SE
M I N A R S I N
P
E R I N A T O L O G Y
37 (2013) 389–400
Available online at www.sciencedirect.com
www.elsevier.com/locate/semperi
Survival of pre-viable preterm infants in the United States: A systematic review and meta-analysis Hamisu M. Salihu, MD, PhDa,b,n, Abraham A. Salinas-Miranda, MD, MPHa,c, Latoya Hill, MPHa, Kristen Chandler, MSW, MPHa,c n
Maternal & Child Health Comparative Effectiveness Research Group, Department of Epidemiology & Biostatistics, College of Public Health, University of South Florida, 13201 Bruce B. Downs Blvd, MDC 56, Tampa, FL 33612 † Division of Maternal Fetal Medicine, Department of Obstetrics and Gynecology, College of Medicine, University of South Florida, Tampa, FL ‡ Department of Community and Family Health, College of Public Health, University of South Florida, Tampa, FL
AR T IC LE INFO
AB ST R A CT The objective of this paper is to review observational studies that addressed the survival of
Key words:
pre-viable gestations in the United States. We searched PubMed, Ovid, CINAHL, and Web of
Survival
Knowledge for studies reporting survival of infants born at o24 gestational weeks and/or
Pre-viable
o500 g in the United States and published between January 2003 and January 2013. The full
Periviable
texts of 70 articles were examined and a total of 15 studies qualified and were selected. We
Extremely preterm birth
analyzed fixed-effect and random-effects models for eight studies on survival to discharge.
Extremely low birth weight
Pooled survival to discharge in the random-effects model was 45.9% (95% CI: 41.1–51.7) and
Pregnancy
39.7% in the fixed-effect model (95% CI: 38.8–40.7). Studies differed by pre-viable survival measures and epochs (1985–2009). Protective factors included antenatal corticosteroids, neonatal resuscitation, and intensive care. The current survival threshold for pre-viable infants warrants reconsideration of the limits of viability. Protective factors that enhance survival should be considered in the management of these infants. & 2013 Elsevier Inc. All rights reserved.
1.
Introduction
The survival of extremely low gestational age newborns (ELGAN, live births o28 weeks of gestation) and extremely low birth weight infants (ELBW, o1000 g, or 2 lb, 3 oz) has improved considerably over the last three decades in developed countries.1–3 Survival is now possible for infants born
before 24 weeks of gestation and for those weighing less than 500 g, who are currently classified as “pre-viable,” “periviable,” “borderline,” or “non-viable” births.4 Despite the growing population of pre-viable survivors reported in published studies over the last decades, the evidence on the survival of pre-viable infants in the United States continues to be scattered and has not been summarized in a systematic
This work has not been published before. H.S. provided the overall leadership and scientific oversight of project, and edited the manuscript. L.H. conducted database searches and participated in initial data abstraction. K.C. conducted data abstraction of full-text articles. A.S. conducted database searches, performed initial data extraction, and abstraction of full texts, analysis, and developed manuscript drafts. All authors participated in reviewing and editing the final manuscript. The authors report no proprietary or commercial interest in any product mentioned or concept discussed in this article. This work was supported primarily by funding from the Agency for Health Care Research and Quality (AHRQ) through a grant on Clinically Enhanced Multi-Purpose Administrative Dataset for Comparative Effectiveness Research” (Award #: 1R0111HS019997). This work represents the views of the author(s) and do not necessarily reflect those of the AHRQ or the University of South Florida. n Corresponding author at: Department of Epidemiology & Biostatistics, College of Public Health, University of South Florida, 13201 Bruce B. Downs Blvd, MDC 56, Tampa, FL 33612. E-mail address: [email protected] (H.M. Salihu). 0146-0005/13/$ - see front matter & 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1053/j.semperi.2013.06.021
390
SE
M I N A R S I N
P
E R I N A T O L O G Y
fashion. Also unclear is the set of protective factors that can potentially enhance survival, as well as information on possible racial/ethnic disparities in the survival of these infants. Thus, the purpose of this analysis was to describe studies on the survival of pre-viable infants in the United States published over the last decade. We also sought to identify factors that confer risk or protection and any disparities in the survival of pre-viable infants in the United States.
2.
Methods
2.1.
