Defining Very Preterm Populations for Systematic Reviews With Meta-analyses

Survival of very preterm (VPT) infants (ie, those born at <32 weeks’ gestation) has improved markedly over recent decades, raising concerns about levels of impairment among survivors. Numerous studies have been conducted on the association between VPT birth and long-term neurodevelopment and health, and this voluminous literature is increasingly synthesized in systematic reviews with meta-analyses. This methodology is considered to provide the highest level of evidence, but its validity depends on appropriate selection of primary studies and management of heterogeneity. Heterogeneity is pervasive in the literature about VPT birth because of differences in criteria for defining preterm populations, study designs, follow-up periods, follow-up rates, and clinical assessments. Furthermore, medical practices, survival, and morbidities vary markedly across countries and hospitals and can affect long-term prognosis. This study aimed to compare the selection criteria, findings, and heterogeneity of systematic reviews with meta-analyses of cognitive outcomes after VPT birth, which are of major concern in this population and measured in most studies.

We searched for systematic reviews with meta-analyses published between January 2000 and August 2019 that were based on observational studies with cohort designs investigating general cognition (IQ) for VPT children compared with a control group. We defined search terms to identify studies (1) on preterm birth (premature OR preterm OR infant, Premature [Medical Subject Headings (MeSH)]), (2) the outcome [Developmental disabilities [MeSH] OR cognition disorders [MeSH] OR intellectual disability [MeSH] OR cognition [MeSH] OR cognit* OR intelligence OR IQ, and (3) the type of study (Meta-analysis OR meta analys* OR metaanaly* OR [systematic AND review* OR overview*] OR Review Literature as Topic [MeSH]). Two of us (M.S. and J.Z.) independently abstracted methods and results related to study selection, pooled analyses, and heterogeneity. We compiled primary studies and identified unique cohorts when several studies originated from the same cohort based on the country, birth year(s), and research group. For analyses, we used R statistical software, version 3.5.0 (The R Foundation).

Five reviews were identified: 1 was published in 2012¹ and 4 were published in 2018 or 2019.^2-5 All investigated the association of birth at less than 32 weeks’ gestation with childhood IQ, although some also considered other outcomes or subgroups. Eligibility criteria varied for birth weight, assessment ages, and study period (Table). We searched MEDLINE,^1-5 Embase,^1,2 PsychInfo,^1,3-5 and Web of Science^3,4 using different search terms.

A total of 156 primary studies reporting results from 114 VPT cohorts were included in all reviews. Among 107 cohorts included in the 3 reviews published in 2018 that did not have upper limits on assessment age, only 8 were included in all 3 reviews (Figure). When these were limited to the 58 cohorts covered by studies published before 2009 to permit comparison with the 2012 review, none was included in all 4 reviews and 35 (60%) were included in only 1 review. All 7 cohorts included in the review on assessments at ages 3 to 5 years⁵ were included in at least 1 other review, but none was included in all 3 reviews.

All reviews reported a lower IQ among VPT children compared with full-term children (pooled effects ranged from −0.77 [95% CI, −0.88 to −0.66] to −0.86 [95% CI, −0.94 to −0.78] for the standardized mean difference or 11.6-12.9 IQ points for results relating to all children born at <32 weeks’ gestation). Most of the reviews reported a moderate to high heterogeneity overall or in at least 1 gestational age subgroup (I², >60%).

Investigators’ methodological choices appeared to affect the primary study selection in systematic reviews with meta-analyses on VPT birth and cognition despite shared objectives. These differences in study selection represent a missed opportunity to synthesize all available information and raise questions about appropriate criteria for selecting studies in this population. Nonetheless, similar pooled estimates affirmed the robust conclusion that VPT birth is associated with a lower IQ, although with substantial interstudy heterogeneity. A limitation of this study is that we could not attribute differences in study selection to specific methodological choices (ie, inclusion criteria, definitions, search terms), which would require reviewing all primary studies.

The exponential increase in systematic reviews with meta-analyses has called attention to the need to organize research efforts to avoid both redundant and contradictory results when evidence is synthesized.⁶ Continued investigation of the long-term consequences of VPT is essential to address concerns about high, stable impairment rates among VPT survivors and to investigate variation related to perinatal management, follow-up services, and the sociocultural environment. Establishing common guidelines to select primary studies, along with open-access repositories of studies from previous reviews, would ensure that future reviews use all relevant information and optimize the analysis of interstudy heterogeneity.

Accepted for Publication: January 10, 2020.

Corresponding Author: Mariane Sentenac, PhD, EPOPé, INSERM 1153, 53 Avenue de l’Observatoire, 75014 Paris 9, France (mariane.sentenac@inserm.fr).

Published Online: June 15, 2020. doi:10.1001/jamapediatrics.2020.0956

Author Contributions: Drs Sentenac and Zeitlin had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Sentenac, Boutron, Kajantie, Maier, Wolke, Zeitlin.

Acquisition, analysis, or interpretation of data: Sentenac, Draper, Kajantie, Maier, Zeitlin.

Drafting of the manuscript: Sentenac, Zeitlin.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Sentenac.

Obtained funding: Draper, Kajantie, Wolke, Zeitlin.

Administrative, technical, or material support: Sentenac, Kajantie.

Supervision: Wolke, Zeitlin.

Conflict of Interest Disclosures: Dr Sentenac reported receiving grants from the European commission EU Horizon 2020 during the conduct of the study. Dr Boutron reported being the director of Cochrane France and co-convenor of the Bias Methods group at Cochrane. Dr Kajantie reported receiving grants from the European Commission, Academy of Finland, and Foundation for Pediatric Research during the conduct of the study and receiving grants from the Novo Nordisk Foundation, Foundation for Cardiovascular Research, Diabetes Research Foundation, and Sigrid Juselius Foundation outside the submitted work. Dr Maier reported receiving grants from the European Union during the conduct of the study. Dr Zeitlin reported receiving grants from the European Commission Horizon 2020 during the conduct of the study. No other disclosures were reported.

Funding/Support: This study is part of research on the consequences of very preterm birth for use in developing a research agenda for the RECAP Preterm project, an EU Horizon 2020 study (supported by grant 733280) to construct a platform combining data from very preterm cohort studies in Europe.

Role of the Funder/Sponsor: The funding organization had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.