Background
Maternal insulin resistance during pregnancy is a normal and essential adaptation that ensures adequate nutrition to support the life of a growing fetus. Reference Hernandez, Friedman, Barbour, Zeitler and Nadeau1 Women with reduced pregravid insulin sensitivity compensate with an increased insulin response (hyperinsulinaemia), which affects early placental growth and gene expression. Maternal metabolism undergoes ongoing changes in insulin sensitivity mediated by circulating placental factors that drive excessive nutrient availability. Reference Catalano and Shankar2 When augmented by maternal insulin resistance, this increases the risk of many complications for both the mother and neonate. Reference Butte3 The global rise of insulin resistance syndromes such as obesity and polycystic ovarian syndrome among women of childbearing age has led to a growing interest in lifestyle-based strategies to mitigate the complications associated with maternal metabolic demands. Reference Hernandez, Friedman, Barbour, Zeitler and Nadeau1 Maternal hyperinsulinaemia is associated with a milieu of metabolic derangements, including hyperglycaemia, hyperlipidaemia and inflammation in the mother. Reference Tinius, Blankenship and Furgal4 These abnormalities are associated with an increased risk of developing gestational diabetes mellitus (GDM), hypertensive disorders, non-alcoholic fatty liver disease, cardiomyopathy, birthing complications, Reference Hernandez, Friedman, Barbour, Zeitler and Nadeau1,Reference Yogev and Catalano5 as well as a life-long increased risk of metabolic syndrome and type 2 diabetes. Reference Daly, Toulis and Thomas6
Maternal hyperinsulinaemia is known to have transgenerational impacts that beget childhood metabolic dysfunction. These genetic and epigenetic exposures in utero have lasting effects on the offspring’s later life metabolic health. Reference Hernandez, Friedman, Barbour, Zeitler and Nadeau1,Reference Juonala, Magnussen and Berenson7 Maternal insulin resistance has been shown to alter placental and fetal metabolism in ways that can lead to fetal overgrowth, endothelial dysfunction, and neurological disorders in the neonate. Reference Sobrevia, Salsoso and Fuenzalida8 Furthermore, alterations in the maternal insulin/insulin-like growth factor axis have been demonstrated in GDM pregnancies, Reference Zhu, Mendola and Albert9 and are thought to have an important role in mediating fetal outcomes. Reference Retnakaran10,Reference Gęca and Kwaśniewska11
The fetal response to maternal overnutrition is fetal hyperinsulinaemia, which mediates developmental pathways involved with growth, body composition, and mitochondrial function in the offspring. Reference Hernandez, Friedman, Barbour, Zeitler and Nadeau1 In pregnancies complicated by diabetes mellitus, the maternal hyperglycaemia-fetal hyperinsulinaemia response leads to an increased risk of developmental and metabolic complications such as neonatal hypoglycaemia and macrosomia (infant born > 4 kg). Although interventions that reduce occurrence of adverse neonatal events are well-established in GDM pregnancies particularly, 12 there is growing interest in the long-term developmental impacts of in utero hyperinsulinaemia for infants born to women with normal-to-borderline hyperglycaemia.
Evidence from the largest blinded multinational study on adverse pregnancy outcomes, the hyperglycaemia and adverse pregnancy outcome (HAPO) cohort study (23,316 participants) highlighted the breadth at which fetal hyperinsulinaemia can be observed in pregnancies across a spectrum of clinical and sub-clinical maternal hyperglycaemia. Reference Metzger, Lowe and Dyer13 The HAPO study showed an independent and continuous linear relationship between non-diabetic hyperglycaemia in pregnancy and cord serum c-peptide. Newborns with higher cord serum c-peptide (>90th percentile) were also larger and fatter and had a higher clinical incidence on neonatal hypoglycaemia. Reference Metzger, Persson and Lowe14 Cord c-peptide has repeatedly been found to mediate the relationship between maternal body mass index (BMI) and infant size, Reference Lee, Barr and Longmore15 and maternal hyperglycaemia and childhood adiposity, Reference Josefson, Scholtens and Kuang16 thus providing a plausible causal link between in utero exposures and later life metabolic diseases. Reference Muhlhausler, Gugusheff, Ong and Vithayathil17 This suggests all women at risk of fetal hyperinsulinaemia, such as those with obesity, may benefit from interventions to mitigate excursions in nutrient excess, not only those affected by GDM.
Various dietary interventions have been implicated for reducing the risk of adverse outcomes in GDM pregnancies. Reference Han, Middleton, Shepherd, Van Ryswyk and Crowther18 A meta-analysis by Yamamoto et al. Reference Yamamoto, Kellett and Balsells19 of 18 studies showed various modified dietary interventions including the low glycaemic index (GI) and the dietary approaches to stop hypertension (DASH) diet were associated with improved maternal glycaemia, lower neonatal birth weight and reduced macrosomia. Outside of GDM, various dietary interventions for pregnant women with overweight or obesity have been implicated for limiting gestational weight gain and preventing GDM incidence. Reference Lamminpää, Vehviläinen-Julkunen and Schwab20 Among them, effective dietary interventions involve nutrient or energy restriction alongside behavioural and physical activity components. It is plausible that such interventions which reduce GDM occurrence also create a more favourable metabolic environment for fetal development.
