Association between maternal hyperglycemia in pregnancy and offspring anthropometry in early childhood: the pandora wave 1 study

Background In-utero hyperglycemia exposure influences later cardiometabolic risk, although few studies include women with pre-existing type 2 diabetes (T2D) or assess maternal body mass index (BMI) as a potential confounder. Objective To explore the association of maternal T2D and gestational diabetes mellitus (GDM) with childhood anthropometry, and the influence of maternal BMI on these associations. Methods The PANDORA cohort comprises women (n = 1138) and children (n = 1163). Women with GDM and T2D were recruited from a hyperglycemia in pregnancy register, and women with normoglycemia from the community. Wave 1 follow-up included 423 children, aged 1.5–5 years (median follow-up age 2.5 years). Multivariable linear regression assessed associations between maternal antenatal variables, including BMI and glycemic status, with offspring anthropometry (weight, height, BMI, skinfold thicknesses, waist, arm and head circumferences). Results Greater maternal antenatal BMI was associated with increased anthropometric measures in offspring independent of maternal glycemic status. After adjustment, including for maternal BMI, children exposed to maternal GDM had lower mean weight (−0.54 kg, 95% CI: −0.99, −0.11), BMI (−0.55 kg/m2, 95% CI: −0.91, −0.20), head (−0.52 cm, 95% CI: −0.88, −0.16) and mid-upper arm (−0.32 cm, 95% CI: −0.63, −0.01) circumferences, and greater mean suprailiac skinfold (0.78 mm, 95% CI: 0.13, 1.43), compared to children exposed to normoglycemia. Adjustment for maternal BMI strengthened the negative association between GDM and child weight, BMI and circumferences. Children exposed to maternal T2D had smaller mean head circumference (−0.82 cm, 95% CI: −1.33, −0.31) than children exposed to normoglycemia. Maternal T2D was no longer associated with greater child mean skinfolds (p = 0.14) or waist circumference (p = 0.18) after adjustment for maternal BMI. Conclusions Children exposed to GDM had greater suprailiac skinfold thickness than unexposed children, despite having lower mean weight, BMI and mid-upper arm circumference, and both GDM and T2D were associated with smaller mean head circumference. Future research should assess whether childhood anthropometric differences influence lifetime cardiometabolic and neurodevelopmental risk.

There was an overlap period of approximately two years following the introduction of the new guidelines during which time women were diagnosed with GDM by either of the above guidelines, with formal change over to the new WHO guidelines across the NT in January 2015.Diagnosis of GDM based on these new guidelines included any of the following, a fasting glucose level ≥5.1mmol/L, 1 hour plasma glucose ≥10.0mmol/L or 2 hour plasma glucose ≥8.5mmol/L.Of the PANDORA cohort, 10.3% satisfied only the ADIPS glucose thresholds, 11.5% satisfied only the WHO glucose thresholds and 76.6% satisfied both.
Women were classified as having pre-existing type 1 or type 2 diabetes in pregnancy using WHO criteria, 3 defined as diagnosis using standard diagnostic criteria (OGTT or HbA1c) before the index pregnancy and confirmed on the medical record.Women were classified as having normoglycemia in pregnancy if they did not meet criteria for either GDM or pre-existing diabetes in pregnancy.
An issue when assessing women with GDM is determining women with true GDM as compared to women who have undiagnosed pre-existing type 2 diabetes who were not tested (and so not diagnosed) until the index pregnancy.As per IADPSG 2 and WHO 3 guidelines, women diagnosed with GDM, but meeting glucose or HbA1c values diagnostic of T2D outside of pregnancy were sub-classified as having "diabetes mellitus in pregnancy".This is defined as fasting plasma glucose ≥ 7.0 mmol/L and/or 2-hour plasma glucose ≥ 11.1 mmol/L, or HbA1c ≥ 48mmol/mol (6.5%).

