenia Is Associated With the Effective Dose to the Circulating Immune Cells in Breast Cancer.

Background: Lymphopenia is a known significant factor for treatment outcome in cancer patients, with underlying risk factor poorly understood in breast cancer. We hypothesize that the effective dose to the circulating immune cells (EDIC) which was related with lymphopenia in lung cancer will also have significant effect for radiation induced lymphopenia (RIL) in patients with breast cancer. Material and Methods: Patients treated with adjuvant radiotherapy (RT) and with complete blood tests within one week from RT end/start (post/preRT) were eligible in this study. Radiation dosimetric factors were collected retrospectively, and EDIC for each patient was calculated based on the doses to lung, heart and total body according to the model description, as previously reported. RIL was defined by the CTCAE5.0 based on postRT peripheral lymphocyte count (PLC). Linear regression was first used to test the correlation between EDIC with post/preRT PLC ratio and postRT PLC, using all these as continuous variables. Normal tissue complication probability (NTCP) was used to develop models that predict the CTCAE graded RIL from EDIC. Results: A total of 735 patients were eligible. The mean post/preRT PLC ratio was 0.66 (95% CI: 0.64-0.68) and mean EDIC of breast cancer was 1.70Gy (95% CI: 1.64-1.75). Both post/preRT PLC ratio and postRT PLC were significantly correlated with EDIC (P<0.001), with R2 of 0.246. For patients with normal preRT PLC, the post/preRT PLC ratio was better associated with EDIC, and postRT PLC was expressed as PLC preRT × (0.89 - 0.16 × EDIC). For patients with preRT lymphopenia, postRT PLC was better associated with EDIC and it was 1.1 - 0.17 × EDIC. Using binned EDIC as the dose variable, the bootstrap validated NTCPs fit the data nicely with R2 of 0.93, 0.96, and 0.94 for grade-1, grade-2, and grade-3 RIL, respectively. The corresponding EDIC to induce 50% of grade-1, grade-2 and grade-3 RIL was 1.2, 2.1 and 3.7 Gy, respectively. Conclusion: EDIC is a significant factor for RIL in patients with breast cancer, and may be used to compute the risk of lymphopenia in each individual patient with the use of the conventional NTCP modeling. External validation is needed before the EDIC can be used to guide RT plan.

Higher Radiation Dose to the Immune Cells Correlates with Worse Tumor Control and Overall Survival in Patients with Stage III NSCLC: A Secondary Analysis of RTOG0617.

Background: We hypothesized that the Effective radiation Dose to the Immune Cells (EDIC) in circulating blood is a significant factor for the treatment outcome in patients with locally advanced non-small-cell lung cancer (NSCLC). Methods: This is a secondary study of a phase III trial, NRG/RTOG 0617, in patients with stage III NSCLC treated with radiation-based treatment. The EDIC was computed as equivalent uniform dose to the entire blood based on radiation doses to all blood-containing organs, with consideration of blood flow and fractionation effect. The primary endpoint was overall survival (OS), and the secondary endpoints were progression-free survival (PFS) and local progression-free survival (LPFS). The EDIC-survival relationship was analyzed with consideration of clinical significant factors. Results: A total of 456 patients were eligible. The median EDIC values were 5.6 Gy (range, 2.1-12.2 Gy) and 6.3 Gy (2.1-11.6 Gy) for the low- and high-dose groups, respectively. The EDIC was significantly associated with OS (hazard ratio [HR] = 1.12, p = 0.005) and LPFS (HR = 1.09, p = 0.02) but PFS (HR = 1.05, p = 0.17) after adjustment for tumor dose, gross tumor volume and other factors. OS decreased with an increasing EDIC in a non-linear pattern: the two-year OS decreased first with a slope of 8%/Gy when the EDIC < 6 Gy, remained relatively unchanged when the EDIC was 6-8 Gy, and followed by a further reduction with a slope of 12%/Gy when the EDIC > 8 Gy. Conclusions: The EDIC is a significant independent risk factor for poor OS and LPFS in RTOG 0617 patients with stage III NSCLC, suggesting that radiation dose to circulating immune cells is critical for tumor control. Organ at risk for the immune system should be considered during RT plan.

