Research
Original Investigation
Noninvasive Ventilation With vs Without Early Surfactant to Prevent Chronic Lung Disease in Preterm Infants A Systematic Review and Meta-analysis Tetsuya Isayama, MD; Chatree Chai-Adisaksopha, MD; Sarah D. McDonald, MD, MSc Editorial IMPORTANCE Controversy exists regarding which of the 2 major strategies currently used to
prevent chronic lung disease (CLD) in preterm infants is optimal: noninvasive continuous positive airway pressure (NCPAP) or intubate-surfactant-extubate (INSURE). Preterm infants often require surfactant administration because of respiratory distress syndrome.
Supplemental content at jamapediatrics.com
OBJECTIVE To evaluate whether early INSURE or NCPAP alone is more effective in preventing CLD, death, or both. DATA SOURCES We searched the MEDLINE, EMBASE, Cochrane Controlled Trials Register, and Cumulative Index to Nursing and Allied Health Literature databases from their inception to January 2, 2015, along with conference proceedings and trial registrations. STUDY SELECTION Randomized clinical trials that compared early INSURE with NCPAP alone in preterm infants who had never been intubated before the study entry were selected. Among 1761 initially identified articles, 9 trials (1551 infants) were included. DATA EXTRACTION AND SYNTHESIS Duplicate study selection and data extraction were performed. Meta-analysis was conducted using random-effects models with quality-of-evidence assessment according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system. MAIN OUTCOMES AND MEASURES Seven main outcomes were selected a priori to be assessed according to GRADE, including a composite outcome of CLD and/or death, CLD alone, death alone, air leakage, severe intraventricular hemorrhage, neurodevelopmental impairment, and a composite outcome of death and/or neurodevelopmental impairment. RESULTS There were no statistically significant differences between early INSURE and NCPAP alone for all outcomes assessed. However, the relative risk (RR) estimates appeared to favor early INSURE over NCPAP alone, with a 12% RR reduction in CLD and/or death (RR, 0.88; 95% CI, 0.76-1.02; risk difference [RD], −0.04; 95% CI, −0.08 to 0.01; moderate quality of evidence), a 14% decrease in CLD (RR, 0.86; 95% CI, 0.71-1.03; RD, −0.03; 95% CI, −0.06 to 0.01; moderate quality of evidence), and a 50% decrease in air leakage (RR, 0.50; 95% CI, 0.24-1.07; RD, −0.03; 95% CI, −0.06 to 0.00; very low quality of evidence). The sample size was less than the optimal information size. CONCLUSIONS AND RELEVANCE Currently, no evidence suggests that either early INSURE or NCPAP alone is superior to the other. INSURE does not appear to increase CLD and/or death, CLD alone, and air leakage and may reduce these adverse outcomes compared with NCPAP alone. Further adequately powered trials are required.
JAMA Pediatr. doi:10.1001/jamapediatrics.2015.0510 Published online June 8, 2015.
Author Affiliations: Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada (Isayama, Chai-Adisaksopha, McDonald); Department of Newborn and Developmental Paediatrics, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada (Isayama); Department of Obstetrics and Gynecology, McMaster University, Hamilton, Ontario, Canada (McDonald); Department of Radiology, McMaster University, Hamilton, Ontario, Canada (McDonald). Corresponding Author: Tetsuya Isayama, MD, Department of Newborn and Developmental Paediatrics, Sunnybrook Health Science Centre, 2075 Bayview Ave, Toronto, ON M4N 3M5, Canada ([email protected]).
(Reprinted) E1
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
Research Original Investigation
Chronic Lung Disease Prevention in Preterm Infants
C
hronic lung disease (CLD), also called bronchopulmonary dysplasia, is one of the most important morbidities in preterm infants, characterized by a prolonged need for supplemental oxygen and/or respiratory support.1 Infants with CLD have higher risks of postdischarge mortality and respiratory morbidity along with neurodevelopmental impairments later in life.2 Approximately 30% of very low-birth-weight infants develop CLD,3 and the rate reaches 40% for those born at 28 weeks’ gestation or less. Regardless of the improvement in survival of very low-birth-weight infants in the last 2 decades, the CLD rate has not decreased3 or even increased.4 Although CLD is a multifactorial disease, prematurity and ventilator-induced lung injuries from volutrauma, barotrauma, and oxygen toxicity accompanied with prolonged mechanical ventilations are a major cause of CLD.1 Approximately two-thirds of preterm infants born at less than 33 weeks’ gestations have respiratory distress syndrome5 shortly after birth and often require intubations for surfactant administration followed by ventilatory support.6 Several previous systematic reviews7-9 of randomized clinical trials of preterm infants found that the early use of noninvasive continuous positive airway pressure (NCPAP) to avoid mechanical ventilation decreased CLD, death, or both compared with the respiratory management using routine intubation. Accordingly, clinical practice guidelines from the European Association of Perinatal Medicine6 and the American Academy of Pediatrics10 added a recommendation for the early use of NCPAP, avoiding intubation, as an alternative to a routine or early surfactant administration recommended in their previous guidelines.11,12 However, one major disadvantage of this NCPAP strategy without intubation is a lack or delay of the administration of surfactant that is generally given via an endotracheal tube after intubation.13 Early surfactant compared with delayed administration for respiratory distress syndrome is effective in preventing CLD14 and recommended.6 Therefore, physicians who treat preterm infants with or at high risk of respiratory distress syndrome currently have a difficult dilemma: avoiding intubation using NCPAP alone without early surfactant administration or intubate to administer surfactant. Given this clinical dilemma, there has been increasing interest in an intermediate strategy called intubate-surfactantextubate (INSURE), in which infants are intubated for surfactant administration and immediately extubated to NCPAP. Early INSURE is a promising strategy because it enables early administration of surfactant while avoiding prolonged mechanical ventilation that can lead to CLD.13 However, it is not clear whether early INSURE is superior to NCPAP. Therefore, this systematic review and meta-analysis of randomized clinical trials aimed to examine whether early INSURE, compared with NCPAP alone, is more effective in preventing death, CLD, or both in preterm infants with or at high risk of respiratory distress syndrome.
