Reinforcement learning for intensive care medicine: actionable clinical insights from novel approaches to reward shaping and off-policy model evaluation

BACKGROUND

Reinforcement learning (RL) holds great promise for intensive care medicine given the abundant availability of data and frequent sequential decision-making. But despite the emergence of promising algorithms, RL driven bedside clinical decision support is still far from reality. Major challenges include trust and safety. To help address these issues, we introduce cross off-policy evaluation and policy restriction and show how detailed policy analysis may increase clinical interpretability. As an example, we apply these in the setting of RL to optimise ventilator settings in intubated covid-19 patients.

METHODS

With data from the Dutch ICU Data Warehouse and using an exhaustive hyperparameter grid search, we identified an optimal set of Dueling Double-Deep Q Network RL models. The state space comprised ventilator, medication, and clinical data. The action space focused on positive end-expiratory pressure (peep) and fraction of inspired oxygen (FiO2) concentration. We used gas exchange indices as interim rewards, and mortality and state duration as final rewards. We designed a novel evaluation method called cross off-policy evaluation (OPE) to assess the efficacy of models under varying weightings between the interim and terminal reward components. In addition, we implemented policy restriction to prevent potentially hazardous model actions. We introduce delta-Q to compare physician versus policy action quality and in-depth policy inspection using visualisations.

RESULTS

We created trajectories for 1118 intensive care unit (ICU) admissions and trained 69,120 models using 8 model architectures with 128 hyperparameter combinations. For each model, policy restrictions were applied. In the first evaluation step, 17,182/138,240 policies had good performance, but cross-OPE revealed suboptimal performance for 44% of those by varying the reward function used for evaluation. Clinical policy inspection facilitated assessment of action decisions for individual patients, including identification of action space regions that may benefit most from optimisation.

CONCLUSION

Cross-OPE can serve as a robust evaluation framework for safe RL model implementation by identifying policies with good generalisability. Policy restriction helps prevent potentially unsafe model recommendations. Finally, the novel delta-Q metric can be used to operationalise RL models in clinical practice. Our findings offer a promising pathway towards application of RL in intensive care medicine and beyond.

Does Reinforcement Learning Improve Outcomes for Critically Ill Patients? A Systematic Review and Level-of-Readiness Assessment

OBJECTIVE

Reinforcement learning (RL) is a machine learning technique uniquely effective at sequential decision-making, which makes it potentially relevant to ICU treatment challenges. We set out to systematically review, assess level-of-readiness and meta-analyze the effect of RL on outcomes for critically ill patients.

DATA SOURCES

A systematic search was performed in PubMed, Embase.com, Clarivate Analytics/Web of Science Core Collection, Elsevier/SCOPUS and the Institute of Electrical and Electronics Engineers Xplore Digital Library from inception to March 25, 2022, with subsequent citation tracking.

DATA EXTRACTION

Journal articles that used an RL technique in an ICU population and reported on patient health-related outcomes were included for full analysis. Conference papers were included for level-of-readiness assessment only. Descriptive statistics, characteristics of the models, outcome compared with clinician's policy and level-of-readiness were collected. RL-health risk of bias and applicability assessment was performed.

DATA SYNTHESIS

A total of 1,033 articles were screened, of which 18 journal articles and 18 conference papers, were included. Thirty of those were prototyping or modeling articles and six were validation articles. All articles reported RL algorithms to outperform clinical decision-making by ICU professionals, but only in retrospective data. The modeling techniques for the state-space, action-space, reward function, RL model training, and evaluation varied widely. The risk of bias was high in all articles, mainly due to the evaluation procedure.

CONCLUSION

In this first systematic review on the application of RL in intensive care medicine we found no studies that demonstrated improved patient outcomes from RL-based technologies. All studies reported that RL-agent policies outperformed clinician policies, but such assessments were all based on retrospective off-policy evaluation.

Red Blood Cell Transfusion in the Intensive Care Unit

Importance

Red blood cell (RBC) transfusion is common among patients admitted to the intensive care unit (ICU). Despite multiple randomized clinical trials of hemoglobin (Hb) thresholds for transfusion, little is known about how these thresholds are incorporated into current practice.

Objective

To evaluate and describe ICU RBC transfusion practices worldwide.

Design, Setting, and Participants

International, prospective, cohort study that involved 3643 adult patients from 233 ICUs in 30 countries on 6 continents from March 2019 to October 2022 with data collection in prespecified weeks.

Exposure

ICU stay.

Main Outcomes and Measures

The primary outcome was the occurrence of RBC transfusion during ICU stay. Additional outcomes included the indication(s) for RBC transfusion (consisting of clinical reasons and physiological triggers), the stated Hb threshold and actual measured Hb values before and after an RBC transfusion, and the number of units transfused.

Results

Among 3908 potentially eligible patients, 3643 were included across 233 ICUs (median of 11 patients per ICU [IQR, 5-20]) in 30 countries on 6 continents. Among the participants, the mean (SD) age was 61 (16) years, 62% were male (2267/3643), and the median Sequential Organ Failure Assessment score was 3.2 (IQR, 1.5-6.0). A total of 894 patients (25%) received 1 or more RBC transfusions during their ICU stay, with a median total of 2 units per patient (IQR, 1-4). The proportion of patients who received a transfusion ranged from 0% to 100% across centers, from 0% to 80% across countries, and from 19% to 45% across continents. Among the patients who received a transfusion, a total of 1727 RBC transfusions were administered, wherein the most common clinical indications were low Hb value (n = 1412 [81.8%]; mean [SD] lowest Hb before transfusion, 7.4 [1.2] g/dL), active bleeding (n = 479; 27.7%), and hemodynamic instability (n = 406 [23.5%]). Among the events with a stated physiological trigger, the most frequently stated triggers were hypotension (n = 728 [42.2%]), tachycardia (n = 474 [27.4%]), and increased lactate levels (n = 308 [17.8%]). The median lowest Hb level on days with an RBC transfusion ranged from 5.2 g/dL to 13.1 g/dL across centers, from 5.3 g/dL to 9.1 g/dL across countries, and from 7.2 g/dL to 8.7 g/dL across continents. Approximately 84% of ICUs administered transfusions to patients at a median Hb level greater than 7 g/dL.

Conclusions and Relevance

RBC transfusion was common in patients admitted to ICUs worldwide between 2019 and 2022, with high variability across centers in transfusion practices.

