A comprehensive search of electronic databases MEDLINE and EMBASE was conducted for studies published between the beginning of each database and March 2021 (details provided as Additional file 1). Reference lists of relevant studies were also examined to identify any additional relevant studies not identified in the search.

All abstracts were screened, and full text assessed for eligibility by two independently reviewers (LC and AD), conflicts were resolved by consensus and, where necessary, through discussion with the other co-authors.

Original research of any study design that fulfilled the criteria below was eligible; this meant we could include population-based register studies, cohort studies, case control studies, cross-sectional studies, and randomised controlled trials if they included relevant data. We sought studies that included a comparator group of non-CHD patients from the same population for aim (a) (see aims at end of introduction), but we also included studies that only included CHD patients. Whilst these studies may not address the question for patients and their families as to whether they are likely to be as successful in school as their peers, our PPI work suggested it was still helpful to know what proportion of those with CHD obtain a university degree or complete secondary education. Furthermore, we identified a source that provided summary data, stratified by age, of the proportion of people in most countries of the world achieving the three educational outcomes explored in this study (see below). Thus, for most studies that only had data in CHD patients we were still able to compare them to overall educational levels in their country.

Studies were therefore eligible if they reported (or provided sufficient data for us to be able to calculate) the rate, odds, or proportion of level of educational attainment in adults (aged 18 years of age or older) with a history of any CHD. They were also eligible if CHD patients had not undergone procedures, whilst those in which patients had undergone procedures were eligible irrespective of the type, timing, or number of repeat procedures. We also included studies irrespective of whether the aim was to explore educational attainment in patients with CHD or not. The cut-off of 18 years was chosen so that we could assess differences in educational attainment at the age of completion of compulsory education in most high-income countries, and with measures (completing a university degree, secondary education, or vocational training) that are likely to influence future life chances. In initial screening we included studies with a lower age threshold (16 years or older) and, in the data extraction process, explored whether it was possible to obtain results for those only 18 years or older.

Whether comparing CHD patients to a control group without CHD or comparing the proportion with an educational outcome in CHD patients to country-level proportions, we studied three outcomes, and studies could be included if they had data on at least one of these:

The outcomes were all analysed as ‘not achieving’ (e.g., not obtaining a university degree).

To avoid double counting data, separate articles reporting educational outcomes in the same patient group were evaluated and the article providing most complete information (largest sample or more recent study) was selected for inclusion.

Data were extracted independently by two reviewers (LC, AD). For each study, we extracted information on the total number of patients with CHD and those without and, where provided, the number of CHD and non-CHD participants who achieved each educational attainment measure. We also extracted information on the age and sex of participants, the geographical region of the study, year of publication and the severity of the disease. Three authors (LC, RC, DAL) a priori defined key confounders of the association between CHD and educational attainment. Confounders are by definition factors that could plausibly affect the risk of having CHD and the educational outcomes [11]. Maternal pregnancy characteristics (e.g. higher early/pre-pregnancy BMI, smoking and alcohol) have been hypothesised to influence CHD risk in offspring, though whether these are all causal factors for offspring educational attainment is debatable [12]. As these are likely to be influenced by maternal/parental education, which is an important determinant of offspring educational attainment, we considered parental education to be a key confounder. CHD risk also varies by parental age at birth and ethnicity, which in turn influence educational attainment. Therefore, we considered the three key confounders to be parental education, age, ethnicity and extracted information on whether studies adjusted for these. In the risk of bias assessment (see below) we considered a study to have minimal risk of bias from residual confounding if they had adjusted for parental education, ethnicity and age, allowing that adjustment to be for either one, or both, of the parents. All relevant results in whatever form were extracted (i.e., any of: adjusted or unadjusted odds ratios, risk ratios, hazard ratios, differences in risk, with relevant standard errors or confidence intervals, or proportion of participants with each educational measure), with information on what analyses were used to obtain the results.

