Interpretable Machine Learning Model for Locoregional Relapse Prediction in Oropharyngeal Cancers

Background: There is no evidence to support surgery or radiotherapy as the best treatment for resectable oropharyngeal cancers with a negative HPV status. Predictive algorithms may help to decide which strategy to choose, but they will only be accepted by caregivers and European authorities if they are interpretable. As a proof of concept, we developed a predictive and interpretable algorithm to predict locoregional relapse at 18 months for oropharyngeal cancers as a first step towards that goal. Methods: The model was based on clinical and Pyradiomics features extracted from the dosimetric CT scan. Intraclass correlation was used to filter out features dependant on delineation. Correlated redundant features were also removed. An XGBoost model was cross-validated and optimised on the HN1 cohort (79 patients), and performances were assessed on the ART ORL cohort (45 patients). The Shapley Values were used to provide an overall and local explanation of the model. Results: On the ART ORL cohort, the model trained on HN1 yielded a precision—or predictive positive value—of 0.92, a recall of 0.42, an area under the curve of the receiver operating characteristic of 0.68 and an accuracy of 0.64. The most contributory features were shape Voxel Volume, grey level size zone matrix Small Area Emphasis (glszmSAE), gldm Dependence Non Uniformity Normalized (gldmDNUN), Sex and Age. Conclusions: We developed an interpretable and generalizable model that could yield a good precision—positive predictive value—for relapse at 18 months on a different test cohort.

13-gene DNA Methylation Analysis from Oral Brushing: A Promising Non Invasive Tool in the Follow-up of Oral Cancer Patients

Background: This study aimed to evaluate the prognostic value of a non-invasive sampling procedure based on 13-gene DNA methylation analysis in the follow-up of patients previously treated for oral squamous cell carcinoma (OSCC). Methods: The study population included 49 consecutive patients treated for OSCC. Oral brushing sample collection was performed at two different times: before any cancer treatment in the tumor mass and during patient follow-up almost 6 months after OSCC treatment, within the regenerative area after OSCC resection. Each sample was considered positive or negative in relation to a predefined cut-off value. Results: Before any cancer treatment, 47/49 specimens exceeded the score and were considered as positive. Six months after OSCC resection, 16/49 specimens also had positive scores in the samples collected from the regenerative area. During the follow-up period, 7/49 patients developed locoregional relapse: 6/7 patients had a positive score in the regenerative area after OSCC resection. The presence of a positive score after oral cancer treatment was the most powerful variable related to the appearance of locoregional relapse. Conclusion: 13-gene DNA methylation analysis by oral brushing may have a clinical application as a prognostic non-invasive tool in the follow-up of patients surgically treated for OSCC.

Survival following Multimodality Treatment Including Surgery for Stage IA–IIIB Small-Cell Lung Cancer

Background Prognosis in limited disease small-cell lung cancer (SCLC) after concurrent chemoradiotherapy is poor. While some studies show better survival after multimodality treatment including surgery, other trials failed to prove a surgery-related survival benefit. Therefore, this study investigated survival in stage IA–IIIB SCLC following surgery combined with chemotherapy and/or thoracic radiotherapy. Methods We retrospectively reviewed all stage IA–IIIB SCLC patients without supraclavicular lymph node involvement at a single institution between January 1999 and August 2016 after multimodality treatment with curative intent. This comprised surgery consisting of primary tumor resection and systematic lymph node dissection combined with chemotherapy, chemoradiotherapy, or thoracic radiotherapy. Survival was determined using the Kaplan–Meier method, and differences were compared using log-rank tests. The risk of locoregional relapse was calculated. Results A total of 47 patients (29 men, 18 women; mean age: 62 years) were included. Thirty-day mortality was 0%. Overall median survival was 56 months, and 2-, 3-, 5-, and 10-year survival rates were 69, 54, 46, and 30%, respectively. The only significant prognostic factor (p = 0.006) was R0 resection (n = 40) increasing median survival to 64 versus 17 months in case of technical inoperability (n = 5). The risk of locoregional relapse was 2.5% (n = 1) after R0 resection. Conclusions Multimodality treatment including surgery was safe and led to considerable survival. R0 resection was the only factor extending survival. It could be achieved in most patients and was associated with a low risk of locoregional relapse. Prospective randomized controlled studies are needed to define best practice in stage IA–IIIB SCLC.