Methods
• This systematic review aimed to compare the effectiveness of the 2 major strategies currently used to prevent chronic lung disease (CLD) of preterm infants: intubate-surfactant-extubate (INSURE) and noninvasive continuous positive airway pressure (NCPAP). • Although there were no statistically significant differences, the relative risk (RR) estimates appeared to favor early INSURE over NCPAP alone for CLD and/or death, CLD alone, and air leakage, with RRs (95% CIs) of 0.88 (0.76-1.02), 0.86 (0.71-1.03), and 0.50 (0.24-1.07), respectively. • There is no evidence suggesting that either early INSURE or NCPAP alone is superior to the other; however, INSURE at least does not appear to increase CLD and/or death, CLD alone, and air leakage and may reduce these adverse outcomes compared with NCPAP alone.
Cochrane Handbook for Systematic Reviews of Interventions15 and the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system.16
Criteria for Eligible Studies for This Systematic Review All published and unpublished randomized clinical trials were included with no language restrictions. This systematic review included studies that compared early INSURE vs NCPAP alone for preterm infants born at less than 37 weeks’ gestational age with or at high risk of respiratory distress syndrome who had never been intubated before the study entry. Infants in the early INSURE group were intubated, given surfactant, and extubated to NCPAP within 1 hour after intubation. Infants in the NCPAP alone group continued to receive NCPAP initially and, when NCPAP was not tolerated, rescued by intubation followed by mechanical ventilation or INSURE. This systematic review excluded the following: studies using surfactant administration methods without intubation, such as nebulized surfactant17 or instillation via a thin catheter inserted directly into the trachea or a laryngeal mask airway18; studies in which infants in both the early INSURE and NCPAP alone groups were not routinely administered NCPAP; and duplicated studies or data and studies without sufficient data regarding the outcomes to be summarized. Seven important patient outcomes were selected a priori to be assessed in this systematic review, including (1) a composite outcome of CLD (the most important respiratory outcome of preterm infants, which was defined as is typical as oxygen use and/or reparatory support at 36 weeks’ postmenstrual age) and/or death at discharge, (2) CLD alone, (3) death at discharge alone, (4) air leakage (pneumothorax and/or pulmonary interstitial emphysema), (5) severe intraventricular hemorrhage (grade 3 or 4),19 (6) neurodevelopmental impairment (cerebral palsy, cognitive deficit, hearing loss, or blindness) at 18 months or older, and (7) a composite outcome of death and/or neurodevelopmental impairment at 18 months or older.
Search Methods for Identification of Studies
The protocol of this systematic review, written before the literature search, is available in the eAppendix in the Supplement. This systematic review was conducted according to the E2
At a Glance
The literature searches were conducted in MEDLINE (1946 to January 2, 2015), EMBASE (1980 to January 2, 2015), Cochrane Central Register of Controlled Trials (CENTRAL, January
JAMA Pediatrics Published online June 8, 2015 (Reprinted)
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
jamapediatrics.com
Chronic Lung Disease Prevention in Preterm Infants
2, 2015), and Cumulative Index to Nursing and Allied Health Literature (1991 to January 2, 2015) and through hand searching of references in narrative and systematic reviews. The search strategy used for MEDLINE, which was modified for other databases, is available in eTable 1 in the Supplement. Search terms included 3 concepts: (1) newborns or infants (population), (2) surfactant (intervention), and (3) randomized clinical trials (study design). Abstracts or conference proceedings of the Pediatric Academic Society (2002-2014) and International Workshop on Surfactant Replacement (2006-2014) were searched. Trial registrations, including ClinicalTrials.gov20 and World Health Organization International Clinical Trials Registry Platform,21 were searched to find unpublished or recent completed relevant trials.
Study Selection and Data Extraction All records found in the literature search were screened by titles and abstracts. Potentially relevant records were selected for full-text review. Data were extracted using a data collection form designed for this systematic review. We contacted authors for missing data or clarifications, if needed.
