How Are Global Diet Quality Scores at 9 Years Associated With Cardiometabolic Disease Risk in Early Adolescence in Mysore, India?

Abstract

Objective

The purpose of the study was to assess Global Diet Quality Score (GDQS) at age 9.5 years and associated risk of cardiometabolic outcomes at 13.5 years in a birth cohort in Mysore, India.

Background

Assessing relationships between diet quality and cardiometabolic outcomes among children is important to inform the targeting and development of interventions to prevent cardiometabolic diseases. At present, this evidence is lacking, particularly in low- and middle-income countries.

Methods

Using data from the Mysore Parthenon Birth Cohort Study when children were 9.5 years of age, GDQS was computed from a 136-item food-frequency questionnaire. Children were categorized as being at low, moderate, or high risk of poor diet quality outcomes based on the GDQS value. At 13.5 years, cardiometabolic risk factor data were collected. Data were analyzed using linear and logistic regression models adjusted for covariates.

Results

Data were available at both time points for 538 children. At 9.5 years, the majority of children (72%) were at moderate risk of poor diet quality outcomes, with 25% and 3% being at low and high risk, respectively. Higher total GDQSs at 9.5 years of age were associated with lower fasting plasma glucose, insulin concentrations, and insulin resistance at 13.5 years of age. There were no associations between GDQS and anthropometric measures, lipids, or blood pressure.

Conclusion

The association between diet quality among children in this cohort and some elements of cardiometabolic risk in early adolescence adds to the case for early interventions to address risk of poor diet quality. Understanding context-specific barriers to a high-quality diet in different settings and developing solutions with communities to overcome these barriers should be a priority for researchers and policymakers.

INTRODUCTION

Background and Rationale for the Study

The dual burden of under- and overnutrition is a major public health problem, with suboptimal diet accounting for an estimated 22% of deaths globally.¹ There is evidence that diet quality is associated with several cardiometabolic health outcomes among adults, both in terms of healthy foods being protective and unhealthy foods increasing risk.^2,3 For example, higher fruit and vegetable intake is associated with reduced risk of cardiovascular mortality,⁴ while higher intakes of sugar-sweetened beverages have been associated with increased risk of type 2 diabetes.⁵

Late childhood and early adolescence are critical windows for physical and cognitive development,⁶ and there is evidence that diet tracks from adolescence into adulthood.⁷ It is important to identify what stage in the life course is most opportune to intervene to prevent cardiometabolic disorders. This requires an understanding of the longitudinal relationships between lifestyle behaviors and risk factors. A recent scoping review reported that there are limited data on adolescent diets, particularly in low- and middle-income countries (LMICs), and the majority of such data are cross-sectional and not sufficiently disaggregated by factors such as age, gender, and locality.⁸ In India, there is evidence that micronutrient intakes among children, adolescents, and women of reproductive age are inadequate.^9,10

Cardiometabolic Disease in India

The prevalence of cardiometabolic diseases and risk factors is increasing in India, particularly in urban areas.¹ A recent population-based survey among adults aged over 20 years reported that 11% were diabetic, 15% were prediabetic, and almost 40% were abdominally obese. Over 80% were dyslipidemic and 35% were hypertensive.¹¹ There is also evidence that adolescents are at risk of developing cardiometabolic disease without being overweight.¹² A study in a rural Konkan region among 1520 girls aged 16–18 years found a prevalence of prediabetes of 39% alongside approximately 30% of participants being stunted and underweight for age.¹³

Dietary metrics enable the following: (1) identification of at-risk groups, (2) examination of diet and outcome associations, and (3) evaluation of dietary interventions. There are few metrics that are designed to assess diet among children in LMICs. The Global Diet Quality Score (GDQS)¹⁴ was designed to study both risk of micronutrient inadequacy and noncommunicable disease among men and nonpregnant, nonlactating women aged 15 years and older at the population level worldwide. A feature of the GDQS is that it is an entirely food-based metric so has less intensive data requirements to enable rapid and timely dietary assessment. The metric requires information on portion size but does not require food-composition data. It has recently been adapted and validated for use among children 2–14 years of age.^15–18 The GDQS has yet to be applied to longitudinal data in India.

