Machine Learning-Based Prediction of Hospitalization During Chemoradiotherapy With Daily Step Counts.

Importance: Toxic effects of concurrent chemoradiotherapy (CRT) can cause treatment interruptions and hospitalizations, reducing treatment efficacy and increasing health care costs. Physical activity monitoring may enable early identification of patients at high risk for hospitalization who may benefit from proactive intervention. Objective: To develop and validate machine learning (ML) approaches based on daily step counts collected by wearable devices on prospective trials to predict hospitalizations during CRT. Design, Setting, and Participants: This study included patients with a variety of cancers enrolled from June 2015 to August 2018 on 3 prospective, single-institution trials of activity monitoring using wearable devices during CRT. Patients were followed up during and 1 month following CRT. Training and validation cohorts were generated temporally, stratifying for cancer diagnosis (70:30). Random forest, neural network, and elastic net-regularized logistic regression (EN) were trained to predict short-term hospitalization risk based on a combination of clinical characteristics and the preceding 2 weeks of activity data. To predict outcomes of activity data, models based only on activity-monitoring features and only on clinical features were trained and evaluated. Data analysis was completed from January 2022 to March 2023. Main Outcomes and Measures: Model performance was evaluated in terms of the receiver operating characteristic area under curve (ROC AUC) in the stratified temporal validation cohort. Results: Step counts from 214 patients (median [range] age, 61 [53-68] years; 113 [52.8%] male) were included. EN based on step counts and clinical features had high predictive ability (ROC AUC, 0.83; 95% CI, 0.66-0.92), outperforming random forest (ROC AUC, 0.76; 95% CI, 0.56-0.87; P = .02) and neural network (ROC AUC, 0.80; 95% CI, 0.71-0.88; P = .36). In an ablation study, the EN model based on only step counts demonstrated greater predictive ability than the EN model with step counts and clinical features (ROC AUC, 0.85; 95% CI, 0.70-0.93; P = .09). Both models outperformed the EN model trained on only clinical features (ROC AUC, 0.53; 95% CI, 0.31-0.66; P < .001). Conclusions and Relevance: This study developed and validated a ML model based on activity-monitoring data collected during prospective clinical trials. Patient-generated health data have the potential to advance predictive ability of ML approaches. The resulting model from this study will be evaluated in an upcoming multi-institutional, cooperative group randomized trial.

Perioperative Immunotherapy for Non-Small Cell Lung Cancer: Practical Application of Emerging Data and New Challenges.

Immune checkpoint inhibition, with or without chemotherapy, is an established standard of care for metastatic non-small cell lung cancer (NSCLC). For locally advanced NSCLC treated with chemoradiotherapy, consolidation immunotherapy has dramatically improved outcomes. Recently, immunotherapy has also been established as a valuable component of treatment for resectable NSCLC with pembrolizumab, atezolizumab, and nivolumab all approved for use in this setting. As more results read out from ongoing perioperative clinical trials, navigating treatment options will likely become increasingly complex for the practicing oncologist. In this paper, we distill key outcomes from major perioperative trials and highlight current knowledge gaps. In addition, we provide practical considerations for incorporating perioperative immunotherapy into the clinical management of operable NSCLC.

Bronchoscopic deployment and implantation of Diffusing alpha-emitters Radiation Therapy into the lung and mediastinum for treatment of lung cancer: a pre-clinical safety and feasibility study.