Original Investigation Research
based on a priori hypotheses that the RR of CLD, death, or both in the early INSURE group compared with the NCPAP group would be lower in the following subgroups: groups that included very premature infants (gestational age of 40% vs ≤40%). The statistical test for the reporting bias was planned for outcomes with at least 10 included studies.15 Missing data were assessed for each outcome in each study, and complete-case analyses were used in this systematic review. Four sensitivity analyses were conducted to evaluate the robustness of the study results: (1) using random-effects models with Knapp-Hartung adjustment (using R, version 3.1.2, with the metaphor package),22,23 (2) using fixedeffects methods (Mantel-Haenszel test), (3) excluding studies with high risk of biases, or (4) using a broader definition of mortality, including death within 7 or 28 days.
Assessment of Quality of Evidence and Summary Tables Risk of Bias Assessment in Included Studies The risk of bias was assessed for each outcome in all included studies using the Cochrane Systematic Review Handbook.15 The risk of bias was evaluated for random sequence generation, allocation concealment, masking of participants and personnel (performance bias), masking of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), and other risk of bias.15 The study selection, data extraction, and risk of bias assessment were independently conducted by 2 authors (T.I. and C.C.-A.). The agreements between these 2 reviewers for the study selection and risk of bias assessment were evaluated using a weighted κ statistic and the percentage of overall agreement. The disagreements between the 2 reviewers were resolved through discussion, with a third reviewer available for adjudication if needed (S.D.M.).
Data Synthesis and Analysis To incorporate heterogeneity among studies and potentially yield more conservative results, primary meta-analyses were conducted with Mantel-Haenszel random-effects methods as planned a priori using Review Manager (RevMan), version 5.1 (The Nordic Cochrane Centre, The Cochrane Collaboration).15 The pooled results were summarized as relative risks (RRs), risk differences (RDs), and 95% CIs along with forest plots for each outcome. A 2-sided P < .05 was used to determine statistical significance, and as is typical of systematic reviews, there was no adjustment for multiple outcomes. Heterogeneity among the included studies was assessed by examination of forest plots, the I2 statistic, and χ2 tests for heterogeneity. Three preplanned subgroup analyses were conducted to explore the potential sources of heterogeneities stratified by (1) mean (or median) gestational age of infants (1 hour). These subgroup analyses were jamapediatrics.com
The quality of evidence was assessed and rated using 4 levels (high, moderate, low, or very low) for each outcome across studies and summarized as an evidence profile table according to GRADE.24 The quality of evidence of randomized clinical trials was preliminarily evaluated as high-quality evidence and was downgraded for a risk of bias, imprecision, inconsistency, indirectness, and publication bias for each outcome.25 For the assessment of imprecision, the sample size required to detect a 20% RR reduction, called optimal information sizes, was calculated using total event rates in the control groups of included studies.26
Results The searches of the electronic databases and gray literature identified 1761 nonduplicate records, among which 62 articles were selected for full-text review (Figure 1) and 9 trials met the inclusion criteria,27-35 with a total of 1551 preterm infants. The agreement of the study selection between the 2 reviewers was excellent, with a κ of 0.828 and a raw agreement of 95.2%. All included studies were randomized clinical trials with 2 parallel groups except for one study,28 which examined 3 parallel groups, from which only 2 groups (early INSURE and NCPAP alone) were included in this systematic review. Authors of 4 included studies28-31,34 provided additional study information related to missing data and outcome definitions for this systematic review. There were variations in the study maternal or infant characteristics (eg, antenatal corticosteroid rate varied from 50% to 99% and gestational age at birth ranged from 25 to 35 weeks), timing of the interventions (from shortly after birth to 72 hours after birth), and back-up measures for NCPAP failure (Table 1). The rates of successful extubation within 1 hour of intubation in the early INSURE group were more than 90% (range, 90.5%-100%) except for one study28 with a slightly lower rate (83.3%). Failure rates (intubation rates) (Reprinted) JAMA Pediatrics Published online June 8, 2015
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
E3
Research Original Investigation
Chronic Lung Disease Prevention in Preterm Infants
Figure 1. Flow Diagram of the Literature Search 2963 Records identified in searching databases (MEDLINE, EMBASE, CENTRAL, CINAHL)
8 Gray literature identified from other sources (eg,conference proceedings and trial registrations)
1761 Records after duplicates removed 1699 Records excluded in title and abstract screening 62 Full-text articles assessed for eligibility 53 Full-text articles excluded 6 Duplicates 4 Not RCTs 43 Not relevant 9 Studies included in qualitative synthesis 9 Studies included in quantitative synthesis (meta-analysis)
CENTRAL indicates Cochrane Central Register of Controlled Trials; CINAHL, Cumulative Index to Nursing and Allied Health Literature; RCT, randomized clinical trial.
in the NCPAP group varied among included studies (15%84.8%) but were mostly high (>40% in 7 of 9 trials). The rates of infants who required reintubation after the successful INSURE in the early INSURE group were low (10%-26%) except for in the trials by Verder et al34 (39%), Dunn et al28 (42%), and Reininger et al31 (50%). The incidence of the outcomes and the number of missing data in included studies are given in eTable 2 in the Supplement.