Despite cord blood metabolites providing a promising marker of in utero exposures and infant metabolic development health, to our knowledge, there have been no systematic reviews published in the last 5 years that summarise current evidence from dietary studies on fetal insulin metabolism outcomes. Since it is unclear whether cord-blood metabolites, specifically insulin and c-peptide, from non-GDM pregnancies can provide a sensitive indicator of sub-clinical in utero exposures, excess infant adiposity is proposed to be both a plausible and measurable mediator of transgenerational metabolic dysfunction. Reference Castillo, Santos and Matijasevich21,Reference Catalano, Thomas, Huston-Presley and Amini22 Neonatal adiposity appears to be tightly linked to in utero fetal insulin levels, Reference Stanley, Fraser, Milner and Bruce23 and altered cord-blood metabolites; reflective of altered fatty acid oxidation and mitochondrial dysfunction. Reference Kadakia, Scholtens and Rouleau24
For this review, we aimed to examine the relationship between different dietary interventions in metabolically healthy and high-risk pregnant populations and sub-clinical fetal hyperinsulinaemia measured as cord blood metabolites and neonatal adiposity.
Methods
This review was prepared in accordance with the preferred reporting items for systematic reviews and meta-analyses guidelines. Reference Liberati, Altman and Tetzlaff25 The review protocol was registered in the PROSPERO International prospective register of systematic reviews (ID CRD42020146453). The population, intervention, comparator, outcomes and study design criteria, used to define the research question and to select the studies, are presented in Table 1.
Table 1. Categories for formulation of the research question for a systematic review on effects of dietary interventions and associations of dietary intake with neonatal outcomes
Search strategy
The literature search was performed using MEDLINE, Web of Science, Cochrane Controlled Register of Controlled Trials, Cumulative Index to Nursing and Allied Health Literature, SCOPUS and SPORTDiscus using the keywords provided in Supplementary Table 1. The search was narrowed using filters of full text, peer-reviewed, journal articles, human, English language and publication day from November 1985 to January 2021.
Inclusion and exclusion criteria
Inclusion criteria were as follows: (i) randomised controlled trials (RCTs) involving a dietary intervention component that included either a macronutrient or dietary pattern modification, (ii) dietary intake assessment ascertained by validated food frequency questionnaire, multiple-day food diary, or 24-h dietary recall method; (iii) participants including pregnant adult women (aged >18 years, recruited at any point during their pregnancy) and their neonates up to 72 h from birth (for cord blood) and up to 6 months of age (for adiposity); (iv) articles from 1985 to present to capture lifestyle patterns of current times; (v) outcome measure examining fetal insulin secretion by cord blood metabolite analysis and/or infant adiposity. Valid outcome measures included cord blood c-peptide, insulin and/or glucose measured at birth, or infant adiposity measured within 6 months of age using four-compartment or two-compartment model methods, that is, air displacement plethysmography (ADP) or skin fold thickness (SFT).
Exclusion criteria were as follows: (i) research investigating food security or malnutrition; (ii) studies examining a nutritional supplement or supplemental food product; (iii) dietary intervention component inadequately described; (iv) full text article not available in the English language; (v) brief communications, case series, editorials, review studies; (vi) pilot intervention studies without a control group.
Study selection
Suitable literature was identified following a three-step screening process (Fig 1). All literature retrieved was collated into Endnote and duplicates removed, then subjected to preliminary (title and abstract) and full-text review. Literature retrieval was performed by the first author (SN). Two independent reviewers (SN and CT) subsequently filtered the identified articles by evaluating titles and abstracts, and subsequently full texts and references based on the inclusion and exclusion criteria. Additional articles identified by hand searching of reference lists from previous review and selected studies were also considered. Any discrepancy in assessment between reviewers was resolved through discussion and rechecking of the full text.
Fig. 1. Flow diagram of literature search.
Data extraction
Data were extracted from the included studies in the following domains: country; author; year; study design; number of participants; study groups; participant characteristics (maternal age, BMI, gestational age, gestational weight gain, neonatal birth weight and prevalence of large-for-gestational age); primary outcomes; intervention protocol including intervention content and compliance assessment, timing, outcome measure of interest and study findings. Due to the heterogeneity of populations included in RCTs, results from women with GDM, BMI ≥ 25 kg/m2, and ‘healthy’ populations were reported separately. Primary outcomes were reported as mean differences between groups (SD) and median (interquartile range), or relative risk if not otherwise available. Where mean differences were not reported in the manuscripts, between group differences were calculated using pooled standard deviations. Reference Altman26 In the instance where all data from included articles was not available, authors were contacted to obtain missing data.