Effect of lung ultrasound-guided fluid deresuscitation on duration of ventilation in intensive care unit patients (CONFIDENCE): protocol for a multicentre randomised controlled trial

BACKGROUND

Fluid therapy is a common intervention in critically ill patients. It is increasingly recognised that deresuscitation is an essential part of fluid therapy and delayed deresuscitation is associated with longer invasive ventilation and length of intensive care unit (ICU) stay. However, optimal timing and rate of deresuscitation remain unclear. Lung ultrasound (LUS) may be used to identify fluid overload. We hypothesise that daily LUS-guided deresuscitation is superior to deresuscitation without LUS in critically ill patients expected to undergo invasive ventilation for more than 24 h in terms of ventilator free-days and being alive at day 28.

METHODS

The "effect of lung ultrasound-guided fluid deresuscitation on duration of ventilation in intensive care unit patients" (CONFIDENCE) is a national, multicentre, open-label, randomised controlled trial (RCT) in adult critically ill patients that are expected to be invasively ventilated for at least 24 h. Patients with conditions that preclude a negative fluid balance or LUS examination are excluded. CONFIDENCE will operate in 10 ICUs in the Netherlands and enrol 1000 patients. After hemodynamic stabilisation, patients assigned to the intervention will receive daily LUS with fluid balance recommendations. Subjects in the control arm are deresuscitated at the physician's discretion without the use of LUS. The primary endpoint is the number of ventilator-free days and being alive at day 28. Secondary endpoints include the duration of invasive ventilation; 28-day mortality; 90-day mortality; ICU, in hospital and total length of stay; cumulative fluid balance on days 1-7 after randomisation and on days 1-7 after start of LUS examination; mean serum lactate on days 1-7; the incidence of reintubations, chest drain placement, atrial fibrillation, kidney injury (KDIGO stadium ≥ 2) and hypernatremia; the use of invasive hemodynamic monitoring, and chest-X-ray; and quality of life at day 28.

DISCUSSION

The CONFIDENCE trial is the first RCT comparing the effect of LUS-guided deresuscitation to routine care in invasively ventilated ICU patients. If proven effective, LUS-guided deresuscitation could improve outcomes in some of the most vulnerable and resource-intensive patients in a manner that is non-invasive, easy to perform, and well-implementable.

TRIAL REGISTRATION

ClinicalTrials.gov NCT05188092. Registered since January 12, 2022.

Concise Versus Extended Lung Ultrasound Score to Monitor Critically Ill Patients With COVID-19

BACKGROUND

Lung ultrasound (LUS) can be used to monitor critically ill patients with COVID-19, but the optimal number of examined lung zones is disputed.

METHODS

This was a prospective observational study. The objective was to investigate whether concise (6 zones) and extended (12 zones) LUS scoring protocols are clinically equivalent in critically ill ICU subjects with COVID-19. The primary outcome of this study was (statistical) agreement between concise and extended LUS score index evaluated in both supine and prone position. Agreement was determined using correlation coefficients and Bland-Altman plots to detect systematic differences between protocols. Secondary outcomes were difference between LUS score index in supine and prone position using similar methods.

RESULTS

We included 130 LUS examinations in 40 subjects (mean age 69.0 ± 8.5y, 75% male). Agreement between concise and extended LUS score index had no clinically relevant constant or proportional bias and limits of agreement were below the smallest detectable change. Across position changes, supine LUS score index was 8% higher than prone LUS score index and had limits above the smallest detectable change, indicating true LUS score index differences between protocols may occur due to the position change itself. Lastly, inter-rater and intra-rater agreement were very good.

CONCLUSIONS

Concise LUS was equally informative as extended LUS for monitoring critically ill subjects with COVID-19 in supine or prone position. Clinicians can monitor patients undergoing position changes but must be wary that LUS score index alterations may result from the position change itself rather than disease progression or clinical improvement.

Diagnosing acute kidney injury ahead of time in critically ill septic patients using kinetic estimated glomerular filtration rate

INTRODUCTION

Accurate and actionable diagnosis of Acute Kidney Injury (AKI) ahead of time is important to prevent or mitigate renal insufficiency. The purpose of this study was to evaluate the performance of Kinetic estimated Glomerular Filtration Rate (KeGFR) in timely predicting AKI in critically ill septic patients.

METHODS

We conducted a retrospective analysis on septic ICU patients who developed AKI in AmsterdamUMCdb, the first freely available European ICU database. The reference standard for AKI was the Kidney Disease: Improving Global Outcomes (KDIGO) classification based on serum creatinine and urine output (UO). Prediction of AKI was based on stages defined by KeGFR and UO. Classifications were compared by length of ICU stay (LOS), need for renal replacement therapy and 28-day mortality. Predictive performance and time between prediction and diagnosis were calculated.

RESULTS

Of 2492 patients in the cohort, 1560 (62.0%) were diagnosed with AKI by KDIGO and 1706 (68.5%) by KeGFR criteria. Disease stages had agreement of kappa = 0.77, with KeGFR sensitivity 93.2%, specificity 73.0% and accuracy 85.7%. Median time to recognition of AKI Stage 1 was 13.2 h faster for KeGFR, and 7.5 h and 5.0 h for Stages 2 and 3. Outcomes revealed a slight difference in LOS and 28-day mortality for Stage 1.

CONCLUSIONS

Predictive performance of KeGFR combined with UO criteria for diagnosing AKI is excellent. Compared to KDIGO, deterioration of renal function was identified earlier, most prominently for lower stages of AKI. This may shift the actionable window for preventing and mitigating renal insufficiency.

Development and validation of a multivariable model predicting the required catheter dwell time among mechanically ventilated critically ill patients in three randomized trials

BACKGROUND

The anatomic site for central venous catheter insertion influences the risk of central venous catheter-related intravascular complications. We developed and validated a predictive score of required catheter dwell time to identify critically ill patients at higher risk of intravascular complications.

METHODS

We retrospectively conducted a cohort study from three multicenter randomized controlled trials enrolling consecutive patients requiring central venous catheterization. The primary outcome was the required catheter dwell time, defined as the period between the first catheter insertion and removal of the last catheter for absence of utility. Predictors were identified in the training cohort (3SITES trial; 2336 patients) through multivariable analyses based on the subdistribution hazard function accounting for death as a competing event. Internal validation was performed in the training cohort by 500 bootstraps to derive the CVC-IN score from robust risk factors. External validation of the CVC-IN score were performed in the testing cohort (CLEAN, and DRESSING2; 2371 patients).

