Impact of differences in acute respiratory distress syndrome randomised controlled trial inclusion and exclusion criteria: systematic review and meta-analysis

Abstract

Background: Control-arm mortality varies between acute respiratory distress syndrome (ARDS) RCTs.

Methods: We systematically reviewed ARDS RCTs that commenced recruitment after publication of the American-European Consensus (AECC) definition (MEDLINE, Embase, and Cochrane central register of controlled trials; January 1994 to October 2020). We assessed concordance of RCT inclusion criteria to ARDS consensus definitions and whether exclusion criteria are strongly or poorly justified. We estimated the proportion of between-trial difference in control-arm 28-day mortality explained by the inclusion criteria and RCT design characteristics using meta-regression.

Results: A literature search identified 43 709 records. One hundred and fifty ARDS RCTs were included; 146/150 (97.3%) RCTs defined ARDS inclusion criteria using AECC/Berlin definitions. Deviations from consensus definitions, primarily aimed at improving ARDS diagnostic certainty, frequently related to duration of hypoxaemia (117/146; 80.1%). Exclusion criteria could be grouped by rationale for selection into strongly or poorly justified criteria. Common poorly justified exclusions included pregnancy related, age, and comorbidities (infectious/immunosuppression, hepatic, renal, and human immunodeficiency virus/acquired immunodeficiency syndrome). Control-arm 28-day mortality varied between ARDS RCTs (mean: 29.8% [95% confidence interval: 27.0-32.7%; I²=88.8%; τ²=0.02; P<0.01]), and differed significantly between RCTs with different Pao₂:FiO₂ ratio inclusion thresholds (26.6-39.9 kPa vs <26.6 kPa; P<0.01). In a meta-regression model, inclusion criteria and RCT design characteristics accounted for 30.6% of between-trial difference (P<0.01).

Conclusions: In most ARDS RCTs, consensus definitions are modified to use as inclusion criteria. Between-RCT mortality differences are mostly explained by the Pao₂:FiO₂ ratio threshold within the consensus definitions. An exclusion criteria framework can be applied when designing and reporting exclusion criteria in future ARDS RCTs.

Keywords: ARDS; exclusion; inclusion; mortality; randomised controlled trial.

Fig 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowsheet showing RCT selection process. ARDS, acute respiratory distress syndrome.

Fig 2

Risk of bias for mortality outcomes in ARDS RCTs. Domains were specified and adjudicated as per the Cochrane risk-of-bias tool. RCTs were divided by intervention type and ordered by year of publication. ARDS, acute respiratory distress syndrome.

Fig 3

(a) Univariate, weighted analysis of the impact of inclusion criteria and RCT characteristics on 28-day mortality. (b) Results of a meta-regression model of the impact of inclusion criteria and RCT characteristics on 28-day mortality. Total variance is made up of between- and within-study variance. Within-study variance cannot be explained using study-level factors. I² is the proportion of variance in ARDS RCTs that is attributable to between-trial variance. The proportion of between-trial variance that can be explained (R²) is 30.6%; inclusion criteria (28.8%), and RCT design characteristics (1.8%); 69.4% of between-trial variance remained unexplained. ARDS, acute respiratory distress syndrome; CI, confidence interval; CXR, chest radiograph; P:F, Pao₂:FiO₂; LA, left atrial; PAWP, pulmonary artery wedge pressure.

Fig 3

(a) Univariate, weighted analysis of the impact of inclusion criteria and RCT characteristics on 28-day mortality. (b) Results of a meta-regression model of the impact of inclusion criteria and RCT characteristics on 28-day mortality. Total variance is made up of between- and within-study variance. Within-study variance cannot be explained using study-level factors. I² is the proportion of variance in ARDS RCTs that is attributable to between-trial variance. The proportion of between-trial variance that can be explained (R²) is 30.6%; inclusion criteria (28.8%), and RCT design characteristics (1.8%); 69.4% of between-trial variance remained unexplained. ARDS, acute respiratory distress syndrome; CI, confidence interval; CXR, chest radiograph; P:F, Pao₂:FiO₂; LA, left atrial; PAWP, pulmonary artery wedge pressure.

Fig 4

Categories of exclusion criteria reported in ARDS RCTs. ARDS, acute respiratory distress syndrome; HIV/AIDS, human immunodeficiency virus/acquired immunodeficiency syndrome.