PANDORA Wave 1 participants
A subgroup of women from the PANDORA birth cohort were invited to participate in the PANDORA Wave 1 follow-up study.Eligible children for Wave 1 (Figure 1) were aged 1.5-5 years and from five groups, classified by maternal glycemic status and ethnicity.Women with type 1 diabetes (n=18) and Europid women with T2D (n=9) were not eligible for Wave 1 due to small numbers, noting that T2D is uncommon in pregnant Europid women across Australia.To enable direct comparison between First Nations and Europid populations, women of other ethnicities (non-Europid, non-Indigenous) were also ineligible.There are 884 women with either type 2 diabetes or GDM within PANDORA, of whom 638 were eligible to participate in Wave 1 based on ethnicity and diabetes type, and 235 women in the comparator group, of whom 222 had consented for ongoing follow-up.Only 489 of the 860 eligible women were invited to participate in Wave 1 for the following reasons; i) women residing in extremely remote locations (n=143) or who moved outside of the NT (n=104) were not invited for Wave 1 as it was not logistically feasible at the time to assess them, (ii) peer-reviewed funding was awarded for Wave 1 sample size of n=400, calculated for maternal and child outcomes (iii) children were eligible to be seen from 1.5 years of age and this age was not reached for all children at the time of Wave 1 visits to certain communities.Among women invited, 416 (85%) participated in Wave 1. Preferential sampling was employed to ensure adequate numbers of women from each of the baseline hyperglycaemia in pregnancy groups participated in Wave 1. Wave 1 was completed in December 2018 and involved 416 mothers and 423 children (255 First Nations and 168 Europid children).
Follow-up child anthropometric assessment (age 1.5-5 years) Weight was measured using digital scales (Seca Infant Scales, Hamburg, Germany), from children wearing light clothing, no nappies and no shoes.Height was measured as either supine length and standing height in light of the age range of children involved.Supine length was measured from the crown of the head (against a headboard) to a footboard held against the plantar surface of the feet, on a standardized length board (Seca, Hamburg, Germany).Standing height was measured using a standard stadiometer on a level floor (Seca, Hamburg, Germany).Circumferences using a plastic tape measure were taken: 1) Head circumference -above the ears and midway between the eyebrows and hairline to the occipital prominence, 2) Waist circumference at the midpoint of the lowest rib cage and the iliac crest, at the line of the umbilicus in a horizontal plane, in a standing position during end-tidal expiration, 3) Mid upper arm circumference at mid-point between left acromium and the olecranon process.Skinfold thickness was measured on the left side at 3 sites using Holtain calipers (Holtain Ltd, Crosswell, Pembrokeshire, United Kingdom): 1) triceps -midway between the acromion and olecranon, 2) subscapular -lower angle of the scapula, 3) suprailiac -mid-axillary line just above the crest of the ilium.The sum of skinfolds outcome was determined from the triceps, subscapular and suprailiac skinfolds in all children.
Method for each measurement was detailed in the standard operating procedure.Training of study personnel was undertaken by a single person so as to maintain consistency between staff.All measurements were repeated at least twice and the average calculated.A further measurement was made if the difference in measures was ≥0.5cm (height), >0.3cm (circumferences), ≥0.2kg (weight), or >0.5mm (skinfolds), with the average of the two closest measures being used in the analysis.

Infant feeding practices
Breastfeeding data were obtained by either direct telephone or email contact with women at 6-8 months post-partum.The same questionnaire was administered by phone or email.In women for whom direct contact could not be established, data on breastfeeding were obtained through data linkage to primary care electronic medical records at 4-7 months.Predominant breastfeeding was defined as an infant being fed human milk as the only form of milk until 6 months of age consistent with the World Health Organization (WHO) definition, allowing for oral intake of water.It was not possible to determine from the electronic medical record whether breastfeeding was exclusive.

Cord blood c-peptide
Venous cord blood was collected from 245 singleton babies (normoglycemia n = 67, GDM n = 137, T2D n= 41) at delivery.The specimen was obtained by free drainage (free flow) of cord blood or drawn by needle aspiration (puncture) from a clamped segment of an umbilical vein.The specimen was stored in a styrofoam cooler for transport to the hospital's laboratory.Collection of cord blood was not feasible for all participants due to the unpredictable nature of birth and delivery and the need to prioritise clinical care.Cord blood c-peptide was measured using the electrochemiluminescence immunoassay on a cobas e602 (Roche Diagnostics, Mannheim, Germany).

Sample size
Prior to data collection, an assumption was made that Wave 1 would comprise at least 100 women with normoglycemia and 200 with hyperglycemia, and, of the presumed 200 women with hyperglycemia, it was thought that approximately 100 would be Aboriginal women and 100 Europid.At study conclusion in fact there were 123 women with normoglycemia and 293 women with hyperglycemia.Of the 123 women with normoglycemia, 61 women (63 children) were Aboriginal and 62 Europid (62 children).Of the 293 women with hyperglycemia, 189 were Aboriginal women (78 children born to women with T2D and 111 born to women with GDM) and 104 Europid (106 children born to women with GDM).With this number of children, the study had 80% power to detect a minimum difference in child BMI and weight z scores equal to 0.45 SD between the 5 study groups, using a two sample t-test (two-tailed alpha = 0.05).