The objective of this study was to analyze data from the Mysore Parthenon birth cohort to answer the question of whether diet quality at 9.5 years is associated with cardiometabolic risk at 13.5 years.

METHODS

Study Participants

All participants were children enrolled in the Mysore Parthenon Study.¹⁹ This was a birth cohort set up at Holdsworth Memorial Hospital (HMH) in Mysore, India, to investigate the long-term cardiovascular risk outcomes associated with maternal gestational diabetes and body composition of the infant at birth. Details of the cohort have been published elsewhere.²⁰

Holdsworth Memorial Hospital is a charitable nonprofit institution and aims to serve those living in the city of Mysore and in the surrounding rural areas. Patients at the hospital pay for treatment but costs are reduced for the poorest. The majority of women who deliver at HMH are from middle- and lower-middle-income groups.

Participant Recruitment and Flow Through the Study

Between June 1997 and August 1998, pregnant women attending the antenatal clinic of HMH were recruited to the study if they fulfilled the following criteria: nondiabetic prior to pregnancy, less than 32 weeks’ gestation at the time of recruitment, and planning to deliver at HMH. A total of 1233 women were eligible for the study; 830 (67%) agreed to participate. Infants were included in the study if they were singletons and had no major congenital anomalies; 663 met the inclusion criteria and entered the birth cohort. After birth, the children were followed up every 6 months until 14 years of age for detailed anthropometry and cardiometabolic investigations.¹⁹ Of the 663 children, 25 died before reaching the age of 9 years and 8 developed major medical conditions. Therefore, 630 children were eligible to be followed up for the present study. At 9.5 years, 539 of these agreed to take part (82 refused and 9 were not traceable). At 13.5 years, 545 of 630 took part in the follow-up study in which cardiometabolic risk factors were assessed (73 refused and 12 were not traceable). Data were available at both time points for 516 participants.

Procedure

Children were invited to attend the research center at HMH for 1 day, accompanied by their parents, as close as possible to the age of 9.5 years. The range of ages at attendance was 9.1–9.7 years, with a mean (SD) value of 9.40 (0.11). Demographic, anthropometric, and dietary data were collected on the same day. At approximately 13.5 years, the children were again invited to attend the research center for a single day with their parents and demographic, anthropometric, and cardiometabolic risk factor data were collected. The range of ages was 13.2–14.0 years, with a mean (SD) value of 13.53 (0.15) years.

Demographic Data

Socioeconomic status (SES) and residence (urban or rural) were recorded when the child was aged 9.5 and at 13.5 years. The SES data were collected using the Standard of Living questionnaire from the Indian National Family Health Survey.²¹ The respondent was the child’s primary caregiver or the head of household. Children were classified as living in an urban or rural area based on their address at 9.5 years. Towns with a population greater than 100 000 were defined as urban areas based on Indian government census data.²²

Development and Administration of the Food-Frequency Questionnaire for the Mysore Parthenon Cohort

A detailed description of the development of the 136-item food-frequency questionnaire (FFQ) with a reference period of a typical month has been published previously.²³ Average portion sizes were estimated using observations of home-prepared foods and recording weights of ingredients, cooked weight, and quantities consumed. Standardized utensil and model kits were then developed for use with the FFQ.²⁴ When the children were aged 9.5 years, the FFQ was validated against serum nutrient concentrations.²⁵

At administration of the FFQ in the present study, the participant and caregiver were present during the assessment and the FFQ was administered by a trained nutritionist. Both child and caregiver responded.

Diet Quality Assessment Using the GDQS

Each food and beverage as well as the component ingredients of “home prepared” foods were coded to the appropriate GDQS food group, and points were assigned based on the gram-weight cutoffs shown in Table 1. The dataset reported in the present study was 1 of 7 datasets used to validate the GDQS for children aged 5–9 years.¹⁷

Table 1.