Assessment of Risk of Bias None of the included studies used masking of the interventions except for one trial.31 The lack of masking of health care professionals could have affected clinical care and was considered an unclear risk of bias for performance bias. Because all the outcomes were objective, the detection bias attributable to the lack of masking was considered low risk for most of the outcomes except for CLD, whose diagnoses might be affected by the different criteria of oxygen administration among physicians.36 The data for CLD in the study by Dilmen et al27 and for severe intraventricular hemorrhage in the study by Verder et al35 were judged as being at high risk of attrition bias due to missing data (>10%). Two studies34,35 were stopped early for significant findings that may overestimate intervention effects37 and, hence, were considered at high risk of bias. Two other studies28,31 were stopped early because of slow patient enrollment and were considered at low risk of bias. The consensus of the risk of bias assessment in the included studies between the 2 reviewers (T.I. and C.C.-A.) is reported in eTable 3 in the Supplement. The overall κ and raw agreements with quadratic weighting for the risk of bias between the 2 authors (T.I. and C.C.-A.) were 0.71 and 0.85, respectively. E4
Effects of Interventions Although no statistically significant differences were detected, the RR estimates and absolute rates appeared to favor early INSURE over NCPAP alone (Figure 2). There was a 12% reduction in RR estimate of CLD and/or death (RR, 0.88; 95% CI, 0.76-1.02; P = .10; I2 = 0%; RD, −0.04; 95% CI, −0.08 to 0.01; 6 trials with 1250 infants), 14% decrease in CLD (RR, 0.86; 95% CI, 0.71-1.03; P = .10; I2 = 0%; RD, −0.03; 95% CI, −0.06 to 0.01; 6 trials with 1128 infants), and 50% decrease in air leakage (RR, 0.50; 95% CI, 0.24-1.07; P = .07; I2 = 28%; RD, −0.03; 95% CI, −0.06 to 0.00; 9 trials with 1547 infants). The 95% CIs of the RR covered widely more than 1.00 for death (RR, 0.94; 95% CI, 0.67-1.32; P = .72; I2 = 0%; RD, 0.01; 95% CI, −0.02 to 0.03; 7 trials with 1396 infants) and severe intraventricular hemorrhage (RR, 0.79; 95% CI, 0.45-1.39; P = .42; I2 = 0%; RD, −0.00; 95% CI, −0.02 to 0.01; 7 trials with 1325 infants). No study reported neurodevelopmental impairment at 18 months or older. Three predefined subgroup analyses, based on the mean or median gestational ages, the NCPAP backup measures, and the timing of interventions, and a post hoc subgroup analysis stratified by the fraction of inspired oxygen found no significant differences among the subgroups for any of the 5 outcomes, with P values ranging from .13 to .98 (eFigure 1 and eFigure 2 in the Supplement). The sensitivity analyses found results similar to the primary analyses except for the significant result for CLD in the first sensitivity analysis with KnappHartung adjustment (RR, 0.86; 95% CI, 0.74-0.99) and the wide variation in 95% CIs for the RRs for air leakage (eTable 4 in the Supplement).
Quality of the Evidence The quality of evidence was evaluated across included studies for each outcome and was summarized in a GRADE evidence profile table (Table 2).24,38 The risk of bias was considered not serious for all outcomes except for air leakage because the sensitivity analyses that excluded studies with a high risk of bias found results similar to those in the primary analyses. For air leakage, the risk of bias was considered serious because the sensitivity analyses changed the meta-analysis result. The imprecision was considered serious for all outcomes because the total sample size for each outcome was less than the sample size required for a 20% RR reduction (optimal information size),26 which would be 1440, 2080, 7212, 11 136, and 17 022 infants for CLD and/or death, CLD alone, death alone, air leakage, and severe intraventricular hemorrhage, respectively (eTable 5 in the Supplement). The inconsistency was judged as serious for air leakage because the point estimate of the RR in the study by Sandri et al 33 was different from the other studies (Figure 2D). Publication bias was not detected in the funnel plots for any of the outcomes (eFigure 3 in the Supplement). Statistical tests were not conducted because this systematic review included fewer than 10 studies and would have too low a power to detect funnel plot asymmetry.15 The serious imprecision downgraded the quality of evidence from high to moderate for all outcomes. The serious risk of bias and inconsistency downgraded the quality of evidence further by 2 levels for air leakage from moderate to very low.
JAMA Pediatrics Published online June 8, 2015 (Reprinted)
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
jamapediatrics.com
Chronic Lung Disease Prevention in Preterm Infants
Original Investigation Research
Table 1. Summary of Main Characteristics of Included Studies
Study Years
Maternal Antenatal Timing of Total No. of No. of Enrollment, Corticosteroid Study Use, % Centers Infants h
Source
Countries
Dilmen et al,27 2014
Turkey
Dunn et al,28 2011
United States 2003- 27 Canada 2009
Imani et al,29 2013
Iran
20082010
1
Kandraju et al,30 2013
India
20082011
20092010
6
Infants’ Gestational Age at Birth, wk
Extubation a in 0.5
FiO2 >0.4, apnea, PCO2 >65 mm Hg, pH
Original Investigation
Noninvasive Ventilation With vs Without Early Surfactant to Prevent Chronic Lung Disease in Preterm Infants A Systematic Review and Meta-analysis Tetsuya Isayama, MD; Chatree Chai-Adisaksopha, MD; Sarah D. McDonald, MD, MSc Editorial IMPORTANCE Controversy exists regarding which of the 2 major strategies currently used to
prevent chronic lung disease (CLD) in preterm infants is optimal: noninvasive continuous positive airway pressure (NCPAP) or intubate-surfactant-extubate (INSURE). Preterm infants often require surfactant administration because of respiratory distress syndrome.