Search strategy
We conducted a systematic review and meta-analysis of observational studies.5 We searched for all observational studies that assessed survival for pre-viable live births for the neonatal or post-neonatal period, published between January 2003 and January 2013, in the United States only. We defined pre-viable births (independent variable) as infants born at less than 24 weeks of gestation and/or whose weight was less than 500 g at birth. Pre-viable infant survival (dependent variable) was defined as an infant surviving beyond the first day of life up to 1 year of life, and was categorized into survival at discharge from neonatal intensive care unit (neonatal survival) or beyond the neonatal period (post-neonatal survival, up to 1 year post-discharge, or 18–22 months survival). Since the interest was to measure survival rates, we focused on cohort studies.6 Studies that did not assess survival or only assessed infant mortality without documenting the denominator for periviable infant live births, those without documentation that infants were alive beyond 24 h after birth (to avoid misclassification with fetal deaths), and clinical studies with small sample size exclusively assessing specific interventions or specific morbidities were excluded. We searched PubMed, Ovid, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and ISI Web of Knowledge. The search was restricted to English language and humans. We used Medical Subject Headings (MeSH terms) and expert-suggested key words to capture variation in terminology, as follows: step 1: pre-viable OR periviable OR marginally viable OR micro-preemies OR extremely preterm OR extremely low birth weight OR extremely low gestational age newborns; step 2: survival OR mortality; step 3: United States OR USA; and step 4: combined the results of steps 1, 2, and 3 using “AND” as Boolean Operator (#1 AND #2 AND #3).
2.2.
Selection of articles
Our search identified 1121 articles for possible inclusion: PubMed provided 286 records, CINAHL yielded 122 records, Ovid 184 records, and Web of Knowledge 35 records (Fig. 1). After removing duplicates, 541 articles were retained and screened to exclude ineligible articles. A large number of excluded articles were case reports, clinical trials, case-controls, editorials, and commentaries that did not refer to US population or were not population-based. Subsequently, 66 articles were selected for full-text review, plus four additional back references. Accordingly, 70 full-text articles were
37 (2013) 389–400
examined by two members of the study team, who finally selected 15 studies. The process was managed with the reference manager software EndNote X5.7,8
2.3.
Assessment of methodological quality
Two reviewers individually rated the articles chosen for this systematic review using the Newcastle–Ottawa Quality Assessment Scale (NOS).9 A “star system” was utilized to assess the overall quality of each study based on three general categories: selection of study groups (i.e., the representativeness of the sample, selection of non-exposed, and ascertainment of exposure); comparability of the groups (i.e., control for confounding factors); and how the researchers determined the outcome of interest for cohort studies specifically (reliability of methods for assessing outcomes, adequacy of follow-up, and attrition).10,11 We computed Krippendorff's alpha reliability coefficients using the software ReCal OIR for reliability analyses.12 The Krippendorff's alpha was 0.7, which is considered moderate agreement.13 Final scores were obtained through consensus (Table 1).
2.4.
Data analysis
We calculated survival rates based on the number of survivors among pre-viable live births. Using 11 data points (based on epochs within studies) from a subset of eight studies that had a common outcome measurement (i.e., reported survival to discharge), we computed a single pooled estimate using Comprehensive Meta-Analysis Software, version 2.2.604.14 For this purpose, we estimated fixed-effect model (FEM) and random-effects model (REM). The survival rates were separated by data collection period within the same study (e.g., Donohue A: 1993–1995; Donohue B: 2001–2003; and Hoekstra, 2004: A–C). Forest plots were constructed to illustrate effect sizes and pooled survival across studies using fixed-effect models (FEM) and random-effects models (REM) to evaluate different assumptions to the data. The remaining studies not included in the meta-analysis were not comparable and we opted to describe them qualitatively. We calculated Q and I2 statistics to determine the total variability in the effect estimate that could be attributed to between-study heterogeneity. Under our estimated FEM, the null hypothesis that all studies were evaluating the same effects was rejected with the Q test for heterogeneity (Q ¼ 161.6; df ¼ 10; p o 0.05), which favored the adequacy of REM over FEM. Additionally, the I2 test, which is a measure of the degree of inconsistency across studies,15 identified that 93.8% of total variation across studies was due to heterogeneity rather than chance in FEM. In contrast, the impact of study heterogeneity was smaller under the REM assumptions (Q ¼ 41, df ¼ 10, p 4 0.05; and I2 ¼ 75% of total variation across studies due to heterogeneity). Possible sources of heterogeneity could come from marked differences in study sample sizes (N range ¼ 75–3631), unbalanced proportions of 21-week vs 24-weeks preterm infants leading to higher mortality or higher survival rates in specific studies, overlapping data collection years, and diversity of populations considered. Sensitivity analyses were conducted by omitting one study