Monte Carlo Dose Calculation Using MRI Based Synthetic CT Generated by Fully Convolutional Neural Network for Gamma Knife Radiosurgery.

The aim of this work is to study the dosimetric effect from generated synthetic computed tomography (sCT) from magnetic resonance (MR) images using a deep learning algorithm for Gamma Knife (GK) stereotactic radiosurgery (SRS). The Monte Carlo (MC) method is used for dose calculations. Thirty patients were retrospectively selected with our institution IRB's approval. All patients were treated with GK SRS based on T1-weighted MR images and also underwent conventional external beam treatment with a CT scan. Image datasets were preprocessed with registration and were normalized to obtain similar intensity for the pairs of MR and CT images. A deep convolutional neural network arranged in an encoder-decoder fashion was used to learn the direct mapping from MR to the corresponding CT. A number of metrics including the voxel-wise mean error (ME) and mean absolute error (MAE) were used for evaluating the difference between generated sCT and the true CT. To study the dosimetric accuracy, MC simulations were performed based on the true CT and sCT using the same treatment parameters. The method produced an MAE of 86.6 ± 34.1 Hundsfield units (HU) and a mean squared error (MSE) of 160.9 ± 32.8. The mean Dice similarity coefficient was 0.82 ± 0.05 for HU > 200. The difference for dose-volume parameter D95 between the ground true dose and the dose calculated with sCT was 1.1% if a synthetic CT-to-density table was used, and 4.9% compared with the calculations based on the water-brain phantom.

Risk factors for radiation induced lymphopenia in patients with breast cancer receiving adjuvant radiotherapy.

BACKGROUND: This study aimed to investigate radiation-induced lymphopenia and its potential risk factors in patients with breast cancer receiving adjuvant radiotherapy. METHODS: Breast cancer patients received adjuvant radiotherapy (RT) at our hospital with peripheral lymphocyte counts (PLC) at pre-and immediately after RT (post-RT) were eligible. The primary endpoints were any grade of lymphopenia post-RT and nadir-PLC/pre-PLC <0.8. Patient characteristics, tumor factors, and treatment factors were collected for risk assessment. Data are presented as mean and 95% confidence interval (CI) unless otherwise specified. Matched analysis was used to compare the statistical significance between different RT techniques. RESULTS: A total of 735 consecutive patients met the study criteria. The mean PLC was 1.58×109/L before and 0.99×109/L post-RT (P<0.001). At the end of RT, 60.5% of patients had lymphopenia. Univariate and multivariable logistic analyses showed that RT technique involving RapidArc, mean lung dose, and chemotherapy were significant risk factors (P<0.05) for lymphopenia. RT technique was the only significant risk factor (P<0.05) for nadir-PLC/pre-PLC <0.8. Patients treated with RapidArc had a significantly greater reduction of PLC along with greater V5 of the lungs, even after matching mean lung dose and radiated volume. CONCLUSIONS: Lymphopenia is common in patients with breast cancer after adjuvant RT. RT technique is the only significant factor for lymphopenia and nadir-PLC/pre-PLC <0.8, suggesting the significance of RT technique choice to minimize lymphopenia and improve treatment outcomes.

Impact of effective dose to immune cells (EDIC) on lymphocyte nadir and survival in limited-stage SCLC.

