Heterogeneous effects of alveolar recruitment in acute respiratory distress syndrome: a machine learning reanalysis of the Alveolar Recruitment for Acute Respiratory Distress Syndrome Trial

Differential Effect of Positive End-Expiratory Pressure Strategies in Patients With ARDS: A Bayesian Analysis of Clinical Subphenotypes

BACKGROUND

ARDS is a heterogeneous condition with two subphenotypes identified by different methodologies. Our group similarly identified two ARDS subphenotypes using nine routinely available clinical variables. However, whether these are associated with differential response to treatment has yet to be explored.

RESEARCH QUESTION

Are there differential responses to positive end-expiratory pressure (PEEP) strategies on 28-day mortality according to subphenotypes in adult patients with ARDS?

STUDY DESIGN AND METHODS

We evaluated data from two prior ARDS trials (Higher vs Lower Positive End-Expiratory Pressures in Patients With the ARDS [ALVEOLI] and the Alveolar Recruitment in ARDS Trial [ART]) that compared different PEEP strategies. We classified patients into one of two subphenotypes as described previously. We assessed the differential effect of PEEP with a Bayesian hierarchical logistic model for the primary outcome of 28-day mortality.

RESULTS

We analyzed data from 1,559 patients with ARDS. Compared with lower PEEP, a higher PEEP strategy resulted in higher 28-day mortality in patients with subphenotype A disease in the ALVEOLI study (OR, 1.61; 95% credible interval [CrI], 0.90-2.94) and ART (OR, 1.73; 95% CrI, 1.01-2.98), with a probability of harm resulting from higher PEEP in this subphenotype of 94.3% and 97.7% in the ALVEOLI and ART studies, respectively. Higher PEEP was not associated with mortality in patients with subphenotype B disease in each trial (OR, 0.95 [95% CrI, 0.51-1.73] and 1.00 [95% CrI, 0.63-1.55], respectively), with probability of benefit of 56.4% and 50.7% in the ALVEOLI and ART studies, respectively. These effects were not modified by Pao2 to Fio2 ratio, driving pressure, or the severity of illness for the cohorts.

INTERPRETATION

We found evidence of differential response to PEEP strategies across two ARDS subphenotypes, suggesting possible harm with a higher PEEP strategy in one subphenotype. These observations may assist with predictive enrichment in future clinical trials.

High PEEP/low FiO2 ventilation is associated with lower mortality in COVID-19

RATIONALE

The positive end-expiratory pressure (PEEP) strategy in patients with coronavirus 2019 (COVID-19) acute respiratory distress syndrome (ARDS) remains debated. Most studies originate from the initial waves of the pandemic. Here we aimed to assess the impact of high PEEP/low FiO2 ventilation on outcomes during the second wave in the Netherlands.

METHODS

Retrospective observational study of invasively ventilated COVID-19 patients during the second wave. Patients were categorized based on whether they received high PEEP or low PEEP ventilation according to the ARDS Network tables. The primary outcome was ICU mortality, and secondary outcomes included hospital and 90-day mortality, duration of ventilation and length of stay, and the occurrence of kidney injury. Propensity matching was performed to correct for factors with a known relationship to ICU mortality.

RESULTS

This analysis included 790 COVID-ARDS patients. At ICU discharge, 32 (22.5%) out of 142 high PEEP patients and 254 (39.2%) out of 848 low PEEP patients had died (HR 0.66 [0.46-0.96]; P = 0.03). High PEEP was linked to improved secondary outcomes. Matched analysis did not change findings.

CONCLUSIONS

High PEEP ventilation was associated with improved ICU survival in patients with COVID-ARDS.

Epidemiology, ventilation management and outcomes of COVID-19 ARDS patients versus patients with ARDS due to pneumonia in the Pre-COVID era

BACKGROUND

Ventilation management may differ between COVID-19 ARDS (COVID-ARDS) patients and patients with pre-COVID ARDS (CLASSIC-ARDS); it is uncertain whether associations of ventilation management with outcomes for CLASSIC-ARDS also exist in COVID-ARDS.

METHODS

Individual patient data analysis of COVID-ARDS and CLASSIC-ARDS patients in six observational studies of ventilation, four in the COVID-19 pandemic and two pre-pandemic. Descriptive statistics were used to compare epidemiology and ventilation characteristics. The primary endpoint were key ventilation parameters; other outcomes included mortality and ventilator-free days and alive (VFD-60) at day 60.

RESULTS

This analysis included 6702 COVID-ARDS patients and 1415 CLASSIC-ARDS patients. COVID-ARDS patients received lower median VT (6.6 [6.0 to 7.4] vs 7.3 [6.4 to 8.5] ml/kg PBW; p < 0.001) and higher median PEEP (12.0 [10.0 to 14.0] vs 8.0 [6.0 to 10.0] cm H2O; p < 0.001), at lower median ΔP (13.0 [10.0 to 15.0] vs 16.0 [IQR 12.0 to 20.0] cm H2O; p < 0.001) and higher median Crs (33.5 [26.6 to 42.1] vs 28.1 [21.6 to 38.4] mL/cm H2O; p < 0.001). Following multivariable adjustment, higher ΔP had an independent association with higher 60-day mortality and less VFD-60 in both groups. Higher PEEP had an association with less VFD-60, but only in COVID-ARDS patients.

CONCLUSIONS

Our findings show important differences in key ventilation parameters and associations thereof with outcomes between COVID-ARDS and CLASSIC-ARDS.

TRIAL REGISTRATION

Clinicaltrials.gov (identifier NCT05650957), December 14, 2022.