GDQS Food-Group Gram Cutoffs, Point Assignment, and Median (IQR) Consumption by the Cohort at 9.5 Years

	GDQS gram cutoffs				Point values				Intake (g/d)
GDQS food group	1	2	3	4	1	2	3	4	Median	25th centile	75th centile
Citrus fruits	<14	14–39	>39	NA	0	1	2	NA	18	7	40
Deep-orange fruits	<14	14–103	>103	NA	0	1	2	NA	207	118	414
Other fruits	<15	15–76	>76	NA	0	1	2	NA	148	88	236
Dark-green leafy vegetables	<7	7–23	>23	NA	0	2	4	NA	8	2	17
Cruciferous vegetables	<7	7–22	>22	NA	0	0.25	0.5	NA	9	1	21
Deep-orange vegetables	<5	5–28	>28	NA	0	0.25	0.5	NA	0	0	0
Other vegetables	<13	13–96	>96	NA	0	0.25	0.5	NA	36	21	59
Legumes	<5	5–26	>26	NA	0	2	4	NA	81	60	109
Deep-orange tubers	<7	7–36	>36	NA	0	0.25	0.5	NA	12	6	17
Nuts and seeds	<4	4–7	>7	NA	0	2	4	NA	1	0	3
Whole grains	<4	4–8	>8	NA	0	1	2	NA	58	36	98
Liquid oils	<1	1–5	>5	NA	0	1	2	NA	0	0	5
Fish and seafood	<8	8–40	>40	NA	0	1	2	NA	0	0	14
Poultry and game meat	<9	9–27	>27	NA	0	1	2	NA	5	2	9
Low-fat dairy	<19	19–93	>93	NA	0	1	2	NA	0	0	118
Eggs	<3	3–20	>20	NA	0	1	2	NA	15	7	30
High-fat dairy (in milk equivalents)	<20	20–101	>101–734	>734	0	1	2	0	9	0	60
Red meat	<5	5–46	>46	NA	0	1	0	NA	11	4	26
Processed meat	<5	5–17	>17	NA	2	1	0	NA	0	0	0
Refined grains and baked goods	<4	4–20	>20	NA	2	1	0	NA	364	221	658
Sweets and ice cream	<7	7–23	>23	NA	2	1	0	NA	72	47	108
Sugar-sweetened beverages	<35	35–150	>150	NA	2	1	0	NA	176	32	334
Juice	<21	21–102	>102	NA	2	1	0	NA	0	0	21
White roots and tubers	<15	15–76	>76	NA	2	1	0	NA	11	7	17
Purchased deep-fried foods	<5	5–25	>25	NA	2	1	0	NA	47	33	69

n = 538.

Abbreviations: GDQS, Global Diet Quality Score; NA, Not applicable.

The development of the GDQS among adults and its 25 component food groups has been described previously.¹⁴ A higher score indicates a higher-quality diet with respect to both nutrient adequacy and cardiometabolic risk. For healthy food groups or GDQS positive foods (n = 16), greater points are scored for higher intakes and a score of 0 is given for low intakes. For the 7 unhealthy food groups (GDQS negative score foods), 0 is scored for high intakes and higher points are given for low intakes. Red meat and high-fat dairy (also GDQS negative foods) are classified as unhealthy when consumed in excessive amounts; therefore, increasing points are given for moderate consumption of these foods and no points are given for low consumption or for the highest category of consumption of these foods (see Table 1 for point values for each food group). The total GDQS is obtained by summing points across all 25 food groups and ranges from 0 to 49. The GDQS positive scores range from 0 to 32 and GDQS negative scores range from 0 to 17.

Strengths of the GDQS include its development and validation from diverse settings across the world, its food-based design does not require nutrient composition data, and it provides information about the consumption of food groups that contribute to nutrient adequacy and noncommunicable disease risk reduction that can be easily communicated to a range of audiences. Although the GDQS is designed to measure diet quality using 1 day of recall for population-level assessment rather than individual-level assessment, our study uses information on habitual diet via an FFQ to assess associations between diet quality and disease risk.^14,26

Anthropometric Measurements

At 9.5 and 13.5 years, height was measured to the nearest 0.1 cm using a microtoise wall-mounted stadiometer (CMS Instruments, London, UK). Weight was measured to the nearest 100 g using an electronic digital weighing scale (Salter, Kent, UK). Body mass index (BMI) was calculated as weight divided by the square of height (kg/m²). BMI was compared with World Health Organization (WHO) international child growth standards²⁷ and BMI z scores were calculated and used as an outcome in the analysis. The change in BMI z score between 9.5 and 13.5 years was also included as an outcome measure. Waist circumference was measured to the nearest millimeter using anthropometric tape. Categorical outcomes were defined as follows based on the WHO standards: underweight (BMI-for-age <1 SD), overweight (BMI-for-age +1–2 SDs), and obese (BMI-for-age >2 SDs).

