Search

Background: Automated analysis of lung computed tomography (CT) scans may help characterize subphenotypes of acute respiratory illness. We integrated lung CT features measured via deep learning with clinical and laboratory data in spontaneously breathing subjects to enhance the identification of COVID-19 subphenotypes., Methods: This is a multicenter observational cohort study in spontaneously breathing patients with COVID-19 respiratory failure exposed to early lung CT within 7 days of admission. We explored lung CT images using deep learning approaches to quantitative and qualitative analyses; latent class analysis (LCA) by using clinical, laboratory and lung CT variables; regional differences between subphenotypes following 3D spatial trajectories., Results: Complete datasets were available in 559 patients. LCA identified two subphenotypes (subphenotype 1 and 2). As compared with subphenotype 2 (n = 403), subphenotype 1 patients (n = 156) were older, had higher inflammatory biomarkers, and were more hypoxemic. Lungs in subphenotype 1 had a higher density gravitational gradient with a greater proportion of consolidated lungs as compared with subphenotype 2. In contrast, subphenotype 2 had a higher density submantellar-hilar gradient with a greater proportion of ground glass opacities as compared with subphenotype 1. Subphenotype 1 showed higher prevalence of comorbidities associated with endothelial dysfunction and higher 90-day mortality than subphenotype 2, even after adjustment for clinically meaningful variables., Conclusions: Integrating lung-CT data in a LCA allowed us to identify two subphenotypes of COVID-19, with different clinical trajectories. These exploratory findings suggest a role of automated imaging characterization guided by machine learning in subphenotyping patients with respiratory failure., Trial Registration: ClinicalTrials.gov Identifier: NCT04395482. Registration date: 19/05/2020., (© 2024. The Author(s).)

Background: Rapidly improving acute respiratory distress syndrome (RIARDS) is an increasingly appreciated subgroup of ARDS in which hypoxemia improves within 24 h after initiation of mechanical ventilation. Detailed clinical and biological features of RIARDS have not been clearly defined, and it is unknown whether RIARDS is associated with the hypoinflammatory or hyperinflammatory phenotype of ARDS. The purpose of this study was to define the clinical and biological features of RIARDS and its association with inflammatory subphenotypes., Methods: We analyzed data from 215 patients who met Berlin criteria for ARDS (endotracheally intubated) and were enrolled in a prospective observational cohort conducted at two sites, one tertiary care center and one urban safety net hospital. RIARDS was defined according to previous studies as improvement of hypoxemia defined as (i) PaO 2 :FiO 2  > 300 or (ii) SpO2: FiO 2  > 315 on the day following diagnosis of ARDS (day 2) or (iii) unassisted breathing by day 2 and for the next 48 h (defined as absence of endotracheal intubation on day 2 through day 4). Plasma biomarkers were measured on samples collected on the day of study enrollment, and ARDS phenotypes were allocated as previously described., Results: RIARDS accounted for 21% of all ARDS participants. Patients with RIARDS had better clinical outcomes compared to those with persistent ARDS, with lower hospital mortality (13% vs. 57%; p value < 0.001) and more ICU-free days (median 24 vs. 0; p value < 0.001). Plasma levels of interleukin-6, interleukin-8, and plasminogen activator inhibitor-1 were significantly lower among patients with RIARDS. The hypoinflammatory phenotype of ARDS was more common among patients with RIARDS (78% vs. 51% in persistent ARDS; p value = 0.001)., Conclusions: This study identifies a high prevalence of RIARDS in a multicenter observational cohort and confirms the more benign clinical course of these patients. We report the novel finding that RIARDS is characterized by lower concentrations of plasma biomarkers of inflammation compared to persistent ARDS, and that hypoinflammatory ARDS is more prevalent among patients with RIARDS. Identification and exclusion of RIARDS could potentially improve prognostic and predictive enrichment in clinical trials., (© 2024. The Author(s).)

