Study Selection
Figure 1 illustrates the article selection process employed in our meta-analysis. The initial database search yielded a total of 3,095 results. Upon removing duplicates, there were 1,243 unique results. During the abstract screening phase, 1,080 articles were excluded based on predefined criteria. I full texts of the remaining 163 articles were reviewed. Additionally, authors were contacted to provide original data with sex-disaggregated outcomes. Nine of the fifteen of the contacted authors responded to our inquiries. After evaluation of the full texts, an additional 148 articles were excluded based on the exclusion criteria including outcomes irrelevant to the meta-analysis and missing sex-disaggregated data after contacting original authors. Table 1 provides a summary of the 15 studies included in the meta-analysis.
Figure 1.
Table 1. Summary characteristics of included studies.
Author
|
Study Period
|
Cardiac Phenotype (Age)
|
Cohort Site
|
Atallah
|
1996-2010
|
HLHS
|
Stollery Children’s Hospital, Canada
|
ElMahrouk
|
2001-2016
|
Complex with ECMO
|
Egypt, Saudi Arabia
|
Griffiths
|
2005-2014
|
DORV or dTGA
|
Utah, USA
|
Hames
|
2002-2017
|
Truncus arteriosus with ECMO
|
ELSO registry
|
Hamzah
|
2002-2017
|
Truncus arteriosus (<1 year)
|
National Inpatient Sample
|
Howard
|
2006-2013
|
Complex with ECMO (<29 days)
|
Boston Children’s Hospital, USA
|
Hoxha
|
2008-2015
|
Aortic arch hypoplasia, IAA, HLHS (neonates)
|
Verona, Italy
|
Lim
|
2012-2016
|
Complex (<10 years)
|
Singapore
|
Manuel
|
2011-2017
|
Single ventricle
|
Angola
|
Martin & Cheung
|
1996-2016
|
Complex
|
Stollery Children’s Hospital, Canada
|
Merkle
|
2008-2016
|
Complex with ECMO
|
Germany
|
Newburger
|
2005-2008
|
HLHS and single ventricle
|
Single Ventricle Reconstruction trial
|
Sznycer-Taub
|
2007-2013
|
Complex with ECMO
|
Michigan, USA
|
Tomotaki
|
2006-2011
|
Complex (VLBW)
|
Japan
|
DORV-double-outlet right ventricle; dTGA-d-Transposition of great arteries; ECMO-extracorporeal membrane oxygenation; HLHS-hypoplastic left heart syndrome; IAA-interrupted aortic arch; VLBW-very low birth weight
Characteristics of Included Studies
The included studies had diverse geographic distribution, spanning multiple countries, including the United States, Canada, Jordan, Angola, Singapore, Germany, and Italy, with two studies examining international patient cohorts28,29. These investigations predominantly adopted a retrospective cohort study design, except for one study that reported data from the Single Ventricle Reconstruction Trial24. Notably, the included cohorts varied significantly in study size, study length, geographical location, racial demographics, and number of clinical sites involved.
The articles also exhibited substantial heterogeneity in CHD phenotypes, type of surgical repair, and age at the time of surgery. Among the 15 articles, five studies (33%) focused on patients who required extracorporeal membrane oxygenation (ECMO) following cardiac surgery28,31-33-. Furthermore, certain studies concentrated on the outcomes of specific cardiac surgical procedures, including the Norwood procedure, first stage of palliative surgery for hypoplastic left heart syndrome24,30,34,35, the arterial switch for transposition of the great arteries or double-outlet right ventricle36, truncus arteriosus repair28,37, hypoplastic or interrupted aortic arch surgery38, or unspecified cardiac surgeries25,39,40. Some studies specifically investigated surgical outcomes in patients born premature, who are known to have heightened risk for mortality and ND impairment25,39. In aggregate, these 15 articles collectively represented 4,999 patients, with a slight male predominance at 54%. Notably, one study reported the largest cohort with 3,009 patients37.
Primary Outcomes of Interest
Mortality
Sex-disaggregated mortality data was available in all included articles. The overall gender distribution aligned with the expected pattern, as many of the cardiac phenotypes examined exhibited a male predominance41,42. Combined rates of all mortality were 424/2690 (15.8%) in male CHD patients and 398/2309 (17.2%) in female CHD patients. No difference was observed in mortality by sex (OR 0.99, 95% CI 0.76-1.30) (Figure 2). No publication bias was observed on the funnel plot (Suplepmental Figure 1).