BACKGROUND: Effective dose to immune cell (EDIC), an estimated radiation dose to the circulating lymphocytes, is of significance for overall survival (OS) in non-small cell lung cancer. This study aimed to validate the EDIC's OS effect on limited-stage small cell lung cancer (LS-SCLC). METHOD AND MATERIALS: This study included LS-SCLC patients received definitive chemo-radiation in one single center from 2012 to 2017. All patients had multiple complete-blood-count tests including lymphocyte count at pre-, during- and end- radiotherapy. EDIC, computed according to doses of the lung, heart, and the total body, was assessed for its correlation with lymphocyte nadir, OS and progression free survival (PFS). RESULTS: Of 503 eligible patients, the mean EDIC was 7.34 Gy. The mean lymphocyte nadir was 0.48 × 109 cells/L, significantly lower than 1.65 × 109 cells/L at pre-radiotherapy (p < 0.001). EDIC was significantly correlated with lymphocyte nadir under both univariate (p < 0.001) and multivariable linear regression (p < 0.001). Multivariable analysis showed EDIC (HR = 0.1072, p = 0.005) and lymphocyte nadir (HR = 0.345, p = 0.003) were both significant for OS. EDIC was also significant for PFS (HR = 1.046, p = 0.026). The C-indexes of OS prediction were 0.593, 0.617, 0.676, and 0.684, for lymphocyte nadir alone, EDIC alone, combined lymphocyte nadir model, and combined EDIC model, respectively. CONCLUSIONS: This study demonstrated that EDIC is an independent predictor for lymphocyte nadir, PFS and OS. EDIC may serve as a predictor for lymphocyte nadir and a surrogate marker for OS in LS-SCLC. More attention should be paid to EDIC to decease the lymphocyte toxicity and improve survival.

Patient-Specific Lymphocyte Loss Kinetics as Biomarker of Spleen Dose in Patients Undergoing Radiation Therapy for Upper Abdominal Malignancies.

PURPOSE: Radiation therapy (RT)-induced lymphopenia (RIL) is linked with inferior survival in esophageal and pancreatic cancers. Previous work has demonstrated a correlation between spleen dose and RIL risk. The present study correlates spleen dose-volume parameters with fractional lymphocyte loss rate (FLL) and total percent change in absolute lymphocyte count (%ΔALC) and suggests spleen dose constraints to reduce RIL risk. METHODS AND MATERIALS: This registry-based study included 140 patients who underwent RT for pancreatic (n = 67), gastroesophageal (n = 61), or biliary tract (n = 12) adenocarcinoma. Patient-specific parameters of lymphocyte loss kinetics, including FLL and %ΔALC, were calculated based on serial ALCs obtained during RT. Spearman's rho was used to correlate spleen dose-volume parameters with %ΔALC, end-treatment ALC, and FLL. Multivariable logistic regression was used to identify predictors of ≥grade 3 and grade 4 RIL. RESULTS: Spleen dose-volume parameters, including mean spleen dose (MSD), all correlated with %ΔALC, end-treatment ALC, and FLL. Controlling for baseline ALC and planning target volume (PTV), an increase in any spleen dose-volume parameter increased the odds of developing ≥grade 3 lymphopenia. Each 1-Gy increase in MSD increased the odds of ≥grade 3 RIL by 18.6%, and each 100-cm3 increase in PTV increased the odds of ≥grade 3 lymphopenia by 20%. Patients with baseline ALC < 1500 cells/µL had a high risk of ≥grade 3 RIL regardless of MSD or PTV. FLL was an equally good predictor of ≥grade 3 lymphopenia as any spleen dose-volume parameter. CONCLUSIONS: In patients undergoing RT for upper abdominal malignancies, higher spleen dose is associated with higher per-fraction lymphocyte loss rates, higher total %ΔALC, and increased odds of severe lymphopenia. Spleen dose constraints should be individualized based on baseline ALC and PTV size to minimize RIL risk, although our findings require validation in larger, ideally prospective data sets.

Risk factors for symptomatic radiation pneumonitis after stereotactic body radiation therapy (SBRT) in patients with non-small cell lung cancer.