Cardiometabolic Outcome Measurements

At 13.5 years, fasting plasma glucose, insulin, total cholesterol, high-density-lipoprotein (HDL) cholesterol, low-density-lipoprotein (LDL) cholesterol, and triglycerides were measured.²⁸ Insulin resistance was estimated using the homeostasis model assessment for insulin resistance (HOMA-IR).

Categorical outcomes were defined as follows: prediabetes (fasting glucose ≥5.6 mmol/L), diabetes (fasting glucose >7 mmol/L), and high total cholesterol (>5.17 mmol/L).

Data Analysis

Variable distributions were assessed and, where normality assumptions were not met, variables were log-transformed for analysis. Continuous data are presented as mean (SD) for normally distributed variables and median (interquartile range [IQR]) for those that were not normally distributed. The GDQS was normally distributed and was included in analyses as a continuous variable and was divided into 3 categories of risk as per previous publications: high risk (GDQS <15), medium risk (GDQS 15 to <23), and low risk (GDQS ≥23).¹⁴ Descriptive demographic characteristics and anthropometric and cardiometabolic data were calculated and are presented by risk category.

To assess associations between diet quality at 9.5 years and outcomes at 13.5 years, linear or logistic regression models with GDQS (as a continuous variable) as the main predictor were used. Continuous cardiometabolic risk factor variables were the outcomes in linear regression models and overweight or obesity and prediabetes or diabetes were outcomes in the logistic regression models. Linear regression models were also run, in which change in BMI z score between 9.5 and 13.5 years was the outcome. All models included age and sex, urban/rural residence, and SES level of household. Two-sided tests were used, and P < .05 was the level of statistical significance used. All analyses were conducted using SPSS version 29.0 (IBM Corporation, Armonk, NY, USA).

RESULTS

Data on dietary intake at 9.5 years were available for 538 children. Table 1 presents the gram per day intakes of each of the 25 GDQS food groups among the 538 children in the cohort at 9.5 years. For several foods, including all types of fruit, legumes, eggs, whole grains, sugar-sweetened beverages, and sweets and ice cream, there was variability in quantity of intake, with children in the top quarter of intakes consuming approximately 3 times as much as those in the bottom quarter. Two food groups (deep-orange vegetables and processed meats) were not reported as being consumed by any of the children. These foods were not included as items on the FFQ, but respondents had the opportunity to report foods not on the FFQ in an “other” section. No participants reported these foods as “other” foods.

The healthy foods that appeared to be contributing most to the GDQS total and GDQS positive scores were deep-orange and other fruits, legumes, eggs, and whole grains. Sugar-sweetened beverages, sweets and ice cream, refined grains and baked goods, and purchased deep-fried foods appeared to be contributing most to GDQS total and GDQS negative scores.

The mean (SD) GDQS score for the whole cohort was 20.9 (3.2), the mean (SD) GDQS positive score was 15.4 (2.8), and the mean (SD) GDQS negative score was 5.5. In Table 2, participant characteristics are presented in categories according to GDQS risk. There were only 16 (3%) children in the high-risk GDQS category (total score <15). It was considered that this number was too small to be appropriate for statistical comparison across the risk categories. The majority of children (72%) were in the moderate-risk category (GDQS ≥15 and <23), with 25% at low risk (GDQS ≥23). Approximately 70% of children in the low-risk category were urban dwelling versus 88% in the high-risk category. Approximately 15% of the children were overweight/obese and just under 10% had plasma glucose measurements that would classify them as having prediabetes.

Table 2.