Rationale: Two molecular phenotypes have been identified in acute respiratory distress syndrome (ARDS). In the ROSE (Reevaluation of Systemic Early Neuromuscular Blockade) trial of cisatracurium in moderate to severe ARDS, we addressed three unanswered questions: 1 ) Do the same phenotypes emerge in a more severe ARDS cohort with earlier recruitment; 2 ) Do phenotypes respond differently to neuromuscular blockade? and 3 ) What biological pathways most differentiate inflammatory phenotypes? Methods: We performed latent class analysis in ROSE using preenrollment clinical and protein biomarkers. In a subset of patients ( n  = 134), we sequenced whole-blood RNA using enrollment and Day 2 samples and performed differential gene expression and pathway analyses. Informed by the differential gene expression analysis, we measured additional plasma proteins and evaluated their abundance relative to gene expression amounts. Measurements and Main Results: In ROSE, we identified the hypoinflammatory (60.4%) and hyperinflammatory (39.6%) phenotypes with similar biological and clinical characteristics as prior studies, including higher mortality at Day 90 for the hyperinflammatory phenotype (30.3% vs. 61.6%; P  < 0.0001). We observed no treatment interaction between the phenotypes and randomized groups for mortality. The hyperinflammatory phenotype was enriched for genes associated with innate immune response, tissue remodeling, and zinc metabolism at Day 0 and collagen synthesis and neutrophil degranulation at Day 2. Longitudinal changes in gene expression patterns differed dependent on survivorship. For most highly expressed genes, we observed correlations with their corresponding plasma proteins' abundance. However, for the class-defining plasma proteins in the latent class analysis, no correlation was observed with their corresponding genes' expression. Conclusions: The hyperinflammatory and hypoinflammatory phenotypes have different clinical, protein, and dynamic transcriptional characteristics. These findings support the clinical and biological potential of molecular phenotypes to advance precision care in ARDS.

Background: In sepsis and acute respiratory distress syndrome (ARDS), heterogeneity has contributed to difficulty identifying effective pharmacotherapies. In ARDS, two molecular phenotypes (hypoinflammatory and hyperinflammatory) have consistently been identified, with divergent outcomes and treatment responses. In this study, we sought to derive molecular phenotypes in critically ill adults with sepsis, determine their overlap with previous ARDS phenotypes, and evaluate whether they respond differently to treatment in completed sepsis trials., Methods: We used clinical data and plasma biomarkers from two prospective sepsis cohorts, the Validating Acute Lung Injury biomarkers for Diagnosis (VALID) study (N=1140) and the Early Assessment of Renal and Lung Injury (EARLI) study (N=818), in latent class analysis (LCA) to identify the optimal number of classes in each cohort independently. We used validated models trained to classify ARDS phenotypes to evaluate concordance of sepsis and ARDS phenotypes. We applied these models retrospectively to the previously published Prospective Recombinant Human Activated Protein C Worldwide Evaluation in Severe Sepsis and Septic Shock (PROWESS-SHOCK) trial and Vasopressin and Septic Shock Trial (VASST) to assign phenotypes and evaluate heterogeneity of treatment effect., Findings: A two-class model best fit both VALID and EARLI (p<0·0001). In VALID, 804 (70·5%) of the 1140 patients were classified as hypoinflammatory and 336 (29·5%) as hyperinflammatory; in EARLI, 530 (64·8%) of 818 were hypoinflammatory and 288 (35·2%) hyperinflammatory. We observed higher plasma pro-inflammatory cytokines, more vasopressor use, more bacteraemia, lower protein C, and higher mortality in the hyperinflammatory than in the hypoinflammatory phenotype (p<0·0001 for all). Classifier models indicated strong concordance between sepsis phenotypes and previously identified ARDS phenotypes (area under the curve 0·87-0·96, depending on the model). Findings were similar excluding participants with both sepsis and ARDS. In PROWESS-SHOCK, 1142 (68·0%) of 1680 patients had the hypoinflammatory phenotype and 538 (32·0%) had the hyperinflammatory phenotype, and response to activated protein C differed by phenotype (p=0·0043). In VASST, phenotype proportions were similar to other cohorts; however, no treatment interaction with the type of vasopressor was observed (p=0·72)., Interpretation: Molecular phenotypes previously identified in ARDS are also identifiable in multiple sepsis cohorts and respond differently to activated protein C. Molecular phenotypes could represent a treatable trait in critical illness beyond the patient's syndromic diagnosis., Funding: US National Institutes of Health., Competing Interests: Declaration of interests PS reports funding from the US National Institutes of Health (NIH) and National Institute of General Medical Sciences; and consulting fees from AstraZeneca. LBW reports funding from NIH, Department of Defense (DoD), Genentech, Boehringer Ingelheim, and CSL Behring; consulting fees from Akebia Therapeutics, Santhera, Global