BACKGROUND AND PURPOSE: Radiation pneumonitis (RP) can be a potential fatal toxicity of stereotactic body radiation therapy (SBRT) for medically inoperable non-small cell lung cancer (NSCLC). This study aimed to examine the risk factors that predict RP and explore dosimetric tolerance for safe practice in a large institutional series of NSCLC patients. MATERIALS AND METHODS: Patients with early-stage and locally recurrent NSCLC who received lung SBRT between 2002 and 2015 formed the study population. The primary endpoint was grade 2 or above radiation pneumonitis (RP2). Lungs were re-contoured consistently by one radiation oncologist according to the RTOG atlas for organs at risk. Dosimetric factors were computed consistently with exclusion of gross tumor volume of either ipsilateral, contralateral, or total lungs. RESULTS: A total of 339 patients were eligible. With a median follow-up of 47 months, RP2 was recorded in 10% patients. History of respiratory comorbidity, previous thoracic radiation, right lung location, mean lung doses of total or ipsilateral lung, and total lung volume receiving 20 Gy were all significantly associated with the risk of RP2. The dosimetric parameters of contralateral lung, including mean dose and volume receiving more than 5, 10, and 20 Gy, were not significantly associated with RP2 (ps > 0.05). A model of combining significant clinical and dosimetric factors had a predictive accuracy AUC of 0.76. According to this model, RP2 can be limited to <10% should the patient have no previous lung radiation and the mean dose of total and ipsilateral lungs be kept less than 6 Gy and 20 Gy, respectively. CONCLUSION: Dosimetric factors of total or ipsilateral lung together with important clinical factors were significant risk factors for symptomatic radiation pneumonitis after SBRT. Constraining mean lung dose can limit clinically significant lung toxicity.

Lymphopenia and Radiation Dose to Circulating Lymphocytes With Neoadjuvant Chemoradiation in Esophageal Squamous Cell Carcinoma.

PURPOSE: We hypothesized that radiation-induced lymphopenia could be predicted by the effective dose to the circulating immune cells (EDIC) in advanced esophageal squamous cell carcinoma treated with trimodality therapy according to the Dutch ChemoRadiotherapy for Oesophageal cancer followed by Surgery Study (CROSS) trial regimen. To test this hypothesis, we examined the effect of EDIC on the degree of lymphocyte drop (lymphocyte nadir). METHODS AND MATERIALS: Patients with advanced nonmetastatic esophageal squamous cell carcinoma treated in a single tertiary cancer center from 2012 to 2018 were eligible for this study. All patients had to have a radiation therapy plan available for EDIC computation and received neoadjuvant chemoradiation according to the Dutch CROSS trial regimen before radical esophagectomy. The EDIC was calculated as a function of integral doses to the lung, heart, and total body with a verified mathematical model. The association between EDIC and lymphocyte nadir was studied, and the relationships of overall survival (OS) with lymphocyte nadir and EDIC were assessed using multivariable Cox regression model. RESULTS: This analysis included 92 eligible consecutive patients (77 men and 15 women). The mean EDIC was 2.8 Gy (range, 0.6-4.4). EDIC was significantly correlated with lymphocyte nadir (Spearman coefficient = -0.505; P < .01), and lymphocyte nadir was a significant independent factor for shorter OS (hazard ratio = 0.63; P < .001). Lymphocyte nadir was also the most significant factor in determining OS among other clinical parameters. Exploratory analysis showed significant OS differences between EDIC groups (<2, 2-4, and >4 Gy). The 2-year OS rates were 66.7%, 42.7%, and 16.7% for EDIC <2, 2 to 4, and >4 Gy, respectively. CONCLUSIONS: There was a significant correlation between radiation dose to circulating immune cells and lymphocyte nadir, which in turn affected OS in patients with advanced nonmetastatic esophageal squamous cell carcinoma treated by trimodality therapy.

Ultra-high dose rate effect on circulating immune cells: A potential mechanism for FLASH effect?

