The search strategy was developed with a medical librarian (19). We searched PubMed, Embase, Cochrane Database (CENTRAL) and Web of Science databases. For each database a specific search strategy was developed using subject headings and free text related to “cardiac function”, “infants”, “children”, and “diabetes”. The search was performed on June 27nd 2023. The complete search strategy is available as Supplementary Table S1.

Screening of titles and abstracts of the identified studies was done independently by two reviewers (CBS and AM). Disagreements were resolved by discussion, or if consensus was not reached, by consultation with a third reviewer (KJK).

Data from eligible studies was extracted using a modified data collection template based on the Cochrane Consumers and Communication Review Group Data Extraction Template (20). Two reviewers (CBS and AM) extracted data independently. Authors were not contacted for missing information.

The following data was extracted from included studies: title, first author, country, journal, publication year, conflicts of interest, study design, methodology, facility, study period, type of ultrasound equipment used, single or repeated outcome acquisition, recruitment, adjustment for confounders, number of cases and controls, in- and exclusion criteria, maternal characteristics (type and treatment of diabetes, HbA1c, body mass index, age at birth), and offspring characteristics (gestational age, birth weight, postnatal age at outcome acquisition, gender, co-morbidities).

Quality assessment was performed by two reviewers (CBS and AM) independently using a modified version of the Newcastle-Ottawa Scale (NOS) specific for quality assessment of observational studies (20, 21). Disagreements were resolved by discussion. If no consensus was reached, a third reviewer (KJK) was consulted.

Based on their design, studies were divided into two groups (case-control, and cohort studies). Each study in the respective groups was awarded points in three categories for a maximum of nine points: (1) Selection (0–4 points) (2) Comparability (0–2 points) (3) Outcome/Exposure (0–3 points). Items in each category are listed in Table 2. Studies scoring 7–9 points were evaluated as high quality, 4–6 points as high risk of bias, and 0–3 points as very high risk of bias.

Within each cardiac function category the most reported echocardiographic parameter was used for synthesis of results across studies. For the four most widely reported ultrasound parameters of cardiac function (IVSd, LVEF, MPI and LV E/A) in the first week of life outcomes were extracted as mean ± standard deviation (SD) for meta-analysis. Where no SD was reported, an average SD was imputed to enable inclusion in the meta-analysis. Where separate means and SDs were listed for subgroups of diabetes a combined data group was made for any type of diabetes by decomposing mean and SD (45, 46). Meta-analysis was performed in Review Manager 5.4.1 1 (The Cochrane Collaboration, 2020) using a random effect model and reported as mean differences between infants of mothers with any type of diabetes vs. infants of healthy control mothers. Between study heterogeneity was assessed using the I² statistic (47). The heterogeneity of remaining results did not allow for meta-analysis. As no summary measure was applicable across all studies, we aggregated results for each cardiac function category. According to Popay et al (48), a tabulated and narrative summarisation of results was done, as in previous systematic reviews by our group and others (49, 50). Study results were summarized for each outcome grouped by age at outcome acquisition (first week of life, 1–6 months, or 2–8 years). Analysis was subgrouped by age group as the included outcome measures are known to change with age. Further, while myocardial hypertrophy is usually thought to be transient with spontaneous resolution in the first months of life, it is not known to what degree functional cardiac changes persist into later life and whether they influence the risk of early onset cardiovascular disease. Echocardiographic outcome assessments for the individual studies were categorized as decreased, increased or no difference. An overall conclusion was made for each outcome in each age group, based on the number of studies reporting an increase, decrease or no difference. If the number of studies was equally distributed among the categories, the highest number of cases determined the overall conclusion. Type of maternal diabetes was described. Treatment of diabetes and levels of HbA1c during pregnancy was summarized as average values and percentages. Publication bias was assessed graphically by funnel plots when 10 or more studies were eligible for meta-analysis (47).

