Effect of surfactant administration on outcomes of adult patients in acute respiratory distress syndrome: a meta-analysis of randomized controlled trials

Databases (PubMed,EMBASE, Medline, Cochrane database, Elsevier, Web of Science and ClinicalTrials.​gov) were searched until December 2017. Searches strategies were used with medical key words:<‘adult respiratory distress syndrome’, ‘acute respiratory distress syndrome’, or ‘ARDS’>; <‘pulmonary surfactant’ or‘lung surfactant’>; and < ‘adult’>. We conducted manual searching techniques to identify appropriate studies and applied no language restrictions. Randomized controlled clinical trials using adult participants (older than 18 years) were included in this meta analysis.

Two reviewers (S-S.M., W.C.) independently screened and extracted titles, abstracts, and citations to evaluate each study and any disagreements were resolved by third reviewer (F-M.G.). The investigators selected and determined the enrolled studies depending on the inclusion and exclusion criteria.

Trials with following features were included: 1) Type of study: Randomized controlled clinical trials; 2) Population: Adult patients (older than 18 years) who were diagnosed with acute respiratory distress syndrome; 3) Intervention: pulmonary surfactant administration; 4) Control: ARDS standard treatment; 5) The following outcomes were included. a) Primary outcomes: mortality at short term (7–10-day), mid-term (28–30-day) and long term (90–180-day); b) Secondary outcomes: PaO₂/FiO₂ (mmHg).

Exclusion criteria were as follows: 1) The age of participants were lower than 18 years old; 2) Trial with insufficient information; 3) The study was a review, case report, letter, or other type of publication and animal trial; 4) The study was non-randomized controlled trial; 5) the study did not include mortality or PaO₂/FiO₂ data; and 6) the full text was unavailable.

According to the Cochrane Handbook, random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting were assessed to research the internal validity of included trials.

Trial sequential analysis (TSA; TSA software version 0.9 Beta; Copenhagen Trial Unit, Copenhagen, Denmark) was applied to help to clarify whether additional trials are needed in the cumulative meta-analysis. TSA also controls the risks of type I and type II errors for meta-analysis [6, 7].

Grading of Recommendations Assessment, Development, and Evaluation methodology (GRADE) pro Guideline Development Tool were conducted to evaluate design, quality, consistency, precision, directness and possible publication bias of the included trials. GRADE was assessed in three levels (high, moderate, low, and very low).

We conducted a meta-analysis on the effect of pulmonary surfactant to adult ARDS patients using the methods recommended by the Cochrane Collaboration software RevMan 5.3 (The Nordic Cochrane Centre, Rigshospitalet, Copenhagen, Denmark). Risk ratio (RR) was reported with 95% confidence interval (CI) for the dichotomous data and weighted mean differences(MD) with 95% CIs for the continuous data. A Z-test was performed to statistically evaluate the treatment effects in different groups (13). We measured and quantified the statistical heterogeneity and inconsistency by the Mantel-Haenszel (M-H) chi-square test and the I² test in RevMan 5.3 [8]. The statistically significant heterogeneity was evaluated as p < 0.10 with the M-H chi-square test. In addition, we assess I² index as heterogeneity. Higgins and colleagues proposed 25, 50 and 75% of I² values would indicate low, medium and high heterogeneity, respectively [8]. A fixed-effect model was used unless there was significant heterogeneity, in which case we applied a random effects model. In cases of obvious heterogeneity (p < 0.10 with M-H test; I² > 50%), the meta-analysis employed the random-effects model; otherwise, the meta-analysis used the fixed-effects model.

A subgroup meta-analysis was performed to determine the effect of surfactant administration on outcomes of acute respiratory distress syndrome patients. The primary outcome of the surfactant effect was selected as mortality. Mortality of ARDS patients were classified into short term mortality (7–10-day), mid-term mortality(28–30-day) and long term mortality (90–180-day). Thus, we performed three subgroups meta-analysis of different terms of ARDS mortality. Acute physiology and chronic health evaluation II (APACHE II) is positive correlation of illness severity. The patients of mean APACHE II > 15 were regarded as more severe ARDS and also investigated to analysis 28–30-day mortality.