Assessment of reporting quality
Risk of bias was assessed using the Cochrane Collaboration’s bias risk assessment tool, where bias was classified into six domains: selection, performance, detection, attrition, reporting and other bias. Reference Higgins, Altman and Gøtzsche27 The ‘other’ domain referred to dietary intervention compliance, which was classified as low risk when the study design included compliance measures to evaluate participant adherence to assigned dietary intervention. Examples of low compliance bias included telephone or face-to-face dietary review sessions throughout the intervention period, and validated forms of dietary assessments by 24-h food recall or 3-day food diaries. No dietary follow up, low participant adherence or the intervention group not achieving the intended dietary change was rated as a high risk of compliance bias.
Results
Study selection
The number of identified studies is shown in the flow diagram of the literature search in Fig. 1. Titles of 868 articles were found from the search result and 733 were retained after removal of duplicates. Among the 733 that were screened for title and abstract, 135 articles were subjected to full text revision. Articles excluded were not dietary RCTs, did not provide an assessment of the outcome measures of interest, were a protocol paper or review article.
A total of 14 articles from 11 RCTs involving 3614 pregnant women were included. The main characteristics from the included studies are summarised in Table 1. Five studies recruited women in their first trimester with a BMI ≥ 25 kg/m2. Reference Ferrara, Hedderson and Brown28–Reference Rhodes, Pawlak and Takoudes32 Mean participant BMI for these studies ranged between 28.0 and 34.4 kg/m2, which was reported pre-pregnancy Reference Ferrara, Hedderson and Brown28,Reference Van Horn, Peaceman and Kwasny31 or from first-trimester body weight. Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29,Reference Zhang, Wang and Yang30,Reference Rhodes, Pawlak and Takoudes32 In the DALI study women were selected with a pre-pregnancy BMI ≥ 29 kg/m2, Reference Harreiter, Simmons and Desoye33 and UPBEAT study, participants had a pre-pregnancy BMI ≥ 30 kg/m2. Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35 Women with a previous infant born weighing > 4 kg were recruited for one study (mean BMI 26.5–27.2 kg/m2 measured in the first trimester). Reference Walsh, Mahony, Culliton, Foley and McAuliffe36–Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38 Kizirian et al. Reference Kizirian, Kong and Muirhead39 included women with at least one risk factors for GDM. Two studies recruited women diagnosed with GDM. Reference Rae, Bond, Evans, North, Roberman and Walters40,Reference Mijatovic, Louie and Buso41 Among the remaining 12 studies, women were recruited in the first trimester and the prevalence of GDM ranged from 2 to 41%. Reference Ferrara, Hedderson and Brown28–Reference Kizirian, Kong and Muirhead39
Characteristics of interventions
The intervention characteristics are summarised in Table 3.
Intervention delivery
In five studies, the intervention was provided over between two and five individual dietitian- or nutritionist-led sessions. Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29–Reference Van Horn, Peaceman and Kwasny31,Reference Walsh, Mahony, Culliton, Foley and McAuliffe36–Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38,Reference Mijatovic, Louie and Buso41 Dietary interventions included home-based counselling to reduce sugar consumption alongside docosahexaenoic acid supplementation (2 × 2 factorial design), Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29 eucaloric low GI diet, Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35,Reference Walsh, Mahony, Culliton, Foley and McAuliffe36–Reference Kizirian, Kong and Muirhead39 eucaloric low GI diet versus a low fat diet, Reference Rhodes, Pawlak and Takoudes32 low glycaemic load (GL) diet with a participant-driven mobile app Reference Zhang, Wang and Yang30 ; calorie-restricted diet (30% restricted), Reference Rae, Bond, Evans, North, Roberman and Walters40 modestly lower carbohydrate diet (135 g/day) Reference Mijatovic, Louie and Buso41 ; and the DASH diet. Reference Van Horn, Peaceman and Kwasny31 Two studies included one-on-one motivational interviewing techniques with participants to facilitate behaviour change for weight management. Reference Ferrara, Hedderson and Brown28,Reference Harreiter, Simmons and Desoye33 The UPBEAT study involved eight weekly health-trainer led sessions on reducing dietary GL without restricting dietary energy intake. Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35 Compliance checks in addition to study visits were food diaries Reference Kizirian, Kong and Muirhead39,Reference Rae, Bond, Evans, North, Roberman and Walters40 or 24-h food recall, Reference Ferrara, Hedderson and Brown28,Reference Kizirian, Kong and Muirhead39,Reference Mijatovic, Louie and Buso41 telephone interviews, Reference Zhang, Wang and Yang30,Reference Van Horn, Peaceman and Kwasny31,Reference Harreiter, Simmons and Desoye33–Reference Patel, Godfrey and Pasupathy35 and a logbook with weekly goals in one study. Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35