RESULTS

The analysis was restricted to patients requiring mechanical ventilation to comply with model assumptions. Immunosuppression (2 points), high creatinine > 100 micromol/L (2 points), use of vasopressor (1 point), obesity (1 point) and older age (40-59, 1 point; ≥ 60, 2 points) were independently associated with the required catheter dwell time. At day 28, area under the ROC curve for the CVC-IN score was 0.69, 95% confidence interval (CI) [0.66-0.72] in the training cohort and 0.64, 95% CI [0.61-0.66] in the testing cohort. Patients with a CVC-IN score ≥ 4 in the overall cohort had a median required catheter dwell time of 24 days (versus 11 days for CVC-IN score < 4 points). The positive predictive value of a CVC-IN score ≥ 4 was 76.9% for > 7 days required catheter dwell time in the testing cohort.

CONCLUSION

The CVC-IN score, which can be used for the first catheter, had a modest ability to discriminate required catheter dwell time. Nevertheless, preference of the subclavian site may contribute to limit the risk of intravascular complications, in particular among ventilated patients with high CVC-IN score. Trials Registration NCT01479153, NCT01629550, NCT01189682.

Development and validation of a multivariable prediction model of central venous catheter-tip colonization in a cohort of five randomized trials

Background

The majority of central venous catheters (CVC) removed in the ICU are not colonized, including when a catheter-related infection (CRI) is suspected. We developed and validated a predictive score to reduce unnecessary CVC removal.

Methods

We conducted a retrospective cohort study from five multicenter randomized controlled trials with systematic catheter-tip culture of consecutive CVCs. Colonization was defined as growth of ≥103 colony-forming units per mL. Risk factors for colonization were identified in the training cohort (CATHEDIA and 3SITES trials; 3899 CVCs of which 575 (15%) were colonized) through multivariable analyses. After internal validation in 500 bootstrapped samples, the CVC-OUT score was computed by attaching points to the robust (> 50% of the bootstraps) risk factors. External validation was performed in the testing cohort (CLEAN, DRESSING2 and ELVIS trials; 6848 CVCs, of which 588 (9%) were colonized).

Results

In the training cohort, obesity (1 point), diabetes (1 point), type of CVC (dialysis catheter, 1 point), anatomical insertion site (jugular, 4 points; femoral 5 points), rank of the catheter (second or subsequent, 1 point) and catheterization duration (≥ 5 days, 2 points) were significantly and independently associated with colonization . Area under the ROC curve (AUC) for the CVC-OUT score was 0.69, 95% confidence interval (CI) [0.67–0.72]. In the testing cohort, AUC for the CVC-OUT score was 0.60, 95% CI [0.58–0.62]. Among 1,469 CVCs removed for suspected CRI in the overall population, 1200 (82%) were not colonized. The negative predictive value (NPV) of a CVC-OUT score < 6 points was 94%, 95% CI [93%–95%].

Conclusion

The CVC-OUT score had a moderate ability to discriminate catheter-tip colonization, but the high NPV may contribute to reduce unnecessary CVCs removal. Preference of the subclavian site is the strongest and only modifiable risk factor that reduces the likelihood of catheter-tip colonization and consequently the risk of CRI.

Clinical Trials Registration: NCT00277888, NCT01479153, NCT01629550, NCT01189682, NCT00875069.

Association between early cumulative fluid balance and successful liberation from invasive ventilation in COVID-19 ARDS patients — insights from the PRoVENT-COVID study: a national, multicenter, observational cohort analysis

Background

Increasing evidence indicates the potential benefits of restricted fluid management in critically ill patients. Evidence lacks on the optimal fluid management strategy for invasively ventilated COVID-19 patients. We hypothesized that the cumulative fluid balance would affect the successful liberation of invasive ventilation in COVID-19 patients with acute respiratory distress syndrome (ARDS).

Methods

We analyzed data from the multicenter observational ‘PRactice of VENTilation in COVID-19 patients’ study. Patients with confirmed COVID-19 and ARDS who required invasive ventilation during the first 3 months of the international outbreak (March 1, 2020, to June 2020) across 22 hospitals in the Netherlands were included. The primary outcome was successful liberation of invasive ventilation, modeled as a function of day 3 cumulative fluid balance using Cox proportional hazards models, using the crude and the adjusted association. Sensitivity analyses without missing data and modeling ARDS severity were performed.

Results

Among 650 patients, three groups were identified. Patients in the higher, intermediate, and lower groups had a median cumulative fluid balance of 1.98 L (1.27–7.72 L), 0.78 L (0.26–1.27 L), and − 0.35 L (− 6.52–0.26 L), respectively. Higher day 3 cumulative fluid balance was significantly associated with a lower probability of successful ventilation liberation (adjusted hazard ratio 0.86, 95% CI 0.77–0.95, P = 0.0047). Sensitivity analyses showed similar results.

Conclusions

In a cohort of invasively ventilated patients with COVID-19 and ARDS, a higher cumulative fluid balance was associated with a longer ventilation duration, indicating that restricted fluid management in these patients may be beneficial.

Trial registration Clinicaltrials.gov (NCT04346342); Date of registration: April 15, 2020.

Graphical abstract

Predicting responders to prone positioning in mechanically ventilated patients with COVID-19 using machine learning

Background

For mechanically ventilated critically ill COVID-19 patients, prone positioning has quickly become an important treatment strategy, however, prone positioning is labor intensive and comes with potential adverse effects. Therefore, identifying which critically ill intubated COVID-19 patients will benefit may help allocate labor resources.

Methods

From the multi-center Dutch Data Warehouse of COVID-19 ICU patients from 25 hospitals, we selected all 3619 episodes of prone positioning in 1142 invasively mechanically ventilated patients. We excluded episodes longer than 24 h. Berlin ARDS criteria were not formally documented. We used supervised machine learning algorithms Logistic Regression, Random Forest, Naive Bayes, K-Nearest Neighbors, Support Vector Machine and Extreme Gradient Boosting on readily available and clinically relevant features to predict success of prone positioning after 4 h (window of 1 to 7 h) based on various possible outcomes. These outcomes were defined as improvements of at least 10% in PaO₂/FiO₂ ratio, ventilatory ratio, respiratory system compliance, or mechanical power. Separate models were created for each of these outcomes. Re-supination within 4 h after pronation was labeled as failure. We also developed models using a 20 mmHg improvement cut-off for PaO₂/FiO₂ ratio and using a combined outcome parameter. For all models, we evaluated feature importance expressed as contribution to predictive performance based on their relative ranking.