Comparing the absolute echocardiographic measurements (Supplementary Table S4) to established normative values (Z-scores) provides a context for interpreting the cardiac function of IDMs. For the purpose of generating example Z-scores we used, for the first week of life: 3.5 kg, 52 cm, BSA 0.23 m²; for 1–6 months of age: 6 kg, 60 cm, BSA 0.3 m², for 1–8 years: 16 kg, 100 cm, BSA 0.66 m². Based on these parameters example Z-score means (Z = 0) were retrieved from the “Cardio Z” App for IVSd (51) LV E/A (52) and MPI (53). First week of life: IVSd (3.7 mm), LV E/A (1.25), MPI (0.34); 1–6 months of age: IVSd (4.0 mm), LV E/A (1.24), MPI (0.34); 1–8 years: IVSd (5.1 mm), LV E/A (1,6), MPI (0.35). Normal pediatric LVEF is 56%–78%.

We identified 31 studies which met the inclusion criteria and were eligible for inclusion in the review (Figure 1) (1–7, 9, 22–44). Ten studies were observational cohort studies with a control group, and 21 were case-control studies. In total, the studies reported on 1,679 diabetic pregnancies and 2,694 healthy control pregnancies. Six studies (5, 26, 27, 31, 36, 44) repeated measurements on the same cases and controls at different ages. Thus, the final dataset included 2023 echocardiographies from cases and 2,901 echocardiographies from healthy controls. The total number of cases and controls per study are presented in Supplementary Table S2.

Of all cases 100 women had T1DM, 28 had T2DM, 1,250 had GDM, and in 301 cases the type of diabetes was not described. Of the 1,679 diabetic mothers, 218 were treated with diet alone (12.5%), 66 were treated with oral anti-diabetic medication (3.8%), 227 were on insulin therapy (13.0%), and 16 received a combination of the three therapy methods (0.9%). For the remaining 1,152 women, diabetes treatment was not specified (69.7%). Two out of three studies reporting on oral anti-diabetic medicine reported the type of anti-diabetic medicine: metformin in 24 women (25), and glyburide in 17 women (4). Both groups of women treated with specific anti-diabetic medication had GDM. In 18 studies, glycaemic control was not reported as individual HbA1c levels, but defined as good control, poor control, or no control at all. One study reported HbA1c as a correlation with aortic stiffness (27). In the 12 remaining studies (3, 4, 7, 23, 25, 26, 28, 30, 32, 34, 36, 38) the median HbA1c in the diabetic women was 7.0% across studies. Data on diabetes treatment and levels of glycaemic control in the individual studies are presented in Supplementary Table S3.

Twenty-three studies performed echocardiographic measurements on infants in the first week of life (1–5, 9, 22–24, 26–38, 40). Four studies performed measurements in the first 1–6 months of life (5, 25, 26, 36). Seven studies performed measurements on children at the age of 2–8 years (6, 7, 39, 41–44). Three studies performed repeated measurements in the first week of life and at 1–6 months (5, 26, 36), and one study repeated measurements two times at 1–8 years (44). All extracted echocardiographic outcome measurements in each age group are presented in Supplementary Table S4.

ADA diabetes classification criteria were used in six studies for PDM and GDM (5, 28, 35, 37, 38, 43), NICE criteria were used in one study for GDM (25), and WHO criteria were used in three studies for GDM (2, 23, 42). Four studies used local diagnostic criteria for PDM and GDM (3, 32, 34, 40), and 17 studies did not address criteria for diabetes diagnosis at all. Evaluation of the differences between types of diabetes and its relation to functional cardiac changes in the offspring was not possible. Seventeen studies made a distinction between types of diabetes, while 14 studies were not consistent in the selection of diabetes types when comparing cases to controls in relation to functional cardiac changes (3–5, 9, 27, 30, 32–34, 39–42, 44).