Comparison group
Most studies described control groups as receiving usual care that included standard periodic antenatal visits and referral for medical care as indicated (i.e. following GDM diagnosis). Reference Ferrara, Hedderson and Brown28,Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29,Reference Van Horn, Peaceman and Kwasny31,Reference Harreiter, Simmons and Desoye33–Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38 Among them, two studies specified providing nutrition advice to the control group as a part of routine care in alignment with national guidelines. Reference Ferrara, Hedderson and Brown28,Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29 Five studies compared two dietary interventions where the intensity of the treatment was similar between both groups. These included a low GL vs non-low GL dietary intervention, Reference Zhang, Wang and Yang30 modestly lower- vs moderate-carbohydrate diet, Reference Mijatovic, Louie and Buso41 low GI vs high fibre, higher GI, Reference Kizirian, Kong and Muirhead39 energy-restricted and non-energy restricted diet, Reference Rae, Bond, Evans, North, Roberman and Walters40 and a low GI vs a low fat diet. Reference Rhodes, Pawlak and Takoudes32
Effects of interventions on dietary intake
Change in dietary intake or behaviour at the end of the intervention are described in Table 3. Women receiving the telehealth-delivered behavioural intervention for weight management in the GLOW trial reported lower energy intake and a lower incidence of excess gestational weight gain (intervention: 41% vs control: 66%, P < 0.001). Reference Ferrara, Hedderson and Brown28 In the DALI trial, Reference Harreiter, Desoye and van Poppel42 the intervention resulted in lower intake of sugar drinks, fat, and carbohydrate, as well as reduced portion sizes, sedentary time, and weight gain. In the GI Baby 4 RCT, Reference Kizirian, Kong and Muirhead39 the Low GI intervention group had a significantly lower dietary GI, while there was no difference in energy and other nutrients. In the MOM FIT RCT, the intervention group presented with significantly higher Dixon DASH and Fung DASH scores, and higher HEI 2010 scores indicating compliance to the prescribed DASH diet. Reference Van Horn, Peaceman and Kwasny31 In the ROLO study, the intervention group was shown to have a significantly lower dietary GI, GL and percentage energy from carbohydrates. Reference Walsh, Mahony, Culliton, Foley and McAuliffe36,Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38,Reference Donnelly, Lindsay, Walsh, Horan, Molloy and McAuliffe43 The study intended to provide eucaloric diets, however, the intervention group had lower energy intake in trimesters two and three. Compared to the control, intervention group from the UPBEAT study reported lower dietary GL, GI, total energy intake, total percentage energy from fat, saturated fat, and high percentage energy intake of protein. Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35 Rhodes et al. Reference Rhodes, Pawlak and Takoudes32 reported significantly lower dietary GI and GL in the Low GL intervention group. However, there was no significant difference in dietary fat (as a percentage of energy intake) when compared against the ‘Low Fat’ control group. Zhang et al. Reference Zhang, Wang and Yang30 showed no significant difference between study groups for dietary GI, however, a total reduction in GL, energy, and carbohydrate intake was observed in both groups over the duration of the study.
The following studies did not achieve the intended dietary change. The moderately energy restricted intervention by Rae et al. Reference Rae, Bond, Evans, North, Roberman and Walters40 indicated no significant difference between groups for total energy intake. In the moderately lower carbohydrate intervention study by Mijatovic et al., Reference Mijatovic, Louie and Buso41 the intervention group had significantly lower total energy, carbohydrate and protein intake. However, groups did not vary in percentage of dietary energy from carbohydrate, protein, fat, GI, GL, and total sugar, starch and fibre. Only 20% of participants in the intervention group were reported meeting the target carbohydrate intake compared with 65% in the control group. Garmendia et al. Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29 did not provide a post-intervention comparison of participant dietary intake in the main study findings. Adherence to the lifestyle intervention was defined as attendance to dietary counselling sessions, for which overall attendance was 69% and 31% attended all sessions.