A systematic review of machine learning models for management, prediction and classification of ARDS

AIM

Acute respiratory distress syndrome or ARDS is an acute, severe form of respiratory failure characterised by poor oxygenation and bilateral pulmonary infiltrates. Advancements in signal processing and machine learning have led to promising solutions for classification, event detection and predictive models in the management of ARDS.

METHOD

In this review, we provide systematic description of different studies in the application of Machine Learning (ML) and artificial intelligence for management, prediction, and classification of ARDS. We searched the following databases: Google Scholar, PubMed, and EBSCO from 2009 to 2023. A total of 243 studies was screened, in which, 52 studies were included for review and analysis. We integrated knowledge of previous work providing the state of art and overview of explainable decision models in machine learning and have identified areas for future research.

RESULTS

Gradient boosting is the most common and successful method utilised in 12 (23.1%) of the studies. Due to limitation of data size available, neural network and its variation is used by only 8 (15.4%) studies. Whilst all studies used cross validating technique or separated database for validation, only 1 study validated the model with clinician input. Explainability methods were presented in 15 (28.8%) of studies with the most common method is feature importance which used 14 times.

CONCLUSION

For databases of 5000 or fewer samples, extreme gradient boosting has the highest probability of success. A large, multi-region, multi centre database is required to reduce bias and take advantage of neural network method. A framework for validating with and explaining ML model to clinicians involved in the management of ARDS would be very helpful for development and deployment of the ML model.

Prospective, randomized, controlled trial assessing the effects of a driving pressure-limiting strategy for patients with acute respiratory distress syndrome due to community-acquired pneumonia (STAMINA trial): protocol and statistical analysis plan

BACKGROUND

Driving pressure has been suggested to be the main driver of ventilator-induced lung injury and mortality in observational studies of acute respiratory distress syndrome. Whether a driving pressure-limiting strategy can improve clinical outcomes is unclear.

OBJECTIVE

To describe the protocol and statistical analysis plan that will be used to test whether a driving pressure-limiting strategy including positive end-expiratory pressure titration according to the best respiratory compliance and reduction in tidal volume is superior to a standard strategy involving the use of the ARDSNet low-positive end-expiratory pressure table in terms of increasing the number of ventilator-free days in patients with acute respiratory distress syndrome due to community-acquired pneumonia.

METHODS

The ventilator STrAtegy for coMmunIty acquired pNeumoniA (STAMINA) study is a randomized, multicenter, open-label trial that compares a driving pressure-limiting strategy to the ARDSnet low-positive end-expiratory pressure table in patients with moderate-to-severe acute respiratory distress syndrome due to community-acquired pneumonia admitted to intensive care units. We expect to recruit 500 patients from 20 Brazilian and 2 Colombian intensive care units. They will be randomized to a driving pressure-limiting strategy group or to a standard strategy using the ARDSNet low-positive end-expiratory pressure table. In the driving pressure-limiting strategy group, positive end-expiratory pressure will be titrated according to the best respiratory system compliance.

OUTCOMES

The primary outcome is the number of ventilator-free days within 28 days. The secondary outcomes are in-hospital and intensive care unit mortality and the need for rescue therapies such as extracorporeal life support, recruitment maneuvers and inhaled nitric oxide.

CONCLUSION

STAMINA is designed to provide evidence on whether a driving pressure-limiting strategy is superior to the ARDSNet low-positive end-expiratory pressure table strategy for increasing the number of ventilator-free days within 28 days in patients with moderate-to-severe acute respiratory distress syndrome. Here, we describe the rationale, design and status of the trial.

The effect of immunosuppressive therapies on the endothelial host response in critically ill COVID-19 patients

While several effective therapies for critically ill patients with COVID-19 have been identified in large, well-conducted trials, the mechanisms underlying these therapies have not been investigated in depth. Our aim is to investigate the association between various immunosuppressive therapies (corticosteroids, tocilizumab and anakinra) and the change in endothelial host response over time in critically ill COVID-19 patients. We conducted a pre-specified multicenter post-hoc analysis in a Dutch cohort of COVID-19 patients admitted to the ICU between March 2020 and September 2021 due to hypoxemic respiratory failure. A panel of 18 immune response biomarkers in the complement, coagulation and endothelial function domains were measured using ELISA or Luminex. Biomarkers were measured on day 0-1, day 2-4 and day 6-8 after start of COVID-19 treatment. Patients were categorized into four treatment groups: no immunomodulatory treatment, corticosteroids, anakinra plus corticosteroids, or tocilizumab plus corticosteroids. The association between treatment group and the change in concentrations of biomarkers was estimated with linear mixed-effects models, using no immunomodulatory treatment as reference group. 109 patients with a median age of 62 years [IQR 54-70] of whom 72% (n = 78) was male, were included in this analysis. Both anakinra plus corticosteroids (n = 22) and tocilizumab plus corticosteroids (n = 38) were associated with an increase in angiopoietin-1 compared to no immune modulator (n = 23) (beta of 0.033 [0.002-0.064] and 0.041 [0.013-0.070] per day, respectively). These treatments, as well as corticosteroids alone (n = 26), were further associated with a decrease in the ratio of angiopoietin-2/angiopoietin-1 (beta of 0.071 [0.034-0.107], 0.060 [0.030-0.091] and 0.043 [0.001-0.085] per day, respectively). Anakinra plus corticosteroids and tocilizumab plus corticosteroids were associated with a decrease in concentrations of complement complex 5b-9 compared to no immunomodulatory treatment (0.038 [0.006-0.071] and 0.023 [0.000-0.047], respectively). Currently established treatments for critically ill COVID-19 patients are associated with a change in biomarkers of the angiopoietin and complement pathways, possibly indicating a role for stability of the endothelium. These results increase the understanding of the mechanisms of interventions and are possibly useful for stratification of patients with other inflammatory conditions which may potentially benefit from these treatments.