Supplemental content at jamapediatrics.com
OBJECTIVE To evaluate whether early INSURE or NCPAP alone is more effective in preventing CLD, death, or both. DATA SOURCES We searched the MEDLINE, EMBASE, Cochrane Controlled Trials Register, and Cumulative Index to Nursing and Allied Health Literature databases from their inception to January 2, 2015, along with conference proceedings and trial registrations. STUDY SELECTION Randomized clinical trials that compared early INSURE with NCPAP alone in preterm infants who had never been intubated before the study entry were selected. Among 1761 initially identified articles, 9 trials (1551 infants) were included. DATA EXTRACTION AND SYNTHESIS Duplicate study selection and data extraction were performed. Meta-analysis was conducted using random-effects models with quality-of-evidence assessment according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system. MAIN OUTCOMES AND MEASURES Seven main outcomes were selected a priori to be assessed according to GRADE, including a composite outcome of CLD and/or death, CLD alone, death alone, air leakage, severe intraventricular hemorrhage, neurodevelopmental impairment, and a composite outcome of death and/or neurodevelopmental impairment. RESULTS There were no statistically significant differences between early INSURE and NCPAP alone for all outcomes assessed. However, the relative risk (RR) estimates appeared to favor early INSURE over NCPAP alone, with a 12% RR reduction in CLD and/or death (RR, 0.88; 95% CI, 0.76-1.02; risk difference [RD], −0.04; 95% CI, −0.08 to 0.01; moderate quality of evidence), a 14% decrease in CLD (RR, 0.86; 95% CI, 0.71-1.03; RD, −0.03; 95% CI, −0.06 to 0.01; moderate quality of evidence), and a 50% decrease in air leakage (RR, 0.50; 95% CI, 0.24-1.07; RD, −0.03; 95% CI, −0.06 to 0.00; very low quality of evidence). The sample size was less than the optimal information size. CONCLUSIONS AND RELEVANCE Currently, no evidence suggests that either early INSURE or NCPAP alone is superior to the other. INSURE does not appear to increase CLD and/or death, CLD alone, and air leakage and may reduce these adverse outcomes compared with NCPAP alone. Further adequately powered trials are required.
JAMA Pediatr. doi:10.1001/jamapediatrics.2015.0510 Published online June 8, 2015.
Author Affiliations: Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada (Isayama, Chai-Adisaksopha, McDonald); Department of Newborn and Developmental Paediatrics, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada (Isayama); Department of Obstetrics and Gynecology, McMaster University, Hamilton, Ontario, Canada (McDonald); Department of Radiology, McMaster University, Hamilton, Ontario, Canada (McDonald). Corresponding Author: Tetsuya Isayama, MD, Department of Newborn and Developmental Paediatrics, Sunnybrook Health Science Centre, 2075 Bayview Ave, Toronto, ON M4N 3M5, Canada ([email protected]).
(Reprinted) E1
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
Research Original Investigation
Chronic Lung Disease Prevention in Preterm Infants
C
hronic lung disease (CLD), also called bronchopulmonary dysplasia, is one of the most important morbidities in preterm infants, characterized by a prolonged need for supplemental oxygen and/or respiratory support.1 Infants with CLD have higher risks of postdischarge mortality and respiratory morbidity along with neurodevelopmental impairments later in life.2 Approximately 30% of very low-birth-weight infants develop CLD,3 and the rate reaches 40% for those born at 28 weeks’ gestation or less. Regardless of the improvement in survival of very low-birth-weight infants in the last 2 decades, the CLD rate has not decreased3 or even increased.4 Although CLD is a multifactorial disease, prematurity and ventilator-induced lung injuries from volutrauma, barotrauma, and oxygen toxicity accompanied with prolonged mechanical ventilations are a major cause of CLD.1 Approximately two-thirds of preterm infants born at less than 33 weeks’ gestations have respiratory distress syndrome5 shortly after birth and often require intubations for surfactant administration followed by ventilatory support.6 Several previous systematic reviews7-9 of randomized clinical trials of preterm infants found that the early use of noninvasive continuous positive airway pressure (NCPAP) to avoid mechanical ventilation decreased CLD, death, or both compared with the respiratory management using routine intubation. Accordingly, clinical practice guidelines from the European Association of Perinatal Medicine6 and the American Academy of Pediatrics10 added a recommendation for the early use of NCPAP, avoiding intubation, as an alternative to a routine or early surfactant administration recommended in their previous guidelines.11,12 However, one major disadvantage of this NCPAP strategy without intubation is a lack or delay of the administration of surfactant that is generally given via an endotracheal tube after intubation.13 Early surfactant compared with delayed administration for respiratory distress syndrome is effective in preventing CLD14 and recommended.6 Therefore, physicians who treat preterm infants with or at high risk of respiratory distress syndrome currently have a difficult dilemma: avoiding intubation using NCPAP alone without early surfactant administration or intubate to administer surfactant. Given this clinical dilemma, there has been increasing interest in an intermediate strategy called intubate-surfactantextubate (INSURE), in which infants are intubated for surfactant administration and immediately extubated to NCPAP. Early INSURE is a promising strategy because it enables early administration of surfactant while avoiding prolonged mechanical ventilation that can lead to CLD.13 However, it is not clear whether early INSURE is superior to NCPAP. Therefore, this systematic review and meta-analysis of randomized clinical trials aimed to examine whether early INSURE, compared with NCPAP alone, is more effective in preventing death, CLD, or both in preterm infants with or at high risk of respiratory distress syndrome.