Blood Therapeutics, and Boehringer Ingelheim; and stock options in Virtuoso Surgical. CSC reports funding from NIH; research grants from Roche Genentech and Quantum Leap Healthcare Collaborative; consulting fees from Vasomune Therapeutics, GEn1E Lifesciences, NGM Bio, Cellenkos, and Janssen; and a patent on metagenomic sequencing for sepsis diagnosis (co-recipient). MAM reports funding from Roche Genentech, Quantum Therapeutics, NIH/National Heart, Lung, and Blood Institute/National Institute of Allergy and Infectious Diseases, DoD, and California Institute for Regenerative Medicine; and consulting fees from Johnson & Johnson, Gilead Sciences, and Novartis. MMC reports funding from NIH and DoD; and intellectual property royalties from an issued patent (#11 410 777). JAR reports an investigator-initiated grant from Grifols provided to and administered by the University of British Columbia, Canadian Institutes of Health Research; three grants from the St Paul's Foundation; patents owned by the University of British Columbia related to the use of PCSK9 inhibitor(s) in sepsis and the use of vasopressin in septic shock, and by Ferring Pharmaceuticals for use of selepressin in septic shock; formerly being a founder, Director, and shareholder in Cyon Therapeutics (now closed); being a shareholder in Molecular You; receiving consulting fees in the last 3 years from SIB, Ferring Pharmaceuticals, and Par Pharmaceutical; and having been a funded member of the Data and Safety Monitoring Board of an NIH-sponsored trial of plasma in COVID-19 (PASS-IT-ON). KDL reports grants from NIH: National Institute of Diabetes and Digestive and Kidney Diseases; consulting fees from bioMérieux, UpToDate, SeaStar Medical, AM-Pharma, and Baxter; and stock or stock options in Amgen. All other authors declare no competing interests., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Importance: Reducing the duration of untreated psychosis (DUP) is essential to improving outcomes for people with first-episode psychosis (FEP). Current US approaches are insufficient to reduce DUP to international standards of less than 90 days., Objective: To determine whether population-based electronic screening in addition to standard targeted clinician education increases early detection of psychosis and decreases DUP, compared with clinician education alone., Design, Setting, and Participants: This cluster randomized clinical trial included individuals aged 12 to 30 years presenting for services between March 2015 and September 2017 at participating sites that included community mental health clinics and school support and special education services. Eligible participants were referred to the Early Diagnosis and Preventative Treatment (EDAPT) Clinic. Data analyses were performed in September and October 2019 for the primary and secondary analyses, with the exploratory subgroup analyses completed in May 2021., Interventions: All sites in both groups received targeted education about early psychosis for health care professionals. In the active screening group, clients also completed the Prodromal Questionnaire-Brief using tablets at intake; referrals were based on those scores and clinical judgment. In the group receiving treatment as usual (TAU), referrals were based on clinical judgment alone., Main Outcomes and Measures: Primary outcomes included DUP, defined as the period from full psychosis onset to the date of the EDAPT diagnostic telephone interview, and the number of individuals identified with FEP or a psychosis spectrum disorder. Exploratory analyses examined differences by site type, completion rates between conditions, and days from service entry to telephone interview., Results: Twenty-four sites agreed to participate, and 12 sites were randomized to either the active screening or TAU group. However, only 10 community clinics and 4 school sites were able to fully implement population screening and were included in the final analysis. The total potentially eligible population size within each study group was similar, with 2432 individuals entering at active screening group sites and 2455 at TAU group sites. A total of 303 diagnostic telephone interviews were completed (178 [58.7%] female individuals; mean [SD] age, 17.09 years [4.57]). Active screening sites reported a significantly higher detection rate of psychosis spectrum disorders (136 cases [5.6%], relative to 65 [2.6%]; P < .001) and referred a higher proportion of individuals with FEP and DUP less than 90 days (13 cases, relative to 4; odds ratio, 0.30; 95% CI, 0.10-0.93; P = .03). There was no difference in mean (SD) DUP between groups (active screening group, 239.0 days [207.4]; TAU group 262.3 days [170.2])., Conclusions and Relevance: In this cluster trial, population-based technology-enhanced screening across community settings detected more than twice as many individuals with psychosis spectrum disorders compared with clinical judgment alone but did not reduce DUP. Screening could identify people undetected in US mental health services. Significant DUP reduction may require interventions to reduce time to the first mental health contact., Trial Registration: ClinicalTrials.gov Identifier: NCT02841956.