Machine learning predicts lung recruitment in acute respiratory distress syndrome using single lung CT scan

BACKGROUND

To develop and validate classifier models that could be used to identify patients with a high percentage of potentially recruitable lung from readily available clinical data and from single CT scan quantitative analysis at intensive care unit admission. 221 retrospectively enrolled mechanically ventilated, sedated and paralyzed patients with acute respiratory distress syndrome (ARDS) underwent a PEEP trial at 5 and 15 cmH2O of PEEP and two lung CT scans performed at 5 and 45 cmH2O of airway pressure. Lung recruitability was defined at first as percent change in not aerated tissue between 5 and 45 cmH2O (radiologically defined; recruiters: Δ45-5non-aerated tissue  > 15%) and secondly as change in PaO2 between 5 and 15 cmH2O (gas exchange-defined; recruiters: Δ15-5PaO2  > 24 mmHg). Four machine learning (ML) algorithms were evaluated as classifiers of radiologically defined and gas exchange-defined lung recruiters using different models including different variables, separately or combined, of lung mechanics, gas exchange and CT data.

RESULTS

ML algorithms based on CT scan data at 5 cmH2O classified radiologically defined lung recruiters with similar AUC as ML based on the combination of lung mechanics, gas exchange and CT data. ML algorithm based on CT scan data classified gas exchange-defined lung recruiters with the highest AUC.

CONCLUSIONS

ML based on a single CT data at 5 cmH2O represented an easy-to-apply tool to classify ARDS patients in recruiters and non-recruiters according to both radiologically defined and gas exchange-defined lung recruitment within the first 48 h from the start of mechanical ventilation.

Effect of intraoperative PEEP with recruitment maneuvers on the occurrence of postoperative pulmonary complications during general anesthesia--protocol for Bayesian analysis of three randomized clinical trials of intraoperative ventilation

Background: Using the frequentist approach, a recent meta-analysis of three randomized clinical trials in patients undergoing intraoperative ventilation during general anesthesia for major surgery failed to show the benefit of ventilation that uses high positive end-expiratory pressure with recruitment maneuvers when compared to ventilation that uses low positive end-expiratory pressure without recruitment maneuvers. Methods: We designed a protocol for a Bayesian analysis using the pooled dataset. The multilevel Bayesian logistic model will use the individual patient data. Prior distributions will be prespecified to represent a varying level of skepticism for the effect estimate. The primary endpoint will be a composite of postoperative pulmonary complications (PPC) within the first seven postoperative days, which reflects the primary endpoint of the original studies. We preset a range of practical equivalence to assess the futility of the intervention with an interval of odds ratio (OR) between 0.9 and 1.1 and assess how much of the 95% of highest density interval (HDI) falls between the region of practical equivalence. Ethics and dissemination: The used data derive from approved studies that were published in recent years. The findings of this current analysis will be reported in a new manuscript, drafted by the writing committee on behalf of the three research groups. All investigators listed in the original trials will serve as collaborative authors.

Heterogeneous treatment effects of dexamethasone 12 mg versus 6 mg in patients with COVID-19 and severe hypoxaemia-Post hoc exploratory analyses of the COVID STEROID 2 trial

BACKGROUND

Corticosteroids improve outcomes in patients with severe COVID-19. In the COVID STEROID 2 randomised clinical trial, we found high probabilities of benefit with dexamethasone 12 versus 6 mg daily. While no statistically significant heterogeneity in treatment effects (HTE) was found in the conventional, dichotomous subgroup analyses, these analyses have limitations, and HTE could still exist.

METHODS

We assessed whether HTE was present for days alive without life support and mortality at Day 90 in the trial according to baseline age, weight, number of comorbidities, category of respiratory failure (type of respiratory support system and oxygen requirements) and predicted risk of mortality using an internal prediction model. We used flexible models for continuous variables and logistic regressions for categorical variables without dichotomisation of the baseline variables of interest. HTE was assessed both visually and with p and S values from likelihood ratio tests.

RESULTS

There was no strong evidence for substantial HTE on either outcome according to any of the baseline variables assessed with all p values >.37 (and all S values <1.43) in the planned analyses and no convincingly strong visual indications of HTE.

CONCLUSIONS

We found no strong evidence for HTE with 12 versus 6 mg dexamethasone daily on days alive without life support or mortality at Day 90 in patients with COVID-19 and severe hypoxaemia, although these results cannot rule out HTE either.

A structured narrative review of clinical and experimental studies of the use of different positive end-expiratory pressure levels during thoracic surgery

OBJECTIVES

This study aimed to present a review on the general effects of different positive end-expiratory pressure (PEEP) levels during thoracic surgery by qualitatively categorizing the effects into detrimental, beneficial, and inconclusive.

DATA SOURCE

Literature search of Pubmed, CNKI, and Wanfang was made to find relative articles about PEEP levels during thoracic surgery. We used the following keywords as one-lung ventilation, PEEP, and thoracic surgery.

RESULTS

We divide the non-individualized PEEP value into five grades, that is, less than 5, 5, 5-10, 10, and more than 10 cmH₂ O, among which 5 cmH₂ O is the most commonly used in clinic at present to maintain alveolar dilatation and reduce the shunt fraction and the occurrence of atelectasis, whereas individualized PEEP, adjusted by test titration or imaging method to adapt to patients' personal characteristics, can effectively ameliorate intraoperative oxygenation and obtain optimal pulmonary compliance and better indexes relating to respiratory mechanics.

