Development and validation of a CT-based deep learning algorithm to augment non-invasive diagnosis of idiopathic pulmonary fibrosis.

RATIONALE: Non-invasive diagnosis of idiopathic pulmonary fibrosis (IPF) involves identification of usual interstitial pneumonia (UIP) pattern by computed tomography (CT) and exclusion of other known etiologies of interstitial lung disease (ILD). However, uncertainty in identification of radiologic UIP pattern leads to the continued need for invasive surgical biopsy. We thus developed and validated a machine learning algorithm using CT scans alone to augment non-invasive diagnosis of IPF. METHODS: The primary algorithm was a deep learning convolutional neural network (CNN) with model inputs of CT images only. The algorithm was trained to predict IPF among cases of ILD, with reference standard of multidisciplinary discussion (MDD) consensus diagnosis. The algorithm was trained using a multi-center dataset of more than 2000 cases of ILD. A US-based multi-site cohort (n = 295) was used for algorithm tuning, and external validation was performed with a separate dataset (n = 295) from European and South American sources. RESULTS: In the tuning set, the model achieved an area under the receiver operating characteristic curve (AUC) of 0.87 (CI: 0.83-0.92) in differentiating IPF from other ILDs. Sensitivity and specificity were 0.67 (0.57-0.76) and 0.90 (0.83-0.95), respectively. By contrast, pre-recorded assessment prior to MDD diagnosis had sensitivity of 0.31 (0.23-0.42) and specificity of 0.92 (0.87-0.95). In the external test set, c-statistic was also 0.87 (0.83-0.91). Model performance was consistent across a variety of CT scanner manufacturers and slice thickness. CONCLUSION: The presented deep learning algorithm demonstrated consistent performance in identifying IPF among cases of ILD using CT images alone and suggests generalization across CT manufacturers.

Estimating the attributable fraction of mortality from acute respiratory distress syndrome to inform enrichment in future randomised clinical trials.

BACKGROUND: Efficiency of randomised clinical trials of acute respiratory distress syndrome (ARDS) depends on the fraction of deaths attributable to ARDS (AFARDS) to which interventions are targeted. Estimates of AFARDS in subpopulations of ARDS could improve design of ARDS trials. METHODS: We performed a matched case-control study using the Large observational study to UNderstand the Global impact of Severe Acute respiratory FailurE cohort. Primary outcome was intensive care unit mortality. We used nearest neighbour propensity score matching without replacement to match ARDS to non-ARDS populations. We derived two separate AFARDS estimates by matching patients with ARDS to patients with non-acute hypoxaemic respiratory failure (non-AHRF) and to patients with AHRF with unilateral infiltrates only (AHRF-UL). We also estimated AFARDS in subgroups based on severity of hypoxaemia, number of lung quadrants involved and hyperinflammatory versus hypoinflammatory phenotypes. Additionally, we derived AFAHRF estimates by matching patients with AHRF to non-AHRF controls, and AFAHRF-UL estimates by matching patients with AHRF-UL to non-AHRF controls. RESULTS: Estimated AFARDS was 20.9% (95% CI 10.5% to 31.4%) when compared with AHRF-UL controls and 38.0% (95% CI 34.4% to 41.6%) compared with non-AHRF controls. Within subgroups, estimates for AFARDS compared with AHRF-UL controls were highest in patients with severe hypoxaemia (41.1% (95% CI 25.2% to 57.1%)), in those with four quadrant involvement on chest radiography (28.9% (95% CI 13.4% to 44.3%)) and in the hyperinflammatory subphenotype (26.8% (95% CI 6.9% to 46.7%)). Estimated AFAHRF was 33.8% (95% CI 30.5% to 37.1%) compared with non-AHRF controls. Estimated AFAHRF-UL was 21.3% (95% CI 312.8% to 29.7%) compared with non-AHRF controls. CONCLUSIONS: Overall AFARDS mean values were between 20.9% and 38.0%, with higher AFARDS seen with severe hypoxaemia, four quadrant involvement on chest radiography and hyperinflammatory ARDS.

Validation and utility of ARDS subphenotypes identified by machine-learning models using clinical data: an observational, multicohort, retrospective analysis.