Sensitivity analyses were used to assess the impact of study quality issues on the overall effect estimate and the effect size of all identified trials when neglecting heterogeneity and publication status. Sensitivity analyses were conducted by STATA11.0 (Stata Corporation, College Station, TX, USA). A statistical test for funnel plot asymmetry was used to investigate the publication bias. Egger’s test and Begger ‘s inspection were also used to assess bias of meta-analysis conducted by STATA11.0 (Stata Corporation).

The process of study selection was presented as flow diagram in Fig. 1. We initially identified 1762 papers and excluded 421 duplicates references and 1730 references after screening the titles and abstracts for the terms “surfactant”, “acute respiratory distress syndrome” and “randomized control trial”. We assessed 32 articles for eligibility and excluded 6 non-randomized references, 9 studies without control, 3 reports, 2 inconformity study design references and 2 incomplete data references. Finally, 10 were included in this meta-analysis [9‐18]. The RCT by Spragg 2004 [13] was conducted including results from both a North American trial (NA) and a European–South African trial (ES). The data from the two trials in this manuscript were assessed independently. Thus, 11 RCTs were enrolled in our meta-analysis. Eleven trials included 3038 patients. 1545 ARDS patients who received surfactant administration were regarded as experiment group, whereas control group (only received ARDS general therapy). The baseline characteristics of the included RCTs were shown in Table 2.

Trial sequential analysis was calculated for mortality of ARDS patients and the value of PaO2/FiO2 after surfactant therapy. TSA was calculated with α = 0.05 and β = 0.20 (power 80%) and a required diversity-adjusted information size based on the intervention effect suggested by the included trials using fixed-effects models. The cumulated Z-curve (blue) doesn’t crosses the traditional boundary and trial sequential monitoring boundary, indicating that lack of reliable and conclusive evidence for beneficial effects of pulmonary surfactant for both mortality (Fig. 2a) and PaO2/FiO2 outcome (Fig. 2b). There is insufficient information to assess the effect of surfactant for ARDS patients.

Among the included studies, eleven RCTs reported the mortality and were included in the primary analysis. We detected no evidence of a publication bias after a funnel plot analysis (Additional file 1: Figure S1a). Egger’s test and Begger’s inspection (p > 0.01) also implied no publication bias in mortality. In Fig. 3, there was not statistically insignificant heterogeneity (p = 0.76) and medium heterogeneity (I² = 0%) among all mortality in our meta-analysis. Test for overall effect of mortality between surfactant group and control group has no differences[RR(95%CI) =1.02(0.93–1.12), p = 0.65]. Moreover, a subgroup analysis showed that 7–10-day, 28–30-day and 90–180-day mortality between surfactant group and control group also showed no statistical significance. In analysis of 28–30-day mortality, test for overall effect of 28–30-day mortality (APACHE II > 15) between surfactant group and control group has also no differences [RR(95%CI) =1.02(0.88–1.18), p = 0.77](Fig. 4). Three studies included 7–10-mortality (RR(95%CI)=)0.89(0.54–1.49), p = 0.66), nine studies included 28–30-mortality(RR(95%CI) =1.00(0.89–1.12), p = 0.98) and two studies included 90–180-day mortality(RR(95%CI) =1.11(0.94–1.32), p = 0.22) (Fig. 3). Sensitive analysis of comparison between surfactant and placebo group showed the result is stable (Additional file 1: Figure S2). Overall, we concluded that surfactant administration could not improve mortality of adult acute respiratory distress syndrome patients.

We further made the meta-analysis of the result of PaO2/FiO2 ratio and three RCTs were included. In Fig. 5, there was not statistically insignificant heterogeneity (p = 0.3) and medium heterogeneity (I² = 16%) among PaO2/FiO2 ratio. Test for overall effect of PaO2/FiO2 ratio between surfactant group and control group had no obvious differences[MD(95%CI) =0.06(− 0.12–0.24), p = 0.5]. Taken together, these suggested that surfactant administration could not improve PaO2/FiO2 ratio of adult acute respiratory distress syndrome patients.

We assessed each enrolled RCT by the mode of randomization, allocation concealment, level of blinding, incomplete outcome data, selective reporting and other bias (Fig. 6).

RCTs are often rated high on the GRADE scale. Variable risks of bias in all the trials lead us to downgrade the quality of the evidence. Allocation concealment was not reported totally, and the sample sizes were all small. Our application of GRADE methodology led us to conclude that the accumulated evidence is of low quality for mortality and PaO2/FiO2 ratio. For a GRADE profile see Additional file 1: Table S1.