Background: The ventilatory ratio (VR, [minute ventilation × PaCO 2 ]/[predicted body weight × 100 × 37.5]) is associated with mortality in ARDS. The aims of this study were to test whether baseline disease severity or neuromuscular blockade (NMB) modified the relationship between VR and mortality., Methods: This was a post hoc analysis of the PETAL-ROSE trial, which randomized moderate-to-severe ARDS patients to NMB or control. Survival among patients with different VR trajectories or VR cutoff above and below the median was assessed by Kaplan-Meier analysis. The relationships between single-day or 48-h VR trajectories with 28- or 90-day mortality were tested by logistic regression. Randomization allocation to NMB and markers of disease severity were tested as confounders by multivariable regression and interaction term analyses., Results: Patients with worsening VR trajectories had significantly lower survival compared to those with improving VR (n = 602, p < 0.05). Patients with VR > 2 (median) at day 1 had a significantly lower 90-day survival compared to patients with VR ≤ 2 (HR 1.36, 95% CI 1.10-1.69). VR at day 1 was significantly associated with 28-day mortality (OR = 1.40, 95% CI 1.15-1.72). There was no interaction between NMB and VR for 28-day mortality. APACHE-III had a significant interaction with VR at baseline for the outcome of 28-day mortality, such that the relationship between VR and mortality was stronger among patients with lower APACHE-III. There was a significant association between rising VR trajectory and mortality that was independent of NMB, baseline PaO 2 /FiO 2 ratio and generalized markers of disease severity (Adjusted OR 1.81, 95% CI 1.28-2.84 for 28-day and OR 2.07 95% CI 1.41-3.10 for 90-day mortality). APACHE-III and NMB were not effect modifiers in the relationship between VR trajectory and mortality., Conclusions: Elevated baseline and day 1 VR were associated with higher 28-day mortality. The relationship between baseline VR and mortality was stronger among patients with lower APACHE-III. APACHE-III was not an effect modifier for the relationship between VR trajectory and mortality, so that the VR trajectory may be optimally suited for prognostication and predictive enrichment. VR was not different between patients randomized to NMB or control, indicating that VR can be utilized without correcting for NMB., (© 2022. The Author(s).)

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., Competing Interests: Declaration of interests MC has a patent (ARCD P0535US.P2) pending to the University of Chicago (IL, USA) related to clinical deterioration risk prediction algorithms for patients admitted to hospital. MAM reports grants from Roche/Genentech, and personal fees from Johnson and Johnson, Novartis Pharmaceuticals, Gilead Pharmaceuticals, and Pliant Therapeutics, outside the submitted work. LBW reports grants and personal fees from Boehringer Ingelheim, outside the submitted work; grants from Genentech and CSL Behring, outside the submitted work; and personal fees from Merck, Citius, Quark, and Foresee, outside the submitted work. JGL reports personal fees from GlaxoSmithKline and Baxter, outside of the submitted work. GB reports grants and personal fees from Draeger Medical, and personal fees from Ge Healthcare, Hamilton Medical, and Flowmeter SPA, outside the submitted work. CSC reports grants and personal fees from Roche/Genentech and Bayer, outside the submitted work; personal fees from Quark Pharmaceuticals, Gen1e Life Sciences, and Vasomune, outside the submitted work; and grants from Quantum Leap Healthcare Collaborative, outside the submitted work. All other authors declare no competing interests., (Copyright © 2022 Elsevier Ltd. All rights reserved.)

Background: Previous latent class analysis of adults with acute respiratory distress syndrome (ARDS) identified two phenotypes, distinguished by the degree of inflammation. We aimed to identify phenotypes in children with ARDS in whom developmental differences might be important, using a latent class analysis approach similar to that used in adults., Methods: This study was a secondary analysis of data aggregated from the Randomized Evaluation of Sedation Titration for Respiratory Failure (RESTORE) clinical trial and the Genetic Variation and Biomarkers in Children with Acute Lung Injury (BALI) ancillary study. We used latent class analysis, which included demographic, clinical, and plasma biomarker variables, to identify paediatric ARDS (PARDS) phenotypes within a cohort of children included in the RESTORE and BALI studies. The association of phenotypes with clinically relevant outcomes and the performance of paediatric data in adult ARDS classification algorithms were also assessed., Findings: 304 children with PARDS were included in this secondary analysis. Using latent class analysis, a