BACKGROUND: Two acute respiratory distress syndrome (ARDS) subphenotypes (hyperinflammatory and hypoinflammatory) with distinct clinical and biological features and differential treatment responses have been identified using latent class analysis (LCA) in seven individual cohorts. To facilitate bedside identification of subphenotypes, clinical classifier models using readily available clinical variables have been described in four randomised controlled trials. We aimed to assess the performance of these models in observational cohorts of ARDS. METHODS: In this observational, multicohort, retrospective study, we validated two machine-learning clinical classifier models for assigning ARDS subphenotypes in two observational cohorts of patients with ARDS: Early Assessment of Renal and Lung Injury (EARLI; n=335) and Validating Acute Lung Injury Markers for Diagnosis (VALID; n=452), with LCA-derived subphenotypes as the gold standard. The primary model comprised only vital signs and laboratory variables, and the secondary model comprised all predictors in the primary model, with the addition of ventilatory variables and demographics. Model performance was assessed by calculating the area under the receiver operating characteristic curve (AUC) and calibration plots, and assigning subphenotypes using a probability cutoff value of 0·5 to determine sensitivity, specificity, and accuracy of the assignments. We also assessed the performance of the primary model in EARLI using data automatically extracted from an electronic health record (EHR; EHR-derived EARLI cohort). In Large Observational Study to Understand the Global Impact of Severe Acute Respiratory Failure (LUNG SAFE; n=2813), a multinational, observational ARDS cohort, we applied a custom classifier model (with fewer variables than the primary model) to determine the prognostic value of the subphenotypes and tested their interaction with the positive end-expiratory pressure (PEEP) strategy, with 90-day mortality as the dependent variable. FINDINGS: The primary clinical classifier model had an area under receiver operating characteristic curve (AUC) of 0·92 (95% CI 0·90-0·95) in EARLI and 0·88 (0·84-0·91) in VALID. Performance of the primary model was similar when using exclusively EHR-derived predictors compared with manually curated predictors (AUC=0·88 [95% CI 0·81-0·94] vs 0·92 [0·88-0·97]). In LUNG SAFE, 90-day mortality was higher in patients assigned the hyperinflammatory subphenotype than in those with the hypoinflammatory phenotype (414 [57%] of 725 vs 694 [33%] of 2088; p<0·0001). There was a significant treatment interaction with PEEP strategy and ARDS subphenotype (p=0·041), with lower 90-day mortality in the high PEEP group of patients with the hyperinflammatory subphenotype (hyperinflammatory subphenotype: 169 [54%] of 313 patients in the high PEEP group vs 127 [62%] of 205 patients in the low PEEP group; hypoinflammatory subphenotype: 231 [34%] of 675 patients in the high PEEP group vs 233 [32%] of 734 patients in the low PEEP group). INTERPRETATION: Classifier models using clinical variables alone can accurately assign ARDS subphenotypes in observational cohorts. Application of these models can provide valuable prognostic information and could inform management strategies for personalised treatment, including application of PEEP, once prospectively validated. FUNDING: US National Institutes of Health and European Society of Intensive Care Medicine.

Latent Class Analysis Reveals COVID-19-related Acute Respiratory Distress Syndrome Subgroups with Differential Responses to Corticosteroids.