We extracted summary data from ‘Education at a Glance’ on the proportion of adults (25–64 years old) with each of the three educational attainment outcomes for the country of residence and years of data collection of each included study. Education at a Glance is the authoritative source for information on the state of education around the world [13]. It produces annual reports with the first being published in 1998 and the most recent 2019. The age strata 25–64 years was chosen because it most closely matched the ages across the studies identified in our systematic search.

Risk of bias was assessed by two independent reviewers (LC, AD) and disagreements were resolved by discussion with all co-authors.

Risk of bias assessment was performed using the risk of bias instrument for non-randomized studies of exposure [14], which is based on seven items: (1) confounding, (2) selection of participants, (3) classification of exposure, (4) departures from intended exposure, (5) missing data, (6) measurement of outcomes and (7) selection of reported results.

To address patient and family concerns (see Patient and Public Involvement) we quantified (i) educational attainment in patients with CHD compared to their peers without CHD and (ii) educational attainment in CHD patients using all available data.

Finally, we estimated the pooled proportions of CHD patients with each measure of educational attainment across all studies (i.e., both studies that included a non-CHD comparison group and those that did not). Pooled proportions for each outcome of interest (i.e., university, secondary and vocational education attainment) were obtained using the inverse variance method, random effects model (DerSimonian and Laird).

Publication bias was evaluated using funnel plots and Egger's test.

All statistical analyses were performed using R (R Core Team (2019). R Foundation for Statistical Computing, Vienna, Austria. URL https://​www.​R-project.​org/​) and meta-R package (Guido Schwarzer (2007), meta: An R package for meta-analysis, R News.

Prior to analyses, we looked at the work carried out by the CHD charity Little Hearts Matter [18], which works continuously with patients and their family to identify areas of public interest. They indicated education as a key concern for patients and families [19, 20] and this represented a key motivation to undertake this review. At completion of the analysis we met with a group of patients and relatives (two male adult patients, two female adult patients, two mothers of adult patients with CHD) who confirmed that education was a very relevant aspect of their life and a key concern when growing up. In some cases, it was suggested that special educational support could have been useful to them, but this was not provided as not perceived to be necessary by the school. Dissemination of the review’s findings amongst relevant audience (e.g., CHD patients and families, but also teachers) was also recommended.

An overview of the included studies is presented in Tables 1 and 2. The association of CHD with educational attainment was the main aim for 14 (33% of total) studies [21‐34], while in the remaining 28 (67% of total) studies [35‐62] it was not. For most of those, information on education attainment was extracted from tables describing study population characteristics. Information on university degree, completing secondary education, and vocational training was available in 39 (93%), 32 (76%) and 15 (36%) studies respectively, with studies able to contribute to more than one outcome. A non-CHD comparison group was included in 12 (29%) of the studies (Table 1) while the remaining 30 reported only on CHD patients (Table 2). The source of the comparison groups varied between studies, but CHD and non-CHD groups were obtained from same underlying population. One study included both a general (unrelated) population comparison group and a sibling (of the CHD patients) comparison group [57]. As none of the other studies had a sibling comparison group we included results from the general population comparison group only in the main meta-analyses and in a sensitivity analysis repeated the meta-analysis with results comparing CHD patients to their siblings.

Educational attainment was evaluated with the same method in the two groups, with two [25, 27] exceptions, where information on the control group was obtained by published national statistics.

The number of patients with CHD ranged from 25 to 7019 across the studies. The unadjusted pooled analyses of the association of CHD with educational attainment included 11 (N = 104,585 participants, 10,487 with CHD), 10 (N = 167,470 participants, 11,820 with CHD), and 8 (N = 150,813 participants, 9817 with CHD) for university degree, completing secondary education or vocational training, respectively. Equivalent studies for the age and sex adjusted analyses were 9 (N = 88,813 participants, with 8880 CHD), 7 (N = 101,429 participants, 10,010 with CHD), and 6 studies (N = 100,544 participants, 9614 with CHD) for university degree, completing secondary education or vocational training respectively.