CONCLUSIONS

Available data suggest that PEEP might play an important role in one-lung ventilation, the understanding of which will help in exploring a simple and economical method to set the appropriate PEEP level.

Phenotypes of sickle cell intensive care admissions: an unsupervised machine learning approach in a single-center retrospective cohort

Sickle cell disease (SCD) is associated with multiple known complications and increased mortality. This study aims to further understand the profile of intensive care unit (ICU) admissions of SCD patients. In this single-center retrospective cohort (approval number 0926-11), we evaluated SCD-related ICU admissions at our hospital in São Paulo, Brazil. Admissions were clustered using clinical data and organ dysfunction at ICU admission. A hierarchical clustering method was used to distinguish phenotypes. From 140 admissions obtained, 125 were included. The mean age was 30 years, 48% were male, and SS genotype was predominant (71.2%). Non-surgical causes of admissions accounted for 85.6% (n = 107). The mean Sequential Organ Failure Assessment score (SOFA) was 4 (IQR 2-7). Vasopressors were required by 12% and mechanical ventilation by 17.6%. After analysis of the average silhouette width, the optimal number of clusters was 3: cluster 1 (n = 69), cluster 2 (n = 25), cluster 3 (n = 31). Cluster 1 had a mean age of 29 years, 87% of SS genotype, and mean SOFA of 4. Cluster 2 had a mean age of 37 years, 80% of SS genotype, and mean SOFA of 8. Cluster 3 had a mean age of 26 years, 29% of SS genotype, and mean SOFA of 3. The need for mechanical ventilation was 11.6%, 44%, and 9.7%, respectively. Mortality was significantly higher in cluster 2 (44%, p = 0.012). This cohort of critical SCD admissions suggested the presence of three different profiles. This can be informative in the ICU setting to identify SCD patients at higher risk of worse outcomes.

Clinical Phenotyping of Out-of-Hospital Cardiac Arrest Patients With Shockable Rhythm - Machine Learning-Based Unsupervised Cluster Analysis

BACKGROUND

The hypothesis of this study is that latent class analysis could identify the subphenotypes of out-of-hospital cardiac arrest (OHCA) patients associated with the outcomes and allow us to explore heterogeneity in the effects of extracorporeal cardiopulmonary resuscitation (ECPR).

METHODS AND RESULTS

This study was a retrospective analysis of a multicenter prospective observational study (CRITICAL study) of OHCA patients. It included adult OHCA patients with initial shockable rhythm. Patients from 2012 to 2016 (development dataset) were included in the latent class analysis, and those from 2017 (validation dataset) were included for evaluation. The association between subphenotypes and outcomes was investigated. Further, the heterogeneity of the association between ECPR implementation and outcomes was explored. In the study results, a total of 920 patients were included for latent class analysis. Three subphenotypes (Groups 1, 2, and 3) were identified, mainly characterized by the distribution of partial pressure of O2(PO2), partial pressure of CO2(PCO2) value of blood gas assessment, cardiac rhythm on hospital arrival, and estimated glomerular filtration rate. The 30-day survival outcomes were varied across the groups: 15.7% in Group 1; 30.7% in Group 2; and 85.9% in Group 3. Further, the association between ECPR and 30-day survival outcomes by subphenotype groups in the development dataset was as varied. These results were validated using the validation dataset.

CONCLUSIONS

The latent class analysis identified 3 subphenotypes with different survival outcomes and potential heterogeneity in the effects of ECPR.

Decision support system to evaluate ventilation in the acute respiratory distress syndrome (DeVENT study)-trial protocol

Background

The acute respiratory distress syndrome (ARDS) occurs in response to a variety of insults, and mechanical ventilation is life-saving in this setting, but ventilator-induced lung injury can also contribute to the morbidity and mortality in the condition. The Beacon Caresystem is a model-based bedside decision support system using mathematical models tuned to the individual patient’s physiology to advise on appropriate ventilator settings. Personalised approaches using individual patient description may be particularly advantageous in complex patients, including those who are difficult to mechanically ventilate and wean, in particular ARDS.

Methods

We will conduct a multi-centre international randomised, controlled, allocation concealed, open, pragmatic clinical trial to compare mechanical ventilation in ARDS patients following application of the Beacon Caresystem to that of standard routine care to investigate whether use of the system results in a reduction in driving pressure across all severities and phases of ARDS.

Discussion

Despite 20 years of clinical trial data showing significant improvements in ARDS mortality through mitigation of ventilator-induced lung injury, there remains a gap in its personalised application at the bedside. Importantly, the protective effects of higher positive end-expiratory pressure (PEEP) were noted only when there were associated decreases in driving pressure. Hence, the pressures set on the ventilator should be determined by the diseased lungs’ pressure-volume relationship which is often unknown or difficult to determine. Knowledge of extent of recruitable lung could improve the ventilator driving pressure. Hence, personalised management demands the application of mechanical ventilation according to the physiological state of the diseased lung at that time. Hence, there is significant rationale for the development of point-of-care clinical decision support systems which help personalise ventilatory strategy according to the current physiology. Furthermore, the potential for the application of the Beacon Caresystem to facilitate local and remote management of large numbers of ventilated patients (as seen during this COVID-19 pandemic) could change the outcome of mechanically ventilated patients during the course of this and future pandemics.

Trial registration

ClinicalTrials.gov identifier NCT04115709. Registered on 4 October 2019, version 4.0

Supplementary Information

The online version contains supplementary material available at 10.1186/s13063-021-05967-2.