Neonatal outcomes for each intervention
Cord blood c-peptide, insulin, and glucose
Six of the included studies examined neonatal cord blood. Among them, c-peptide was measured in five studies, Reference Ferrara, Hedderson and Brown28,Reference Zhang, Wang and Yang30,Reference Harreiter, Simmons and Desoye33,Reference Patel, Hellmuth and Uhl34,Reference Walsh, Mahony, Culliton, Foley and McAuliffe36 insulin was measured in three studies, Reference Ferrara, Hedderson and Brown28,Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29,Reference Patel, Hellmuth and Uhl34 and four studies reported glucose. Reference Ferrara, Hedderson and Brown28,Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29,Reference Harreiter, Simmons and Desoye33,Reference Patel, Hellmuth and Uhl34
In two studies which delivered a behaviour change intervention to facilitate healthy eating and physical activity for appropriate weight gain, there was no significant difference between groups for cord blood c-peptide, glucose Reference Harreiter, Simmons and Desoye33 or insulin. Reference Ferrara, Hedderson and Brown28 Garmendia et al. Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29 found no significant difference in cord blood glucose and insulin between participants who received home-based dietary counselling to reduce sugar consumption and controls who received no dietary education. In the RCT by Zhang et al., Reference Zhang, Wang and Yang30 when compared to a dietitian-prepared control diet, the low GL intervention had no significant effect on neonatal cord blood c-peptide. In the secondary analysis from the UK UPBEAT low GI study, Reference Patel, Hellmuth and Uhl34 there was no significant difference between groups in cord blood insulin or c-peptide. Similarly, the Irish ROLO study found no significant difference in median cord blood c-peptide among participants who received a eucaloric low GI diet provided over three dietitian-led education sessions. Reference Walsh, Mahony, Culliton, Foley and McAuliffe36
Neonatal adiposity
Eight of the included studies examined neonatal adiposity. Among them, two studies measured SFT at delivery, Reference Rhodes, Pawlak and Takoudes32,Reference Donnelly, Walsh, Byrne, Molloy and McAuliffe37 two studies used ADP at delivery, Reference Van Horn, Peaceman and Kwasny31,Reference Mijatovic, Louie and Buso41 two studies measured SFT at 6 months of age, Reference Patel, Godfrey and Pasupathy35,Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38 and one study used ADP at both delivery and 6 months of age. Reference Kizirian, Kong and Muirhead39
Among the four studies that investigated a low GL dietary intervention, the UPBEAT UK RCT, showed a reduction in subscapular SFT at 6 months of age (mean difference −0.38 mm (−0.70 to −0.06), P = 0.021), but not triceps SFT. Reference Patel, Godfrey and Pasupathy35 There were no significant associations between the intervention and adiposity outcomes in the ROLO low GI study, Reference Donnelly, Walsh, Byrne, Molloy and McAuliffe37,Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38 GI Baby 4 study Reference Kizirian, Kong and Muirhead39 ; MOM FIT DASH dietary intervention Reference Van Horn, Peaceman and Kwasny31 ; the MAMI 1 modestly lower carbohydrate intervention, Reference Mijatovic, Louie and Buso41 moderately energy restricted diet Reference Rae, Bond, Evans, North, Roberman and Walters40 and pilot low GL RCT. Reference Rhodes, Pawlak and Takoudes32
Quality assessment and Risk of Bias
Based on the Cochrane risk of bias assessment tool, Reference Higgins, Altman and Gøtzsche27 the majority of the included studies had a low risk of selection bias due to randomisation and blinded allocation to treatment groups Reference Ferrara, Hedderson and Brown28–Reference Rae, Bond, Evans, North, Roberman and Walters40 (Supplementary Table 2).
Due to the natural of dietary intervention studies, true binding of participants is not feasible. Performance bias also refers to personal involved in delivering the intervention. A low risk of performance bias was suggested in the four studies where the intervention and comparison group were provided with matched intensity. Reference Zhang, Wang and Yang30,Reference Kizirian, Kong and Muirhead39–Reference Mijatovic, Louie and Buso41 These studies did not inform participants on their intervention arm. A high risk of performance bias was identified in six studies where the control group received standard antenatal care and specific dietary input was not explicitly described. Reference Ferrara, Hedderson and Brown28,Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29,Reference Van Horn, Peaceman and Kwasny31,Reference Harreiter, Simmons and Desoye33–Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38
Most of the included studies reported blinding for outcome assessments (neonatal cord blood analyses and adiposity). Reference Ferrara, Hedderson and Brown28–Reference Kizirian, Kong and Muirhead39 Blinding during SFT assessments was unclear in one study. Reference Rae, Bond, Evans, North, Roberman and Walters40
A low risk of attention bias was identified in thirteen studies that used intention-to-treat analysis and adequately described participant withdrawals. Reference Ferrara, Hedderson and Brown28–Reference Kizirian, Kong and Muirhead39,Reference Mijatovic, Louie and Buso41 There was a high risk for attention bias in one study where missing data or drop-outs were not reported. Reference Rae, Bond, Evans, North, Roberman and Walters40 There was a low risk of reporting bias in 11 studies Reference Ferrara, Hedderson and Brown28–Reference Harreiter, Simmons and Desoye33,Reference Patel, Godfrey and Pasupathy35–Reference Mijatovic, Louie and Buso41 which reported all present outcomes from their protocol. One study did not specifically report cord blood insulin and c-peptide outcomes by intervention groups and the authors were contacted for these data, from which the unpublished results are included in Table 2. Reference Patel, Hellmuth and Uhl34 A low risk of compliance bias was identified in 12 studies which provided multiple points of participant contact and dietary assessments to ensure the intervention was adequately followed. Reference Ferrara, Hedderson and Brown28–Reference Zhang, Wang and Yang30,Reference Patel, Hellmuth and Uhl34–Reference Kizirian, Kong and Muirhead39,Reference Mijatovic, Louie and Buso41
Table 2. Study characteristics and results from randomised controlled trials
BMI, body mass index; CI, confidence interval; c-peptide, cord blood c-peptide measured on delivery day; DHA, docosahexaenoic acid; FM, estimated fat mass; GI, glycaemic index; GL, glycaemic load; g, grams; h, hour; IQR, interquartile range; LGA, large-for-gestational-age, defined as infants born >90th percentile; SFT-#, skin fold thickness and number of sites.
