Deep neural networks are revolutionizing the way complex systems are designed. Consequently, there is a pressing need for tools and techniques for network analysis and certification. To help in addressing that need, we present Marabou, a framework for verifying deep neural networks. Marabou is an SMT-based tool that can answer queries about a network's properties by transforming these queries into constraint satisfaction problems. It can accommodate networks with different activation functions and topologies, and it performs high-level reasoning on the network that can curtail the search space and improve performance. It also supports parallel execution to further enhance scalability. Marabou accepts multiple input formats, including protocol buffer files generated by the popular TensorFlow framework for neural networks. We describe the system architecture and main components, evaluate the technique and discuss ongoing work.

OBJECTIVETo evaluate the efficacy and adverse events of prone positioning in nonintubated adult patients with acute hypoxemia and COVID-19.DESIGN, SETTING, AND PARTICIPANTS Pragmatic, unblinded randomized clinical trial conducted at 21 hospitals in Canada, Kuwait, Saudi Arabia, and the US. Eligible adult patients with COVID-19 were not intubated and required oxygen (Ն40%) or noninvasive ventilation. A total of 400 patients were enrolled between May 19, 2020, and May 18, 2021, and final follow-up was completed in July 2021.INTERVENTION Patients were randomized to awake prone positioning (n = 205) or usual care without prone positioning (control; n = 195). MAIN OUTCOMES AND MEASURESThe primary outcome was endotracheal intubation within 30 days of randomization. The secondary outcomes included mortality at 60 days, days free from invasive mechanical ventilation or noninvasive ventilation at 30 days, days free from the intensive care unit or hospital at 60 days, adverse events, and serious adverse events. RESULTS Among the 400 patients who were randomized (mean age, 57.6 years [SD, 12.83 years]; 117 [29.3%] were women), all (100%) completed the trial. In the first 4 days after randomization, the median duration of prone positioning was 4.8 h/d (IQR, 1.8 to 8.0 h/d) in the awake prone positioning group vs 0 h/d (IQR, 0 to 0 h/d) in the control group. By day 30, 70 of 205 patients (34.1%) in the prone positioning group were intubated vs 79 of 195 patients (40.5%) in the control group (hazard ratio, 0.81 [95% CI, 0.59 to 1.12], P = .20; absolute difference, −6.37% [95% CI, −15.83% to 3.10%]). Prone positioning did not significantly reduce mortality at 60 days (hazard ratio, 0.93 [95% CI, 0.62 to 1.40], P = .54; absolute difference, −1.15% [95% CI, −9.40% to 7.10%]) and had no significant effect on days free from invasive mechanical ventilation or noninvasive ventilation at 30 days or on days free from the intensive care unit or hospital at 60 days. There were no serious adverse events in either group. In the awake prone positioning group, 21 patients (10%) experienced adverse events and the most frequently reported were musculoskeletal pain or discomfort from prone positioning (13 of 205 patients [6.34%]) and desaturation (2 of 205 patients [0.98%]). There were no reported adverse events in the control group. CONCLUSIONS AND RELEVANCEIn patients with acute hypoxemic respiratory failure from COVID-19, prone positioning, compared with usual care without prone positioning, did not significantly reduce endotracheal intubation at 30 days. However, the effect size for the primary study outcome was imprecise and does not exclude a clinically important benefit.

Background: The role of remdesivir inThe primary outcome was in-hospital 24.8% and 28.2%, respectively (95% CI the treatment of patients in hospital mortality. Secondary outcomes 0.72 to 1.07). For patients not mechanwith COVID-19 remains ill defined in a included changes in clinical severity, ically ventilated at baseline, the need for global context. The World Health Organ-oxygen-and ventilator-free days (at mechanical ventilation was 8.0% in those ization Solidarity randomized controlled 28 d), incidence of new oxygen or assigned remdesivir, and 15.0% in those trial (RCT) evaluated remdesivir in mechanical ventilation use, duration of receiving standard of care (RR 0.53, 95% CI patients across many countries, with hospital stay, and adverse event rates. 0.38 to 0.75). Mean oxygen-free and Canada enrolling patients using anWe performed a priori subgroup analy-ventilator-free days at day 28 were 15.9 expanded data collection format in the ses according to duration of symptoms (± standard deviation [SD] 10.5) and 21.4 Canadian Treatments for COVID-19 before enrolment, age, sex and severity (± SD 11.3) in those receiving remdesivir (CATCO) trial. We report on the Canad-of symptoms on presentation. and 14.2 (± SD 11) and 19.5 (± SD 12.3) in ian findings, with additional demo-those receiving standard of care (p = 0.006 graphics, characteristics and clinical Results: Across 52 Canadian hospitals, and 0.007, respectively). There was no difoutcomes, to explore the potential for we randomized 1282 patients between ference in safety events of new dialysis, differential effects across different Aug. 14, 2020, and Apr. 1, 2021, to remde-change in creatinine, or new hepatic dyshealth care systems.sivir (n = 634) or standard of care (n = function between the 2 groups. 648). Of these, 15 withdrew consent or Methods: We performed an open-label, were still in hospital, for a total sample of Interpretation: Remdesivir, when compragmatic RCT in Canadian hospitals, in 1267 patients. Among patients assigned pared with standard of care, has a modest conjunction with the Solidarity trial. We to receive remdesivir, in-hospital mortal-but significant efect on outcomes imporrandomized patients to 10 days of rem-ity was 18.7%, compared with 22.6% in tant to patients and health systems, such desivir (200 mg intravenously [IV] on day the standard-of-care arm (relative risk as the need for mechanical ventilation. 0, followed by 100 mg IV daily), plus[RR] 0.83 (95% confidence interval [CI] Trial registration: ClinicalTrials.gov, no. standard care, or standard care alone. 0.67 to 1.03), and 60-day mortality was NCT04330690.

This is the author manuscript accepted for publication and has undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as