Background: Radiotherapy is a standard treatment modality in cancer therapy, particularly for lung cancer. Diffusing alpha-emitters Radiation Therapy sources (hereafter, "Alpha DaRTs") are fixed with Ra-244 (half-life =3.6 days) that releases alpha-emitting atoms into the tumor tissue to an effective range of a few millimeters. Methods: The feasibility, usability, and safety of Alpha DaRTs deployment and implantation via bronchoscopy into the lung parenchyma and mediastinum in a big animal model of healthy swine was studied in two phases: (I) inert and (II) active Alpha DaRTs deployment. The Alpha DaRTs were inserted in both individual and cluster patterns based on a predefined plan. Swine health was monitored throughout the study. The usability of bronchoscopic deployment and implantation was evaluated using a user questionnaire. The movement and migration of the Alpha DaRTs were assessed. Necropsy was performed, and lungs were evaluated via gross pathology and histopathology. Results: A total of 158 Alpha DaRTs were inserted successfully in the lung parenchyma and mediastinum of five swine in two phases. It was possible to deliver and place the Alpha DaRTs in clusters of no more than 4 mm distance between the Alpha DaRTs. No adverse event or change in the health and general condition of animals was observed. Hematologic evaluation did not show any clinically significant abnormality related to the Alpha DaRTs. Histopathology demonstrated local mild inflammatory changes, minimal fibrosis, and dystrophic mineralization with giant cells. Minimal movement and no migration of Alpha DaRTs were observed. Conclusions: Bronchoscopic deployment of Alpha DaRTs in the lung parenchyma and mediastinum of the porcine animal is feasible, precise, and safe.

A proposed multi-domain, digital model for capturing functional status and health-related quality of life in oncology.

Whereas traditional oncology clinical trial endpoints remain key for assessing novel treatments, capturing patients' functional status is increasingly recognized as an important aspect for supporting clinical decisions and assessing outcomes in clinical trials. Existing functional status assessments suffer from various limitations, some of which may be addressed by adopting digital health technologies (DHTs) as a means of collecting both objective and self-reported outcomes. In this mini-review, we propose a device-agnostic multi-domain model for oncology capturing functional status, which includes physical activity data, vital signs, sleep variables, and measures related to health-related quality of life enabled by connected digital tools. By using DHTs for all aspects of data collection, our proposed model allows for high-resolution measurement of objective data as patients navigate their daily lives outside of the hospital setting. This is complemented by electronic questionnaires administered at intervals appropriate for each instrument. Preliminary testing and practical considerations to address before adoption are also discussed. Finally, we highlight multi-institutional pre-competitive collaborations as a means of successfully transitioning the proposed digitally enabled data collection model from feasibility studies to interventional trials and care management.

Selective Personalized RadioImmunotherapy for Locally Advanced Non-Small-Cell Lung Cancer Trial (SPRINT).

PURPOSE: Standard therapy for locally advanced non-small-cell lung cancer (LA-NSCLC) is concurrent chemoradiotherapy followed by adjuvant durvalumab. For biomarker-selected patients with LA-NSCLC, we hypothesized that sequential pembrolizumab and risk-adapted radiotherapy, without chemotherapy, would be well-tolerated and effective. METHODS: Patients with stage III NSCLC or unresectable stage II NSCLC and an Eastern Cooperative Oncology Group performance status of 0-1 were eligible for this trial. Patients with a PD-L1 tumor proportion score (TPS) of ≥50% received three cycles of induction pembrolizumab (200 mg, once every 21 days), followed by a 20-fraction course of risk-adapted thoracic radiotherapy (55 Gy delivered to tumors or lymph nodes with metabolic volume exceeding 20 cc, 48 Gy delivered to smaller lesions), followed by consolidation pembrolizumab to complete a 1-year treatment course. The primary study end point was 1-year progression-free survival (PFS). Secondary end points included response rates after induction pembrolizumab, overall survival (OS), and adverse events. RESULTS: Twenty-five patients with a PD-L1 TPS of ≥50% were enrolled. The median age was 71, most patients (88%) had stage IIIA or IIIB disease, and the median PD-L1 TPS was 75%. Two patients developed disease progression during induction pembrolizumab, and two patients discontinued pembrolizumab after one infusion because of immune-related adverse events. Using RECIST criteria, 12 patients (48%) exhibited a partial or complete response after induction pembrolizumab. Twenty-four patients (96%) received definitive thoracic radiotherapy. The 1-year PFS rate is 76%, satisfying our efficacy objective. One- and 2-year OS rates are 92% and 76%, respectively. The most common grade 3 adverse events were colitis (n = 2, 8%) and esophagitis (n = 2, 8%), and no higher-grade treatment-related adverse events have occurred. CONCLUSION: Pembrolizumab and risk-adapted radiotherapy, without chemotherapy, are a promising treatment approach for patients with LA-NSCLC with a PD-L1 TPS of ≥50%.