Statistical analysis
Multiple linear regression models were used for continuous outcomes, results are reported as regression coefficients (ß estimates) with 95% confidence intervals.Multiple models were considered for all child anthropometric outcomes, in an additive stepwise approach developed to assess whether maternal hyperglycemia and BMI were independent predictors of each child outcome.
Model 1 included maternal glycemia in pregnancy (T2D/ GDM/ normoglycemia), child age and sex.Model 2 was as for Model 1, plus adjustment for maternal variables where p value was ≤0.2 on univariate analysis (such as age, smoking in pregnancy (yes/no), parity (0, 1+), alcohol in pregnancy (yes/no), maternal education (≤10 years or > 10 years of schooling), maternal height, gestational age at birth, maternal anemia on first antenatal bloods (yes/no), predominant breastfeeding at six months (yes/no)).All variables with p value ≤0.2 on univariate analysis were included in the multivariable model building process.Only variables with p value ≤0.1 on stepwise multivariable analysis were included in the final model for each outcome.Maternal ethnicity was included regardless of p-value, acknowledging both that Europid women with T2D were excluded and that ethnicity likely represents unmeasured socioeconomic factors.The p value of ≤0.1 was chosen to include variables that may have an important confounding effect on other exposures, and to explore variables that, although nonsignificant, have a beta coefficient that might indicate a significant effect with a larger sample.Therefore, the final model for each child anthropometric outcome included different covariates (see footnote Table 4).As maternal educational attainment was the socioeconomic measure most consistently associated with outcomes on univariable analyses, it was used in multivariable regression analyses as a marker of socioeconomic status.Model 3 was as for Model 2, plus inclusion of maternal BMI at first antenatal visit, adjusted for gestational age at the time it was assessed.Analyses were stratified by ethnicity as the study design only included First Nations women with T2D.Interactions were also assessed between maternal glycemic status and ethnicity, and maternal BMI and ethnicity.

Sensitivity analyses
Only 6 Europid women had a First Nations partner.Sensitivity analyses with these women excluded from analysis demonstrated no difference in outcomes, and data are therefore presented by maternal ethnicity.
Women with glycemic results consistent with the T2D range outside of pregnancy, but diagnosed for the first time in pregnancy (n=26, 23%) were included with the GDM group for the main analysis.This is consistent with the current classification guidelines (Australasian Diabetes In Pregnancy Society 4 and World Health Organization 3 ), because diagnosis of T2D cannot be definitively confirmed until post-partum.Sensitivity analyses were performed instead including these women in the T2D group.As a number of women did not have weight measured in the third trimester, multivariable models were repeated with the addition of gestational weight gain as a covariate, as well as replacing maternal BMI with gestational weight gain.
In addition, of the 416 women in Wave 1 follow-up, only 317 had adequate breastfeeding data at six months post-partum (108 born to mothers with normoglycemia during pregnancy, 168 born to mothers with GDM, and 41 born to mothers with T2D in pregnancy), reducing sample size.Sensitivity analyses were performed including predominant breastfeeding at 6 months in modelling.

Sensitivity analyses
Sensitivity analyses including predominant breastfeeding at 6 months in modelling demonstrated no difference in outcomes.Of note, the p-values for predominant breastfeeding at six months for the child outcomes of weight, height, BMI, waist, head and mid-upper arm circumferences, and triceps skinfold were <0.2 on univariate analysis.However, on stepwise multivariable analysis, the p value for predominant breastfeeding at six months was >0.1 for each of the above outcomes and thus not included in the final model of any outcome.To avoid reduced sample size, data are therefore presented without breastfeeding included in modelling.
Women with glycemic results consistent with the T2D range outside of pregnancy, but diagnosed for the first time in pregnancy, were included with the GDM group for the main analysis.Sensitivity analyses, firstly excluding these women, and secondly including these women within the T2D group, demonstrated no differences in outcomes (data not shown).There was no difference in results when gestational weight gain was included in modelling instead of maternal BMI (data not shown).

Table S1 : Comparison of demographic characteristics of women in Wave 1 sub-study to those who did not participate (among those eligible from PANDORA cohort)
Of those eligible who did not participate, n= 65 for First Nations women, n=61 for Europid women, who were invited and declined.All others who did not participate were not invited due to study design and feasibility, as outlined in Methods.