PURPOSE: "FLASH" radiotherapy (RT) is a potential paradigm-changing RT technology with marked tumor killing and normal tissue sparing. However, the mechanism of the FLASH effect is not well understood. We hypothesize that the ultra-high dose rate FLASH-RT significantly reduces the killing of circulating immune cells which may partially contribute to the reported FLASH effect. METHODS: This computation study directly models the effect of radiation dose rate on the killing of circulating immune cells. The model considers an irradiated volume that takes up A% of cardiac output and contains B% of total blood. The irradiated blood volume and dose were calculated for various A%, B%, blood circulation time, and irradiation time (which depends on the dose rate). The linear-quadratic model was used to calculate the extent of killing of circulating immune cells at ultra-high vs. conventional dose rates. RESULTS: A strong sparing effect on circulating blood cells by FLASH-RT was noticed; i.e., killing of circulating immune cells reduced from 90% to 100% at conventional dose rates to 5-10% at ultra-high dose rates. The threshold FLASH dose rate was determined to be ~40 Gy/s for mice in an average situation (A% = 50%), consistent with the reported FLASH dose rate in animal studies, and it was approximately one order of magnitude lower for humans than for mice. The magnitude of this sparing effect increased with the dose/fraction, reached a plateau at 30-50 Gy/fraction, and almost completely vanished at 2 Gy/fraction. CONCLUSION: We have calculated a strong sparing effect on circulating immune cells by FLASH-RT, which may contribute to the reported FLASH effects in animal studies.

Central Airway Toxicity After High Dose Radiation: A Combined Analysis of Prospective Clinical Trials for Non-Small Cell Lung Cancer.

PURPOSE: To study the dosimetric risk factors for radiation-induced proximal bronchial tree (PBT) toxicity in patients treated with radiation therapy for non-small cell lung cancer (NSCLC). METHODS AND MATERIALS: Patients with medically inoperable or unresectable NSCLC treated with conventionally fractionated 3-dimensional conformal radiation therapy (3DCRT) in prospective clinical trials were eligible for this study. Proximal bronchial tree (PBT) and PBT wall were contoured consistently per RTOG 1106 OAR-Atlas. The dose-volume histograms (DVHs) of physical prescription dose (DVHp) and biological effective dose (α/ß = 2.5; DVH2.5) were generated, respectively. The primary endpoint was PBT toxicities, defined by CTCAE 4.0 under the terminology of bronchial stricture/atelectasis. RESULTS: Of 100 patients enrolled, with a median follow-up of 64 months (95% confidence interval [CI], 50-78), 73% received 70 Gy or greater and 17% developed PBT toxicity (grade 1, 8%; grade 2, 6%; grade 3, 0%; and grade 4, 3%). The median time interval between RT initiation and onset of PBT toxicity was 8.4 months (95% CI, 4.7-44.1). The combined DVHs showed that no patient with a PBT maximum physical dose <65 Gy developed any PBT toxicity. Cox proportional hazards analysis and receiver operating characteristic analysis demonstrated that V75 of PBT was the most significant dosimetric parameter for both grade 1+ (P = .035) and grade 2+ (P = .037) PBT toxicities. The dosimetric thresholds for V75 of PBT were 6.8% and 11.9% for grade 1+ and grade 2+ PBT toxicity, respectively. CONCLUSIONS: V75 of PBT appeared be the most significant dosimetric parameter for PBT toxicity after conventionally fractionated thoracic 3DCRT. Constraining V75 of PBT can limit clinically significant PBT toxicity.

The impact of the effective dose to immune cells on lymphopenia and survival of esophageal cancer after chemoradiotherapy.