Results

The median duration of prone episodes was 17 h (11–20, median and IQR, N = 2632). Despite extensive modeling using a plethora of machine learning techniques and a large number of potentially clinically relevant features, discrimination between responders and non-responders remained poor with an area under the receiver operator characteristic curve of 0.62 for PaO₂/FiO₂ ratio using Logistic Regression, Random Forest and XGBoost. Feature importance was inconsistent between models for different outcomes. Notably, not even being a previous responder to prone positioning, or PEEP-levels before prone positioning, provided any meaningful contribution to predicting a successful next proning episode.

Conclusions

In mechanically ventilated COVID-19 patients, predicting the success of prone positioning using clinically relevant and readily available parameters from electronic health records is currently not feasible. Given the current evidence base, a liberal approach to proning in all patients with severe COVID-19 ARDS is therefore justified and in particular regardless of previous results of proning.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13613-022-01070-0.

Right dose, right now: bedside, real-time, data-driven, and personalised antibiotic dosing in critically ill patients with sepsis or septic shock—a two-centre randomised clinical trial

Background

Adequate antibiotic dosing may improve outcomes in critically ill patients but is challenging due to altered and variable pharmacokinetics. To address this challenge, AutoKinetics was developed, a decision support system for bedside, real-time, data-driven and personalised antibiotic dosing. This study evaluates the feasibility, safety and efficacy of its clinical implementation.

Methods

In this two-centre randomised clinical trial, critically ill patients with sepsis or septic shock were randomised to AutoKinetics dosing or standard dosing for four antibiotics: vancomycin, ciprofloxacin, meropenem, and ceftriaxone. Adult patients with a confirmed or suspected infection and either lactate > 2 mmol/L or vasopressor requirement were eligible for inclusion. The primary outcome was pharmacokinetic target attainment in the first 24 h after randomisation. Clinical endpoints included mortality, ICU length of stay and incidence of acute kidney injury.

Results

After inclusion of 252 patients, the study was stopped early due to the COVID-19 pandemic. In the ciprofloxacin intervention group, the primary outcome was obtained in 69% compared to 3% in the control group (OR 62.5, CI 11.4–1173.78, p < 0.001). Furthermore, target attainment was faster (26 h, CI 18–42 h, p < 0.001) and better (65% increase, CI 49–84%, p < 0.001). For the other antibiotics, AutoKinetics dosing did not improve target attainment. Clinical endpoints were not significantly different. Importantly, higher dosing did not lead to increased mortality or renal failure.

Conclusions

In critically ill patients, personalised dosing was feasible, safe and significantly improved target attainment for ciprofloxacin.

Trial registration: The trial was prospectively registered at Netherlands Trial Register (NTR), NL6501/NTR6689 on 25 August 2017 and at the European Clinical Trials Database (EudraCT), 2017-002478-37 on 6 November 2017.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-022-04098-7.

Mid-regional proadrenomedullin, C-terminal proendothelin-1 values, and disease course are not different in critically ill SARS-CoV-2 pneumonia patients with obesity

Background/objectives

Patients affected by obesity and Coronavirus disease 2019, the disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), appear to have a higher risk for intensive care (ICU) admission. A state of low-grade chronic inflammation in obesity has been suggested as one of the underlying mechanisms. We investigated whether obesity is associated with differences in new inflammatory biomarkers mid-regional proadrenomedullin (MR-proADM), C-terminal proendothelin-1 (CT-proET-1), and clinical outcomes in critically ill patients with SARS-CoV-2 pneumonia.

Subjects/methods

A total of 105 critically ill patients with SARS-CoV-2 pneumonia were divided in patients with obesity (body mass index (BMI) ≥ 30 kg/m², n = 42) and patients without obesity (BMI < 30 kg/m², n = 63) and studied in a retrospective observational cohort study. MR-proADM, CT-proET-1 concentrations, and conventional markers of white blood count (WBC), C-reactive protein (CRP), and procalcitonin (PCT) were collected during the first 7 days.

Results

BMI was 33.5 (32–36.1) and 26.2 (24.7–27.8) kg/m² in the group with and without obesity. There were no significant differences in concentrations MR-proADM, CT-proET-1, WBC, CRP, and PCT at baseline and the next 6 days between patients with and without obesity. Only MR-proADM changed significantly over time (p = 0.039). Also, BMI did not correlate with inflammatory biomarkers (MR-proADM rho = 0.150, p = 0.125, CT-proET-1 rho = 0.179, p = 0.067, WBC rho = −0.044, p = 0.654, CRP rho = 0.057, p = 0.564, PCT rho = 0.022, p = 0.842). Finally, no significant differences in time on a ventilator, ICU length of stay, and 28-day mortality between patients with or without obesity were observed.

Conclusions

In critically ill patients with confirmed SARS-CoV-2 pneumonia, obesity was not associated with differences in MR-proADM, and CT-proET-1, or impaired outcome.

Trial registration

Netherlands Trial Register, NL8460.

C-Terminal Proarginine Vasopressin is Associated with Disease Outcome and Mortality, but not with Delayed Cerebral Ischemia in Critically Ill Patients with an Aneurysmal Subarachnoid Hemorrhage: A Prospective Cohort Study

BACKGROUND

Aneurysmal subarachnoid hemorrhage (aSAH) is an important indication for intensive care unit admission and may lead to significant morbidity and mortality. We assessed the ability of C-terminal proarginine vasopressin (CT-proAVP) to predict disease outcome, mortality, and delayed cerebral ischemia (DCI) in critically ill patients with aSAH compared with the World Federation of Neurological Surgeons (WFNS) score and Acute Physiological and Chronic Health Evaluation IV (APACHE IV) model.

METHODS

C-terminal proarginine vasopressin was collected on admission in this single-center, prospective, observational cohort study. The primary aim was to investigate the relationship between CT-proAVP and poor functional outcome at 1 year (Glasgow Outcome Scale score 1-3) in a multivariable logistic regression model adjusted for WFNS and APACHE IV scores. Secondary aims were mortality and DCI. The multivariable logistic regression model for DCI was also adjusted for the modified Fisher scale.