two-class model was a better fit for the cohort than a one-class model (p<0·001). Latent class analysis identified two classes: class 1 (181 [60%] of 304 patients with PARDS) and class 2 (123 [40%] of 304 patients with PARDS), referred to as phenotype 1 and 2 hereafter. Phenotype 2 was characterised by higher concentrations of inflammatory biomarkers, a higher incidence of vasopressor use, and more frequent diagnosis of sepsis, consistent with the adult hyperinflammatory phenotype. All levels of severity of PARDS were observed across both phenotypes. Children with the hyperinflammatory phenotype (phenotype 2) had worse clinical outcomes than those with the hypoinflammatory phenotype (phenotype 1), with a longer duration of mechanical ventilation (median 10·0 days [IQR 6·3-21·0] for phenotype 2 vs 6·6 days [4·1-10·8] for phenotype 1, p<0·0001), and higher incidence of mortality (17 [13·8%] of 123 patients vs four [2·2%] of 181 patients, p=0·0001). When using adult phenotype classification algorithms in children, the soluble tumour necrosis factor receptor-1 (sTNFr1), vasopressor use, and interleukin (IL)-6 variables gave an area under the curve (AUC) of 0·956, and the sTNFr1, vasopressor use, and IL-8 variables gave an AUC of 0·954, compared with the gold standard of latent class analysis., Interpretation: Latent class analysis identified two phenotypes in children with ARDS with characteristics similar to those in adults, including worse outcomes among patients with the hyperinflammatory phenotype. PARDS phenotypes should be considered in design and analysis of future clinical trials in children., Funding: US National Institutes of Health., Competing Interests: Declaration of interests MKD, MZ, MAQC, AS, KLD, and HF report grants from the US National Institutes of Health. CSC reports grants from the US National Institutes of Health, Roche/Genentech, Bayer, and Quantum Leap Healthcare Collaborative; and consultancy fees from Quark Pharmaceuticals, Prometic, Gen1e Life Sciences, and Vasomune. MWQ reports consultancy fees from Mitchell Leeds, Tanoury, Nauts, and McKinney and Barbarino. All other authors declare no competing interests., (Copyright © 2022 Elsevier Ltd. All rights reserved.)

Rationale: Using latent class analysis (LCA), two subphenotypes of acute respiratory distress syndrome (ARDS) have consistently been identified in five randomised controlled trials (RCTs), with distinct biological characteristics, divergent outcomes and differential treatment responses to randomised interventions. Their existence in unselected populations of ARDS remains unknown. We sought to identify subphenotypes in observational cohorts of ARDS using LCA., Methods: LCA was independently applied to patients with ARDS from two prospective observational cohorts of patients admitted to the intensive care unit, derived from the Validating Acute Lung Injury markers for Diagnosis (VALID) (n=624) and Early Assessment of Renal and Lung Injury (EARLI) (n=335) studies. Clinical and biological data were used as class-defining variables. To test for concordance with prior ARDS subphenotypes, the performance metrics of parsimonious classifier models (interleukin 8, bicarbonate, protein C and vasopressor-use), previously developed in RCTs, were evaluated in EARLI and VALID with LCA-derived subphenotypes as the gold-standard., Results: A 2-class model best fit the population in VALID (p=0.0010) and in EARLI (p<0.0001). Class 2 comprised 27% and 37% of the populations in VALID and EARLI, respectively. Consistent with the previously described 'hyperinflammatory' subphenotype, Class 2 was characterised by higher proinflammatory biomarkers, acidosis and increased shock and worse clinical outcomes. The similarities between these and prior RCT-derived subphenotypes were further substantiated by the performance of the parsimonious classifier models in both cohorts (area under the curves 0.92-0.94). The hyperinflammatory subphenotype was associated with increased prevalence of chronic liver disease and neutropenia and reduced incidence of chronic obstructive pulmonary disease. Measurement of novel biomarkers showed significantly higher levels of matrix metalloproteinase-8 and markers of endothelial injury in the hyperinflammatory subphenotype, whereas, matrix metalloproteinase-9 was significantly lower., Conclusion: Previously described subphenotypes are generalisable to unselected populations of non-trauma ARDS., Competing Interests: Competing interests: PS declares no competing interests. CSC has received grant funding from the NIH, US Food and Drug Administration, Roche-Genentech, Quantum Leap Healthcare Collaborative, and Bayer Pharmaceuticals and has served as a consultant to Vasomune, Quark, and Gen1e Life Sciences., (© Author(s) (or their employer(s)) 2022. No commercial re-use. See rights and permissions. Published by BMJ.)