Safety and efficacy of stereotactic body proton therapy for high-risk lung tumors.

Purpose: Stereotactic body proton therapy (SBPT) is an emerging treatment strategy for lung tumors that aims to combine the excellent local control benefits of ultra-hypofractionation with the physical advantages of protons, which reduce the integral dose to organs at risk (OARs) compared to photons. To date, however, very little data delivering SBPT in 5 or fewer fractions to lung tumors have been reported. Given that photon stereotactic body radiation therapy can struggle to deliver ablative doses to high-risk tumors (i.e., central/ultra-central location, prior in-field radiation, tumor size >5 cm, or the presence of severe pulmonary comorbidities) while adhering to OAR dose constraints, we hypothesized that SBPT would be an effective alternative for patients with high-risk tumors. Methods and Materials: Twenty-seven high-risk patients with 29 lung tumors treated with SBPT at the New York Proton Center between December 2019 and November 2022 were retrospectively identified. Patients were divided into three major subgroups: early-stage non-small cell lung cancer (NSCLC), locally recurrent NSCLC, and metastatic cancer from lung cancer or other histologies. Patient characteristics were reported using descriptive statistics, actuarial methods were used to quantify disease control rates, and toxicities were scored using CTCAE v 5.0. Results: The most common high-risk indications for SBPT were central/ultra-central tumor location (69.0%), severe COPD (48.1%), reirradiation (44.4%), significant pulmonary fibrosis (22.2%), and large tumor size > 5 cm (18.5%). In total, 96.6% of tumors were fully covered by the prescription dose without compromising target coverage. Three-year actuarial rates of local control for early-stage NSCLC, locally recurrent NSCLC, and metastatic patients were 89%, 100%, and 43%, respectively. Three-year actuarial rates of regional control were 89%, 67%, and 86%. Three-year actuarial rates of distant metastasis-free survival were 79%, 100%, and 0%. Two patients (7.4%), both of whom had clinically significant baseline interstitial lung disease and pre-treatment continuous oxygen demand, experienced grade ≥2 pulmonary toxicity (1 grade 3, 1 grade 5). There were no acute or late grade ≥2 toxicities related to esophagitis, cardiac injury, airway injury, pulmonary fibrosis, bronchopulmonary hemorrhage or brachial plexopathy. Conclusions: In the largest study of proton SBRT reported to date, SBPT has a favorable toxicity profile while being an effective approach for treating most high-risk tumors without requiring dose de-escalation or compromising tumor coverage and warrants further investigation.

Advances in treatment planning and management for the safety and accuracy of lung stereotactic body radiation therapy using proton pencil beam scanning: Simulation, planning, quality assurance, and delivery recommendations.

This study presents the clinical experiences of the New York Proton Center in employing proton pencil beam scanning (PBS) for the treatment of lung stereotactic body radiation therapy. It encompasses a comprehensive examination of multiple facets, including patient simulation, delineation of target volumes and organs at risk, treatment planning, plan evaluation, quality assurance, and motion management strategies. By sharing the approaches of the New York Proton Center and providing recommendations across simulation, treatment planning, and treatment delivery, it is anticipated that the valuable experience will be provided to a broader proton therapy community, serving as a useful reference for future clinical practice and research endeavors in the field of stereotactic body proton therapy for lung tumors.

Proton Therapy for Unresectable and Medically Inoperable Locally Advanced Pancreatic Cancer: Results From a Multi-Institutional Prospective Registry.