Rationale: Two distinct subphenotypes have been identified in acute respiratory distress syndrome (ARDS), but the presence of subgroups in ARDS associated with coronavirus disease (COVID-19) is unknown. Objectives: To identify clinically relevant, novel subgroups in COVID-19-related ARDS and compare them with previously described ARDS subphenotypes. Methods: Eligible participants were adults with COVID-19 and ARDS at Columbia University Irving Medical Center. Latent class analysis was used to identify subgroups with baseline clinical, respiratory, and laboratory data serving as partitioning variables. A previously developed machine learning model was used to classify patients as the hypoinflammatory and hyperinflammatory subphenotypes. Baseline characteristics and clinical outcomes were compared between subgroups. Heterogeneity of treatment effect for corticosteroid use in subgroups was tested. Measurements and Main Results: From March 2, 2020, to April 30, 2020, 483 patients with COVID-19-related ARDS met study criteria. A two-class latent class analysis model best fit the population (P = 0.0075). Class 2 (23%) had higher proinflammatory markers, troponin, creatinine, and lactate, lower bicarbonate, and lower blood pressure than class 1 (77%). Ninety-day mortality was higher in class 2 versus class 1 (75% vs. 48%; P < 0.0001). Considerable overlap was observed between these subgroups and ARDS subphenotypes. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RT-PCR cycle threshold was associated with mortality in the hypoinflammatory but not the hyperinflammatory phenotype. Heterogeneity of treatment effect to corticosteroids was observed (P = 0.0295), with improved mortality in the hyperinflammatory phenotype and worse mortality in the hypoinflammatory phenotype, with the caveat that corticosteroid treatment was not randomized. Conclusions: We identified two COVID-19-related ARDS subgroups with differential outcomes, similar to previously described ARDS subphenotypes. SARS-CoV-2 PCR cycle threshold had differential value for predicting mortality in the subphenotypes. The subphenotypes had differential treatment responses to corticosteroids.

New Strategies in Clinical Guideline Delivery: Randomized Trial of Online, Interactive Decision Support Versus Guidelines for Human Immunodeficiency Virus Treatment Selection by Trainees.

BACKGROUND: Support for clinicians in human immunodeficiency virus (HIV) medicine is critical given national HIV-provider shortages. The US Department of Health and Human Services (DHHS) guidelines are comprehensive but complex to apply for antiretroviral therapy (ART) selection. Human immunodeficiency virus antiretroviral selection support and interactive search tool (HIV-ASSIST) (www.hivassist.com) is a free tool providing ART decision support that could augment implementation of clinical practice guidelines. METHODS: We conducted a randomized study of medical trainees at Johns Hopkins University, in which participants were asked to select an ART regimen for 10 HIV case scenarios through an electronic survey. Participants were randomized to receive either DHHS guidelines alone, or DHHS guidelines and HIV-ASSIST to support their decision making. ART selections were graded "appropriate" if consistent with DHHS guidelines, or concordant with regimens selected by HIV experts at 4 academic institutions. RESULTS: Among 118 trainees, participants randomized to receive HIV-ASSIST had a significantly higher percentage of appropriate ART selections compared to those receiving DHHS guidelines alone (percentage of appropriate responses in DHHS vs HIV-ASSIST arms: median [Q1, Q3], 40% [30%, 50%] vs 90% [80%, 100%]; P < .001). The effect was seen for all case types, but most pronounced for complex cases involving ART-experienced patients with ongoing viremia (DHHS vs HIV-ASSIST: median [Q1, Q3], 0% [0%, 33%] vs 100% [66%, 100%]). CONCLUSIONS: Trainees using HIV-ASSIST were significantly more likely to choose appropriate ART regimens compared to those using guidelines alone. Interactive decision support tools may be important to ensure appropriate implementation of HIV guidelines. CLINICAL TRIALS REGISTRATION: NCT04080765.

Carcinoid Crisis-Induced Acute Systolic Heart Failure.

Carcinoid crisis is a life-threatening manifestation of carcinoid syndrome characterized by profound autonomic instability in the setting of catecholamine release from stress, tumor manipulation, or anesthesia. Here, we present an unusual case of carcinoid crisis leading to acute systolic heart failure requiring mechanical circulatory support. (Level of Difficulty: Intermediate.).

Evaluating the Concordance of Clinician Antiretroviral Prescribing Practices and HIV-ASSIST, an Online Clinical Decision Support Tool.