RESULTS

In 100 patients, the median CT-proAVP level was 24.9 pmol/L (interquartile range 11.5-53.8); 45 patients had a poor 1-year functional outcome, 19 patients died within 30 days, 25 patients died within 1 year, and DCI occurred in 28 patients. Receiver operating characteristics curves revealed high accuracy for CT-proAVP to identify patients with poor 1-year functional outcome (area under the curve [AUC] 0.84, 95% confidence interval [CI] 0.77-0.92, p < 0.001), 30-day mortality (AUC 0.84, 95% CI 0.76-0.93, p < 0.001), and 1-year mortality (AUC 0.79, 95% CI 0.69-0.89, p < 0.001). CT-proAVP had a low AUC for identifying patients with DCI (AUC 0.67, 95% CI 0.55-0.79, p 0.008). CT-proAVP ≥ 24.9 pmo/L proved to be a significant predictor for poor 1-year functional outcome (odds ratio [OR] 8.04, 95% CI 2.97-21.75, p < 0.001), and CT-proAVP ≥ 29.1 pmol/L and ≥ 27.7 pmol/L were significant predictors for 30-day and 1-year mortality (OR 9.31, 95% CI 1.55-56.07, p 0.015 and OR 5.15, 95% CI 1.48-17.93, p 0.010) in multivariable models with WFNS and APACHE IV scores. CT-proAVP ≥ 29.5 pmol/L was not a significant predictor for DCI in a multivariable model adjusted for the modified Fisher scale (p = 0.061).

CONCLUSIONS

C-terminal proarginine vasopressin was able to predict poor functional outcome and mortality in critically ill patients with aSAH. Its prognostic ability to predict DCI was low.

TRIAL REGISTRATION

Nederlands Trial Register: NTR4118.

Prediction of Acute Kidney Injury in the Intensive Care Unit: Preliminary Findings in a European Open Access Database

Acute kidney injury (AKI) is a common complication in critically ill patients and is associated with long-term complications and an increased mortality. This work presents preliminary findings from the first freely available European intensive care database released by Amsterdam UMC. A machine learning (ML) model was developed to predict AKI in the intensive care unit 12 hours before the actual event. Main features of the model included medications and hemodynamic parameters. Our models perform with an accuracy of 81.8% on moderate to severe AKI and 79.8% on all AKI patients. Those results can compete with models reported in the literature and introduce an ML model for AKI based on European patient data.

Incidence, Clinical Characteristics and Outcomes of Early Hyperbilirubinemia in Critically Ill Patients: Insights From the MARS Study

OBJECTIVE

To investigate the incidence, clinical characteristics and outcomes of early hyperbilirubinemia in critically ill patients.

DESIGN AND SETTING

This is a post hoc analysis of a prospective multicenter cohort study.

PATIENTS

Patients with measured bilirubin levels within the first 2 days after ICU admission were eligible. Patients with liver cirrhosis were excluded.

ENDPOINTS

The primary endpoint was the incidence of early hyperbilirubinemia, defined as bilirubin ≥33 μmol/L within 2 days after ICU admission. Secondary endpoints included clinical characteristics of patients with versus patients without early hyperbilirubinemia, and outcomes up to day 30.

RESULTS

Of 4,836 patients, 559 (11.6%) patients had early hyperbilirubinemia. Compared to patients without early hyperbilirubinemia, patients with early hyperbilirubinemia presented with higher severity of illness scores, and higher incidences of sepsis and organ failure. After adjustment for confounding variables, early hyperbilirubinemia remained associated with mortality at day 30 (odds ratio, 1.31 [95%-confidence interval 1.06-1.60]; P = 0.018). Patients with early hyperbilirubinemia and thrombocytopenia (interaction P-value = 0.005) had a higher likelihood of death within 30 days (odds ratio, 2.61 [95%-confidence interval 2.08-3.27]; P < 0.001) than patients with early hyperbilirubinemia and a normal platelet count (odds ratio, 1.09 [95%-confidence interval 0.75-1.55]; P = 0.655).

CONCLUSIONS

Early hyperbilirubinemia occurs frequently in the critically ill, and these patients present with higher disease severity and more often with sepsis and organ failures. Early hyperbilirubinemia has an association with mortality, albeit this association was only found in patients with concomitant thrombocytopenia.

Effect of Low-Normal vs High-Normal Oxygenation Targets on Organ Dysfunction in Critically Ill Patients: A Randomized Clinical Trial

IMPORTANCE

Hyperoxemia may increase organ dysfunction in critically ill patients, but optimal oxygenation targets are unknown.

OBJECTIVE

To determine whether a low-normal Pao2 target compared with a high-normal target reduces organ dysfunction in critically ill patients with systemic inflammatory response syndrome (SIRS).

DESIGN, SETTING, AND PARTICIPANTS

Multicenter randomized clinical trial in 4 intensive care units in the Netherlands. Enrollment was from February 2015 to October 2018, with end of follow-up to January 2019, and included adult patients admitted with 2 or more SIRS criteria and expected stay of longer than 48 hours. A total of 9925 patients were screened for eligibility, of whom 574 fulfilled the enrollment criteria and were randomized.

INTERVENTIONS

Target Pao2 ranges were 8 to 12 kPa (low-normal, n = 205) and 14 to 18 kPa (high-normal, n = 195). An inspired oxygen fraction greater than 0.60 was applied only when clinically indicated.

MAIN OUTCOMES AND MEASURES

Primary end point was SOFARANK, a ranked outcome of nonrespiratory organ failure quantified by the nonrespiratory components of the Sequential Organ Failure Assessment (SOFA) score, summed over the first 14 study days. Participants were ranked from fastest organ failure improvement (lowest scores) to worsening organ failure or death (highest scores). Secondary end points were duration of mechanical ventilation, in-hospital mortality, and hypoxemic measurements.

