Childhood Cognitive Ability: Relationship to Gestational Diabetes Mellitus in India

Abstract

Aims/Hypothesis

To test the hypothesis that maternal gestational diabetes mellitus (GDM) is associated with poorer cognitive ability in children born to mothers with GDM compared to children born to non-GDM mothers in India.

Methods

During 1997-98 maternal GDM status was assessed at 30±2 weeks of gestation. Between 2007-2008, at a mean age of 9.7 years, 515 children (32-offspring of GDM mothers (ODM’s); 483-offspring of non-GDM mothers (controls)) from the Mysore Parthenon birth cohort underwent cognitive function assessment using tests from the Kaufman Assessment Battery for children-second edition and additional tests measuring learning, long-term storage/retrieval, short-term memory, reasoning, attention and concentration, visuo-spatial and verbal abilities.

Results

Compared to controls, ODM’S scored higher in tests for learning, long-term retrieval/storage (p=0.008), reasoning (p=0.02), verbal ability (p=0.01) and attention and concentration (p=0.003). In multiple regression, adjusted for the child’s age, sex, gestation, neonatal weight and head circumference, maternal age, parity, BMI, parent’s socio-economic status, education and rural/urban residence, this difference remained significant only for learning, long-term retrieval/storage (β=0.4SD (95% CI: 0.01, 0.75); p=0.042) and verbal ability (β=0.5SD (95% CI: 0.09, 0.83); p=0.015) and not with other test scores.

Conclusions/interpretation

In this population of healthy Indian children, there was no evidence of lower cognitive ability in ODM’s. In fact some cognitive scores were higher in ODM’s.

Introduction

The prevalence of gestational diabetes mellitus (GDM) is increasing in India and is estimated to be as high as 10-20% in urban populations [1]. Some studies have reported language impairment, inattention, motor delays and lower intelligence score among offspring of GDM mothers (ODM’s) [2,3]. The Mysore Parthenon cohort was established to study the effects of GDM on offspring body composition and glucose/insulin metabolism [4], and enabled us to examine cognitive ability in children (aged 9-10 years) in relation to maternal GDM status.

RESEARCH DESIGN AND METHODS

The cohort methodology has been reported in detail elsewhere [4]. Of 785 pregnant women recruited in 1997-98, who completed an OGTT at 30±2 week’s gestation, 630 delivered live, normal babies at Holdsworth Memorial Hospital, Mysore, India. GDM (defined by Carpenter-Coustan criteria) was diagnosed in 49 (6.9%) women. Following the OGTT, women diagnosed as having GDM were managed by their consultant obstetricians and therefore no information is available about the treatment and metabolic control of these women. The children had detailed anthropometry at birth, annually until the age of 5 and every 6 months thereafter. We extracted data from labour ward records concerning complications in pregnancy (pre-term delivery, pregnancy induced hypertension, maternal infection), mode of delivery and neonatal Apgar score.

During 2007-2008, cognitive function was assessed among 515 available children (82% of the original cohort; aged 9-10 years; 32-ODM’s and 483-offspring of non-GDM mothers (controls) using 3 core tests from Kaufman’s assessment battery for children-second edition [5] and additional tests [6,7] measuring learning, long-term retrieval/storage (Atlantis), short-term memory (word order), reasoning (pattern reasoning), attention and concentration (coding-Wechsler intelligence scale for children-third edition), visuo-spatial (Kohs block design) and verbal abilities. Detailed descriptions of these tests are described in electronic supporting material (ESM Table1) and also reported in our earlier publication [8].

We also collected data on maternal age, parity, BMI, height and urban/rural residence in pregnancy, the child’s sex, gestation, neonatal weight and head circumference and current age, BMI and height, the parent’s educational level and socio-economic status (SES) measured using the Standard of Living Index [9].

The hospital research ethics committee approved the study and informed consent was obtained from parents and children.

Statistical methods

Variables with skewed distributions were either log-transformed (maternal BMI and visuo-spatial ability) or square-root transformed (pattern reasoning). Cognitive test scores (transformed and non-transformed) were z-standardized for interpretation of regression models. Differences between ODM’s and controls were analyzed using unpaired t-tests and Chi-square tests. Multiple regression analysis, adjusting for potential covariates/confounders, was performed to examine offspring cognitive ability in relation to GDM status. In order to examine the association of GDM with cognitive function among children with similar SES and similar rural/urban residence, controls (n=32) were selected from the entire control sample (n=483), blind to the children’s cognitive scores, by exactly matching manually the sex, standard of living index score and rural/urban dwelling. If more than one control was available the one with the lowest study number was selected. Differences in the cognitive test scores in these matched pairs were analysed using paired t-tests. Stata v 10 (Stata corporation, TX, USA) was used for all analyses.

RESULTS

As previously described [4] GDM women were older, shorter, had higher BMI in pregnancy, and better educational attainment than non-GDM women (Table 1). SES and the proportion living in an urban rather than rural setting were higher among the GDM group, though these differences were not statistically significant. ODM’s had higher birthweight, larger neonatal head circumference and higher current BMI than controls.

Table 1.