Of the 1,679 cases identified in the review, only 100 (6%) infants were born to mothers with T1DM and 28 (1,7%) infants were born to mothers with T2DM. The majority of infants, 1,250 (74%), were born to mothers with GDM, while in 301 (18%) infants the type of maternal diabetes was not specified. The total number of cases used for different types of diabetes are summarized in Supplementary Table S2.

Points awarded to each study by NOS are summarized in Table 2. The representativeness of infants born to mothers with diabetes (IDMs) was appropriately described in 29 of 31 studies and the controls were selected from the same hospitals as the cases in 29 of 31 studies. The echocardiographic measures used were reported in all 31 studies. Eighteen studies reported maternal HbA1c and/or diabetes therapy. Analysis of echocardiography was blinded in 12 studies (1, 4, 6, 7, 25, 28, 31, 37, 38, 41, 43, 44). Five studies adjusted for potential confounders such as race, birth weight, maternal age and time elapsed from delivery to postnatal outcome acquisition (3, 25, 41, 43, 44). IDMs were comparable with the control group with respect to postnatal age at outcome acquisition in all studies. Twenty-eight studies were scored as high quality, and three studies were scored as high risk of bias. No studies scored as very high risk of bias. A random effects model was used to incorporate heterogeneity in meta-analysis. Funnel plots based on IVSd and LVEF metanalysis were symmetric but with horisontal scatter, suggesting that the heterogeneity fit with the assumptions of the model, and raising no major concern about publication bias (54). The sources of heterogeneity could not be discerned from the funnel plots. Funnel plots are included as Supplementary Figures S1 and S2.

The most reported echocardiographic parameter within each cardiac function category are summarized according to postnatal age in Tables 3, 4. The calculations used for summarized conclusions are available in Supplementary Table S7. Increased IVSd, decreased LVEF, increased MPI, decreased LV E/A ratio and/or increased PAP indicate abnormal cardiac function.

Of note, outside of the most reported parameters used for formal synthesis of results, five studies reported on LV global strain by 2D speckle-tracking echocardiography (2, 5, 24, 25, 41), two of these also on RV global strain (5, 24). Both LV and RV global strain was reduced in IDMs compared to controls in the 1st week of life and at 1–6 months.

Meta-analysis was possible for the primary outcome category of IVSd, and for LVEF, LV E/A-ratio and MPI in the first week of life. Twenty studies were included in the meta-analysis (1–5, 9, 22–24, 26–35, 40). Forest plots comparing each outcome between IDMs and healthy control mothers were made (Figures 2–5). IVSd was significantly increased in IDMs compared to controls [mean difference 1.0 mm (95% CI, 0.54–1.46 mm)]. LVEF in IDMs was similar to controls [mean difference 0.2% (95% CI, −1.0 to 1.4%)]. MPI was significantly higher in IDMs than in controls [mean difference, 0.06 (CI 95%, 0.02–0.09)]. LV E/A ratio was significantly decreased in IDMs compared to controls [mean difference −0.15 (CI 95%, −0.27 to −0.04)]. In the analysis of all four parameters considerable heterogeneity between studies was found by I² statistic.

The absolute echocardiographic measurements (Supplementary Table S4) for IDMs compared to the example Z-score means calculated in the methods section were: In the first week of life: IVSd Z > 0 in 14/17 studies, LV E/A Z < 0 in 12/12 studies, MPI Z > 0 in 4/5 studies, LVEF in the normal range in 10/11 studies. At 1–6 months: IVSd Z > 0 in 5/5 studies, LV E/A Z < 0 in 1/3 studies, MPI Z > 0 in 2/2 studies, LVEF in the normal range in 5/5 studies. At 1–8 years: IVSd Z > 0 in 2/6 studies, LV E/A Z < 0 in 1/5 studies, MPI Z > 0 in 2/2 studies, LVEF in the normal range in 2/4 studies.