Purpose: Compared with photon-based techniques, proton beam radiation therapy (PBT) may improve the therapeutic ratio of radiation therapy (RT) for locally advanced pancreatic cancer (LAPC), but available data have been limited to single-institutional experiences. This study examined the toxicity, survival, and disease control rates among patients enrolled in a multi-institutional prospective registry study and treated with PBT for LAPC. Methods and Materials: Between March 2013 and November 2019, 19 patients with inoperable disease across 7 institutions underwent PBT with definitive intent for LAPC. Patients received a median radiation dose/fractionation of 54 Gy/30 fractions (range, 50.4-60.0 Gy/19-33 fractions). Most received prior (68.4%) or concurrent (78.9%) chemotherapy. Patients were assessed prospectively for toxicities using National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0. Kaplan-Meier analysis was used to analyze overall survival, locoregional recurrence-free survival, time to locoregional recurrence, distant metastasis-free survival, and time to new progression or metastasis for the adenocarcinoma cohort (17 patients). Results: No patients experienced grade ≥3 acute or chronic treatment-related adverse events. Grade 1 and 2 adverse events occurred in 78.7% and 21.3% of patients, respectively. Median overall survival, locoregional recurrence-free survival, distant metastasis-free survival, and time to new progression or metastasis were 14.6, 11.0, 11.0, and 13.9 months, respectively. Freedom from locoregional recurrence at 2 years was 81.7%. All patients completed treatment with one requiring a RT break for stent placement. Conclusions: Proton beam RT for LAPC offered excellent tolerability while still maintaining disease control and survival rates comparable with dose-escalated photon-based RT. These findings are consistent with the known physical and dosimetric advantages offered by proton therapy, but the conclusions are limited owing to the patient sample size. Further clinical studies incorporating dose-escalated PBT are warranted to evaluate whether these dosimetric advantages translate into clinically meaningful benefits.

Exploring causes and consequences of early discontinuation of durvalumab after chemoradiotherapy for non-small cell lung cancer.

Introduction: For most locally advanced non-small cell lung cancer (LA-NSCLC) patients who complete definitive chemoradiotherapy (CRT) and do not experience disease progression, one year of adjuvant durvalumab is recommended. Here, we explore causes and consequences of early durvalumab discontinuation. Materials and Methods: We reviewed patients treated for LA-NSCLC with definitive CRT who began adjuvant durvalumab between 2017 and 2021. Duration of durvalumab receipt and causes for early discontinuation were tabulated. Logistic regression models were utilized to evaluate predictors of early durvalumab discontinuation. Landmark analyses were performed to explore associations between early durvalumab discontinuation and clinical outcomes (progression-free survival (PFS), overall survival (OS)). Results: Fifty-nine patients were included. Forty-one patients (69%) discontinued durvalumab early, most commonly for disease progression (n = 14) or lung toxicity (n = 10). Multivariable analysis revealed mean heart radiotherapy dose (MHD) was associated with risk of durvalumab discontinuation from progression (HR = 2.34 per 10 Gy, p = 0.052), and there was a trend suggesting an association between MHD and risk of durvalumab discontinuation from lung toxicity (HR = 2.16 per 10 Gy, p = 0.126). Median PFS duration following durvalumab initiation was 14 months, and median OS duration was 32 months. Landmark analyses that excluded patients with progression or death within one year of durvalumab initiation demonstrated improved outcomes for patients who completed one year of durvalumab (2-year PFS 100% v. 40%, p < 0.001; 2-year OS 100% v. 67%, p = 0.862). Improved outcomes were observed for patients who received MHD below the cohort median (9.3 Gy) compared to patients with higher MHD (median PFS 32 months v. 8 months, p < 0.001; 2-year OS 69% v. 44%, p = 0.088). Conclusion: For LA-NSCLC patients treated with CRT followed by immunotherapy, extent of cardiac irradiation may be a risk factor for immunotherapy discontinuation, disease recurrence, and death.

Bacterial Decolonization for Prevention of Radiation Dermatitis: A Randomized Clinical Trial.