Figure 2.
Studies varied in reported timing of mortality, so mortality rates were also analyzed by sub-groups: in-hospital mortality (n=5 studies), post-ECMO mortality (n=5 studies), and post-Stage I mortality (n=4 studies). This analysis did show females had a decreased odds for mortality (OR 0.85; 95% CI 0.77-0.97) (Figure 3).
Figure 3.
Neurodevelopment Outcomes
Only three studies included ND outcomes in male and female patients after surgery for CHD. Two of the articles assessed children with the BSID-III with BSID-III assessments, defining a cognitive score less than 70 as severe NDI24,25. These two studies had overlapping cohorts, so results were merged, and duplicates removed for the meta-analysis. The Single Ventricle Reconstruction Trial assessed children with the BSID-II30. BSID-II Motor Development scores less than 70 was defined as severe ND impairment. In another study children were assessed with the Kyoto Scale of Psychological Development, which has been well-correlated to BSID39. The article defined a developmental quotient (DQ) score less than 70 as severe ND impairment36.
A total of 475 patients had ND outcomes available, 63% male and 37% female. Rates of severe ND impairment after surgery for congenital heart defects were 15% in male patients and 12% in female. No sex differences were noted for this outcome (Figure 4). No publication bias was observed on the funnel plot (Supplemental Figure 2).
Figure 4.
Methodological Quality (Risk of Bias)
All the studies included in this analysis were cohort studies, and the CLARITY tool was employed to assess for risk of bias27. This comprehensive tool evaluates the potential sources of bias in cohort studies across eight distinct domains, which encompass study participation, study attrition, measurement of prognostic factors, outcome measurement, management of study confounding, the rigor of statistical analysis, intervention integrity, and an overall assessment of the risk of bias. Each domain is scored using a Likert scale, with four possible choices ranging from low risk (1) to high risk (4). Consequently, a study that receives all low-risk scores across all domains would accumulate a total score of eight, while the maximum attainable score is 32.
Total scores obtained by the reviewed articles fell within the range of 10 to 16, indicating varying degrees of potential bias across the included studies (Table 2). Notably, the three domains with the highest observed risk of bias were related to the measurement of prognostic factors, the management of study confounding, and the assessment of follow-up outcomes.
Table 2. Risk of Bias Assessment of studies included in meta-analysis. Q1-Q8 defined as in the CLARITY risk of bias tool for cohort studies21.
Author
|
Q1
|
Q2
|
Q3
|
Q4
|
Q5
|
Q6
|
Q7
|
Q8
|
Total
|
Howard, T
|
1
|
1
|
1
|
1
|
2
|
1
|
1
|
2
|
10
|
Griffiths, ER
|
1
|
2
|
1
|
1
|
2
|
1
|
1
|
1
|
10
|
Hames, DL
|
1
|
1
|
1
|
1
|
2
|
1
|
3
|
2
|
12
|
ElMahrouk, AF
|
2
|
1
|
1
|
2
|
2
|
1
|
2
|
2
|
13
|
Hamzah, M
|
1
|
2
|
2
|
2
|
2
|
2
|
3
|
2
|
16
|
Atallah, J
|
1
|
2
|
2
|
2
|
2
|
1
|
1
|
2
|
13
|
Tomotaki
|
1
|
1
|
2
|
2
|
2
|
1
|
2
|
2
|
13
|
Sznycer-Taub
|
1
|
1
|
1
|
1
|
2
|
1
|
3
|
2
|
12
|
Martin, BJ
|
1
|
1
|
2
|
2
|
2
|
1
|
2
|
2
|
13
|
Manuel, V.
|
1
|
2
|
1
|
2
|
2
|
1
|
2
|
2
|
13
|
Lim CY
|
1
|
1
|
1
|
1
|
2
|
2
|
1
|
1
|
10
|
Newburger, JW
|
1
|
1
|
1
|
1
|
2
|
1
|
1
|
2
|
10
|
Merkle, J
|
1
|
1
|
1
|
1
|
2
|
1
|
2
|
1
|
10
|
Hoxha, S
|
1
|
1
|
1
|
2
|
2
|
2
|
2
|
1
|
12
|
Cheung, PY
|
1
|
1
|
2
|
2
|
2
|
1
|
1
|
2
|
12
|