This systematic review summarizes the association between maternal diabetes and altered cardiac function in offspring assessed by echocardiography. Based on observational studies comparing IDMs and control infants in the first week of life there is an association between maternal diabetes and the primary outcome category of morphology/myocardial hypertrophy (increased IVSd), as well as impaired diastolic function (decreased LV E/A), impaired global ventricular function (increased MPI), and impaired pulmonary hemodynamics (increased PAP and PDA). No change was found in systolic function (LVEF). The summarized conclusions on IVSd, LV E/A, MPI, and LVEF in the first week of life were confirmed in the meta-analysis. Results on cardiac function after the first week of life are based on comparatively small datasets and must therefore be interpreted with caution. For children aged 1–6 months there was an association between maternal diabetes and cardiac function assessed by ultrasound for the primary outcome of morphology/hypertrophy measured by increased IVSd as well as for MPI. No studies reported data for children from 6 months up to 1 years. For children from 1 to 8 years there was no difference between groups in IVSd or MPI, LVEF or LV E/A ratio.

Increased IVSd is rare in infants from normal pregnancies (9, 23, 34), but cardiomyopathy and septal hypertrophy has previously been reported to be present in about 30% of IDMs born with cardiac function abnormalities (2). The pathophysiology of maternal diabetes in relation to cardiac function of the infant is multifactorial and complex, and relates to the maternal glycaemic control (55). One of the main effects of maternal diabetes is the hyperglycaemic environment which leads to foetal and neonatal hyperinsulinism. Postnatal hyperinsulinism is usually a transient event for IDMs (56). This may explain why IVSd was significantly increased in both the meta-analysis and the overall conclusions of cardiac function in IDMs in the first week of life, but not later in life, a finding which is supported by a previous review of the role of hyperinsulinism in neonatal cardiac hypertrophy (56).

MPI is an indicator of overall cardiac performance including both systolic and diastolic function (2, 23). In our meta-analysis MPI was found to be higher in cases than in controls indicating a global worsening of the ventricular function. The clinical utility for MPI in childhood has not been fully established, and caution must be taken into consideration when characterizing MPI. The interpreting of MPI changes is loading dependent, as MPI increases with increased afterload and reduced preload. Still, long isovolumic phases generally imply worse function of the heart. High MPI at birth can decrease and normalize with maturation and reduced afterload over time, as seen in neonates born to a mother without diabetes (57). Samanth et al. found that MPI in infants of poorly controlled diabetic mothers was significantly higher than in infants of well controlled diabetic mothers indicating a correlation between the maternal glycaemic control and the ventricular function in IDMs. However, Al-Biltagi et al., did not find this correlation between increased MPI and maternal HbA1c.

LVEF was not found to be different for IDMs compared to controls. LVEF is very dependent on preload and afterload. In heart failure, LV dysfunction may frequently be identified by markers of deformation, when LVEF is normal. EF is further influenced by ventricular geometry, and increased wall thickness may contribute to preserved EF despite reduced longitudinal and circumferential shortening (58). MPI, stroke volume and cardiac output, together with indices of myocardial deformation, are likely to be more reliable quantifiers of LV systolic function.

Decreased mitral E/A ratio was seen for IDMs in their first week of life and was significantly decreased compared to healthy controls in the meta-analysis. Diastolic dysfunction in infants involves a reduction of the E wave and an increase of the A wave contributing to the inversion of the E/A ratio. The result is a value lower than 1, and could represent the first manifestation of diabetic cardiomyopathy (30). Even for women with good glycaemic control LV E/A ratio is found be decreased in IDMs (23, 30). The E/A waves are influenced by several factors including age, preload, afterload, and heart rate. In adults E/A is used as an indicator in the context of heart failure, its role as function index in children and neonates awaits further clarification (57). We were unable to investigate an association between glycaemic control and LV E/A ratio in this review due to lack of data.