Lower versus higher oxygenation targets in critically ill patients with severe hypoxaemia: secondary Bayesian analysis to explore heterogeneous treatment effects in the Handling Oxygenation Targets in the Intensive Care Unit (HOT-ICU) trial

BACKGROUND

In the Handling Oxygenation Targets in the Intensive Care Unit (HOT-ICU) trial, a lower (8 kPa) vs a higher (12 kPa) PaO2 target did not affect mortality amongst critically ill adult patients. We used Bayesian statistics to evaluate any heterogeneity in the effect of oxygenation targets on mortality between different patient groups within the HOT-ICU trial.

METHODS

We analysed 90-day all-cause mortality using adjusted Bayesian logistic regression models, and assessed heterogeneous treatment effects according to four selected baseline variables using both hierarchical models of subgroups and models with interactions on the continuous scales. Results are presented as mortality probability (%) and relative risk (RR) with 95% credibility intervals (CrI).

RESULTS

All 2888 patients in the intention-to-treat cohort of the HOT-ICU trial were included. The adjusted 90-day mortality rates were 43.0% (CrI: 38.3-47.8%) and 42.3% (CrI: 37.7-47.1%) in the lower and higher oxygenation groups, respectively (RR 1.02 [CrI: 0.93-1.11]), with 36.5% probability of an RR <1.00. Analyses of heterogeneous treatment effects suggested a dose-response relationship between baseline norepinephrine dose and increased mortality with the lower oxygenation target, with 95% probability of increased mortality associated with the lower oxygenation target as norepinephrine doses increased.

CONCLUSIONS

A lower oxygenation target was unlikely to affect overall mortality amongst critically ill adult patients with acute hypoxaemic respiratory failure. However, our results suggest an increasing mortality risk for patients with a lower oxygen target as the baseline norepinephrine dose increases. These findings warrant additional investigation.

CLINICAL TRIAL REGISTRATION

NCT03174002.

Ventilator-Associated Lung Injury

Ventilatory support, while life saving, can also cause or aggravate lung injury through several mechanisms which are encompassed within ventilator-associated lung injury (VALI). The important realizationin the acute respiratory distress syndrome that the “baby” lung resided in non-dependent areas led to the conceptualization of “lung rest” to reduce stress and strain to exposed alveolar units. We discuss concepts and mechanisms within VALI that ultimately induce maladaptive lung responses, as well as, current and future management strategies to detect and mitigate VALI at the bedside.

Clustering out‐of‐hospital cardiac arrest patients with non‐shockable rhythm by machine learning latent class analysis

Aim

We aimed to identify subphenotypes among patients with out‐of‐hospital cardiac arrest (OHCA) with initial non‐shockable rhythm by applying machine learning latent class analysis and examining the associations between subphenotypes and neurological outcomes.

Methods

This study was a retrospective analysis within a multi‐institutional prospective observational cohort study of OHCA patients in Osaka, Japan (the CRITICAL study). The data of adult OHCA patients with medical causes and initial non‐shockable rhythm presenting with OHCA between 2012 and 2016 were included in machine learning latent class analysis models, which identified subphenotypes, and patients who presented in 2017 were included in a dataset validating the subphenotypes. We investigated associations between subphenotypes and 30‐day neurological outcomes.

Results

Among the 12,594 patients in the CRITICAL study database, 4,849 were included in the dataset used to classify subphenotypes (median age: 75 years, 60.2% male), and 1,465 were included in the validation dataset (median age: 76 years, 59.0% male). Latent class analysis identified four subphenotypes. Odds ratios and 95% confidence intervals for a favorable 30‐day neurological outcome among patients with these subphenotypes, using group 4 for comparison, were as follows; group 1, 0.01 (0.001–0.046); group 2, 0.097 (0.051–0.171); and group 3, 0.175 (0.073–0.358). Associations between subphenotypes and 30‐day neurological outcomes were validated using the validation dataset.

Conclusion

We identified four subphenotypes of OHCA patients with initial non‐shockable rhythm. These patient subgroups presented with different characteristics associated with 30‐day survival and neurological outcomes.

Is artificial intelligence ready to solve mechanical ventilation? Computer says blow

Artificial intelligence (AI) has the potential to identify treatable phenotypes, optimise ventilation strategies, and provide clinical decision support for patients who require mechanical ventilation. Gallifant and colleagues performed a systematic review to identify studies using AI to solve a diverse range of clinical problems in the ventilated patient. They identify 95 studies, the majority of which were reported in the last 5 yr. Their findings indicate that the majority of studies have significant methodological bias and are a long way from deployment.

Randomised clinical trials in critical care: past, present and future

Randomised clinical trials (RCTs) are the gold standard for providing unbiased evidence of intervention effects. Here, we provide an overview of the history of RCTs and discuss the major challenges and limitations of current critical care RCTs, including overly optimistic effect sizes; unnuanced conclusions based on dichotomization of results; limited focus on patient-centred outcomes other than mortality; lack of flexibility and ability to adapt, increasing the risk of inconclusive results and limiting knowledge gains before trial completion; and inefficiency due to lack of re-use of trial infrastructure. We discuss recent developments in critical care RCTs and novel methods that may provide solutions to some of these challenges, including a research programme approach (consecutive, complementary studies of multiple types rather than individual, independent studies), and novel design and analysis methods. These include standardization of trial protocols; alternative outcome choices and use of core outcome sets; increased acceptance of uncertainty, probabilistic interpretations and use of Bayesian statistics; novel approaches to assessing heterogeneity of treatment effects; adaptation and platform trials; and increased integration between clinical trials and clinical practice. We outline the advantages and discuss the potential methodological and practical disadvantages with these approaches. With this review, we aim to inform clinicians and researchers about conventional and novel RCTs, including the rationale for choosing one or the other methodological approach based on a thorough discussion of pros and cons. Importantly, the most central feature remains the randomisation, which provides unparalleled restriction of confounding compared to non-randomised designs by reducing confounding to chance.