RESULTS

Among the 574 patients who were randomized, 400 (70%) were enrolled within 24 hours (median age, 68 years; 140 women [35%]), all of whom completed the trial. The median Pao2 difference between the groups was -1.93 kPa (95% CI, -2.12 to -1.74; P < .001). The median SOFARANK score was -35 points in the low-normal Pao2 group vs -40 in the high-normal Pao2 group (median difference, 10 [95% CI, 0 to 21]; P = .06). There was no significant difference in median duration of mechanical ventilation (3.4 vs 3.1 days; median difference, -0.15 [95% CI, -0.88 to 0.47]; P = .59) and in-hospital mortality (32% vs 31%; odds ratio, 1.04 [95% CI, 0.67 to 1.63]; P = .91). Mild hypoxemic measurements occurred more often in the low-normal group (1.9% vs 1.2%; median difference, 0.73 [95% CI, 0.30 to 1.20]; P < .001). Acute kidney failure developed in 20 patients (10%) in the low-normal Pao2 group and 21 patients (11%) in the high-normal Pao2 group, and acute myocardial infarction in 6 patients (2.9%) in the low-normal Pao2 group and 7 patients (3.6%) in the high-normal Pao2 group.

CONCLUSIONS AND RELEVANCE

Among critically ill patients with 2 or more SIRS criteria, treatment with a low-normal Pao2 target compared with a high-normal Pao2 target did not result in a statistically significant reduction in organ dysfunction. However, the study may have had limited power to detect a smaller treatment effect than was hypothesized.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT02321072.

Targeting coenzyme Q10 synthesis overcomes bortezomib resistance in multiple myeloma

During the development of drug resistance, multiple myeloma (MM) cells undergo changes to their metabolism. However, how these metabolic changes can be exploited to improve treatment efficacy is not known. Here we demonstrate that targeting coenzyme Q10 (CoQ) biosynthesis through the mevalonate pathway works in synergy with the proteasome inhibitor bortezomib (BTZ) in MM. We show that gene expression signatures relating to the mitochondrial tricarboxylic acid (TCA) cycle and electron transport chain (ETC) predispose to clinical BTZ resistance and poor prognosis in MM patients. Mechanistically, BTZ-resistant cells show increased activity of glutamine-driven TCA cycle and oxidative phosphorylation, together with an increased vulnerability towards ETC inhibition. Moreover, BTZ resistance is accompanied by high levels of the mitochondrial electron carrier CoQ, while the mevalonate pathway inhibitor simvastatin increases cell death and decreases CoQ levels, specifically in BTZ-resistant cells. Both in vitro and in vivo, simvastatin enhances the effect of bortezomib treatment. Our study links CoQ synthesis to drug resistance in MM and provides a novel avenue for improving BTZ responses through statin-induced inhibition of mitochondrial metabolism.

Early high-dose vitamin C in post-cardiac arrest syndrome (VITaCCA): study protocol for a randomized, double-blind, multi-center, placebo-controlled trial

BACKGROUND

High-dose intravenous vitamin C directly scavenges and decreases the production of harmful reactive oxygen species (ROS) generated during ischemia/reperfusion after a cardiac arrest. The aim of this study is to investigate whether short-term treatment with a supplementary or very high-dose intravenous vitamin C reduces organ failure in post-cardiac arrest patients.

METHODS

This is a double-blind, multi-center, randomized placebo-controlled trial conducted in 7 intensive care units (ICUs) in The Netherlands. A total of 270 patients with cardiac arrest and return of spontaneous circulation will be randomly assigned to three groups of 90 patients (1:1:1 ratio, stratified by site and age). Patients will intravenously receive a placebo, a supplementation dose of 3 g of vitamin C or a pharmacological dose of 10 g of vitamin C per day for 96 h. The primary endpoint is organ failure at 96 h as measured by the Resuscitation-Sequential Organ Failure Assessment (R-SOFA) score at 96 h minus the baseline score (delta R-SOFA). Secondary endpoints are a neurological outcome, mortality, length of ICU and hospital stay, myocardial injury, vasopressor support, lung injury score, ventilator-free days, renal function, ICU-acquired weakness, delirium, oxidative stress parameters, and plasma vitamin C concentrations.

DISCUSSION

Vitamin C supplementation is safe and preclinical studies have shown beneficial effects of high-dose IV vitamin C in cardiac arrest models. This is the first RCT to assess the clinical effect of intravenous vitamin C on organ dysfunction in critically ill patients after cardiac arrest.

TRIAL REGISTRATION

ClinicalTrials.gov NCT03509662. Registered on April 26, 2018. https://clinicaltrials.gov/ct2/show/NCT03509662 European Clinical Trials Database (EudraCT): 2017-004318-25. Registered on June 8, 2018. https://www.clinicaltrialsregister.eu/ctr-search/trial/2017-004318-25/NL.

Breathing variability-implications for anaesthesiology and intensive care

The respiratory system reacts instantaneously to intrinsic and extrinsic inputs. This adaptability results in significant fluctuations in breathing parameters, such as respiratory rate, tidal volume, and inspiratory flow profiles. Breathing variability is influenced by several conditions, including sleep, various pulmonary diseases, hypoxia, and anxiety disorders. Recent studies have suggested that weaning failure during mechanical ventilation may be predicted by low respiratory variability. This review describes methods for quantifying breathing variability, summarises the conditions and comorbidities that affect breathing variability, and discusses the potential implications of breathing variability for anaesthesia and intensive care.

Lung ultrasound and computed tomography to monitor COVID-19 pneumonia in critically ill patients: a two-center prospective cohort study

BACKGROUND

Lung ultrasound can adequately monitor disease severity in pneumonia and acute respiratory distress syndrome. We hypothesize lung ultrasound can adequately monitor COVID-19 pneumonia in critically ill patients.

METHODS

Adult patients with COVID-19 pneumonia admitted to the intensive care unit of two academic hospitals who underwent a 12-zone lung ultrasound and a chest CT examination were included. Baseline characteristics, and outcomes including composite endpoint death or ICU stay > 30 days were recorded. Lung ultrasound and CT images were quantified as a lung ultrasound score involvement index (LUSI) and CT severity involvement index (CTSI). Primary outcome was the correlation, agreement, and concordance between LUSI and CTSI. Secondary outcome was the association of LUSI and CTSI with the composite endpoints.

RESULTS

We included 55 ultrasound examinations in 34 patients, which were 88% were male, with a mean age of 63 years and mean P/F ratio of 151. The correlation between LUSI and CTSI was strong (r = 0.795), with an overall 15% bias, and limits of agreement ranging - 40 to 9.7. Concordance between changes in sequentially measured LUSI and CTSI was 81%. In the univariate model, high involvement on LUSI and CTSI were associated with a composite endpoint. In the multivariate model, LUSI was the only remaining independent predictor.