Methods
• This systematic review aimed to compare the effectiveness of the 2 major strategies currently used to prevent chronic lung disease (CLD) of preterm infants: intubate-surfactant-extubate (INSURE) and noninvasive continuous positive airway pressure (NCPAP). • Although there were no statistically significant differences, the relative risk (RR) estimates appeared to favor early INSURE over NCPAP alone for CLD and/or death, CLD alone, and air leakage, with RRs (95% CIs) of 0.88 (0.76-1.02), 0.86 (0.71-1.03), and 0.50 (0.24-1.07), respectively. • There is no evidence suggesting that either early INSURE or NCPAP alone is superior to the other; however, INSURE at least does not appear to increase CLD and/or death, CLD alone, and air leakage and may reduce these adverse outcomes compared with NCPAP alone.
Cochrane Handbook for Systematic Reviews of Interventions15 and the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system.16
Criteria for Eligible Studies for This Systematic Review All published and unpublished randomized clinical trials were included with no language restrictions. This systematic review included studies that compared early INSURE vs NCPAP alone for preterm infants born at less than 37 weeks’ gestational age with or at high risk of respiratory distress syndrome who had never been intubated before the study entry. Infants in the early INSURE group were intubated, given surfactant, and extubated to NCPAP within 1 hour after intubation. Infants in the NCPAP alone group continued to receive NCPAP initially and, when NCPAP was not tolerated, rescued by intubation followed by mechanical ventilation or INSURE. This systematic review excluded the following: studies using surfactant administration methods without intubation, such as nebulized surfactant17 or instillation via a thin catheter inserted directly into the trachea or a laryngeal mask airway18; studies in which infants in both the early INSURE and NCPAP alone groups were not routinely administered NCPAP; and duplicated studies or data and studies without sufficient data regarding the outcomes to be summarized. Seven important patient outcomes were selected a priori to be assessed in this systematic review, including (1) a composite outcome of CLD (the most important respiratory outcome of preterm infants, which was defined as is typical as oxygen use and/or reparatory support at 36 weeks’ postmenstrual age) and/or death at discharge, (2) CLD alone, (3) death at discharge alone, (4) air leakage (pneumothorax and/or pulmonary interstitial emphysema), (5) severe intraventricular hemorrhage (grade 3 or 4),19 (6) neurodevelopmental impairment (cerebral palsy, cognitive deficit, hearing loss, or blindness) at 18 months or older, and (7) a composite outcome of death and/or neurodevelopmental impairment at 18 months or older.
Search Methods for Identification of Studies
The protocol of this systematic review, written before the literature search, is available in the eAppendix in the Supplement. This systematic review was conducted according to the E2
At a Glance
The literature searches were conducted in MEDLINE (1946 to January 2, 2015), EMBASE (1980 to January 2, 2015), Cochrane Central Register of Controlled Trials (CENTRAL, January
JAMA Pediatrics Published online June 8, 2015 (Reprinted)
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
jamapediatrics.com
Chronic Lung Disease Prevention in Preterm Infants
2, 2015), and Cumulative Index to Nursing and Allied Health Literature (1991 to January 2, 2015) and through hand searching of references in narrative and systematic reviews. The search strategy used for MEDLINE, which was modified for other databases, is available in eTable 1 in the Supplement. Search terms included 3 concepts: (1) newborns or infants (population), (2) surfactant (intervention), and (3) randomized clinical trials (study design). Abstracts or conference proceedings of the Pediatric Academic Society (2002-2014) and International Workshop on Surfactant Replacement (2006-2014) were searched. Trial registrations, including ClinicalTrials.gov20 and World Health Organization International Clinical Trials Registry Platform,21 were searched to find unpublished or recent completed relevant trials.
Study Selection and Data Extraction All records found in the literature search were screened by titles and abstracts. Potentially relevant records were selected for full-text review. Data were extracted using a data collection form designed for this systematic review. We contacted authors for missing data or clarifications, if needed.