Precision Medicine and Heterogeneity of Treatment Effect in Therapies for Acute Respiratory Distress Syndrome

Acute respiratory distress syndrome (ARDS) is a clinically heterogenous syndrome, rather than a distinct disease. This heterogeneity at least partially explains the difficulty in studying treatments for these patients and contributes to the numerous trials of therapies for the syndrome that have not shown benefit. Recent studies have identified different subphenotypes within the heterogenous patient population. These different subphenotypes likely have variable clinical responses to specific therapies, a concept known as heterogeneity of treatment effect (HTE). Recognizing different subphenotypes and HTE has important implications for the clinical management of patients with ARDS. In this review, we will present studies that have identified different subphenotypes and discuss how they can modify the effects of therapies evaluated in trials that are commonly considered to have demonstrated no overall benefit in patients with ARDS.

Coagulation phenotypes in sepsis and effects of recombinant human thrombomodulin: an analysis of three multicentre observational studies

Background

A recent randomised trial showed that recombinant thrombomodulin did not benefit patients who had sepsis with coagulopathy and organ dysfunction. Several recent studies suggested presence of clinical phenotypes in patients with sepsis and heterogenous treatment effects across different sepsis phenotypes. We examined the latent phenotypes of sepsis with coagulopathy and the associations between thrombomodulin treatment and the 28-day and in-hospital mortality for each phenotype.

Methods

This was a secondary analysis of multicentre registries containing data on patients (aged ≥ 16 years) who were admitted to intensive care units for severe sepsis or septic shock in Japan. Three multicentre registries were divided into derivation (two registries) and validation (one registry) cohorts. Phenotypes were derived using k-means with coagulation markers, platelet counts, prothrombin time/international normalised ratios, fibrinogen, fibrinogen/fibrin-degradation-products (FDP), D-dimer, and antithrombin activities. Associations between thrombomodulin treatment and survival outcomes (28-day and in-hospital mortality) were assessed in the derived clusters using a generalised estimating equation.

Results

Four sepsis phenotypes were derived from 3694 patients in the derivation cohort. Cluster dA (n = 323) had severe coagulopathy with high FDP and D-dimer levels, severe organ dysfunction, and high mortality. Cluster dB had severe disease with moderate coagulopathy. Clusters dC and dD had moderate and mild disease with and without coagulopathy, respectively. Thrombomodulin was associated with a lower 28-day (adjusted risk difference [RD]: − 17.8% [95% CI − 28.7 to − 6.9%]) and in-hospital (adjusted RD: − 17.7% [95% CI − 27.6 to − 7.8%]) mortality only in cluster dA. Sepsis phenotypes were similar in the validation cohort, and thrombomodulin treatment was also associated with lower 28-day (RD: − 24.9% [95% CI − 49.1 to − 0.7%]) and in-hospital mortality (RD: − 30.9% [95% CI − 55.3 to − 6.6%]).

Conclusions

We identified four coagulation marker-based sepsis phenotypes. The treatment effects of thrombomodulin varied across sepsis phenotypes. This finding will facilitate future trials of thrombomodulin, in which a sepsis phenotype with high FDP and D-dimer can be targeted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-021-03541-5.

Mechanical ventilation in septic shock

PURPOSE OF REVIEW

The aim of this study was to review the most recent literature on mechanical ventilation strategies in patients with septic shock.

RECENT FINDINGS

Indirect clinical trial evidence has refined the use of neuromuscular blocking agents, positive end-expiratory pressure (PEEP) and recruitment manoeuvres in septic shock patients with acute respiratory distress syndrome. Weaning strategies and devices have also been recently evaluated. The role of lung protective ventilation in patients with healthy lungs, while recognized, still needs to be further refined. The possible detrimental effects of spontaneous breathing in patients who develop acute respiratory distress syndrome is increasingly recognized, but clinical trial evidence is still lacking to confirm this hypothesis. A new concept of lung and diaphragm protective is emerging in the critical care literature, but its application will need a complex intervention implementation approach to allow adequate scrutiny of this concept and uptake by clinicians.

SUMMARY

Many advances in the management of the mechanically ventilated patient with sepsis and septic shock have occurred in recent years, but clinical trial evidence is still necessary to translate new hypotheses to the bedside and find the right balance between benefits and risks of these new strategies.

Using Bayesian Methods to Augment the Interpretation of Critical Care Trials. An Overview of Theory and Example Reanalysis of the Alveolar Recruitment for Acute Respiratory Distress Syndrome Trial

Most randomized trials are designed and analyzed using frequentist statistical approaches such as null hypothesis testing and P values. Conceptually, P values are cumbersome to understand, as they provide evidence of data incompatibility with a null hypothesis (e.g., no clinical benefit) and not direct evidence of the alternative hypothesis (e.g., clinical benefit). This counterintuitive framework may contribute to the misinterpretation that the absence of evidence is equal to evidence of absence and may cause the discounting of potentially informative data. Bayesian methods provide an alternative, probabilistic interpretation of data. The reanalysis of completed trials using Bayesian methods is becoming increasingly common, particularly for trials with effect estimates that appear clinically significant despite P values above the traditional threshold of 0.05. Statistical inference using Bayesian methods produces a distribution of effect sizes that would be compatible with observed trial data, interpreted in the context of prior assumptions about an intervention (called "priors"). These priors are chosen by investigators to reflect existing beliefs and past empirical evidence regarding the effect of an intervention. By calculating the likelihood of clinical benefit, a Bayesian reanalysis can augment the interpretation of a trial. However, if priors are not defined a priori, there is a legitimate concern that priors could be constructed in a manner that produces biased results. Therefore, some standardization of priors for Bayesian reanalysis of clinical trials may be desirable for the critical care community. In this Critical Care Perspective, we discuss both frequentist and Bayesian approaches to clinical trial analysis, introduce a framework that researchers can use to select priors for a Bayesian reanalysis, and demonstrate how to apply our proposal by conducting a novel Bayesian trial reanalysis.