Characteristics of Children and Their Families by GDQS Risk Category at 9.5 Years (n = 538) and at 13.5 Years (n = 516)

	GDQS risk category at 9.5 years a
Characteristic	High (n = 16)	Moderate (n = 389)	Low (n = 133)
Demographic factors (9.5 y)
Sex (% female)	50.0	54.2	48.9
Residence (% urban)	87.5	75.8	69.9
Maternal education duration >10 years (%)	43.8	31.9	30.8
Head of household education duration >10 y (%)	75.1	45.3	41.3
Head of household occupation (% in professional role)	25.0	11.3	13.5
Standard of Living Index score,b mean (SD)	39.5 (6.7)	36.4 (8.1)	35.6 (8.7)
Diet quality at 9.5 y,c mean (SD)
GDQS total	14.0 (0.8)	19.7 (2.0)	30.0 (1.8)
GDQS positive	10.0 (1.0)	14.6 (2.1)	18.4 (2.5)
GDQS negative	3.9 (1.0)	5.2 (1.6)	6.6 (1.9)
Anthropometry (9.5 y), median (IQR)
BMI (kg/m2)	13.8 (12.4, 16.1)	14.4 (13.4, 15.7)	14.3 (13.3, 15.2)
Waist circumference (cm)	51.1 (47.8, 57.8)	53.5 (50.8, 57.2)	53.6 (50.5, 56.2)
	High (n = 16)	Moderate (n = 371)	Low (n = 128)
Anthropometry (13.5 y), median (IQR)
BMI (kg/m2)	17.2 (14.3, 19.6)	17.1 (15.5, 19.3)	17.0 (15.6, 19.1)
Waist circumference (cm)	64.3 (57.7, 70.1)	65.2 (60.5, 70.9)	64.4 (60.2, 69.6)
NCD outcomes at 13.5 y, median (IQR)
Plasma glucose (mmol/L)	5.14 (4.86, 5.32)	5.08 (4.83, 5.33)	5.00 (4.78, 5.28)
Plasma insulin (pmol/L)	36.7 (26.9, 45.6)	41.3 (30.6, 54.4)	35.8 (26.0, 52.9)
HOMA-IR	1.48 (0.96, 1.74)	1.56 (1.07, 2.01)	1.34 (0.98, 1.85)
Total cholesterol (mmol/L)	3.96 (3.34, 4.29)	3.52 (3.08, 3.94)	3.60 (3.06, 4.01)
HDL cholesterol (mmol/L)	1.20 (1.02, 1.35)	1.06 (0.91, 1.22)	1.06 (0.93, 1.24)
LDL cholesterol (mmol/L)	2.26 (1.96, 2.53)	2.06 (1.72, 2.44)	2.10 (1.71, 2.41)
Triglycerides (mmol/L)	0.63 (0.54, 1.07)	0.75 (0.55, 1.01)	0.72 (0.49, 1.01)
Categorical risk outcomes (13.5 y), n (%)
Underweight, BMI < –1 SD	4 (0.8)	47 (9.1)	19 (3.7)
Overweight, BMI +1–2 SDs	1 (0.1)	40 (7.8)	21 (4.1)
Obese, BMI > +2 SDs	0 (0.0)	19 (3.6)	3 (0.6)
Prediabetes, fasting serum glucose ≥5.6 mmol/L	2 (0.3)	46 (8.9)	4 (0.8)
Diabetes, fasting serum glucose >7mmol/L	0 (0.0)	1 (0.1)	0 (0.0)
Cholesterol, >5.172 mmol/L	0 (0.0)	3 (0.6)	0 (0.0)

^aCategories of risk for nutrient inadequacy and NCDs defined as follows: high risk, ≥23; medium risk, ≥15 and <23; low risk, <15.

^bNational Family Health Survey.²¹

^cGDQS total score range: 0–49; GDQS positive sub-metric range: 0–32; GDQS negative sub-metric range: 0–17.

Abbreviations: BMI, body mass index; GDQS, Global Diet Quality Score; HDL, high-density-lipoprotein; HOMA-IR, homeostasis model assessment for insulin resistance; LDL, low-density-lipoprotein; NCD, noncommunicable disease.

Results of adjusted linear regression models with GDQS, GDQS positive, and GDQS negative scores as predictor variables and cardiometabolic outcomes are presented in Table 3. Children with lower total GDQS and GDQS negative scores had higher plasma glucose concentrations, but there was no such finding with GDQS positive scores. Plasma insulin concentration and HOMA-IR were inversely associated with total GDQS and GDQS positive scores, and these findings were statistically significant. There was no such finding with GDQS negative scores and plasma insulin and HOMA-IR. None of the 3 GDQS metrics predicted risk of adverse lipid or anthropometric outcomes, nor did they predict change in BMI between 9.5 and 13.5 years.