Importance: Evidence-based approaches for the prevention of acute radiation dermatitis (ARD) are limited, and additional strategies are necessary to optimize care. Objective: To determine the efficacy of bacterial decolonization (BD) to reduce ARD severity compared with standard of care. Design, Setting, and Participants: This phase 2/3 randomized clinical trial was conducted from June 2019 to August 2021 with investigator blinding at an urban academic cancer center and enrolled patients with breast cancer or head and neck cancer receiving radiation therapy (RT) with curative intent. Analysis was performed on January 7, 2022. Interventions: Intranasal mupirocin ointment twice daily and chlorhexidine body cleanser once daily for 5 days prior to RT and repeated for 5 days every 2 weeks through RT. Main Outcomes and Measures: The primary outcome as planned prior to data collection was the development of grade 2 or higher ARD. Based on wide clinical variability of grade 2 ARD, this was refined to grade 2 ARD with moist desquamation (grade 2-MD). Results: Of 123 patients assessed for eligibility via convenience sampling, 3 were excluded, and 40 refused to participate, with 80 patients in our final volunteer sample. Of 77 patients with cancer (75 patients with breast cancer [97.4%] and 2 patients with head and neck cancer [2.6%]) who completed RT, 39 were randomly assigned BC, and 38 were randomly assigned standard of care; the mean (SD) age of the patients was 59.9 (11.9) years, and 75 (97.4%) were female. Most patients were Black (33.7% [n = 26]) or Hispanic (32.5% [n = 25]). Among patients with breast cancer and patients with head and neck cancer (N = 77), none of the 39 patients treated with BD and 9 of the 38 patients (23.7%) treated with standard of care developed ARD grade 2-MD or higher (P = .001). Similar results were observed among the 75 patients with breast cancer (ie, none treated with BD and 8 [21.6%] receiving standard of care developed ARD grade ≥2-MD; P = .002). The mean (SD) ARD grade was significantly lower for patients treated with BD (1.2 [0.7]) compared with patients receiving standard of care (1.6 [0.8]) (P = .02). Of the 39 patients randomly assigned to BD, 27 (69.2%) reported regimen adherence, and only 1 patient (2.5%) experienced an adverse event related to BD (ie, itch). Conclusions and Relevance: The results of this randomized clinical trial suggest that BD is effective for ARD prophylaxis, specifically for patients with breast cancer. Trial Registration: ClinicalTrials.gov Identifier: NCT03883828.

Association of Staphylococcus aureus Colonization With Severity of Acute Radiation Dermatitis in Patients With Breast or Head and Neck Cancer.

Importance: Pathogenesis of acute radiation dermatitis (ARD) is not completely understood. Pro-inflammatory cutaneous bacteria may contribute to cutaneous inflammation after radiation therapy. Objective: To evaluate whether nasal colonization with Staphylococcus aureus (SA) before radiation therapy is associated with ARD severity in patients with breast or head and neck cancer. Design, Setting, and Participants: This prospective cohort study with observers blinded to colonization status was conducted from July 2017 to May 2018 at an urban academic cancer center. Patients aged 18 years or older with breast or head and neck cancer and plans for fractionated radiation therapy (≥15 fractions) with curative intent were enrolled via convenience sampling. Data were analyzed from September to October 2018. Exposures: Staphylococcus aureus colonization status before radiation therapy (baseline). Main Outcomes and Measures: The primary outcome was ARD grade using the Common Terminology Criteria for Adverse Event Reporting, version 4.03. Results: Among 76 patients analyzed, mean (SD) age was 58.5 (12.6) years and 56 (73.7%) were female. All 76 patients developed ARD: 47 (61.8%) with grade 1, 22 (28.9%) with grade 2, and 7 (9.2%) with grade 3. The prevalence of baseline nasal SA colonization was higher among patients who developed grade 2 or higher ARD compared with those who developed grade 1 ARD (10 of 29 [34.5%] vs 6 of 47 [12.8%]; P = .02, by χ2 test). Conclusions and Relevance: In this cohort study, baseline nasal SA colonization was associated with development of grade 2 or higher ARD in patients with breast or head and neck cancer. The findings suggest that SA colonization may play a role in the pathogenesis of ARD.

Remote Activity Monitoring and Electronic Patient-Reported Outcomes Collection During Radiotherapy for Head and Neck Cancer: A Pilot Study.