PEEP in thoracic anesthesia: pros and cons

Protective ventilation includes a strategy with low tidal volume, Plateau pressure, driving pressure, positive end-expiratory pressure (PEEP), and recruitment maneuvers on the ventilated lung. The rationale for the application of PEEP during one-lung ventilation (OLV) is that PEEP may contribute to minimize atelectrauma, preventing airway closure and alveolar collapse and improving the ventilation/perfusion to the ventilated lung. However, in case of high partial pressure of oxygen the application of PEEP may cause increased pulmonary vascular resistance, thus diverting blood flow to the non-ventilated lung, and worsening ventilation/perfusion. Further, PEEP may be associated with higher risk of hemodynamic impairment, increased need for fluids and vasoactive drugs. Positive effects on outcome have been reported by titrating PEEP according to driving pressure, targeted to obtain the optimum respiratory as well as pulmonary system compliance. This may vary according to the method employed for titration and should be performed individually for each patient. In summary, the potential for harm combined with the lack of evidence for improved outcome suggest that PEEP must be judiciously used during OLV even when titrated to a safe target, and only as much as necessary to maintain an appropriate gas exchange under low protective tidal volumes and driving pressures.

Alveolar recruitment manoeuvre results in improved pulmonary function in obese patients undergoing bariatric surgery: a randomised trial

Perioperative ventilation is an important challenge of anaesthesia, especially in obese patients: body mass index is correlated with reduction of the pulmonary volume and they develop significantly more perioperative atelectasis and pulmonary complications. The alveolar recruitment manoeuvre is the most effective technique to reverse atelectasis. However, the clinical benefit on lung function in the perioperative period is not clear. The aim of the present study is to assess the perioperative clinical results of systematic alveolar recruitment manoeuvre associated with protective ventilation in patients undergoing laparoscopic bariatric surgery. It was a single-centre, randomised, double blind, superiority trial: control group with standard protective ventilation and recruitment group with protective ventilation and systematic recruitment manoeuvre. The primary outcome was a composite clinical criterion of pulmonary dysfunction including oxygen saturation, oxygen needs and dyspnoea in recovery room and at day 1. Secondary outcomes were recruitment manoeuvre tolerance, pulmonary and non-pulmonary complications, length of hospital stay and proportion of Intensive Care Unit admission. Two hundred and thirty patients were included: 115 in the recruitment manoeuvre group and 115 in the control group, 2 patients were excluded from the analysis in the control group. Patients in the recruitment manoeuvre group had significantly lower rate of pulmonary dysfunction in the recovery room (73% versus 84% (p =  0.043) and 77% versus 88% at postoperative day 1 (p =  0.043)). No significant differences were found for secondary outcomes. No patient was excluded from the recruitment manoeuvre group for intolerance to the manoeuvre. Recruitment manoeuvre is safe and effective in reducing early pulmonary dysfunction in obese patients undergoing bariatric surgery.

Protective mechanical ventilation in suspected influenza infection

INTRODUCTION:

Patients with acute respiratory failure due to influenza require ventilatory support. However, mechanical ventilation itself can exacerbate lung damage and increase mortality.

METHODS:

The aim of this study was to describe a feasible and protective ventilation protocol, with limitation of the tidal volume to ≤6 mL/kg of the predicted weight and a driving pressure ≤15 cmH₂O after application of the alveolar recruitment maneuver and PEEP titration.

RESULTS:

Initial improvement in oxygenation and respiratory mechanics were observed in the four cases submitted to the proposed protocol.

CONCLUSIONS:

Our results indicate that the mechanical ventilation strategy applied could be optimized.

Heterogeneity of Effect of Net Ultrafiltration Rate among Critically Ill Adults Receiving Continuous Renal Replacement Therapy

INTRODUCTION

In continuous renal replacement therapy (CRRT)-treated patients, a net ultrafiltration (NUF) rate >1.75 mL/kg/h has been associated with increased mortality. However, there may be heterogeneity of effect of NUF rate on mortality, according to patient characteristics.

METHODS

To investigate the presence and impact of heterogeneity of effect, we performed a secondary analysis of the "Randomized Evaluation of Normal versus Augmented Level of Renal Replacement Therapy" (RENAL) trial. Exposure was NUF rate (weight-adjusted fluid volume removed per hour) stratified into tertiles (<1.01 mL/kg/h; 1.01-1.75 mL/kg/h; or >1.75 mL/kg/h). Primary outcome was 90-day mortality. Patients were clustered according to baseline characteristics. Heterogeneity of effect was assessed according to clusters and baseline edema and related to the additional impact of baseline cardiovascular Sequential Organ Failure Assessment (SOFA) score. We excluded patients with missing values for baseline weight and/or treatment duration.

RESULTS

We identified 2 clusters. The largest (cluster 1; n = 941) included more severely ill patients, with more sepsis, more edema, and more vasopressor therapy (all p < 0.001). Compared to the middle tertile, the probability of harm was greater with the high tertile of NUF rate in patients in cluster 1 and in patients with baseline edema (probability of harm, cluster 1: 99.9%; edema: 99.1%). Moreover, higher baseline cardiovascular SOFA score also increased mortality risk with both high and low compared to middle NUF rates in cluster 1 patients and in patients with edema.

