Machine learning for dose-volume histogram based clinical decision-making support system in radiation therapy plans for brain tumors.

PURPOSE: To create and investigate a novel, clinical decision-support system using machine learning (ML). METHODS AND MATERIALS: The ML model was developed based on 79 radiotherapy plans of brain tumor patients that were prescribed a total dose of 60 Gy delivered with volumetric-modulated arc therapy (VMAT). Structures considered for analysis included planning target volume (PTV), brainstem, cochleae, and optic chiasm. The model aimed to classify the target variable that included class-0 corresponding to plans for which the PTV treatment planning objective was met and class-1 that was associated with plans for which the PTV objective was not met due to the priority trade-off to meet one or more organs-at-risk constraints. Several models were evaluated using double-nested cross-validation and an area-under-the-curve (AUC) metric, with the highest performing one selected for further investigation. The model predictions were explained with Shapely additive explanation (SHAP) interaction values. RESULTS: The highest-performing model was Logistic Regression achieving an accuracy of 93.8 ± 4.1% and AUC of 0.98 ± 0.02 on the testing data. The SHAP analysis indicated that the ΔD99% metric for PTV had the greatest influence on the model predictions. The least important feature was ΔDMAX for the left and right cochleae. CONCLUSIONS: The trained model achieved satisfactory accuracy and can be used by medical physicists in a data-driven quality assurance program as well as by radiation oncologists to support their decision-making process in terms of treatment plan approval and potential plan modifications. Model explanation analysis showed that the model relies on clinically valid logic when making predictions.

Trimodality therapy for esophageal cancer: The role of surgical and radiation treatment parameters in the development of anastomotic complications.

BACKGROUND: Here, we investigated the relationship between clinical parameters, including the site of surgical anastomosis and radiation dose to the anastomotic region, and anastomotic complications in esophageal cancer patients treated with trimodality therapy. METHODS: Between 2007 and 2016, esophageal cancer patients treated with trimodality therapy at a tertiary academic cancer center were identified. Patient, treatment, and outcome parameters were collected. Radiation dose to the gastric regions were extracted. Anastomotic complication was defined as leak and/or stricture. We used Fisher's exact and Wilcoxon rank-sum tests to compare the association between clinical parameters and anastomotic complications. RESULTS: Of 89 patients identified, the median age was 63 years, 82% (n = 73) were male, and 82% had distal (n = 47) or gastroesophageal junction (n = 26) tumors. Median follow-up was 25.8 months. Esophagectomies were performed with cervical (65%, n = 58) or thoracic anastomoses (35%, n = 31). Anastomotic complications developed in 60% (n = 53). Cervical anastomosis was associated with anastomotic complications (83%, n = 44/53, p < 0.01). Radiation to any gastric substructure was not associated with anastomotic complications (p > 0.05). In the subset of patients with distal/gastroesophageal junction tumors undergoing esophagectomy with cervical anastomosis where radiation was delivered to the future neoesophagus, 80% (n = 35/44) developed anastomotic complications. In this high-risk subgroup, radiation was not associated with anastomotic complications (p > 0.05). CONCLUSIONS: Our analysis did not demonstrate an association between radiation dose to gastric substructures and anastomotic complications. However, it showed an association between esophagectomy with cervical anastomosis and anastomotic complications. Patients with distal/gastroesophageal junction tumors who undergo esophagectomy with cervical anastomosis have higher rates of anastomotic complications unrelated to radiation to gastric substructures.

Relationship Between Prior Radiotherapy and Checkpoint-Inhibitor Pneumonitis in Patients With Advanced Non-Small-Cell Lung Cancer.

PURPOSE: To investigate the relationship between radiotherapy (RT), in particular chest RT, and development of immune-related (IR) pneumonitis in non-small-cell lung cancer (NSCLC) patients treated with anti-programmed cell death 1 (PD-1)/programmed death ligand 1 (PD-L1). PATIENTS AND METHODS: Between June 2011 and July 2017, NSCLC patients treated with anti-PD-1/PD-L1 at a tertiary-care academic cancer center were identified. Patient, treatment, prior RT (intent, technique, timing, courses), and IR pneumonitis details were collected. Treating investigators diagnosed IR pneumonitis clinically. Diagnostic IR pneumonitis scans were overlaid with available chest RT plans to describe IR pneumonitis in relation to prior chest RT. We evaluated associations between patient, treatment, RT details, and development of IR pneumonitis by Fisher exact and Wilcoxon rank-sum tests. RESULTS: Of the 188 NSCLC patients we identified, median follow-up was 6.78 (range, 0.30-79.3) months and median age 66 (range, 39-91) years; 54% (n = 102) were male; and 42% (n = 79) had stage I-III NSCLC at initial diagnosis. Patients received anti-PD-1/PD-L1 monotherapy (n = 127, 68%) or PD-1/PD-L1-based combinations (n = 61, 32%). In the entire cohort, 70% (132/188) received any RT, 53% (100/188) chest RT, and 37% (70/188) curative-intent chest RT. Any grade IR pneumonitis occurred in 19% (36/188; 95% confidence interval, 13.8-25.6). Of those who developed IR pneumonitis and received chest RT (n = 19), patients were more likely to have received curative-intent versus palliative-intent chest RT (17/19, 89%, vs. 2/19, 11%; P = .051). Predominant IR pneumonitis appearances were ground-glass opacities outside high-dose chest RT regions. CONCLUSION: No RT parameter was significantly associated with IR pneumonitis. On subset analysis of patients who developed IR pneumonitis and who had received prior chest RT, IR pneumonitis was more common in patients who received curative-intent chest RT. Attention should be paid to NSCLC patients receiving curative-intent RT followed by anti-PD-1/PD-L1 agents.