Background: Heterogeneity in Acute Respiratory Distress Syndrome (ARDS), as a consequence of its non-specific definition, has led to a multitude of negative randomised controlled trials (RCTs). Investigators have sought to identify heterogeneity of treatment effect (HTE) in RCTs using clustering algorithms. We evaluated the proficiency of several commonly-used machine-learning algorithms to identify clusters where HTE may be detected., Methods: Five unsupervised: Latent class analysis (LCA), K-means, partition around medoids, hierarchical, and spectral clustering; and four supervised algorithms: model-based recursive partitioning, Causal Forest (CF), and X-learner with Random Forest (XL-RF) and Bayesian Additive Regression Trees were individually applied to three prior ARDS RCTs. Clinical data and research protein biomarkers were used as partitioning variables, with the latter excluded for secondary analyses. For a clustering schema, HTE was evaluated based on the interaction term of treatment group and cluster with day-90 mortality as the dependent variable., Findings: No single algorithm identified clusters with significant HTE in all three trials. LCA, XL-RF, and CF identified HTE most frequently (2/3 RCTs). Important partitioning variables in the unsupervised approaches were consistent across algorithms and RCTs. In supervised models, important partitioning variables varied between algorithms and across RCTs. In algorithms where clusters demonstrated HTE in the same trial, patients frequently interchanged clusters from treatment-benefit to treatment-harm clusters across algorithms. LCA aside, results from all other algorithms were subject to significant alteration in cluster composition and HTE with random seed change. Removing research biomarkers as partitioning variables greatly reduced the chances of detecting HTE across all algorithms., Interpretation: Machine-learning algorithms were inconsistent in their abilities to identify clusters with significant HTE. Protein biomarkers were essential in identifying clusters with HTE. Investigations using machine-learning approaches to identify clusters to seek HTE require cautious interpretation., Funding: NIGMS R35 GM142992 (PS), NHLBI R35 HL140026 (CSC); NIGMS R01 GM123193, Department of Defense W81XWH-21-1-0009, NIA R21 AG068720, NIDA R01 DA051464 (MMC)., Competing Interests: Declaration of Competing Interest Dr Sinha received institutional grants from the National Institute of Health (NIH), Ms Spicer received institutional grant funding from the NIH and Department of Defense. Prof Delucchi declares no relevant conflicts of interest. Prof McAuley received institutional grants from the NIHR, Innovate UK, MRC, Novavax, Northern Ireland HSC R&D Division, Wellcome Trust; royalties/licenses from Queen's University Belfast for a patent of novel treatment for inflammatory diseases (USB962032); consulting fees from Bayer, GlaxoSmithKline, Boehringer Ingelhelm, Novartis, Eli Lilly, SOBI; payment from GlaxoSmithKline as an educational seminary speaker; travel support from Vir Biotechnology Inc and Faron Pharmaceuticals as a member of Data Safety Monitoring Board. His spouse received consultancy fees from Insmed and payments for participation in grant review panel from California Institute for Regenerative Medicine. Prof Calfee received institutional grants from the NIH, Bayer, Roche-Genentech, Quantum Leap Healthcare Collaborative; and personal fees from Quark, Vasomune, and Gen1e Life Science. Dr Churpek received institutional grants from the NIH, Department of Defense, and EarlySense., (Copyright © 2021 The Authors. Published by Elsevier B.V. All rights reserved.)

Introduction: This study examined the impact of a San Francisco City and County ban on all flavoured tobacco products, including menthol cigarettes, among clients in residential substance use disorder (SUD) treatment., Methods: We conducted cross-sectional surveys of clients at two residential SUD programmes before the County began enforcing the ban (n=160) and twice after enforcement began (n=102, n=120). The samples were compared on demographic characteristics, smoking status, smoking behaviours and the proportion reporting menthol as their usual cigarette. Menthol smokers were asked whether they smoked only menthol cigarettes, mostly menthol, both menthol and non-menthol or mostly non-menthol. Post-ban samples were asked about awareness of the ban and access to menthol cigarettes., Results: In multivariate analyses, we found no evidence that the ban was associated with decreased number of cigarettes per day or increased readiness to quit among current smokers. However, odds were lower post-ban for reporting menthol as the usual cigarette (OR=0.80, 95% CI 0.72 to 0.90), and for smoking only menthol cigarettes (OR=0.19, 95% CI 0.18 to 0.19). Perhaps most importantly, and with the ability to influence all other findings, 50% of self-identified menthol smokers reported purchasing menthol cigarettes in San Francisco nearly 1 year after the ban was implemented., Conclusion: In subgroups where smoking has remained elevated, like those receiving SUD treatment, local menthol bans may have only modest impacts on smoking behaviour. Broader regional, state or national bans, that effectively restrict access to menthol products, may be needed to show stronger effects on smoking behaviour., Competing Interests: Competing interests: No, there are no competing interests., (© Author(s) (or their employer(s)) 2021. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.)