CONCLUSIONS

In CRRT patients, both high and low NUF rates may be harmful, especially in those with edema, sepsis, and greater illness severity. Cardiovascular SOFA scores modulate this association. Additional studies are needed to test these hypotheses, and targeted trials of NUF rates based on risk stratification appear justified.

TRIAL REGISTRATION

ClinicalTrials.gov identifier: NCT00221013.

Data-driven ICU management: Using Big Data and algorithms to improve outcomes

The digitalization of the Intensive Care Unit (ICU) led to an increasing amount of clinical data being collected at the bedside. The term "Big Data" can be used to refer to the analysis of these datasets that collect enormous amount of data of different origin and format. Complexity and variety define the value of Big Data. In fact, the retrospective analysis of these datasets allows to generate new knowledge, with consequent potential improvements in the clinical practice. Despite the promising start of Big Data analysis in medical research, which has seen a rising number of peer-reviewed articles, very limited applications have been used in ICU clinical practice. A close future effort should be done to validate the knowledge extracted from clinical Big Data and implement it in the clinic. In this article, we provide an introduction to Big Data in the ICU, from data collection and data analysis, to the main successful examples of prognostic, predictive and classification models based on ICU data. In addition, we focus on the main challenges that these models face to reach the bedside and effectively improve ICU care.

Machine Learning Predicts Prolonged Acute Hypoxemic Respiratory Failure in Pediatric Severe Influenza

Supplemental Digital Content is available in the text.

Influenza virus is a major cause of acute hypoxemic respiratory failure. Early identification of patients who will suffer severe complications can help stratify patients for clinical trials and plan for resource use in case of pandemic.

Machine learning to assist clinical decision-making during the COVID-19 pandemic

BACKGROUND

The number of cases from the coronavirus disease 2019 (COVID-19) global pandemic has overwhelmed existing medical facilities and forced clinicians, patients, and families to make pivotal decisions with limited time and information.

MAIN BODY

While machine learning (ML) methods have been previously used to augment clinical decisions, there is now a demand for "Emergency ML." Throughout the patient care pathway, there are opportunities for ML-supported decisions based on collected vitals, laboratory results, medication orders, and comorbidities. With rapidly growing datasets, there also remain important considerations when developing and validating ML models.

CONCLUSION

This perspective highlights the utility of evidence-based prediction tools in a number of clinical settings, and how similar models can be deployed during the COVID-19 pandemic to guide hospital frontlines and healthcare administrators to make informed decisions about patient care and managing hospital volume.

A 23-year-old man with left lung atelectasis treated with a targeted segmental recruitment maneuver: a case report

Background

Lung atelectasis are nonventilated parts of lung tissue and occur as a result of the collapse of the pulmonary parenchyma (alveoli). Various therapeutic procedures for inflating the collapsed pulmonary parenchyma, such as bronchial aspiration and/or standard recruitment maneuvers, are not always successful.

Case presentation

We report a case of a 23-year-old Croatian man with a parapharyngeal abscess on the left side of the neck with spreading of infection in the mediastinum and left side of the thorax and consequent major atelectasis of the left lung. The patient was mechanically ventilated. We decided to apply a new method in which a pulmonary artery catheter was placed (guided by bronchoscope) on the entrance to the lower left bronchus. The pulmonary artery catheter balloon was inflated to achieve bronchial closure. Using another respirator, we ventilated the affected lobe separately with continuously high pressure of 30 cmH₂O. After 30 minutes, we removed the pulmonary artery catheter from the lower left bronchus and placed it in the upper left bronchus and repeated the procedure. Our method allowed a significantly longer duration (30 minutes) of continuously high pressure of 30 cmH₂O separately to only one of the total of five lobes of the lungs while the other four lobes were simultaneously ventilated continuously with protective ventilation mode.

Conclusion

Use of a pulmonary artery catheter and two respirators in our patient’s case proved to be a successful method for recruiting the atelectatic lung while maintaining protective ventilation of the lung segments without atelectasis.

Heterogeneity of treatment effect of stress ulcer prophylaxis in ICU patients: A secondary analysis protocol

BACKGROUND

In the Stress Ulcer Prophylaxis in the Intensive Care Unit (SUP-ICU) trial, 3291 adult ICU patients at risk for gastrointestinal (GI) bleeding were randomly allocated to intravenous pantoprazole 40 mg or placebo once daily in the ICU. No difference was observed between the groups in the primary outcome 90-day mortality or the secondary outcomes, except for clinically important gastrointestinal bleeding. However, heterogeneity of treatment effect (HTE) not detected by conventional subgroup analyses could be present.

METHODS

This is a protocol and statistical analysis plan for a secondary, post hoc, exploratory analysis of the SUP-ICU trial. We will explore HTE in one set of subgroups based on severity of illness (using the Simplified Acute Physiology Score [SAPS] II) and another set of subgroups based on the total number of risk factors for GI bleeding in each patient using Bayesian hierarchical models. We will summarise posterior probability distributions using medians and 95% credible intervals and present probabilities for different levels of benefit and harm of the intervention in each subgroup. Finally, we will assess if the treatment effect interacts with SAPS II and the number of risk factors separately on the continuous scale using marginal effects plots.

CONCLUSIONS

The outlined post hoc analysis will explore whether HTE was present in the SUP-ICU trial and may help answer some of the remaining questions regarding the balance between benefits and harms of pantoprazole in ICU patients at risk of GI bleeding. CLINICALTRIALS.

GOV REGISTRATION

NCT02467621.