Study names: DALI, Vitamin D and lifestyle intervention for GDM prevention; GLOW, gestational weight gain and optimal wellness RCT; LIMIT, limiting weight gain in overweight and obese women during pregnancy to improve health outcomes; MAMI 1, macronutrient adjustments in mothers with gestational diabetes study 1; MIGHT, maternal obesity/overweight control through healthy nutrition; MOMFIT, maternal offspring metabolics family intervention trial; ROLO, randomised control trial of low glycaemic index diet in pregnancy; UPBEAT, UK pregnancies better eating and activity trial.
Values reported as mean ± SD, median (IQR) or otherwise indicated.
†Reported from an earlier publication on the same population.
‡Participant demographics reported from Donnelly et al. Reference Donnelly, Walsh, Byrne, Molloy and McAuliffe37 DOI: 10.1111/j.2047-6310.2013.00216.x.
¶Participants with a glycated haemoglobin >6% at 36 weeks gestation.
¶¶Participants treated with insulin in studies that included only women with GDM.
The dietary assessment tool used by Harreiter et al. Reference Harreiter, Simmons and Desoye33 introduced a potential risk of compliance bias; although there was frequent participant contact with a personal lifestyle coach as well as telephone contact, dietary assessments were self-reported according to a short (12-item) questionnaire and not analysed for nutritional intakes. The analysis was also not appropriate to compare adiposity outcomes due to low subject number. The UPBEAT RCT was considered to be low risk for compliance bias, however, it should be noted that the involvement of any dietitian in the control group by standard antenatal referral pathways was not reported although suggested to be minimal. Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35 Four studies which reported non-significant differences between groups for the intended dietary change at the end of the intervention received an overall ‘high’ risk for other bias. Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29,Reference Rhodes, Pawlak and Takoudes32,Reference Rae, Bond, Evans, North, Roberman and Walters40,Reference Mijatovic, Louie and Buso41 Although reporting no differences between intervention groups, Zhang Reference Zhang, Wang and Yang30 indicated a net effect of both study arms on dietary GI.
Discussion
Principal findings
This review aimed to systematically examine the effect of maternal diet on fetal hyperinsulinaemia measured at birth. Due to limited data assessing cord blood insulin or c-peptide (to indicate fetal hyperinsulinaemia), we also included neonatal adiposity as a surrogate marker for in utero hyperinsulinaemia. To our knowledge, this is the only review published in the last 5 years to summarise experimental evidence on the impact of maternal diet on fetal insulin metabolism and neonatal (up until 6 months of age) adiposity.
The main results from the included studies do not show any effect of an intervention on neonatal cord blood insulin, c-peptide, or glucose. Among the studies that measured neonatal adiposity, the only intervention that showed a significant reduction in adiposity was the Low GL diet from the UPBEAT study (subscapular SFT −0.38 mm, P = 0.021, subscapular SFT z-score −0.26, P = 0.031). Reference Patel, Godfrey and Pasupathy35 There was a trend toward improved neonatal adiposity outcomes in the ROLO study (low GI intervention) Reference Donnelly, Walsh, Byrne, Molloy and McAuliffe37 and DASH diet study. Reference Van Horn, Peaceman and Kwasny31
From the studies which showed no effect of the intended dietary intervention on the outcomes examined in this review, it was noteworthy that the majority of these did not achieve all aspects of the expected dietary change, Reference Zhang, Wang and Yang30,Reference Rhodes, Pawlak and Takoudes32,Reference Rae, Bond, Evans, North, Roberman and Walters40,Reference Mijatovic, Louie and Buso41 or did not report follow up dietary assessment information Reference Garmendia, Casanello, Flores, Kusanovic and Uauy29 (Table 3). Further, two studies were pilot studies with small participant group numbers, suggesting inadequate statistical power to detect differences between groups in neonatal adiposity. Reference Kizirian, Kong and Muirhead39,Reference Mijatovic, Louie and Buso41 Data included from the DALI trial was a secondary analysis, which authors reported was inadequately powered to detect differences in neonatal outcomes. Reference Harreiter, Desoye and van Poppel42
Table 3. Dietary intervention characteristics from randomised controlled trials
IoM guidelines: Institute of Medicine guidelines no more than 7 kg for women with pre-pregnancy BMI 25 · 0–29 · 9 kg/m2 or 5 kg for women with pre-pregnancy BMI 30 · 0 kg/m2 or higher.
DASH, dietary approaches to stop hypertension; FFQ, food frequency questionnaire; GI, glycaemic index; GL, glycaemic load; h, hour; IQR, interquartile range; SD, standard deviation; SEM, standard error from the mean.
Low GI dietary interventions are well-recognised as a feasible and effective strategy in the medical nutritional management of GDM. Reference Han, Middleton, Shepherd, Van Ryswyk and Crowther18 The most recent meta-analysis and systematic review of low GI diets provided to women with GDM clearly demonstrated that these diets improve glycaemic control, Reference Xu and Ye44 which is key for reducing the risk of maternal and neonatal complications. 12 Consistent with findings in our current review, the benefits of a low GI diet appear to exist wider than GDM pregnancies. In a meta-analysis by Zhang et al. Reference Zhang, Han and Chen45 authors concluded low GI diets provided during healthy pregnancies and to those with GDM were associated with a reduced incidence of infants born large-for-gestational-age, as well as improved maternal fasting and 2-h postprandial glucose levels.