Preoperative contralateral lung radiation dose is associated with postoperative pulmonary toxicity in patients with locally advanced non-small cell lung cancer treated with trimodality therapy.

PURPOSE: In patients with non-small cell lung cancer (NSCLC) who undergo trimodality therapy (chemoradiation followed by surgical resection), it is unknown whether limiting preoperative radiation dose to the uninvolved lung reduces postsurgical morbidity. This study evaluated whether radiation fall-off dose parameters to the contralateral lung that is unaffected by NSCLC are associated with postoperative complications in NSCLC patients treated with trimodality therapy. METHODS AND MATERIALS: We retrospectively reviewed NSCLC patients who underwent trimodality therapy between March 2008 and October 2016, with available restored digital radiation plans. Fischer's exact test was used to assess associations between patient and treatment characteristics and the development of treatment-related toxicity. Spearman rank correlation was used to measure the strength of association between dosimetric parameters. RESULTS: Forty-six patients were identified who received trimodality therapy with intensity modulated radiation (median, 59.4 Gy; range, 45-70) and concurrent platinum doublet chemotherapy, followed by surgical resection. The median age was 64.9 years (range, 45.6-81.6). The median follow-up time was 1.9 years (range, 0.3-8.4). Twenty-four (52.2%) patients developed any-grade pulmonary toxicity and 14 (30.4%) patients developed grade 2+ pulmonary toxicity. There was an increased incidence of any-grade pulmonary toxicity in patients with contralateral lung volume receiving at least 20 Gy (V20) ≥7% compared with <7% (90%, n = 9 vs 41.7%, n = 15; P = .01). Similarly, contralateral lung V10 ≥20% was associated with an increased rate of any-grade pulmonary toxicity compared with V10 <20% (80%, n = 12 vs 38.7%, n = 12; P = .01). Pneumonectomy/bilobectomy was associated with grade 2+ pulmonary toxicity (P = .04). CONCLUSIONS: Patients who received a higher radiation fall-off dose volume parameter (V20 ≥7% and V10 ≥20%) to the contralateral uninvolved lung had a higher incidence of any-grade postoperative pulmonary toxicity. Limiting radiation fall-off dose to the uninvolved lung may be an important modifiable radiation parameter in limiting postoperative toxicity in trimodality patients.

Pancreatic cancer in Saudi patients treated at tertiary institution. Ten years retrospective study.

OBJECTIVES: To describe presentation, management, and outcome, and determine prognostic factors for pancreatic cancer patients. METHODS: A retrospective review of patients diagnosed with pancreatic cancer at King Abdulaziz Medical City, Riyadh, Saudi Arabia during the period from January 2000 to December 2010. Descriptive statistics were conducted on the collected data and survival was estimated using the Kaplan Meier estimate. Univariate and multiple regression analyses were carried out. RESULTS: The medical records of 179 patients were reviewed. The patients' median age was 63 years ranging from 15-96 years, and 116 (64.8%) of them were male. The one-year survival rate was 39% and the 5-year survival was 10%. The median overall survival (OS) was 6.9 months. Age at diagnosis, grade, T stage, N stage, M stage, TNM stage group, and the combined stage group (stage III/IV versus others), site of distant metastasis, carcinoembryonic antigen (CEA), carbohydrate antigen 19-9, surgery and chemotherapy were significant predictors for OS on an univariate Cox proportional hazards regression analysis. A multiple regression model including all the significant predictors was conducted. Age at the time of diagnosis and M stage were significant variables. CONCLUSION: Our patients present at a younger age and have better 5-year survival compared with the United States Surveillance Epidemiology and End Results data, which deserves further evaluation. Age and disease stage were identified as independent prognostic factors for survival in this patient population.

Risk factors of diabetic foot in central Saudi Arabia.

OBJECTIVE: To identify the risk factors of diabetic foot (DF) in diabetic patients. METHODS: In a case-control study, medical records of 50 patients with DF, and 50 diabetic controls without DF were selected randomly from the patients seen at King Abdulaziz Medical City (KAMC), Riyadh, Kingdom of Saudi Arabia. Selected vascular, neuropathic, metabolic, health care, and lifestyle risk factors were investigated. Multiple logistic regression was used to relate these potential risk factors to the odds of DF. RESULTS: Diabetic foot was significantly associated with: gender, age, education, type of diabetes, duration of disease, level of erythrocyte sedimentation rate (ESR), presence of peripheral neuropathy, peripheral vascular disease, chronic renal diseases, ischemic heart diseases, hypertension, and previous history of diabetic foot. After adjusting for the potentially confounding effects of age and gender by using the logistic regression analysis, independent predictors of DF were: the duration of diabetes, presence of neuropathy, and ESR level. In the prediction of DF, receiver operating characteristic curves (ROC) were applied to identify the most valid cut-off points of the duration of diabetes (11 years), and ESR level (54 mm/hr). CONCLUSION: These findings could help diabetologists recognize early, and manage DF, and thus reduce the risk of limb amputation, and the cost that accompanies limb loss in this prevalent condition.