BACKGROUND: Individualized selection of antiretroviral (ARV) therapy is complex, considering drug resistance, comorbidities, drug-drug interactions, and other factors. HIV-ASSIST (www.hivassist.com) is a free, online tool that provides ARV decision support. HIV-ASSIST synthesizes patient and virus-specific attributes to rank ARV combinations based upon a composite objective of achieving viral suppression and maximizing tolerability. OBJECTIVE: To evaluate concordance of HIV-ASSIST recommendations with ARV selections of experienced HIV clinicians. DESIGN: Retrospective cohort study. PATIENTS: New and established patients at the Johns Hopkins Bartlett HIV Clinic and San Francisco Veterans Affairs HIV Clinic completing clinic visits were included. Chart reviews were conducted of the most recent clinic visit to generate HIV-ASSIST recommendations, which were compared to prescribed regimens. MAIN MEASURES: For each provider-prescribed regimen, we assessed its corresponding HIV-ASSIST "weighted score" (scale of 0 to 10 +, scores of < 2.0 are preferred), rank within HIV-ASSIST's ordered listing of ARV regimens, and concordance with the top five HIV-ASSIST ranked outputs. KEY RESULTS: Among 106 patients (16% female), 23 (22%) were ARV-naïve. HIV-ASSIST outputs for ARV-naïve patients were 100% concordant with prescribed regimens (median rank 1 [IQR 1-3], median weighted score 1.1 [IQR 1-1.2]). For 18 (17%) ARV-experienced patients with ongoing viremia, HIV-ASSIST outputs were 89% concordant with prescribed regimens (median rank 2 [IQR 1-3], median weighted score 1 [IQR 1-1.2]). For 65 (61.3%) patients that were suppressed on a current ARV regimen, HIV-ASSIST recommendations were concordant 88% of the time (median rank 1 [IQR 1-1], median weighted score 1.1 [IQR 1-1.6]). In 18% of cases, HIV-ASSIST weighted score suggested that the prescribed regimen would be considered "less preferred" (score > 2.0) than other available alternatives. CONCLUSION: HIV-ASSIST is an educational decision support tool that provides ARV recommendations concordant with experienced HIV providers from two major academic centers for a diverse set of patient scenarios.

Development and Validation of HIV-ASSIST, an Online, Educational, Clinical Decision Support Tool to Guide Patient-Centered ARV Regimen Selection.

BACKGROUND: Multiple antiretroviral (ARV) regimens are effective at achieving HIV viral suppression, but differ in pill burden, side effects, barriers to resistance, and impact on comorbidities. Current guidelines advocate for an individualized approach to ARV regimen selection, but synthesizing these modifying factors is complex and time-consuming. METHODS: We describe the development of HIV-ASSIST (https://www.hivassist.com), a free, online decision support tool for ARV selection and HIV education. HIV-ASSIST ranks potential ARV options for any given patient scenario using a composite objective of achieving viral suppression while maximizing tolerability and adherence. We used a multiple-criteria decision analysis framework to construct mathematical algorithms and synthesize various patient-specific (eg, comorbidities and treatment history) and virus-specific (eg, HIV mutations) attributes. We then conducted a validation study to evaluate HIV-ASSIST with prescribing practices of experienced HIV providers at 4 large academic centers. We report on concordance of provider ARV selections with the 5 top-ranked HIV-ASSIST regimens for 10 diverse hypothetical patient-case scenarios. RESULTS: In the validation cohort of 17 experienced HIV providers, we found 99% concordance between HIV-ASSIST recommendations and provider ARV selections for 4 case-scenarios of ARV-naive patients. Among 6 cases of ARV-experienced patients (3 with and 3 without viremia), there was 84% and 88% concordance, respectively. Among 3 cases of ARV-experienced patients with viremia, providers reported 20 different ARV selections, suggesting substantial heterogeneity in ARV preferences in clinical practice. CONCLUSIONS: HIV-ASSIST is a novel patient-centric educational decision support tool that provides ARV recommendations concordant with experienced HIV providers for a diverse set of patient scenarios.

Author Correction: Division of labour between Myc and G1 cyclins in cell cycle commitment and pace control.

This Article contains errors in Supplementary Table 3, which are described in the Author Correction associated with this Article. The simulation results in the Article were based on the correct formula and thus the results are not affected by this correction. The errors have not been fixed in the original Article.

Epidemiological impact of achieving UNAIDS 90-90-90 targets for HIV care in India: a modelling study.