Studies were carried out in Europe (n = 21) [21, 22, 24‐30, 32, 34, 36, 38, 39, 42, 43, 45, 54, 57, 58, 61], North America (n = 11) [31, 33, 35, 40, 41, 47, 49, 52, 53, 59, 62], South America (n = 1) [60], the Middle East (n = 3) [23, 37, 44], Asia (n = 4) [46, 48, 50, 56], New Zealand (n = 1) [51] and International (n = 1) [55]. Data on educational attainment were obtained by self‐report questionnaires in the majority (39 (93%)), with the remaining three obtaining this from linkage to national registers [22, 49, 57].

The pooled OR from studies comparing educational outcomes between those with and without CHD showed that patients with CHD had higher odds of not obtaining a university degree (OR = 1.38, 95% CI [1.16, 1.65]) (Fig. 2a), not completing secondary education (OR = 1.33, 95% CI [1.09, 1.61) (Fig. 2b) and not completing vocational training (OR = 1.11, 95% CI [0.98–1.26]) (Fig. 2c). For all three outcomes there was evidence of between study heterogeneity and the predictive interval for the odds ratios were 0.81 to 2.37, 0.75 to 2.33, and 0.83 to 1.50, for not obtaining a university degree, completing secondary education, and completing vocational training, respectively. Similar findings were found in an analysis restricted to the 8 studies that had controlled for sex and age and including the study with siblings as control group (Additional file 1: Fig. S2 and S3a–c).

Subgroup analyses did not suggest that between study heterogeneity was driven by differences in disease severity or year of publication (Table 3). There was some evidence that the increased odds of not obtaining a university degree or completing secondary education was more marked in studies from the Middle East compared to studies from Europe and North America, and that associations for these two outcomes were also stronger in women (Table 3). However, number of studies for subgroup analysis were limited. There was no strong evidence of publication bias (Additional file 1: Fig. S1a–c, Egger's P = 0.74; 0.94; 0.50 respectively for not obtaining a university degree, secondary education and vocational training).

The proportions with each educational outcome by country, in studies that did not report a peer non-CHD group, are compared to the summary data from ‘Education at a Glance’ in Table 4. For the majority, the proportions of each outcome in CHD patients were similar to the country level data for adults.

The pooled proportion of patients with CHD who completed a university degree, secondary education and vocational training was 36% [95% CI 30–43], 84% [95% CI 76–90] and 25% [95% CI 16–36] across all studies (Additional file 1: Fig. S4a–c). There was substantial between study heterogeneity and the predictive interval was 0.08 to 0.78 for obtaining a university degree, 0.23 to 0.99 for completing secondary education and 0.03 to 0.75 for completing vocational education.

The item most identified at risk of bias was confounding, due to parental ethnicity, education, or age, as studies either controlled only for patient age and sex or nothing (Additional file 1: Table S2).

The key strength of this study is our attempt, for the first time, to obtain and review all relevant data, including studies where the aim was to assess the association of having a CHD with educational attainment and those where this was not the aim. We acknowledge for the latter that our search strategy may have missed some studies where a description of educational attainment in patients with CHD was somewhere in the paper. We have presented predictive intervals, as well as odds ratios and confidence intervals, which are recommended when undertaking random effects meta-analyses because of assumed between study heterogeneity, but rarely undertaken [16, 73]. We attempted to standardize academic levels achieved whilst focusing on key measures that are related to future employment, socioeconomic position and health (university, secondary and vocational training). However, we acknowledge that across different educational systems the level of knowledge and skills required is likely to vary. Our results are limited by the sparsity of studies and the lack of any studies that have controlled for key confounding factors.

Research in this area is affected by the rarity of the conditions, which limits the possibility to undertake a robust analyses within the single birth cohorts. On the other hand, linkage between educational and health data has not been systematically performed across all countries. Despite recent advances in multidimensional data repositories, which may facilitate research in this area, large registries are unlikely to allow discrimination between the large spectrum of different types of CHD and their different impact on neurological development and educational attainment. Large birth cohort collaborations such as LifeCycle [74] can potentially offer the advantage of achieving a larger sample of patients with CHD [12] with granular longitudinal data and the possibility to investigate variability related to different countries and educational systems.