Machine learning methods offer great promise for fast and accurate detection and prognostication of coronavirus disease 2019 (COVID-19) from standard-of-care chest radiographs (CXR) and chest computed tomography (CT) images. Many articles have been published in 2020 describing new machine learning-based models for both of these tasks, but it is unclear which are of potential clinical utility. In this systematic review, we consider all published papers and preprints, for the period from 1 January 2020 to 3 October 2020, which describe new machine learning models for the diagnosis or prognosis of COVID-19 from CXR or CT images. All manuscripts uploaded to bioRxiv, medRxiv and arXiv along with all entries in EMBASE and MEDLINE in this timeframe are considered. Our search identified 2,212 studies, of which 415 were included after initial screening and, after quality screening, 62 studies were included in this systematic review. Our review finds that none of the models identified are of potential clinical use due to methodological flaws and/or underlying biases. This is a major weakness, given the urgency with which validated COVID-19 models are needed. To address this, we give many recommendations which, if followed, will solve these issues and lead to higher-quality model development and well-documented manuscripts.

BackgroundActivated phosphoinositide 3-kinase δ syndrome (APDS) is a recently described combined immunodeficiency resulting from gain-of-function mutations in PIK3CD, the gene encoding the catalytic subunit of phosphoinositide 3-kinase δ (PI3Kδ).ObjectiveWe sought to review the clinical, immunologic, histopathologic, and radiologic features of APDS in a large genetically defined international cohort.MethodsWe applied a clinical questionnaire and performed review of medical notes, radiology, histopathology, and laboratory investigations of 53 patients with APDS.ResultsRecurrent sinopulmonary infections (98%) and nonneoplastic lymphoproliferation (75%) were common, often from childhood. Other significant complications included herpesvirus infections (49%), autoinflammatory disease (34%), and lymphoma (13%). Unexpectedly, neurodevelopmental delay occurred in 19% of the cohort, suggesting a role for PI3Kδ in the central nervous system; consistent with this, PI3Kδ is broadly expressed in the developing murine central nervous system. Thoracic imaging revealed high rates of mosaic attenuation (90%) and bronchiectasis (60%). Increased IgM levels (78%), IgG deficiency (43%), and CD4 lymphopenia (84%) were significant immunologic features. No immunologic marker reliably predicted clinical severity, which ranged from asymptomatic to death in early childhood. The majority of patients received immunoglobulin replacement and antibiotic prophylaxis, and 5 patients underwent hematopoietic stem cell transplantation. Five patients died from complications of APDS.ConclusionAPDS is a combined immunodeficiency with multiple clinical manifestations, many with incomplete penetrance and others with variable expressivity. The severity of complications in some patients supports consideration of hematopoietic stem cell transplantation for severe childhood disease. Clinical trials of selective PI3Kδ inhibitors offer new prospects for APDS treatment.

IntroductionPneumothorax and pneumomediastinum have both been noted to complicate cases of COVID-19 requiring hospital admission. We report the largest case series yet described of patients with both these pathologies that includes non-ventilated patients.MethodsCases were collected retrospectively from UK hospitals with inclusion criteria limited to a diagnosis of COVID-19 and the presence of either pneumothorax or pneumomediastinum. Patients included in the study presented between March and June 2020. Details obtained from the medical record included demographics, radiology, laboratory investigations, clinical management and survival.ResultsSeventy-one patients from 16 centres were included in the study, of whom 60 patients had pneumothoraces (six also with pneumomediastinum), whilst 11 patients had pneumomediastinum alone. Two of these patients had two distinct episodes of pneumothorax, occurring bilaterally in sequential fashion, bringing the total number of pneumothoraces included to 62. Clinical scenarios included patients who had presented to hospital with pneumothorax, patients who had developed pneumothorax or pneumomediastinum during their inpatient admission with COVID-19 and patients who developed their complication whilst intubated and ventilated, either with or without concurrent extracorporeal membrane oxygenation. Survival at 28 days was not significantly different following pneumothorax (63.1%±6.5%) or isolated pneumomediastinum (53.0%±18.7%; p=0.854). The incidence of pneumothorax was higher in males. The 28-day survival was not different between the sexes (males 62.5%±7.7% versus females 68.4%±10.7%; p=0.619). Patients above the age of 70 had a significantly lower 28-day survival than younger individuals (≥70 years 41.7%±13.5% survival versus <70 years 70.9%±6.8% survival; p=0.018 log-rank).ConclusionThese cases suggest that pneumothorax is a complication of COVID-19. Pneumothorax does not seem to be an independent marker of poor prognosis and we encourage active treatment to be continued where clinically possible.

A proportion of people living with common variable immunodeficiency disorders develop granulomatous-lymphocytic interstitial lung disease (GLILD). We aimed to develop a consensus statement on the definition, diagnosis, and management of GLILD. All UK specialist centers were contacted and relevant physicians were invited to take part in a 3-round online Delphi process. Responses were graded as Strongly Agree, Tend to Agree, Neither Agree nor Disagree, Tend to Disagree, and Strongly Disagree, scored +1, +0.5, 0, -0.5, and -1, respectively. Agreement was defined as greater than or equal to 80% consensus. Scores are reported as mean ± SD. There was 100% agreement (score, 0.92 ± 0.19) for the following definition: "GLILD is a distinct clinico-radio-pathological ILD occurring in patients with [common variable immunodeficiency disorders], associated with a lymphocytic infiltrate and/or granuloma in the lung, and in whom other conditions have been considered and where possible excluded." There was consensus that the workup of suspected GLILD requires chest computed tomography (CT) (0.98 ± 0.01), lung function tests (eg, gas transfer, 0.94 ± 0.17), bronchoscopy to exclude infection (0.63 ± 0.50), and lung biopsy (0.58 ± 0.40). There was no consensus on whether expectant management following optimization of immunoglobulin therapy was acceptable: 67% agreed, 25% disagreed, score 0.38 ± 0.59; 90% agreed that when treatment was required, first-line treatment should be with corticosteroids alone (score, 0.55 ± 0.51).

Pulmonary involvement is frequent and among 140 patients with AAV who underwent a thoracic CT study, almost 80% have pulmonary abnormalities on thoracic CT. Central airway disease occurs exclusively among patients with PR3-ANCA while UIP were mainly seen in those with MPO-ANCA. These findings may have important implications for the investigation, management, and pathogenesis of AAV.