Original Investigation Research
based on a priori hypotheses that the RR of CLD, death, or both in the early INSURE group compared with the NCPAP group would be lower in the following subgroups: groups that included very premature infants (gestational age of 40% vs ≤40%). The statistical test for the reporting bias was planned for outcomes with at least 10 included studies.15 Missing data were assessed for each outcome in each study, and complete-case analyses were used in this systematic review. Four sensitivity analyses were conducted to evaluate the robustness of the study results: (1) using random-effects models with Knapp-Hartung adjustment (using R, version 3.1.2, with the metaphor package),22,23 (2) using fixedeffects methods (Mantel-Haenszel test), (3) excluding studies with high risk of biases, or (4) using a broader definition of mortality, including death within 7 or 28 days.
Assessment of Quality of Evidence and Summary Tables Risk of Bias Assessment in Included Studies The risk of bias was assessed for each outcome in all included studies using the Cochrane Systematic Review Handbook.15 The risk of bias was evaluated for random sequence generation, allocation concealment, masking of participants and personnel (performance bias), masking of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), and other risk of bias.15 The study selection, data extraction, and risk of bias assessment were independently conducted by 2 authors (T.I. and C.C.-A.). The agreements between these 2 reviewers for the study selection and risk of bias assessment were evaluated using a weighted κ statistic and the percentage of overall agreement. The disagreements between the 2 reviewers were resolved through discussion, with a third reviewer available for adjudication if needed (S.D.M.).
Data Synthesis and Analysis To incorporate heterogeneity among studies and potentially yield more conservative results, primary meta-analyses were conducted with Mantel-Haenszel random-effects methods as planned a priori using Review Manager (RevMan), version 5.1 (The Nordic Cochrane Centre, The Cochrane Collaboration).15 The pooled results were summarized as relative risks (RRs), risk differences (RDs), and 95% CIs along with forest plots for each outcome. A 2-sided P < .05 was used to determine statistical significance, and as is typical of systematic reviews, there was no adjustment for multiple outcomes. Heterogeneity among the included studies was assessed by examination of forest plots, the I2 statistic, and χ2 tests for heterogeneity. Three preplanned subgroup analyses were conducted to explore the potential sources of heterogeneities stratified by (1) mean (or median) gestational age of infants (1 hour). These subgroup analyses were jamapediatrics.com
The quality of evidence was assessed and rated using 4 levels (high, moderate, low, or very low) for each outcome across studies and summarized as an evidence profile table according to GRADE.24 The quality of evidence of randomized clinical trials was preliminarily evaluated as high-quality evidence and was downgraded for a risk of bias, imprecision, inconsistency, indirectness, and publication bias for each outcome.25 For the assessment of imprecision, the sample size required to detect a 20% RR reduction, called optimal information sizes, was calculated using total event rates in the control groups of included studies.26
Results The searches of the electronic databases and gray literature identified 1761 nonduplicate records, among which 62 articles were selected for full-text review (Figure 1) and 9 trials met the inclusion criteria,27-35 with a total of 1551 preterm infants. The agreement of the study selection between the 2 reviewers was excellent, with a κ of 0.828 and a raw agreement of 95.2%. All included studies were randomized clinical trials with 2 parallel groups except for one study,28 which examined 3 parallel groups, from which only 2 groups (early INSURE and NCPAP alone) were included in this systematic review. Authors of 4 included studies28-31,34 provided additional study information related to missing data and outcome definitions for this systematic review. There were variations in the study maternal or infant characteristics (eg, antenatal corticosteroid rate varied from 50% to 99% and gestational age at birth ranged from 25 to 35 weeks), timing of the interventions (from shortly after birth to 72 hours after birth), and back-up measures for NCPAP failure (Table 1). The rates of successful extubation within 1 hour of intubation in the early INSURE group were more than 90% (range, 90.5%-100%) except for one study28 with a slightly lower rate (83.3%). Failure rates (intubation rates) (Reprinted) JAMA Pediatrics Published online June 8, 2015
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
E3
Research Original Investigation
Chronic Lung Disease Prevention in Preterm Infants
Figure 1. Flow Diagram of the Literature Search 2963 Records identified in searching databases (MEDLINE, EMBASE, CENTRAL, CINAHL)
8 Gray literature identified from other sources (eg,conference proceedings and trial registrations)
1761 Records after duplicates removed 1699 Records excluded in title and abstract screening 62 Full-text articles assessed for eligibility 53 Full-text articles excluded 6 Duplicates 4 Not RCTs 43 Not relevant 9 Studies included in qualitative synthesis 9 Studies included in quantitative synthesis (meta-analysis)
CENTRAL indicates Cochrane Central Register of Controlled Trials; CINAHL, Cumulative Index to Nursing and Allied Health Literature; RCT, randomized clinical trial.
in the NCPAP group varied among included studies (15%84.8%) but were mostly high (>40% in 7 of 9 trials). The rates of infants who required reintubation after the successful INSURE in the early INSURE group were low (10%-26%) except for in the trials by Verder et al34 (39%), Dunn et al28 (42%), and Reininger et al31 (50%). The incidence of the outcomes and the number of missing data in included studies are given in eTable 2 in the Supplement.
