Managing patients with acute respiratory distress syndrome (ARDS) requires frequent changes in mechanical ventilator respiratory settings to optimize arterial oxygenation assessed by arterial oxygen partial pressure (PaO2) and saturation (SaO2). Pulse oxymetry (SpO2) has been suggested as a non-invasive surrogate for arterial oxygenation however its accuracy in COVID-19 patients is unknown. In this study, we aimed to investigate the influence of COVID-19 status on the association between SpO2 and arterial oxygenation. We prospectively included patients with ARDS and compared COVID-19 to non-COVID-19 patients, regarding SpO2 and concomitant arterial oxygenation (SaO2 and PaO2) measurements, and their association. Bias was defined as mean difference between SpO2 and SaO2 measurements. Occult hypoxemia was defined as a SpO2 ≥ 92% while concomitant SaO2 < 88%. Multiple linear regression models were built to account for confounders. We also assessed concordance between positive end-expiratory pressure (PEEP) trial-induced changes in SpO2 and in arterial oxygenation. We included 55 patients, among them 26 (47%) with COVID-19. Overall, SpO2 and SaO2 measurements were correlated (r = 0.70; p < 0.0001), however less so in COVID-19 than in non-COVID-19 patients (r = 0.55, p < 0.0001 vs. r = 0.84, p < 0.0001, p = 0.002 for intergroup comparison). Bias was + 1.1%, greater in COVID-19 than in non-COVID-19 patients (2.0 vs. 0.3%; p = 0.02). In multivariate analysis, bias was associated with COVID-19 status (unstandardized β = 1.77, 95%CI = 0.38–3.15, p = 0.01), ethnic group and ARDS severity. Occult hypoxemia occurred in 5.5% of measurements (7.7% in COVID-19 patients vs. 3.4% in non-COVID-19 patients, p = 0.42). Concordance rate between PEEP trial-induced changes in SpO2 and SaO2 was 84%, however less so in COVID-19 than in non-COVID-19 patients (69% vs. 97%, respectively). Similar results were observed for PaO2 regarding correlations, bias, and concordance with SpO2 changes. In patients with ARDS, SpO2 was associated with arterial oxygenation, but COVID-19 status significantly altered this association.

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.

No abstract