OBJECTIVE: Recent UNAIDS '90-90-90' targets propose that to end the HIV epidemic by 2030, 90% of persons living with HIV (PLWH) worldwide should know their diagnosis, 90% of diagnosed PLWH should be on antiretroviral therapy (ART) and 90% of PLWH on ART should be virally suppressed by 2020. We sought to quantify the epidemiological impact of achieving these targets in India. METHODS: We constructed a dynamic-transmission model of the Indian HIV epidemic to project HIV infections and AIDS-related deaths that would occur in India over 15 years. We considered several scenarios: continuation of current care engagement (with early ART initiation), achieving 90-90-90 targets on time and delaying achievement by 5 or 10 years. RESULTS: In the base case, assuming continuation of current care engagement, we project 794 000 (95% uncertainty range (UR) 571 000-1 104 000) HIV infections and 689 000 (95% UR 468 000-976 000) AIDS-related deaths in India over 15 years. In this scenario, nearly half of PLWH diagnosed would fail to achieve viral suppression by 2030. With achievement of 90-90-90 targets, India could avert 392 000 (95% UR 248 000-559 000) transmissions (48% reduction) and 414 000 (95% UR 260 000-598 000) AIDS-related deaths (59% reduction) compared to the base-case scenario. Furthermore, fewer than 20 000 (95% UR 12 000-30 000) HIV infections would occur in 2030. Delaying achievement of targets resulted in a similar reduction in HIV incidence by 2030 but at the cost of excess overall infections and mortality. CONCLUSIONS: India can halve the epidemiological burden of HIV over 15 years with achievement of the UNAIDS 90-90-90 targets. Reaching the targets on time will require comprehensive healthcare strengthening, especially in early diagnosis and treatment, expanded access to second-line and third-line ART and long-term retention in care.

Economic and epidemiological impact of early antiretroviral therapy initiation in India.

INTRODUCTION: Recent WHO guidance advocates for early antiretroviral therapy (ART) initiation at higher CD4 counts to improve survival and reduce HIV transmission. We sought to quantify how the cost-effectiveness and epidemiological impact of early ART strategies in India are affected by attrition throughout the HIV care continuum. METHODS: We constructed a dynamic compartmental model replicating HIV transmission, disease progression and health system engagement among Indian adults. Our model of the Indian HIV epidemic compared implementation of early ART initiation (i.e. initiation above CD4 ≥350 cells/mm(3)) with delayed initiation at CD4 ≤350 cells/mm(3); primary outcomes were incident cases, deaths, quality-adjusted-life-years (QALYs) and costs over 20 years. We assessed how costs and effects of early ART initiation were impacted by suboptimal engagement at each stage in the HIV care continuum. RESULTS: Assuming "idealistic" engagement in HIV care, early ART initiation is highly cost-effective ($442/QALY-gained) compared to delayed initiation at CD4 ≤350 cells/mm(3) and could reduce new HIV infections to <15,000 per year within 20 years. However, when accounting for realistic gaps in care, early ART initiation loses nearly half of potential epidemiological benefits and is less cost-effective ($530/QALY-gained). We project 1,285,000 new HIV infections and 973,000 AIDS-related deaths with deferred ART initiation with current levels of care-engagement in India. Early ART initiation in this continuum resulted in 1,050,000 new HIV infections and 883,000 AIDS-related deaths, or 18% and 9% reductions (respectively), compared to current guidelines. Strengthening HIV screening increases benefits of earlier treatment modestly (1,001,000 new infections; 22% reduction), while improving retention in care has a larger modulatory impact (676,000 new infections; 47% reduction). CONCLUSIONS: Early ART initiation is highly cost-effective in India but only has modest epidemiological benefits at current levels of care-engagement. Improved retention in care is needed to realize the full potential of earlier treatment.

Division of labour between Myc and G1 cyclins in cell cycle commitment and pace control.

A body of evidence has shown that the control of E2F transcription factor activity is critical for determining cell cycle entry and cell proliferation. However, an understanding of the precise determinants of this control, including the role of other cell-cycle regulatory activities, has not been clearly defined. Here, recognizing that the contributions of individual regulatory components could be masked by heterogeneity in populations of cells, we model the potential roles of individual components together with the use of an integrated system to follow E2F dynamics at the single-cell level and in real time. These analyses reveal that crossing a threshold amplitude of E2F accumulation determines cell cycle commitment. Importantly, we find that Myc is critical in modulating the amplitude, whereas cyclin D/E activities have little effect on amplitude but do contribute to the modulation of duration of E2F activation, thereby affecting the pace of cell cycle progression.