S
E M I N A R S I N
P
E R I N A T O L O G Y
37 (2013) 389–400
391
1121 articles identified from database searches (PubMed=286; Ovid=184; CINAHL=122; Web of Knowledge=35)
580 duplicate articles excluded
541 articles identified for screening and abstract review 475 articles excluded: Did not satisfy inclusion criteria • Single-case series/report, commentary, editorials, review articles, and randomized trials • Did not measure survival as outcome • Did not measure gestational age
M I N A R S I N
P
E R I N A T O L O G Y
37 (2013) 389–400
Available online at www.sciencedirect.com
www.elsevier.com/locate/semperi
Survival of pre-viable preterm infants in the United States: A systematic review and meta-analysis Hamisu M. Salihu, MD, PhDa,b,n, Abraham A. Salinas-Miranda, MD, MPHa,c, Latoya Hill, MPHa, Kristen Chandler, MSW, MPHa,c n
Maternal & Child Health Comparative Effectiveness Research Group, Department of Epidemiology & Biostatistics, College of Public Health, University of South Florida, 13201 Bruce B. Downs Blvd, MDC 56, Tampa, FL 33612 † Division of Maternal Fetal Medicine, Department of Obstetrics and Gynecology, College of Medicine, University of South Florida, Tampa, FL ‡ Department of Community and Family Health, College of Public Health, University of South Florida, Tampa, FL
AR T IC LE INFO
AB ST R A CT The objective of this paper is to review observational studies that addressed the survival of
Key words:
pre-viable gestations in the United States. We searched PubMed, Ovid, CINAHL, and Web of
Survival
Knowledge for studies reporting survival of infants born at o24 gestational weeks and/or
Pre-viable
o500 g in the United States and published between January 2003 and January 2013. The full
Periviable
texts of 70 articles were examined and a total of 15 studies qualified and were selected. We
Extremely preterm birth
analyzed fixed-effect and random-effects models for eight studies on survival to discharge.
Extremely low birth weight
Pooled survival to discharge in the random-effects model was 45.9% (95% CI: 41.1–51.7) and
Pregnancy
39.7% in the fixed-effect model (95% CI: 38.8–40.7). Studies differed by pre-viable survival measures and epochs (1985–2009). Protective factors included antenatal corticosteroids, neonatal resuscitation, and intensive care. The current survival threshold for pre-viable infants warrants reconsideration of the limits of viability. Protective factors that enhance survival should be considered in the management of these infants. & 2013 Elsevier Inc. All rights reserved.
1.
Introduction
The survival of extremely low gestational age newborns (ELGAN, live births o28 weeks of gestation) and extremely low birth weight infants (ELBW, o1000 g, or 2 lb, 3 oz) has improved considerably over the last three decades in developed countries.1–3 Survival is now possible for infants born
before 24 weeks of gestation and for those weighing less than 500 g, who are currently classified as “pre-viable,” “periviable,” “borderline,” or “non-viable” births.4 Despite the growing population of pre-viable survivors reported in published studies over the last decades, the evidence on the survival of pre-viable infants in the United States continues to be scattered and has not been summarized in a systematic
This work has not been published before. H.S. provided the overall leadership and scientific oversight of project, and edited the manuscript. L.H. conducted database searches and participated in initial data abstraction. K.C. conducted data abstraction of full-text articles. A.S. conducted database searches, performed initial data extraction, and abstraction of full texts, analysis, and developed manuscript drafts. All authors participated in reviewing and editing the final manuscript. The authors report no proprietary or commercial interest in any product mentioned or concept discussed in this article. This work was supported primarily by funding from the Agency for Health Care Research and Quality (AHRQ) through a grant on Clinically Enhanced Multi-Purpose Administrative Dataset for Comparative Effectiveness Research” (Award #: 1R0111HS019997). This work represents the views of the author(s) and do not necessarily reflect those of the AHRQ or the University of South Florida. n Corresponding author at: Department of Epidemiology & Biostatistics, College of Public Health, University of South Florida, 13201 Bruce B. Downs Blvd, MDC 56, Tampa, FL 33612. E-mail address: [email protected] (H.M. Salihu). 0146-0005/13/$ - see front matter & 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1053/j.semperi.2013.06.021
390
SE
M I N A R S I N
P
E R I N A T O L O G Y
fashion. Also unclear is the set of protective factors that can potentially enhance survival, as well as information on possible racial/ethnic disparities in the survival of these infants. Thus, the purpose of this analysis was to describe studies on the survival of pre-viable infants in the United States published over the last decade. We also sought to identify factors that confer risk or protection and any disparities in the survival of pre-viable infants in the United States.
2.
Methods
2.1.
