Radiomics analysis to predict pulmonary nodule malignancy using machine learning approaches.

BACKGROUND: Low-dose CT screening can reduce lung cancer-related mortality. However, most screen-detected pulmonary abnormalities do not develop into cancer and it often remains challenging to identify malignant nodules, particularly among indeterminate nodules. We aimed to develop and assess prediction models based on radiological features to discriminate between benign and malignant pulmonary lesions detected on a baseline screen. METHODS: Using four international lung cancer screening studies, we extracted 2060 radiomic features for each of 16 797 nodules (513 malignant) among 6865 participants. After filtering out low-quality radiomic features, 642 radiomic and 9 epidemiological features remained for model development. We used cross-validation and grid search to assess three machine learning (ML) models (eXtreme Gradient Boosted Trees, random forest, least absolute shrinkage and selection operator (LASSO)) for their ability to accurately predict risk of malignancy for pulmonary nodules. We report model performance based on the area under the curve (AUC) and calibration metrics in the held-out test set. RESULTS: The LASSO model yielded the best predictive performance in cross-validation and was fit in the full training set based on optimised hyperparameters. Our radiomics model had a test-set AUC of 0.93 (95% CI 0.90 to 0.96) and outperformed the established Pan-Canadian Early Detection of Lung Cancer model (AUC 0.87, 95% CI 0.85 to 0.89) for nodule assessment. Our model performed well among both solid (AUC 0.93, 95% CI 0.89 to 0.97) and subsolid nodules (AUC 0.91, 95% CI 0.85 to 0.95). CONCLUSIONS: We developed highly accurate ML models based on radiomic and epidemiological features from four international lung cancer screening studies that may be suitable for assessing indeterminate screen-detected pulmonary nodules for risk of malignancy.

Maternal and childhood medical history and the risk of childhood brain tumours: a case-control study in Ontario, Canada.

BACKGROUND: Studies to date have yielded inconclusive results as to whether maternal medical history during pregnancy, and a child's early-life medical history contribute to the development of childhood brain tumours (CBTs). This study examined associations between maternal and childhood medical history and the risk of CBTs. METHODS: The Childhood Brain Tumour Epidemiology Study of Ontario (CBREO) examined children 0-15 years of age with newly diagnosed CBTs from 1997 to 2003. Multivariable logistic regression analysis determined associations for prenatal medications and childhood medical history, adjusted for child's demographics, and maternal education. Analyses were stratified by histology. A latency period analysis was conducted using 12- and 24-month lead times. RESULTS: Maternal intake of immunosuppressants during the prenatal period was significantly associated with glial tumours (OR 2.73, 95% CI 1.17-6.39). Childhood intake of anti-epileptics was significantly associated with CBTs overall, after accounting for 12-month (OR 8.51, 95% CI 3.35-21.63) and 24-month (OR 6.04, 95% CI 2.06-17.70) lead time before diagnosis. No associations for other medications were found. CONCLUSIONS: This study underscores the need to examine potential carcinogenic effects of the medication classes highlighted and of the indication of medication use. Despite possible reverse causality, increased CBT surveillance for children with epilepsy might be warranted.

Childhood head trauma and the risk of childhood brain tumours: A case-control study in Ontario, Canada.

Head trauma in early childhood has been hypothesized as a potential risk factor for childhood brain tumours (CBTs). However, head trauma has not been extensively studied in the context of CBTs and existing studies have yielded conflicting results. A population-based and hospital-based case-control study of children 0 to 15 years with newly diagnosed CBTs from 1997 to 2003 recruited across Ontario through paediatric oncology centres was conducted. Controls were frequency-matched with cases by age, sex and geographical region. The association was assessed based on multivariable logistic regressions, accounting for child's age, sex, ethnicity, highest level of maternal education and maternal pack-years of smoking during the pregnancy. Analyses were conducted separately based on age of first head trauma, sex and histology. A latency period analysis was conducted. Overall, based on 280 cases and 919 controls, CBTs were not significantly associated with previous history of head trauma (OR 1.34, 95% CI 0.96, 1.86), head trauma severity, number of head injuries, or head or neck X-rays or computed tomography (CT) examinations. Results were consistent across sexes and histological subtypes. However, head trauma within the first year of life was significantly associated with CBTs (OR 2.00, 95% CI 1.01, 3.98), but the association diminished when adjusted for X-ray or CT occurring during the same time period (OR 1.62, 95% CI 0.75, 3.49), albeit limited sample size. Overall, no association was observed between head trauma and CBTs among all children, while head trauma occurring within first year of life may warrant further investigation in future research.

Hyperprogressive disease in early-phase immunotherapy trials: Clinical predictors and association with immune-related toxicities.

BACKGROUND: A subset of patients treated with immune checkpoint inhibitors experience an accelerated tumor growth rate (TGR) in comparison with pretreatment kinetics; this is known as hyperprogression. This study assessed the relation between hyperprogressive disease (HPD) and treatment-related toxicity and clinical factors. METHODS: This study reviewed patients with solid tumors who were enrolled in early-phase immunotherapy trials at Princess Margaret Cancer Centre between August 2012 and September 2016 and had computed tomography scans in the pre-immunotherapy (reference) and on-immunotherapy (experimental) periods. HPD was defined as progression according to Response Evaluation Criteria in Solid Tumors 1.1 at the first on-treatment scan and a ≥2-fold increase in TGR between the reference and experimental periods. Treatment-related toxicities requiring systemic therapy, drug delays, or discontinuation were considered clinically significant adverse events (CSAEs). RESULTS: Of 352 patients, 182 were eligible for analysis. The median age was 60 years, and 54% were male. The Eastern Cooperative Oncology Group performance status was 0 (32%) or 1 (68%). The Royal Marsden Hospital (RMH) prognostic score was 0/1 in 59%. Single-agent immunotherapy was given to 80% of the patients. Most patients (89%) received anti-programmed death (ligand) 1 antibodies alone or in combination with other therapies. HPD occurred in 12 of 182 patients (7%). A higher proportion of females was seen among HPD patients (P = .01), but no association with age, performance status, tumor type, RMH prognostic score, combination immunotherapy, or CSAEs was found. The 1-year overall survival rate was 28% for HPD patients and 53% for non-HPD patients (hazard ratio, 1.7; 95% confidence interval, 0.9-3.3; P = .11). CONCLUSIONS: HPD was observed in 7% of patients with solid tumors treated with immunotherapy. HPD was not associated with CSAEs, age, tumor type, or the type of immunotherapy but was more common in females.