PURPOSE: Modern wearable devices provide objective and continuous activity data that could be leveraged to enhance cancer care. We prospectively studied the feasibility of monitoring physical activity using a commercial wearable device and collecting electronic patient-reported outcomes (ePROs) during radiotherapy (RT) for head and neck cancer (HNC). METHODS: Patients planned for a course of external beam RT with curative intent for HNC were instructed to use a commercial fitness tracker throughout the RT course. During weekly clinic visits, physician-scored adverse events were recorded during using Common Terminology Criteria for Adverse Events version 4.0, and patients completed ePRO surveys using a clinic tablet or computer. Feasibility of activity monitoring was defined as collection of step data for at least 80% of the RT course for at least 80% of patients. Exploratory analyses described associations between step counts, ePROs, and clinical events. RESULTS: Twenty-nine patients with HNC were enrolled and had analyzable data. Overall, step data were recorded on 70% of the days during patients' RT courses, and there were only 11 patients (38%) for whom step data were collected on at least 80% of days during RT. Mixed effects linear regression models demonstrated declines in daily step counts and worsening of most PROs during RT. Cox proportional hazards models revealed a potential association between high daily step counts and both reduced risk of feeding tube placement (hazard ratio [HR], 0.87 per 1,000 steps, P < .001) and reduced risk of hospitalization (HR, 0.60 per 1,000 steps, P < .001). CONCLUSION: We did not achieve our feasibility end point, suggesting that rigorous workflows are required to achieve continuous activity monitoring during RT. Although limited by a modest sample size, our findings are consistent with previous reports indicating that wearable device data can help identify patients who are at risk for unplanned hospitalization.

Predictors of Financial Toxicity in Patients Receiving Concurrent Radiation Therapy and Chemotherapy.

Purpose: Financial toxicity (FT) is a significant concern for patients with cancer. We reviewed prospectively collected data to explore associations with FT among patients undergoing concurrent, definitive chemoradiation therapy (CRT) within a diverse, urban, academic radiation oncology department. Methods and Materials: Patients received CRT in 1 of 3 prospective trials. FT was evaluated before CRT (baseline) and then weekly using the European Organization for Research and Treatment of Cancer Core Quality of Life Questionnaire Core-30 questionnaire. Patients were classified as experiencing FT if they answered ≥2 on a Likert scale question (1-4 points) asking if they experienced FT. Rate of change of FT was calculated using linear regression; worsening FT was defined as increase ≥1 point per month. χ2, t tests, and logistic regression were used to assess predictors of FT. Results: Among 233 patients, patients attended an average of 9 outpatient and 4 radiology appointments over the 47 days between diagnosis and starting CRT. At baseline, 52% of patients reported experiencing FT. Advanced T stage (odds ratio, 2.47; P = .002) was associated with baseline FT in multivariate analysis. The mean rate of FT change was 0.23 Likert scale points per month. In total, 26% of patients demonstrated worsening FT during CRT. FT at baseline was not associated with worsening FT (P = .98). Hospitalization during treatment was associated with worsening FT (odds ratio, 2.30; P = .019) in multivariate analysis. Conclusions: Most patients reported FT before CRT. These results suggest that FT should be assessed (and, potentially, addressed) before starting definitive treatment because it develops early in a patient's cancer journey. Reducing hospitalizations may mitigate worsening FT. Further research is warranted to design interventions to reduce FT and avoid hospitalizations.

Continuous Long-Term Physical Activity Monitoring in Hemodialysis Patients.