Table 3.

Multiple Linear Regression Models With Diet Quality Scores at 9.5 Years as Predictors and NCD Risk Factors at 13.5 Years as Outcomes

	Diet quality metric outcome
Predictor	GDQS total	P	GDQS positive	P	GDQS negative	P
Plasma glucose (mmol/L)	−0.02 (−0.02, 0.00)*	.047	−0.01 (−0.02, 0.01)	.384	−0.02 (−0.04, -0.01)*	.035
Plasma insulin (pmol/L)	−0.43 (−1.12, -0.05)*	.037	−0.62 (−1.50, -0.03)*	.044	0.04 (−0.97, 0.18)	.354
HOMA-IR	−0.02 (−0.04, -0.01)*	.041	−0.03 (−0.06, -0.01)*	.031	0.01 (−0.05, 0.07)	.382
Total cholesterol (mmol/L)	−0.00 (−0.02, 0.02)	.818	0.00 (−0.02, 0.03)	.747	−0.02 (−0.05, 0.02)	.357
HDL cholesterol (mmol/L)	−0.00 (−0.01, 0.00)	.212	0.00 (−0.01, 0.01)	.910	−0.02 (−0.03, 0.00)	.150
LDL cholesterol (mmol/L)	0.00 (−0.01, 0.02)	.621	0.01 (−0.01, 0.02)	.549	−0.00 (−0.03, 0.02)	.957
Triglycerides (mmol/L)	−0.00 (−0.02, 0.01)	.570	−0.00 (−0.02, 0.01)	.507	−0.00 (−0.02, 0.02)	.982
Waist circumference (cm)	−0.02 (−0.19, 0.23)	.857	−0.09 (−0.15, 0.33)	.447	−0.17 (−0.54, 0.21)	.380
BMI (kg/m2)	0.02 (−0.06, 0.10)	.632	0.03 (−0.06, 0.12)	.529	−0.01 (−0.16, 0.14)	.894
Change in BMI from 9.5 to 13.5 y (kg/m2)	0.04 (−0.02, 0.09)	.158	0.03 (−0.03, 0.09)	.298	0.04 (−0.05, 0.14)	.372

Values are β-coefficients (95% CIs). Models were adjusted for age, sex, urban/rural residence, and socioeconomic status level of household head.

*Statistically significant at P < .05 level.

Abbreviations: BMI, body mass index; GDQS, Global Diet Quality Score; HDL, high-density-lipoprotein; HOMA-IR, homeostasis model assessment for insulin resistance; LDL, low-density-lipoprotein; NCD, noncommunicable disease.

Table 4 shows that the GDQS metrics did not predict the odds of being overweight or obese. When looking at prediabetes or diabetes as the outcome, the odds ratios were in the expected direction but did not reach statistical significance.

Table 4.

Logistic Regression Models With Diet Quality Scores at 9.5 Years as Predictors and NCD Risk Factors at 13.5 Years as Outcomes

Predictor	Overweight or obesity	P	Prediabetes or diabetesa	P
GDQS total	1.04 (0.97, 1.12)	.259	0.97 (0.88, 1.01)	.091
GDQS positive	1.05 (0.97, 1.14)	.204	0.95 (0.86, 1.02)	.128
GDQS negative	1.00 (0.88, 1.14)	.981	0.98 (0.85, 1.05)	.277

Values are odds ratios (95% CIs). Models were adjusted for age, sex, urban/rural residence, educational level of household head, socioeconomic status level of household, and occupation category of household head. Overweight was defined as BMI > +1 SD WHO growth reference.²⁷

^aPrediabetes (fasting serum glucose ≥5.6 mmol/L); diabetes (fasting serum glucose >7mmol/L).

Abbreviations: BMI, body mass index; GDQS, Global Diet Quality Score; NCD, noncommunicable disease; WHO, World Health Organization.