Search strategy
We conducted a systematic review and meta-analysis of observational studies.5 We searched for all observational studies that assessed survival for pre-viable live births for the neonatal or post-neonatal period, published between January 2003 and January 2013, in the United States only. We defined pre-viable births (independent variable) as infants born at less than 24 weeks of gestation and/or whose weight was less than 500 g at birth. Pre-viable infant survival (dependent variable) was defined as an infant surviving beyond the first day of life up to 1 year of life, and was categorized into survival at discharge from neonatal intensive care unit (neonatal survival) or beyond the neonatal period (post-neonatal survival, up to 1 year post-discharge, or 18–22 months survival). Since the interest was to measure survival rates, we focused on cohort studies.6 Studies that did not assess survival or only assessed infant mortality without documenting the denominator for periviable infant live births, those without documentation that infants were alive beyond 24 h after birth (to avoid misclassification with fetal deaths), and clinical studies with small sample size exclusively assessing specific interventions or specific morbidities were excluded. We searched PubMed, Ovid, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and ISI Web of Knowledge. The search was restricted to English language and humans. We used Medical Subject Headings (MeSH terms) and expert-suggested key words to capture variation in terminology, as follows: step 1: pre-viable OR periviable OR marginally viable OR micro-preemies OR extremely preterm OR extremely low birth weight OR extremely low gestational age newborns; step 2: survival OR mortality; step 3: United States OR USA; and step 4: combined the results of steps 1, 2, and 3 using “AND” as Boolean Operator (#1 AND #2 AND #3).
2.2.
Selection of articles
Our search identified 1121 articles for possible inclusion: PubMed provided 286 records, CINAHL yielded 122 records, Ovid 184 records, and Web of Knowledge 35 records (Fig. 1). After removing duplicates, 541 articles were retained and screened to exclude ineligible articles. A large number of excluded articles were case reports, clinical trials, case-controls, editorials, and commentaries that did not refer to US population or were not population-based. Subsequently, 66 articles were selected for full-text review, plus four additional back references. Accordingly, 70 full-text articles were
37 (2013) 389–400
examined by two members of the study team, who finally selected 15 studies. The process was managed with the reference manager software EndNote X5.7,8
2.3.
Assessment of methodological quality
Two reviewers individually rated the articles chosen for this systematic review using the Newcastle–Ottawa Quality Assessment Scale (NOS).9 A “star system” was utilized to assess the overall quality of each study based on three general categories: selection of study groups (i.e., the representativeness of the sample, selection of non-exposed, and ascertainment of exposure); comparability of the groups (i.e., control for confounding factors); and how the researchers determined the outcome of interest for cohort studies specifically (reliability of methods for assessing outcomes, adequacy of follow-up, and attrition).10,11 We computed Krippendorff's alpha reliability coefficients using the software ReCal OIR for reliability analyses.12 The Krippendorff's alpha was 0.7, which is considered moderate agreement.13 Final scores were obtained through consensus (Table 1).
2.4.
Data analysis
We calculated survival rates based on the number of survivors among pre-viable live births. Using 11 data points (based on epochs within studies) from a subset of eight studies that had a common outcome measurement (i.e., reported survival to discharge), we computed a single pooled estimate using Comprehensive Meta-Analysis Software, version 2.2.604.14 For this purpose, we estimated fixed-effect model (FEM) and random-effects model (REM). The survival rates were separated by data collection period within the same study (e.g., Donohue A: 1993–1995; Donohue B: 2001–2003; and Hoekstra, 2004: A–C). Forest plots were constructed to illustrate effect sizes and pooled survival across studies using fixed-effect models (FEM) and random-effects models (REM) to evaluate different assumptions to the data. The remaining studies not included in the meta-analysis were not comparable and we opted to describe them qualitatively. We calculated Q and I2 statistics to determine the total variability in the effect estimate that could be attributed to between-study heterogeneity. Under our estimated FEM, the null hypothesis that all studies were evaluating the same effects was rejected with the Q test for heterogeneity (Q ¼ 161.6; df ¼ 10; p o 0.05), which favored the adequacy of REM over FEM. Additionally, the I2 test, which is a measure of the degree of inconsistency across studies,15 identified that 93.8% of total variation across studies was due to heterogeneity rather than chance in FEM. In contrast, the impact of study heterogeneity was smaller under the REM assumptions (Q ¼ 41, df ¼ 10, p 4 0.05; and I2 ¼ 75% of total variation across studies due to heterogeneity). Possible sources of heterogeneity could come from marked differences in study sample sizes (N range ¼ 75–3631), unbalanced proportions of 21-week vs 24-weeks preterm infants leading to higher mortality or higher survival rates in specific studies, overlapping data collection years, and diversity of populations considered. Sensitivity analyses were conducted by omitting one study
S
E M I N A R S I N
P
E R I N A T O L O G Y
37 (2013) 389–400
391
1121 articles identified from database searches (PubMed=286; Ovid=184; CINAHL=122; Web of Knowledge=35)
580 duplicate articles excluded
541 articles identified for screening and abstract review 475 articles excluded: Did not satisfy inclusion criteria • Single-case series/report, commentary, editorials, review articles, and randomized trials • Did not measure survival as outcome • Did not measure gestational age