CONCLUSIONS

Lung ultrasound can be used as an alternative for chest CT in monitoring COVID-19 pneumonia in critically ill patients as it can quantify pulmonary involvement, register changes over the course of the disease, and predict death or ICU stay > 30 days.

TRIAL REGISTRATION

NTR, NL8584. Registered 01 May 2020-retrospectively registered, https://www.trialregister.nl/trial/8584.

Time to stop randomized and large pragmatic trials for intensive care medicine syndromes: the case of sepsis and acute respiratory distress syndrome

In this paper we discuss the limitations of large randomized controlled trials with mortality endpoints in patients with critical illness associated diagnoses such as sepsis. When patients with the same syndrome diagnosis do not share the pathways that lead to death (the attributable risk), any therapy can only lead to small effects in these populations. Using Monte Carlo simulations, we show how the syndrome-attributable risks of critical illness-associated diagnoses are likely overestimated using common statistical methods. This overestimation of syndrome-attributable risks leads to a corresponding overestimation of attainable treatment effects and an underestimation of required sample sizes. We demonstrate that larger and more 'pragmatic' randomized trials are not the solution because they decrease therapeutic and diagnostic precision, the therapeutic effect size and the probability of finding a beneficial effect. Finally, we argue that the most logical solution is a renewed focus on mechanistic research into the complexities of critical illness syndromes.

Why physiology will continue to guide the choice between balanced crystalloids and normal saline: a systematic review and meta-analysis

BACKGROUND

Crystalloids are the most frequently prescribed drugs in intensive care medicine and emergency medicine. Thus, even small differences in outcome may have major implications, and therefore, the choice between balanced crystalloids versus normal saline continues to be debated. We examined to what extent the currently accrued information size from completed and ongoing trials on the subject allow intensivists and emergency physicians to choose the right fluid for their patients.

METHODS

Systematic review and meta-analysis with random effects inverse variance model. Published randomized controlled trials enrolling adult patients to compare balanced crystalloids versus normal saline in the setting of intensive care medicine or emergency medicine were included. The main outcome was mortality at the longest follow-up, and secondary outcomes were moderate to severe acute kidney injury (AKI) and initiation of renal replacement therapy (RRT). Trial sequential analyses (TSA) were performed, and risk of bias and overall quality of evidence were assessed. Additionally, previously published meta-analyses, trial sequential analyses and ongoing large trials were analysed for included studies, required information size calculations and the assumptions underlying those calculations.

RESULTS

Nine studies (n = 32,777) were included. Of those, eight had data available on mortality, seven on AKI and six on RRT. Meta-analysis showed no significant differences between balanced crystalloids versus normal saline for mortality (P = 0.33), the incidence of moderate to severe AKI (P = 0.37) or initiation of RRT (P = 0.29). Quality of evidence was low to very low. Analysis of previous meta-analyses and ongoing trials showed large differences in calculated required versus accrued information sizes and assumptions underlying those. TSA revealed the need for extremely large trials based on our realistic and clinically relevant assumptions on relative risk reduction and baseline mortality.

CONCLUSIONS

Our meta-analysis could not find significant differences between balanced crystalloids and normal saline on mortality at the longest follow-up, moderate to severe AKI or new RRT. Currently accrued information size is smaller, and the required information size is larger than previously anticipated. Therefore, completed and ongoing trials on the topic may fail to provide adequate guidance for choosing the right crystalloid. Thus, physiology will continue to play an important role for individualizing this choice.

Estimating Vitamin C Status in Critically Ill Patients with a Novel Point-of-Care Oxidation-Reduction Potential Measurement

Vitamin C deficiency is common in critically ill patients. Vitamin C, the most important antioxidant, is likely consumed during oxidative stress and deficiency is associated with organ dysfunction and mortality. Assessment of vitamin C status may be important to identify patients who might benefit from vitamin C administration. Up to now, vitamin C concentrations are not available in daily clinical practice. Recently, a point-of-care device has been developed that measures the static oxidation-reduction potential (sORP), reflecting oxidative stress, and antioxidant capacity (AOC). The aim of this study was to determine whether plasma vitamin C concentrations were associated with plasma sORP and AOC. Plasma vitamin C concentration, sORP and AOC were measured in three groups: healthy volunteers, critically ill patients, and critically ill patients receiving 2- or 10-g vitamin C infusion. Its association was analyzed using regression models and by assessment of concordance. We measured 211 samples obtained from 103 subjects. Vitamin C concentrations were negatively associated with sORP (R² = 0.816) and positively associated with AOC (R² = 0.842). A high concordance of 94–100% was found between vitamin C concentration and sORP/AOC. Thus, plasma vitamin C concentrations are strongly associated with plasma sORP and AOC, as measured with a novel point-of-care device. Therefore, measuring sORP and AOC at the bedside has the potential to identify and monitor patients with oxidative stress and vitamin C deficiency.

Unexplained mortality differences between septic shock trials: a systematic analysis of population characteristics and control-group mortality rates

Purpose

Although the definition of septic shock has been standardized, some variation in mortality rates among clinical trials is expected. Insights into the sources of heterogeneity may influence the design and interpretation of septic shock studies. We set out to identify inclusion criteria and baseline characteristics associated with between-trial differences in control group mortality rates.

Methods

We conducted a systematic review of RCTs published between 2006 and 2018 that included patients with septic shock. The percentage of variance in control-group mortality attributable to study heterogeneity rather than chance was measured by I². The association between control-group mortality and population characteristics was estimated using linear mixed models and a recursive partitioning algorithm.

Results

Sixty-five septic shock RCTs were included. Overall control-group mortality was 38.6%, with significant heterogeneity (I² = 93%, P < 0.0001) and a 95% prediction interval of 13.5–71.7%. The mean mortality rate did not differ between trials with different definitions of hypotension, infection or vasopressor or mechanical ventilation inclusion criteria. Population characteristics univariately associated with mortality rates were mean Sequential Organ Failure Assessment score (standardized regression coefficient (β) = 0.57, P = 0.007), mean serum creatinine (β = 0.48, P = 0.007), the proportion of patients on mechanical ventilation (β = 0.61, P < 0.001), and the proportion with vasopressors (β = 0.57, P = 0.002). Combinations of population characteristics selected with a linear model and recursive partitioning explained 41 and 42%, respectively, of the heterogeneity in mortality rates.

Conclusions

Among 65 septic shock trials, there was a clinically relevant amount of heterogeneity in control group mortality rates which was explained only partly by differences in inclusion criteria and reported baseline characteristics.