Background: Physical inactivity is common in patients receiving hemodialysis, but activity patterns throughout the day and in relation to dialysis are largely unknown. This knowledge gap can be addressed by long-term continuous activity monitoring, but this has not been attempted and may not be acceptable to patients receiving dialysis. Methods: Ambulatory patients with end-stage kidney disease receiving thrice-weekly hemodialysis wore commercially available wrist-worn activity monitors for 6 months. Step counts were collected every 15 minutes and were linked to dialysis treatments. Physical function was assessed using the Short Physical Performance Battery (SPPB). Fast time to recovery from dialysis was defined as ≤2 hours. Mixed effects models were created to estimate step counts over time. Results: Of 52 patients enrolled, 48 were included in the final cohort. The mean age was 60 years, and 75% were Black or Hispanic. Comorbidity burden was high, 38% were transported to and from dialysis by paratransit, and 79% had SPPB <10. Median accelerometer use (199 days) and adherence (95%) were high. Forty-two patients (of 43 responders) reported wearing the accelerometer every day, and few barriers to adherence were noted. Step counts were lower on dialysis days (3991 [95% CI, 3187 to 4796] versus 4561 [95% CI, 3757 to 5365]), but step-count intensity was significantly higher during the hour immediately after dialysis than during the corresponding time on nondialysis days (188 steps per hour increase [95% CI, 171 to 205]); these levels were the highest noted at any time. Postdialysis increases were more pronounced among patients with fast recovery time (225 [95% CI, 203 to 248] versus 134 [95% CI, 107 to 161] steps per hour) or those with SPPB ≥7. Estimates were unchanged after adjustment for demographics, diabetes status, and ultrafiltration rate. Conclusions: Long-term continuous monitoring of physical activity is feasible in patients receiving hemodialysis. Highly granular data collection and analysis yielded new insights into patterns of activity after dialysis treatments.

Does HCC Etiology Impact the Efficacy of Stereotactic Body Radiation Therapy for Hepatocellular Carcinoma? An Asian Liver Radiation Therapy Group Study.

Background/Purpose: The Asian Liver Radiation Therapy Study Group has formed a large and detailed multinational database of outcomes following stereotactic body radiation therapy (SBRT) for hepatocellular carcinoma (HCC). Here, we explored the potential impact of HCC etiology on SBRT efficacy. Tumor control probability (TCP) models were established to estimate the likelihood of local control (LC). Methods: Data from 415 patients who were treated with SBRT for HCC were reviewed. Cox proportional hazards models were used to identify key predictors of LC. TCP models accounting for biologic effective dose (BED) and tumor diameter were generated to quantify associations between etiology and LC. Results: Cox models demonstrated that hepatitis C virus (HCV) infection was associated with favorable LC following SBRT (HR=0.52, 95% CI 0.04-0.96, p=0.036). The 2-year LC rate for patients with HCV etiology was 88%, compared to 78% for other patients. Small tumor and high BED were also associated with favorable LC. TCP models demonstrated a 10-20% absolute increase in predicted LC across the range of SBRT doses and tumor sizes. Conclusion: We found a novel association between HCV status and LC after SBRT for HCC that warrants further exploration. If validated in other datasets, our findings could help clinicians tailor SBRT schedules.

Improving Adjuvant Liver-Directed Treatment Recommendations for Unresectable Hepatocellular Carcinoma: An Artificial Intelligence-Based Decision-Making Tool.

PURPOSE: Liver-directed therapy after transarterial chemoembolization (TACE) can lead to improvement in survival for selected patients with unresectable hepatocellular carcinoma (HCC). However, there is uncertainty in the appropriate application and modality of therapy in current clinical practice guidelines. The aim of this study was to develop a proof-of-concept, machine learning (ML) model for treatment recommendation in patients previously treated with TACE and select patients who might benefit from additional treatment with combination stereotactic body radiotherapy (SBRT) or radiofrequency ablation (RFA). METHODS: This retrospective observational study was based on data from an urban, academic hospital system selecting for patients diagnosed with stage I-III HCC from January 1, 2008, to December 31, 2018, treated with TACE, followed by adjuvant RFA, SBRT, or no additional liver-directed modality. A feedforward, ML ensemble model provided a treatment recommendation on the basis of pairwise assessments evaluating each potential treatment option and estimated benefit in survival. RESULTS: Two hundred thirty-seven patients met inclusion criteria, of whom 54 (23%) and 49 (21%) received combination of TACE and SBRT or TACE and RFA, respectively. The ML model suggested a different consolidative modality in 32.7% of cases among patients who had previously received combination treatment. Patients treated in concordance with model recommendations had significant improvement in progression-free survival (hazard ratio 0.5; P = .007). The most important features for model prediction were cause of cirrhosis, stage of disease, and albumin-bilirubin grade (a measure of liver function). CONCLUSION: In this proof-of-concept study, an ensemble ML model was able to provide treatment recommendations for HCC who had undergone prior TACE. Additional treatment in line with model recommendations was associated with significant improvement in progression-free survival, suggesting a potential benefit for ML-guided medical decision making.