PURPOSE: To test the hypothesis that effective dose to circulating immune cells (EDIC) impacts the severity of radiation-induced lymphopenia and clinical outcomes of esophageal cancer patients treated with concurrent chemoradiotherapy (CCRT). MATERIAL AND METHODS: 488 esophageal cancer patients treated with CCRT with and without surgery were analyzed. The EDIC model considers the exposure of circulating immune cells as the proportion of blood flow to lung, heart, liver, and the volume of the exposed area of the body, with the basis of mean lung dose (MLD), mean heart dose (MHD), mean liver dose (MlD), and integral dose (ITD) of the body region scanned, calculated as: EDIC=0.12∗MLD+0.08∗MHD+0.15∗0.85∗MlD∗n451/2+0.45+0.35∗0.85∗nk1/2∗ITD62∗103 Where n is the fraction number. Correlations of EDIC with overall survival (OS), progression free survival (PFS), distant metastasis free survival (DMFS), and locoregional control (LRC) rates were analyzed using both univariable and multivariable Cox models. Lymphopenia during CCRT was graded according to Common Terminology Criteria for Adverse Events version 4.0. RESULTS: Grade 4 lymphopenia resulted in inferior clinical outcomes, including OS, PFS, and DMFS. The median EDIC was 3.6 Gy (range, 0.8-6.0 Gy). Higher EDIC was strongly associated with severe lymphopenia, particularly when EDIC was above 4 Gy. Patients with EDIC > 4.0 Gy had more G4 lymphopenia than those with EDIC ≤ 4.0 Gy (67.3% vs. 40.8%; P < 0.001). On multivariate analysis, increasing EDIC was independently and inversely associated with worse OS, PFS, and DMFS. CONCLUSION: EDIC can be recommended as a useful tool to predict lymphopenia and inferior clinical outcomes, and it should be minimized below 4 Gy.

Weighted-Support Vector Machine Learning Classifier of Circulating Cytokine Biomarkers to Predict Radiation-Induced Lung Fibrosis in Non-Small-Cell Lung Cancer Patients.

BACKGROUND: Radiation-induced lung fibrosis (RILF) is an important late toxicity in patients with non-small-cell lung cancer (NSCLC) after radiotherapy (RT). Clinically significant RILF can impact quality of life and/or cause non-cancer related death. This study aimed to determine whether pre-treatment plasma cytokine levels have a significant effect on the risk of RILF and investigate the abilities of machine learning algorithms for risk prediction. METHODS: This is a secondary analysis of prospective studies from two academic cancer centers. The primary endpoint was grade≥2 (RILF2), classified according to a system consistent with the consensus recommendation of an expert panel of the AAPM task for normal tissue toxicity. Eligible patients must have at least 6 months' follow-up after radiotherapy commencement. Baseline levels of 30 cytokines, dosimetric, and clinical characteristics were analyzed. Support vector machine (SVM) algorithm was applied for model development. Data from one center was used for model training and development; and data of another center was applied as an independent external validation. RESULTS: There were 57 and 37 eligible patients in training and validation datasets, with 14 and 16.2% RILF2, respectively. Of the 30 plasma cytokines evaluated, SVM identified baseline circulating CCL4 as the most significant cytokine associated with RILF2 risk in both datasets (P = 0.003 and 0.07, for training and test sets, respectively). An SVM classifier predictive of RILF2 was generated in Cohort 1 with CCL4, mean lung dose (MLD) and chemotherapy as key model features. This classifier was validated in Cohort 2 with accuracy of 0.757 and area under the curve (AUC) of 0.855. CONCLUSIONS: Using machine learning, this study constructed and validated a weighted-SVM classifier incorporating circulating CCL4 levels with significant dosimetric and clinical parameters which predicts RILF2 risk with a reasonable accuracy. Further study with larger sample size is needed to validate the role of CCL4, and this SVM classifier in RILF2.

A Validation Study on IDO Immune Biomarkers for Survival Prediction in Non-Small Cell Lung Cancer: Radiation Dose Fractionation Effect in Early-Stage Disease.