PAP and PAAT were investigated as markers of pulmonary hemodynamics. PAAT was only reported in one study and was not included in our analysis. PAP can be estimated from tricuspid regurgitant flow velocity and shunt flows through ductus arteriosus (24). The pulmonary hemodynamics can be complicated by the presence of a PDA and/or atrial shunting. Increased atrial flow from left to right may suggest increased pulmonary flow if pulmonary pressures are normal (40). Increased PAP was found for IDMs within the first four days of life. PAP decreased to normal values in all cases at 72 h of life, but it was seen that the normalization was slower in cases than in controls (2, 35, 40). This transient disturbance in the adaptation of pulmonary circulation postnatally may be related to hypoglycemia in IDMs, delayed pulmonary maturation with respiratory morbidity and myocardial hypertrophy.

A delayed decrease in PAP may be contributed to by an increase of C-peptide concentration and beta-cell hyperplasia as seen in macrosomic IDMs. High leptin levels in the umbilical cord could also delay a fall in PAP, as leptin is positively associated with birthweight, indicating that macrosomic IDMs have a delayed normalization of the hemodynamic balance due to leptin-increased sympathetic activity in the kidneys and stimulation of the vascular smooth muscle proliferation (35).

PDA was more frequent in cases (21,4%) compared with controls (11,9%) and persisted longer in cases. Usually, a PDA closes within the first 12 h of life, but for IDMs PDA was frequently registered open until 72 h of life. Persistence of the ductus arteriosus could also contribute to the delayed transition of pulmonary hemodynamics and changes in myocardial compliance caused by an alteration of diastolic filling with an increased intraventricular pressure due to increased blood flow to the pulmonary circulation and left atrium.

Outside of the most reported parameters used for formal synthesis of results, myocardial strain was assessed in five studies which found reduced LV and RV global strain in IDMs compared to controls. Strain measures myocardial deformation, and varies with age, sex and loading conditions of LV and RV (5).

Comparing absolute echocardiographic measurements in IDMs to established normative values (Z-scores) in the first week of life we found IVSd Z > 0, LV E/A Z < 0 and MPI Z > 0 for the majority of studies. At 1–6 months we found IVSd Z > 0, and MPI Z > 0 for the majority of studies. At 1–8 years only MPI Z > 0. For all age groups LVEF was primarily in the normal range. Thus, comprison with example Z-scores for each age group supported the overall conclusions on cardiac function according to postnatal age in Table 3.

Diabetes in pregnancy is associated with offspring cardiac hypertrophy, the primary outcome category in this systematic review, and with impaired diastolic and global cardiac function in infants up to one week old. Most of the studies including older children found either no difference or had small numbers of cases and control. There is very little data on pre-gestational diabetes, which may have significantly different impact on offspring cardiac function compared with GDM. Further, there is little to no information on right ventricle function, which may be where the most important short- and long-term effects are seen. Limitations to the current evidence means that more knowledge is required to understand the association between maternal diabetes in pregnancy and long-term cardiac health. To develop strategies to mitigate the risk of future cardiovascular complications we need longitudinal studies with larger sample sizes and more accurate information on diagnostic criteria, maternal diabetes type, diabetes therapy and glycaemic control. Better glycemic regulation and optimized medical treatment of mothers with diabetes will improve infant outcomes, however there is likely an effect of maternel diabetes regardless of glycemic control. Some studies point towards insulin as a metabolic effector targeting the cardiomyocytes and fetal echocardiography to screen for possible birth defects is advised regardless of the type of diabetes. Maternel diet, tailored antihypertensive treatment and patient education may further improve pregnancy outcomes. We still need to learn more about infant cardiac function, metabolic status, growth and possible modifiable factors after birth. Applying more advanced echocardiographic measures or cardiac MRI may detect subtle changes with potential important long-term consequences, e.g., in myocardial remodeling and contractility. Increasing our knowledge on cardiovascular health in infants born to women with diabetes is important to improve pregnancy care, and to tailor the best possible follow up for these children.