A trend toward reduced adiposity was also observed in the UPBEAT trial during which women with obesity (BMI ≥ 30 kg/m2) were provided with a low GL intervention. Among participants where over a quarter of the women developed GDM, the intervention was associated with reduced neonatal subscapular SFT. Reference Patel, Godfrey and Pasupathy35 Similarly, the ROLO study showed a trend toward lower neonatal SFTs at delivery. Reference Donnelly, Walsh, Byrne, Molloy and McAuliffe37 Participants were women with a previous macrosomia pregnancy, but only 2% of the women developed GDM. It is plausible that low GL dietary strategies could improve glycaemic control in women with pre-existing insulin resistance, reducing the risk of fetal over-nutrition. The greater incidence of GDM among participants from the UPBEAT trial may explain significantly better outcomes among the participants who received this specific low GL intervention. This could be interpreted in two ways; women with greater baseline insulin resistance at the beginning of pregnancy have larger capacity to improve metabolic markers following lifestyle changes. Secondly, interventions provided through usual GDM care may have an independent treatment effect on fetal development thereby being additive to low GL dietary advice. Interventions for women with GDM including glucose monitoring, pharmaceutical interventions (oral hypoglycaemics or insulin) and increased obstetric monitoring all aim to reduce the risk of fetal hyperinsulinaemia and related obstetric complications. 12
A trend towards a lower neonatal body fat percent among women receiving a DASH dietary intervention is consistent with results from the recent systematic review and meta-analysis. Reference Li, Gan and Chen46 Among six studies providing a DASH dietary intervention to pregnant women, the DASH diet was associated with a decreased risk of pre-eclampsia, macrosomia, large-for-gestational-age infants and overall lower ponderal index. Although the DASH diet was found to have a significant lowering effect on maternal fasting plasma glucose, the maternal homeostasis model assessment of insulin resistance was unaffected. Consumption of a DASH diet may improve maternal glucose levels by having a lower glycaemic impact. Improvements in maternal glycaemia may prevent nutrient excess, fetal hyperinsulinaemia and resulting excess growth. Reference Ornoy47 This supports findings from our systematic review to suggest that adherence to a glycaemic DASH diet in women with an elevated pre-pregnancy BMI could reduce neonatal adipose development that is associated with the risk of macrosomia and large-for-gestational-age babies.
Despite limited associations between any one type of intervention and neonatal outcomes, our findings point to an important role of dietary education. There were associations between interventions and lower estimated neonatal fat mass percent and SFT in studies that compared a personalised lifestyle intervention to standard antenatal care (no specific nutrition or lifestyle advice). Reference Van Horn, Peaceman and Kwasny31,Reference Patel, Godfrey and Pasupathy35,Reference Donnelly, Walsh, Byrne, Molloy and McAuliffe37 Common elements from these interventions included dietary personalisation with a dietitian or ‘health trainer’, written education resources, and multiple points of contact to reinforce the provided advice and promote adherence. However, in studies that provided the comparison group with a similar level of support, for example, personalised meal plan, written resources, compliance checks, Reference Zhang, Wang and Yang30,Reference Kizirian, Kong and Muirhead39,Reference Mijatovic, Louie and Buso41 no significant effect of a low GI, low GL, or moderately carbohydrate restricted diet was observed. What this suggests is that personalised dietary advice, including follow up support and goal setting throughout gestation, may be just as, if not more, important than the specific dietary intervention alone. From the included studies, the interventions did not differ significantly in terms of practical dietary changes recommended (i.e. reduce sugar, increase fruit and vegetable intake, unrefined carbohydrate choices), making it difficult to distinguish any specific effect of one dietary change. These findings also highlight that general healthy eating education may be overlooked during usual antenatal care, particularly for women who carry risk factors for insulin resistance but are not diagnosed with GDM. Individualised advice allows women to understand their risk factors, respond to feedback throughout pregnancy, set measurable goals and make realistic behavioural adjustments. Communicating health advice within the context of an individual’s lifestyle is key to supporting healthy behaviours which could impact maternal, infant, and life-long family metabolic health. Reference Procter and Campbell48
A further explanation for the lack of associations between any of the included dietary interventions and cord blood metabolites (c-peptide, insulin, glucose) may be due to challenges associated with the metabolic analyses used. Although umbilical cord blood is considered one of the most useful samples in neonates instead of early peripheral blood examination, Reference Wang, Eerdun, Dong, Hao and Li49 insulin measurement has limitations since degradation increases in the presence of slight haemolysis. Reference O’Rahilly, Burnett, Smith, Darley and Turner50 While c-peptide is a more stable and useful marker of fetal metabolic exposures, Reference Lee, Barr and Longmore15 no difference in c-peptide between intervention and control groups was observed in any of the five studies with this outcome measure. This suggests that these biomarkers may not be suitable to determine the specific metabolic impact of maternal dietary modifications in women without gestational hyperglycaemia.