Covariates/confounders according to maternal gestational diabetes mellitus status. Values are mean (SD) unless otherwise stated

Covariates	ODM (n=32)	Controls (n=483)	P b
Maternal factors in pregnancy
Age (years)	28.7 (4.9)	23.6 (4.0)	<0.0001
Parity 2+ (No (%))	6 (18.8)	82 (17.0)	0.8
Height (cm)	152.2 (5.8)	154.4 (5.3)	0.02
Body mass index (kg/m2)a	27.1 (25.0, 29.1)	23.1 (20.9, 25.9)	<0.0001
Fasting glucose concentration (mmol/L)	5.8 (1.7)	4.5 (0.4)	<0.0001
120-min glucose concentration (mmol/L)	9.8 (4.0)	5.8 (1.1)	<0.0001
Infant factors at birth
Gestational age (weeks)	39.3 (1.2)	39.3 (1.6)	0.9
Birthweight (kg)	3.270 (0.459)	2.854 (0.441)	<0.0001
Birthweight (SD score)
Boys	0.6(0.9)	−0.002 (1.01	0.08
Girls	1.2 (1.1)	−0.07 (1.0)	<0.0001
Head circumference (cm)	34.3 (1.2)	33.8 (1.4)	0.04
Head circumference (SD score)
Boys	0.4 (0.8)	−0.01 (1.0)	0.2
Girls	1.04 (0.5)	−0.04 (1.0)	0.03
Children’s factors at the time of cognitive testing
Age (years)	9.7 (0.3)	9.7 (0.3)	0.3
Height (cm)	131.3 (6.2)	130.7 (5.7)	0.6
Height-SD score
Boys	−0.70 (1.07)	−0.63 (0.88)	0.8
Girls	−0.65 (0.97)	−0.85 (0.95)	0.3
Body mass index (kg/m2)	15.9 (1.8)	14.5 (1.8)	<0.0001
Body mass index –SD score
Boys	−0.75(1.13)	−1.36 (1.24)	0.1
Girls	−0.18 (1.04)	−1.26 (1.22)	0.0001
Parents socio-economic status
Standard of living index (score)	37.4 (7.8)	36.2 (8.3)	0.4
Maternal education (No (%))
<10 completed years	3 (9.4)	186 (38.6)
−10 completed years	10 (31.2)	149 (30.9)	0.001
>10 completed years	19 (59.4)	147 (30.5)
Paternal education (No (%))
<10 completed years	7 (21.9)	186 (38.6)
−10 completed years	20 (62.5)	183 (38.0)	0.02
>10 completed years	5 (15.6)	113 (23.4)
Residence: Rural (No (%))	5 (15.6)	134 (27.7)	0.1
: Urban (No (%))	27 (84.4)	349 (72.3)	0.1

^aTransformed variable; values are median (Inter-quartile range).

^bP values are for the difference between ODM’s and controls derived using t-tests or Chi² tests.

All the cognitive scores increased with increasing SES and parental educational level (p<0.001 for all; data not shown). Children of primiparous mothers performed better than children of multiparous mothers in tests of learning, long-term retrieval/storage (p=0.004), short-term memory (p=0.03), reasoning (p=0.001) and verbal fluency (p=0.001). The cognitive performance of urban children was better than rural children in tests of learning, long-term retrieval/storage (p<0.001), short-term memory (p=0.004), reasoning (p=0.001) and visuo-spatial ability (p=0.001). Visuo-spatial ability and reasoning score increased with increasing maternal age (p<0.001 for both).

Mean cognitive scores were higher in ODM’s than controls, statistically significant for long-term retrieval/storage, reasoning, verbal ability and attention and concentration (Table 2). In multiple regression analysis the differences remained significant for learning, long-term retrieval/storage and verbal ability adjusted for the child’s current age, sex, gestation, neonatal weight and head circumference, maternal age, parity and BMI and the parents’ SES, education and rural/urban residence. Although cognitive performance was better in girls than boys there were no interactions between sex and GDM in relation to cognitive ability. Among controls, maternal 120-minutes glucose concentrations, but not fasting glucose, were positively associated with the children’s verbal ability (p=0.02) and Atlantis score (p=0.07).

Table 2.

Association between maternal gestational diabetes mellitus status and children’s cognitive test scores