Assessment of Risk of Bias None of the included studies used masking of the interventions except for one trial.31 The lack of masking of health care professionals could have affected clinical care and was considered an unclear risk of bias for performance bias. Because all the outcomes were objective, the detection bias attributable to the lack of masking was considered low risk for most of the outcomes except for CLD, whose diagnoses might be affected by the different criteria of oxygen administration among physicians.36 The data for CLD in the study by Dilmen et al27 and for severe intraventricular hemorrhage in the study by Verder et al35 were judged as being at high risk of attrition bias due to missing data (>10%). Two studies34,35 were stopped early for significant findings that may overestimate intervention effects37 and, hence, were considered at high risk of bias. Two other studies28,31 were stopped early because of slow patient enrollment and were considered at low risk of bias. The consensus of the risk of bias assessment in the included studies between the 2 reviewers (T.I. and C.C.-A.) is reported in eTable 3 in the Supplement. The overall κ and raw agreements with quadratic weighting for the risk of bias between the 2 authors (T.I. and C.C.-A.) were 0.71 and 0.85, respectively. E4
Effects of Interventions Although no statistically significant differences were detected, the RR estimates and absolute rates appeared to favor early INSURE over NCPAP alone (Figure 2). There was a 12% reduction in RR estimate of CLD and/or death (RR, 0.88; 95% CI, 0.76-1.02; P = .10; I2 = 0%; RD, −0.04; 95% CI, −0.08 to 0.01; 6 trials with 1250 infants), 14% decrease in CLD (RR, 0.86; 95% CI, 0.71-1.03; P = .10; I2 = 0%; RD, −0.03; 95% CI, −0.06 to 0.01; 6 trials with 1128 infants), and 50% decrease in air leakage (RR, 0.50; 95% CI, 0.24-1.07; P = .07; I2 = 28%; RD, −0.03; 95% CI, −0.06 to 0.00; 9 trials with 1547 infants). The 95% CIs of the RR covered widely more than 1.00 for death (RR, 0.94; 95% CI, 0.67-1.32; P = .72; I2 = 0%; RD, 0.01; 95% CI, −0.02 to 0.03; 7 trials with 1396 infants) and severe intraventricular hemorrhage (RR, 0.79; 95% CI, 0.45-1.39; P = .42; I2 = 0%; RD, −0.00; 95% CI, −0.02 to 0.01; 7 trials with 1325 infants). No study reported neurodevelopmental impairment at 18 months or older. Three predefined subgroup analyses, based on the mean or median gestational ages, the NCPAP backup measures, and the timing of interventions, and a post hoc subgroup analysis stratified by the fraction of inspired oxygen found no significant differences among the subgroups for any of the 5 outcomes, with P values ranging from .13 to .98 (eFigure 1 and eFigure 2 in the Supplement). The sensitivity analyses found results similar to the primary analyses except for the significant result for CLD in the first sensitivity analysis with KnappHartung adjustment (RR, 0.86; 95% CI, 0.74-0.99) and the wide variation in 95% CIs for the RRs for air leakage (eTable 4 in the Supplement).
Quality of the Evidence The quality of evidence was evaluated across included studies for each outcome and was summarized in a GRADE evidence profile table (Table 2).24,38 The risk of bias was considered not serious for all outcomes except for air leakage because the sensitivity analyses that excluded studies with a high risk of bias found results similar to those in the primary analyses. For air leakage, the risk of bias was considered serious because the sensitivity analyses changed the meta-analysis result. The imprecision was considered serious for all outcomes because the total sample size for each outcome was less than the sample size required for a 20% RR reduction (optimal information size),26 which would be 1440, 2080, 7212, 11 136, and 17 022 infants for CLD and/or death, CLD alone, death alone, air leakage, and severe intraventricular hemorrhage, respectively (eTable 5 in the Supplement). The inconsistency was judged as serious for air leakage because the point estimate of the RR in the study by Sandri et al 33 was different from the other studies (Figure 2D). Publication bias was not detected in the funnel plots for any of the outcomes (eFigure 3 in the Supplement). Statistical tests were not conducted because this systematic review included fewer than 10 studies and would have too low a power to detect funnel plot asymmetry.15 The serious imprecision downgraded the quality of evidence from high to moderate for all outcomes. The serious risk of bias and inconsistency downgraded the quality of evidence further by 2 levels for air leakage from moderate to very low.
JAMA Pediatrics Published online June 8, 2015 (Reprinted)
Copyright 2015 American Medical Association. All rights reserved.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Otago Library User on 07/01/2015
jamapediatrics.com
Chronic Lung Disease Prevention in Preterm Infants
Original Investigation Research
Table 1. Summary of Main Characteristics of Included Studies
Study Years
Maternal Antenatal Timing of Total No. of No. of Enrollment, Corticosteroid Study Use, % Centers Infants h
Source
Countries
Dilmen et al,27 2014
Turkey
Dunn et al,28 2011
United States 2003- 27 Canada 2009
Imani et al,29 2013
Iran
20082010
1
Kandraju et al,30 2013
India
20082011
20092010
6
Infants’ Gestational Age at Birth, wk
Extubation a in 0.5
FiO2 >0.4, apnea, PCO2 >65 mm Hg, pH
