We identified 526 records through electronic searches and three records through other sources. We excluded 477 records based on title and abstract screening and 18 duplicates. We assessed 34 full-text articles, of which 15 were possibly eligible for inclusion. In six studies, we contacted the authors for additional information [21, 23, 32-35]; authors of four studies responded [21, 32-34]. Additional relevant data were provided by one author . Finally, 11 studies were included in this review (Table 1) [33, 34, 36-44]. Fig. 1 outlines the study selection.
We excluded 23 studies [21-24, 32, 35, 45-61] after full text retrieval. Reasons for exclusion are listed in Supplement Table 3.
Characteristics of included studies
Of the 11 studies [33, 34, 36-44], ten had a retrospective design, and one had a prospective cohort design. Among the multicenter studies, one was a registry-based international CDHSG collaboration . Of the included articles, five reported the association of VF with survival, two with ECMO need, and four with both outcomes. Two single-center studies [40,43] had overlapping cohorts with two multicentric studies [34,42]. These were included for qualitative synthesis as they reported various echocardiographic parameters and outcomes. However, they were not included in the meta-analysis. One study involved survival analysis , and two studies involved adjusted analysis [34, 42].
Most studies included CDH cases ≥ 34 gestational age (GA). Mean GA was 38 weeks, and mean birth weight (BW) was 2.8 kg. Comorbidities are summarized in Supplement Table 4. Dao et al. included only low-risk CDH cases (type A and B) . Wehrmann et al.  exclusively studied patients with measurable right-to-left or left-to-right atrial shunts. Lawrence et al. limited their analysis to iNO-treated CDH .
There were significant variations in echocardiographic parameters among studies (Table 2). Two studies reported RV function [38, 39], four reported LV function [33, 34, 37, 43], and five reported both RV and LV function [36, 40-42, 44]. Two studies defined LVD a priori [33, 41]. LVD  and RV dysfunction (RVD)  were defined based on ROC curve analysis in two studies. One study defined abnormal VF as < 1 standard deviation below the mean derived from control data . One study reported VD as a dichotomous variable (present or absent) based on qualitative and quantitative echocardiographic assessments . This study specified neither echocardiographic parameters nor the systolic or diastolic nature of the dysfunction. Only one study described isolated RVD and LVD separately . Studies mostly reported systolic function.
The methodological quality was evaluated using QUIPS [30, 31] (Supplement Fig. 1A & 1B). Overall, ten studies had moderate RoB, and one had low RoB. This was commonly due to selection, confounding factors, statistical analysis and reporting, and prognostic factor measurement biases. Only one study had low RoB in all six domains.
LV function and survival
Nine studies reported the association of VF (any reported VF measurement) with survival [33, 34, 36-42]. Four studies with 899 participants (range, 51–674), including 856 who survived, were pooled. We included only LV function data in the pooled analysis if a study reported both RV and LV function. The pooled sensitivity and specificity of normal LV function for survival prediction were 86% (95% CI, 77%–92%) and 44% (95% CI, 25%–65%), respectively. Summary LR+ and LR- were 1.5 (95% CI, 1.1–2.1) and 0.32 (95% CI, 0.21–0.50), respectively. Pooled analysis-derived AUC and DOR were 75% (95% CI, 71%–78%) and 5 (95% CI, 2–9), respectively (Fig. 2A & 2B and Supplement Table 5).
The proportion of heterogeneity likely due to the threshold effect was 100. To explore other potential heterogeneities, meta-regression was conducted (Supplement Fig. 2). Overall, the test performances did not vary by GA, BW, sex, and liver herniation. Two studies provided data for predicting survival using RV function [38, 42]. Both studies showed a significant survival-related predictive value of normal VF (Supplement Fig. 3).
VF and ECMO requirement
Six studies reported an association between LV function and ECMO requirement [33, 40-44]. Three studies (111 out of 815 participants requiring ECMO) could be pooled for meta-analysis [33, 41, 42]. ECMO use ranged from 9.2%–40% among three studies. The overall sensitivity and specificity of LVD for ECMO requirement prediction were 39.8% (95% CI, 27%–52%) and 88% (95% CI, 80%–96%), respectively. Sensitivity ranged from 35%–62%, and specificity from 71%–93%. Summary LR+ and LR- were 2.9 (95% CI, 1.8–4.1) and 0.68 (95% CI, 0.58–0.78), respectively. Pooled analysis-derived DOR was 5.6 (95% CI, 2.5–8.8) (Fig. 3A & 3B).
In one study , RVD was predictive of ECMO need with sensitivity and specificity of 0.45 (95% CI, 0.32%–0.58%) and 0.82 (95% CI, 0.79%–0.85%), respectively (Supplement Fig. 4).
Among studies pooled for meta-analysis, one study  included participants with low-risk CDHSG type A and B, and another  included only severe CDH cases. Four studies had moderate RoB, and one had low RoB. LVD can be complicated by concomitant RVD.
Results of the remaining seven studies were not pooled because of high heterogeneity in the reported echocardiographic markers and continuous variable-related results or because of overlapping cohorts (Supplement Table 6).
Among the reported echocardiographic parameters, higher RV outflow tract velocity time integer (RVOT VTI) (MD, 3.30; 95% CI, 0.54%–6.06%), RV fractional area change (FAC) (MD, 14.40; 95% CI, 8.69%–20.11%), and tricuspid annular plane systolic excursion (TAPSE) (MD, 0.30; 95% CI, 0.14%–0.46%) were predictive of survival. Low LV outflow tract VTI (LVOT VTI) (MD, -4.15; 95% CI, -6.18% to -2.12%) and LV cardiac index (MD, -0.47; 95% CI, -0.91% to -0.03%) were predictive of ECMO need.
Survival without ECMO vs. death/ECMO
Two studies reported survival without ECMO and death/ECMO [39, 40]. Aggarwal et al.  reported associations of lower RV FAC, TAPSE, and RVOT VTI with the combined outcome of death/ECMO, and Patel et al.  reported the association of lower LV global longitudinal strain with this outcome.
We additionally performed a GRADE assessment of our review. All studies were observational. There were serious concerns due to RoB, inconsistency, and survival-related imprecision. The overall certainty of the evidence was very low. For ECMO requirements, imprecision was not detected, and the certainty of evidence was low.