Electronic supplementary material

The online version of this article (10.1007/s00134-018-5134-8) contains supplementary material, which is available to authorized users.

Vitamin C Pharmacokinetics in Critically Ill Patients: A Randomized Trial of Four IV Regimens

BACKGROUND

Early high-dose IV vitamin C is being investigated as adjuvant therapy in patients who are critically ill, but the optimal dose and infusion method are unclear. The primary aim of this study was to describe the dose-plasma concentration relationship and safety of four different dosing regimens.

METHODS

This was a four-group randomized pharmacokinetic trial. Patients who were critically ill with multiple organ dysfunction were randomized to receive 2 or 10 g/d vitamin C as a twice daily bolus infusion or continuous infusion for 48 h. End points were plasma vitamin C concentrations during 96 h, 12-h urine excretion of vitamin C, and oxalate excretion and base excess. A population pharmacokinetic model was developed using NONMEM.

RESULTS

Twenty patients were included. A two-compartment pharmacokinetic model with creatinine clearance and weight as independent covariates described all four regimens best. With 2 g/d bolus, plasma vitamin C concentrations at 1 h were 29 to 50 mg/L and trough concentrations were 5.6 to 16 mg/L. With 2 g/d continuous, steady-state concentrations were 7 to 37 mg/L at 48 h. With 10 g/d bolus, 1-h concentrations were 186 to 244 mg/L and trough concentrations were 14 to 55 mg/L. With 10 g/d continuous, steady-state concentrations were 40 to 295 mg/L at 48 h. Oxalate excretion and base excess were increased in the 10 g/d dose. Forty-eight hours after discontinuation, plasma concentrations declined to hypovitaminosis levels in 15% of patients.

CONCLUSIONS

The 2 g/d dose was associated with normal plasma concentrations, and the 10 g/d dose was associated with supranormal plasma concentrations, increased oxalate excretion, and metabolic alkalosis. Sustained therapy is needed to prevent hypovitaminosis.

TRIAL REGISTRY

ClinicalTrials.gov; No.: NCT02455180; URL: www.clinicaltrials.gov.

SOFA and mortality endpoints in randomized controlled trials: a systematic review and meta-regression analysis

BACKGROUND

The sequential organ failure assessment score (SOFA) is increasingly used as an endpoint in intensive care randomized controlled trials (RCTs). Although serially measured SOFA is independently associated with mortality in observational cohorts, the association between treatment effects on SOFA vs. effects on mortality has not yet been quantified in RCTs. The aim of this study was to quantify the relationship between SOFA and mortality in RCTs and to identify which SOFA derivative best reflects between-group mortality differences.

METHODS

The review protocol was prospectively registered (Prospero CRD42016034014). We performed a literature search (up to May 1, 2016) for RCTs reporting both SOFA and mortality, and analyzed between-group differences in these outcomes. Treatment effects on SOFA and mortality were calculated as the between-group SOFA standardized difference and log odds ratio (OR), respectively. We used random-effects meta-regression to (1) quantify the linear relationship between RCT treatment effects on mortality (logOR) and SOFA (i.e. responsiveness) and (2) quantify residual heterogeneity (i.e. consistency, expressed as I ²).

RESULTS

Of 110 eligible RCTs, 87 qualified for analysis. Using all RCTs, SOFA was significantly associated with mortality (slope = 0.49 (95% CI 0.17; 0.82), p = 0.006, I ² = 5%); the overall mortality effect explained by SOFA score (R ²) was 9%. Fifty-eight RCTs used Fixed-day SOFA as an endpoint (i.e. the score on a fixed day after randomization), 25 studies used Delta SOFA as an endpoint (i.e. the trajectory from baseline score) and 15 studies used other SOFA derivatives as an endpoint. Fixed-day SOFA was not significantly associated with mortality (slope = 0.35 (95% CI -0.04; 0.75), p = 0.08, I ² = 12%) and explained 3% of the overall mortality effect (R ²). Delta SOFA was significantly associated with mortality (slope = 0.70 (95% CI 0.26; 1.14), p = 0.004, I ² = 0%) and explained 32% of the overall mortality effect (R ²).

CONCLUSIONS

Treatment effects on Delta SOFA appear to be reliably and consistently associated with mortality in RCTs. Fixed-day SOFA was the most frequently reported outcome among the reviewed RCTs, but was not significantly associated with mortality. Based on this study, we recommend using Delta SOFA rather than Fixed-day SOFA as an endpoint in future RCTs.

Prevention or Treatment of Ards With Aspirin: A Review of Preclinical Models and Meta-Analysis of Clinical Studies

BACKGROUND

The acute respiratory distress syndrome (ARDS) is a life-threating disorder that contributes significantly to critical illness. No specific pharmacological interventions directed at lung injury itself have proven effective in improving outcome of patients with ARDS. Platelet activation was identified as a key component in ARDS pathophysiology and may provide an opportunity for preventive and therapeutic strategies. We hypothesize that use of acetyl salicylic acid (ASA) may prevent and/or attenuate lung injury.

METHODS

We conducted a systematic review of preclinical studies and meta-analysis of clinical studies investigating the efficacy of ASA in the setting of lung injury. Medline, embase, and cochrane databases were searched.

RESULTS

The literature search yielded 1,314 unique articles. Fifteen preclinical studies and eight clinical studies fulfilled the in- and exclusion criteria. In the animal studies, the overall effect of ASA was positive, e.g., ASA improved survival and attenuated inflammation and pulmonary edema. Mechanisms of actions involved, among others, are interference with the neutrophil-platelets interaction, reduction of leukotrienes, neutrophil extracellular traps, and prostaglandins. High-dose ASA may be the drug of choice. A meta-analysis of three clinical studies showed an association between ASA use and a reduced incidence of ARDS (OR 0.59, 95% CI 0.36-0.98), albeit with substantial between-study heterogeneity. All studies had their own shortcomings in methodological quality.

CONCLUSION

This systematic review of preclinical studies and meta-analysis of clinical studies suggests a beneficial role for ASA in ARDS prevention and treatment. However, the currently available data is insufficient to justify an indication for ASA in ARDS. The body of literature does support further studies in humans. We suggest clinical trials in which the mechanisms of action of ASA in lung injury models are being evaluated to guide optimal timing and dose, before prospective randomized trials.