Organ-at-risk dose prediction using a machine learning algorithm: Clinical validation and treatment planning benefit for lung SBRT.

OBJECTIVE: To quantify the clinical performance of a machine learning (ML) algorithm for organ-at-risk (OAR) dose prediction for lung stereotactic body radiation therapy (SBRT) and estimate the treatment planning benefit from having upfront access to these dose predictions. METHODS: ML models were trained using multi-center data consisting of 209 patients previously treated with lung SBRT. Two prescription levels were investigated, 50 Gy in five fractions and 54 Gy in three fractions. Models were generated using a gradient-boosted regression tree algorithm using grid searching with fivefold cross-validation. Twenty patients not included in the training set were used to test OAR dose prediction performance, ten for each prescription. We also performed blinded re-planning based on OAR dose predictions but without access to clinically delivered plans. Differences between predicted and delivered doses were assessed by root-mean square deviation (RMSD), and statistical differences between predicted, delivered, and re-planned doses were evaluated with one-way analysis of variance (ANOVA) tests. RESULTS: ANOVA tests showed no significant differences between predicted, delivered, and replanned OAR doses (all p ≥ 0.36). The RMSD was 2.9, 3.9, 4.3, and 1.7Gy for max dose to the spinal cord, great vessels, heart, and trachea, respectively, for 50 Gy in five fractions. Average improvements of 1.0, 1.4, and 2.0 Gy were seen for spinal cord, esophagus, and trachea max doses in blinded replans compared to clinically delivered plans with 54 Gy in three fractions, and 1.8, 0.7, and 1.5 Gy, respectively, for the esophagus, heart and bronchus max doses with 50 Gy in five fractions. Target coverage was similar with an average PTV V100% of 94.7% for delivered plans compared to 97.3% for blinded re-plans for 50 Gy in five fractions, and respectively 98.4% versus 99.2% for 54 Gy in three fractions. CONCLUSION: This study validated ML-based OAR dose prediction for lung SBRT, showing potential for improved OAR dose sparing and more consistent plan quality using dose predictions for patient-specific planning guidance.

Disease Burden on PET Predicts Outcomes for Advanced NSCLC Patients Treated with First-Line Immunotherapy.

BACKGROUND: First-line immunotherapy (IMT), with or without cytotoxic chemotherapy, is now recommended for most patients with advanced non-small cell lung cancer (NSCLC) with no targetable mutations. We reviewed outcomes for NSCLC patients treated with first-line IMT at our institution to test the hypothesis that measures of disease burden on staging FDG-PET/CT have prognostic value. MATERIALS AND METHODS: Patient, disease, and treatment details were collected. A gradient-based segmentation tool was used to delineate each PET-avid extracranial lesion. Numbers of extrathoracic lesions and metabolic tumor volumes were tabulated. Oligometastatic disease (OMD) was defined as having ≤3 extrathoracic lesions, with any number of thoracic lesions. Progression-free survival (PFS) and overall survival (OS) rates following initiation of IMT were evaluated using the Kaplan-Meier method, and predictors of PFS and OS were assessed using Cox proportional hazards models and logrank tests. RESULTS: One hundred twenty-four patients met inclusion criteria, and 1143 lesions were contoured. The presence of OMD was associated with favorable PFS (median 13.1 vs. 6.9 months; P = .016) and favorable OS (median 36.5 vs. 15.4 months; P = .002). In multivariable models, OMD was associated with favorable PFS (HR = 0.64; P = .034) and favorable OS (HR = 0.61; P = .063), and metabolic tumor volumes exceeding the cohort median (88 cc) was associated with inferior OS (HR = 1.85; P = .028). CONCLUSION: For advanced NSCLC patients receiving first-line IMT, the presence of extrathoracic OMD and low volumetric disease burden on PET are favorable prognostic factors that could be useful stratification factors in clinical trials and may influence clinical decisions about local and systemic therapy.