PURPOSE: We recently reported that indoleamine 2, 3-dioxygenase (IDO) activity is significantly correlated with more distant metastasis and worse survival. The present study examined whether radiotherapy (RT) dose fractionation correlates with IDO-mediated immune activity in patients with early-stage NSCLC.Methods: Patients with newly diagnosed stage I-II NSCLC treated with either conventionally fractionated 3-dimensional conformal radiotherapy (3DCRT) or stereotactic body radiotherapy (SBRT) were analyzed. Levels of two key molecules associated with the IDO immune checkpoint, serum kynurenine and the kynurenine:tryptophan ratio (K:T ratio), were measured at pre-RT, during-RT, and 3-month post-RT. The relationship between disease control outcomes [overall survival (OS), progression free survival, and local/regional/distant failure rates] and absolute levels of these markers, as well as dynamic changes in their levels during RT, was studied. RESULTS: Fifty-six patients (SBRT = 28, 3DCRT = 28) with early-stage NSCLC were studied. In all patients, higher kynurenine post-RT was significantly associated with worse OS ([HR, 1.25; 95% confidence interval (CI), 1.01-1.55; P = 0.044). No statistically significant differences in absolute kynurenine levels or the K:T ratio were observed in patients treated with 3DCRT or SBRT at any of the three time points. However, the absolute kynurenine levels rose significantly more post-RT in the 3DCRT patients with a median increase 0.721 ng/mL, compared to that of SBRT patients (0.115 ng/mL); P = 0.022. CONCLUSIONS: This study validated that elevated IDO activity correlated with worse survival outcomes in patients with early-stage NSCLC treated with definitive RT. Hypofractionated SBRT may have less immunosuppressive effect than 3DCRT, as measured by IDO.

Comprehensive Analysis of the Kinetics of Radiation-Induced Lymphocyte Loss in Patients Treated with External Beam Radiation Therapy.

Radiation-induced lymphopenia (RIL) is associated with worse survival in patients with solid tumors, as well as lower response rates to checkpoint inhibitors. While single-fraction total-body irradiation is known to result in exponential decreases in the absolute lymphocyte count (ALC), the kinetics of lymphocyte loss after focal fractionated exposures have not previously been characterized. In the current study, lymphocyte loss kinetics was analyzed among patients undergoing focal fractionated radiotherapy for clinical indications. This registry-based study included 419 patients who received either total-body irradiation (TBI; n = 30), stereotactic body radiation therapy (SBRT; n = 73) or conventionally fractionated chemoradiation therapy (CFRT; n = 316). For each patient, serial ALCs were plotted against radiotherapy fraction number. The initial three weeks of treatment for CFRT patients and the entirety of treatment for SBRT and TBI patients were fit to exponential decay in the form ALC(x) = ae-bx, where ALC(x) is the ALC after x fractions. From those fits, fractional lymphocyte loss (FLL) was calculated as FLL = (1 - e-b) * 100, and multivariable regression was performed to identify significant correlates of FLL. Median linearized R2 when fitting the initial fractions was 0.98, 0.93 and 0.97 for patients receiving TBI, SBRT and CFRT, respectively. In CFRT patients, apparent ALC loss rate slowed after week 3. Fitting ALC loss over the entire CFRT course therefore required the addition of a constant term, "c". For TBI and SBRT patients, treatment ended during the pure exponential decay phase. Initial FLL varied significantly with treatment technique. Mean FLL was 35.5%, 24.3% and 10.77% for patients receiving TBI, SBRT and CFRT, respectively (P < 0.001). Significant correlates of FLL varied by site and included field size, dose per fraction, mean spleen dose, chemotherapy backbone and age. Finally, total percentage ALC loss during radiotherapy was highly correlated with FLL (P < 0.001). Lymphocyte depletion kinetics during the initial phase of fractionated radiotherapy are characterized by pure exponential decay. Initial FLL is strongly correlated with radiotherapy planning parameters and total percentage ALC loss. The two groups with the highest FLL received no concurrent chemotherapy, suggesting that ALC loss can be a consequence of radiotherapy alone. This work may assist in selecting patients for adaptive radiotherapy approaches to mitigate RIL risk.

A framework for modeling radiation induced lymphopenia in radiotherapy.