Previous research suggests adipokines may be a more sensitive marker of placental-fetal nutrient transfer. In the HAPO study, lower cord blood levels of adiponectin and c-reactive protein were associated with a higher neonatal adiposity. Reference Lowe, Metzger, Lowe, Dyer, McDade and McIntyre51 Adiponectin has been negatively associated with birth weight and estimated percentage fat mass. Reference Lowe, Metzger, Lowe, Dyer, McDade and McIntyre51 Leptin has also been linked to adipose development and insulin metabolism, Reference Telschow, Ferrari and Deibert52 which was identified in a secondary analysis from the ROLO study, as being associated with greater neonatal adiposity, while fetal c-peptide was not significant after adjustments. Reference Donnelly, Lindsay, Walsh, Horan, Molloy and McAuliffe43
Strengths and Limitations
A strength of this review is the inclusion of studies from diverse populations of pregnant women from varying cultural backgrounds. Studies reviewed included pregnant women with varying risk factors including GDM and obesity. The potential for replication bias introduced from five individual references from two study populations (UPBEAT, Reference Patel, Hellmuth and Uhl34,Reference Patel, Godfrey and Pasupathy35 ROLO Reference Walsh, Mahony, Culliton, Foley and McAuliffe36–Reference Horan, McGowan, Gibney, Byrne, Donnelly and McAuliffe38 ) is a limitation. Heterogeneity among the sampled participants may also limit the practical application of diet advice to specific populations. It is also possible that participant heterogeneity may have confounded a specific effect of individual dietary changes. Among the included studies, maternal insulin resistance was not exclusively examined in the recruitment criteria. This is important because maternal insulin resistance before pregnancy is a fundamental determinant of placental-fetal metabolism, nutrition status, and subsequent fetal development. Reference Hernandez, Friedman, Barbour, Zeitler and Nadeau1 Variability in the prevalence of GDM diagnosis among the participants in the included studies is a limitation. GDM diagnosis and treatment has the potential to influence lifestyle behaviours through additional monitoring and interventions. A higher prevalence of GDM in the included studies where a significant effect of the intervention on neonatal adiposity was observed limits our interpretation. Also noteworthy is the possibility for confounding obstetric variables such as time in labour, placental attachment, cord blood clamping to influence cord blood analysis.
Recommendations for future research
Future research investigating the impact of lifestyle interventions should consider a multi-modal approach to address maternal insulin resistance syndrome as it impacts placental-fetal metabolism. This means dietary factors must be considered alongside other environmental determinants such as physical activity, gestational weight gain, sleep quality, smoking and maternal stress. Reference Gaillard, Wright and Jaddoe53 To reduce heterogeneity from lifestyle factors within studies, RCTs may consider a multi-level intervention design to provide women relevant, personalised interventions based on pre-pregnancy behaviours.
Dietary intervention studies require specific dietary prescriptions that are measurable and repeatable in clinical practice. This means that research examining maternal dietary intake requires thorough, valid dietary assessments, such as 3-day food diaries or the use of diet tracking apps throughout gestation with attention to macronutrient intake, types of carbohydrates and micronutrient adequacy. We recommend that future studies examining neonatal metabolic outcomes should be assessed alongside first trimester maternal insulin resistance to better understand early gestation developmental impacts among mothers with varying patterns of gestational hyperinsulinaemia. Reference North, Zinn and Crofts54 Consistent methodology for examining neonatal metabolites at birth must be considered. This includes accounting for technical variables such as neonatal cord clamping and time in labour.
Conclusions
This systematic review indicates gaps in experimental evidence to demonstrate a specific relationship between dietary interventions during pregnancy to prevent subclinical fetal hyperinsulinaemia, measured by cord-blood metabolites and neonatal adiposity. Limited evidence from RCTs suggests dietary strategies that are known to improve glycaemic control in GDM pregnancies may have a protective effect against excess adiposity by a similar mechanism. Additionally, findings from RCTs suggest that antenatal dietary counselling appears to have a protective effect and should be offered to all pregnant women with overweight or obesity. The reviewed studies identified challenges with dietary research where an impact on sub-clinical metabolic changes may be confounded by lifestyle factors and participant adherence issues. Future large dietary RCTs should consider a multi-modal design to explore the specific effect of dietary modification, particularly in the context of environmental risk factors for maternal insulin resistance such as physical activity, stress and sleep. Further research is needed to confirm if specific nutrient modifications, independent of energy balance, in non-GDM pregnancies reduce the risk of fetal hyperinsulinaemia and fetal overgrowth in women with underlying insulin resistance.
Supplemental Tables
To view supplementary material for this article, please visit https://doi.org/10.1017/S2040174421000623
Acknowledgements
None.
Financial support
None.
Conflicts of Interest
The authors declare no conflicts of interest.
Ethical standards
None.