n=515 (whole cohort)
Cognitive tests	ODM (n=32) Mean (SD)	Controls (n=483) Mean (SD)	P b	β (95% CI)	P c	β (95% CI)	P d	β (95% CI)	P e
Learning, long-term retrieval/storage (Atlantis)	75.9 (16.2)	67.5 (17.1)	0.008	0.5(0.12,0.84)	0.009	0.4(0.02,0.71)	0.04	0.4(0.01,0.75)	0.04
Short-term memory (Word order)	17.0 (2.7)	16.4 (2.6)	0.2	0.2(−0.19, 0.54)	0.3	0.02(−0.33,0.37)	0.9	0.03(−0.35,0.40)	0.8
Reasoning ability a (Pattern reasoning)	12.5 (9.0, 16.0)	10.0 (4.0, 14.0)	0.02	0.4(0.01,0.73)	0.04	0.2(−0.10,0.58)	0.2	0.1(−0.23,0.51)	0.4
Verbal ability-animals (Verbal fluency)	13.1 (3.0)	12.0 (3.3)	0.07	0.3(−0.05,0.67)	0.09	0.2(−0.12,0.59)	0.2	0.2(−0.16,0.61)	0.2
Verbal ability-names (Verbal fluency)	18.2 (5.2)	16.0 (4.9)	0.01	0.4(0.01,0.71)	0.04	0.3(−0.05,0.64)	0.09	0.5(0.09,0.83)	0.02
Visuo-spatial ability a (Koh’s block design)	85.3 (68.1, 97.3)	76.2 (62.6, 87.8)	0.1	0.3(−0.06,0.66)	0.1	0.2(−0.18,0.52)	0.4	0.01(−0.36,0.39)	0.9
Attention and concentration (Coding-WISC-III)	36.8 (8.0)	32.4 (8.1)	0.003	0.4(0.09,0.75)	0.01	0.3(0.01,0.67)	0.04	0.3(−0.04,0.67)	0.08

^aTransformed variable; values are median (Inter-quartile range).

^bP values are for the difference between ODM’s and controls derived using t-tests

β is the effect size (SD) on the outcome (non transformed and transformed) per unit change in the exposure (maternal gestational diabetes mellitus status) derived using multiple linear regression.

^cP value derived using multiple linear regression adjusted for child’s sex, gestation and age

^dP value derived using multiple linear regression adjusted for child’s sex, gestation, age, SES, parents’ education and rural/urban residence

^eP value derived using multiple linear regression adjusted for child’s sex, gestation, age, SES, parents’ education, rural/urban residence, maternal age, BMI and parity in pregnancy and child’s weight and head circumference at birth.

We repeated the analysis by restricting the sample to 32 ODM’s and 32 sex, SES and rural/urban residence matched controls. Characteristics of these matched pairs are described in electronic supporting material (ESM Table 2). There were still differences in maternal education, age and BMI, child’s birthweight and current BMI. Mean cognitive scores were higher in ODM’s than controls, but significant only for verbal ability.

Discussion

In this study of healthy Indian school-age children born in one maternity unit, cognitive scores were higher among ODM’s than controls. Both maternal GDM and higher cognitive performance in the children were associated with higher parental education and/or SES. However, the associations of GDM with better offspring learning, long-term storage/retrieval and verbal ability remained statistically significant after adjusting for these and all other potential confounding factors measured.

Strengths of the study were that we measured a battery of cognitive function tests specifically adapted for, and validated in, a South-Indian population and also collected data on a range of potential covariates/confounders. Limitations in our study were a relatively small number of ODM’s, lack of data on the severity of GDM and treatment, maternal diet and no information on parental intelligence and/or the home environment.

A recent case-control study reported a higher risk of language impairment in ODM’s compared to controls [2]. A recent review has reported no difference in cognitive ability among children born to mothers with or without GDM; however, compared to controls, ODM’s performed less well in fine and gross motor functions. Further, this review reported inverse associations of offspring intelligence scores, attention, language development, learning, memory-span and mental and psychomotor development with the severity of GDM assessed by glycosylated haemoglobin level and ketonuria, suggesting that offspring cognitive performance could be within normal limits in well-controlled GDM [3]. The reasons for our finding of higher cognitive function in ODM’s may be that there were no cases of severe or uncontrolled GDM in our study and/or that we were unable to completely adjust for the fact that in our population GDM was associated with higher maternal education, urban residence (and thus better schooling) and better childhood nutritional status (ODM’s had higher BMI). This effect may not be evident in studies in industrialized populations, where GDM tends to be associated with lower SES and lower maternal education [2]. Alternatively, the fetus of a mother with GDM is exposed to higher concentrations of glucose and fatty acids transferred across the placenta from the maternal circulation [10]. Theoretically these could enhance brain development. This is perhaps supported by our findings of positive association of maternal 120-minutes glucose concentrations with verbal ability and long-term retrieval/storage among controls; unfortunately we do not have data on maternal circulating fatty acids.

In conclusion, in this population of healthy Indian children, there was no evidence of lower cognitive ability in ODM’s. In fact some cognitive scores (two of the six cognitive parameters tested) were higher in ODM’s. This may be due to residual confounding. Although we adjusted for a number of confounding factors, no SES score can perfectly capture all the effects of SES, especially in India, which has a wide range of SES. Even in the SES matched analysis there were significant differences in other important factors, such as maternal education, age and BMI and child’s birthweight and current BMI, between cases and controls. Alternatively the difference may be due to biological effect. The study suggests for further research to examine the relationship between GDM and offspring cognitive ability in larger studies.

Supplementary Material

01

ESM Table 1. Description of the cognitive tests used in the study

ESM Table 2. Covariates/confounders among ODMs and controls in matched pairs. Values are mean (SD) unless otherwise stated

Abbreviations

GDM

Gestational diabetes mellitus

ODM

Offspring of GDM mothers

SES

Socio-economic status