INTRODUCTION: Associations between radiation-induced lymphopenia (RIL) and survival have been extensively reported. However, the immune system is not considered as an organ-at-risk (OAR) in radiotherapy. This study aimed to develop the framework of an immune OAR model that may be utilized to predict and minimize RIL. METHODS: A dynamic model was first developed for lymphocyte trafficking among 5 compartments of the immune system. Radiation dose to the circulating lymphocytes in each compartment was calculated based on the doses to fixed structures of each immune compartment and blood flow patterns. A RIL model was developed based on lymphocyte dynamics, lymphocyte radiosensitivity and reproductivity, and the dose to the lymphocytes. The model was tested in 51 patients by fitting it to weekly-measured absolute lymphocyte counts (ALC) for each patient, considering lymphocyte radiosensitivity and reproductivity as patient-dependent fitting parameters. RESULTS: The fitting was almost perfect for 20 patients, with sum of square of errors (SSE) between measured and predicted ALCs < 0.5. It was acceptable for another 27 patients, with SSE = 0.5~4.0. Only 4 patients had SSE > 4.0. The fitting also provided a method of in vivo estimation of radiosensitivity (α) for each patient. The median α was 0.40 Gy-1 for the 51 patients, consistent with in vitro measured data of 0.41 Gy-1 in the literature. CONCLUSION: We have presented a framework of developing an immune OAR model that has the potential to predict and minimize RIL in radiotherapy.

Machine Learning to Build and Validate a Model for Radiation Pneumonitis Prediction in Patients with Non-Small Cell Lung Cancer.

PURPOSE: Radiation pneumonitis is an important adverse event in patients with non-small cell lung cancer (NSCLC) receiving thoracic radiotherapy. However, the risk of radiation pneumonitis grade ≥ 2 (RP2) has not been well predicted. This study hypothesized that inflammatory cytokines or the dynamic changes during radiotherapy can improve predictive accuracy for RP2. EXPERIMENTAL DESIGN: Levels of 30 inflammatory cytokines and clinical information in patients with stages I-III NSCLC treated with radiotherapy were from our prospective studies. Statistical analysis was used to select predictive cytokine candidates and clinical covariates for adjustment. Machine learning algorithm was used to develop the generalized linear model for predicting risk RP2. RESULTS: A total of 131 patients were eligible and 17 (13.0%) developed RP2. IL8 and CCL2 had significantly (Bonferroni) lower expression levels in patients with RP2 than without RP2. But none of the changes in cytokine levels during radiotherapy was significantly associated with RP2. The final predictive GLM model for RP2 was established, including IL8 and CCL2 at baseline level and two clinical variables. Nomogram was constructed based on the GLM model. The model's predicting ability was validated in the completely independent test set (AUC = 0.863, accuracy = 80.0%, sensitivity = 100%, specificity = 76.5%). CONCLUSIONS: By machine learning, this study has developed and validated a comprehensive model integrating inflammatory cytokines with clinical variables to predict RP2 before radiotherapy that provides an opportunity to guide clinicians.

An Integrated Framework Based on Full Monte Carlo Simulations for Double-Scattering Proton Therapy.

PURPOSE: We developed an integrated framework that employs a full Monte Carlo (MC) model for treatment-plan simulations of a passive double-scattering proton system. MATERIALS AND METHODS: We have previously validated a virtual machine source model for full MC proton-dose calculations by comparing the percentage of depth-dose curves, spread-out Bragg peaks, and lateral profiles against measured commissioning data. This study further expanded our previous work by developing an integrate framework that facilitates its clinical use. Specifically, we have (1) constructed patient-specific applicator and compensator numerically from the plan data and incorporated them into the beamline, (2) created the patient anatomy from the computed tomography image and established the transformation between patient and machine coordinate systems, and (3) developed a graphical user interface to ease the whole process from importing the treatment plan in the Digital Imaging and Communications in Medicine format to parallelization of the MC calculations. End-to-end tests were performed to validate the functionality, and 3 clinical cases were used to demonstrate clinical utility of the framework. RESULTS: The end-to-end tests demonstrated that the framework functioned correctly for all tested functionality. Comparisons between the treatment planning system calculations and MC results in 3 clinical cases revealed large dose difference up to 17%, especially in the beam penumbra and near the end of beam range. The discrepancy likely originates from a variety of sources, such as the dose algorithms, modeling of the beamline, and the dose metric. The agreement for other regions was acceptable. CONCLUSION: An integrated framework was developed for full MC simulations of double-scattering proton therapy. It can be a valuable tool for dose verification and plan evaluation.