DISCUSSION

Our findings indicate that poor diet quality in mid-childhood, as measured by the GDQS, is associated with increased risk of higher plasma glucose, plasma insulin concentrations, and insulin resistance as measured by HOMA-IR in early adolescence. These findings remained when adjusted for covariates. There was only 1 child with glucose levels that indicated diabetes, but 10% of the cohort were prediabetic at 13.5 years. Our finding related to total GDQS score and odds of prediabetes or diabetes may warrant further investigation in a larger cohort with greater numbers of prediabetic adolescents. The GDQS did not predict risk of overweight or obese, change in BMI z score between 9.5 and 13.5 years, high waist circumference, or elevated lipid biomarkers at 13.5 years. A possible explanation for the differences in findings for each of the cardiometabolic risk factors is the length of time between exposure and the onset of the risk factor. Several studies have looked at exposures and outcomes in adults, as described below.

It has been proposed that, in contrast to other cardiovascular disease (CVD) risk factors, overall diet patterns may be more salient for type 2 diabetes outcomes.^29,30 A longitudinal study in Greece among adults found that higher GDQS and GDQS positive scores were inversely associated with cardiovascular outcomes.³¹ The authors concluded that, for CVD risk factors, specific foods or nutrients such as green leafy vegetables (GLVs) might be important to reduce risk. In our study, GLV intake was relatively low across the cohort, which may contribute to the lack of effect of GDQS on lipid outcomes.

There was some nuance in the associations between GDQS positive and negative scores with glucose and insulin outcomes. For glucose, the negative component of the score was a stronger predictor indicating that children who consumed smaller quantities of “unhealthy” foods (higher GDQS negative score) were more likely to have lower plasma glucose. The unhealthy foods largely driving the GDQS score were sugar-sweetened beverages, sweets and ice cream, refined grains and baked goods, and purchased deep-fried foods, whereas for insulin and HOMA-IR, children with higher intakes of “healthy” foods (higher GDQS positive) were more likely to have lower levels. Healthy foods driving the GDQS scores in this cohort were fruit, legumes, and whole grains, which have been reported to be protective against type 2 diabetes.^32,33 The study in Greece found that, after 20 years of follow-up, total GDQS and GDQS positive scores both predicted a lower risk of type 2 diabetes while a GDQS negative score was not associated with risk.³¹

Our findings are, to some extent, aligned with those reported in a Mexican observational cohort study of children and adolescents aged 8–14 years at recruitment and aged 12–21 years at follow-up.³⁴ Dietary Approaches to Stop Hypertension (DASH) index scores at recruitment were inversely associated with serum insulin and HOMA-IR at follow-up. The same study also reported an inverse association between diet quality and serum triglycerides as well as a positive association with HDL cholesterol. The foods that were found to be driving the associations with these outcomes (ie, fruit, vegetables, and whole grains) are similar to those in our study.³⁵

In the Nurses’ Health Study in the United States, total GDQS was inversely associated with type 2 diabetes risk among women of reproductive age and older; this was found to be mainly attributable to lower intakes of unhealthy foods as opposed to higher intakes of healthy or protective foods.³⁶ This aligns with our finding that children with a higher GDQS negative score had lower plasma glucose concentrations.

There is evidence that interventions to change diet quality among women have beneficial effects on weight and waist circumference gain in Mexico and the United States.^37,38 Such changes over time in diet will be important to study in future longitudinal analyses investigating associations between diet quality and cardiometabolic risk outcomes, particularly in LMICs and among adolescents, with little evidence reported to date in large longitudinal studies.⁸

Given the enormity of the public health problem in LMICs such as India, it will be important to use valid diet quality metrics such as the GDQS^15–18 to study associations with cardiometabolic risk and other health outcomes in larger populations of adolescents and young adults. The need to develop effective lifestyle interventions that are optimally targeted is evident.

Strengths and Limitations

A limitation of dietary data is that they are susceptible to reporting errors, but this was mitigated by including both the child and caregiver as respondents and by using a detailed FFQ with realistic utensils and models for portion-size estimation. The GDQS was initially developed for assessment among adults but has recently been validated among children in several sites globally.^15–18 A strength of this study therefore is the use of a dietary metric validated among children.

There are other factors known to contribute to cardiometabolic disease risk, such as fitness, for which data were unavailable. There was a small proportion of children in the high-risk category for GDQS, which may mean that the results pertaining to children in this category were not representative and caution should be used in their interpretation. Strengths of our study include that our data have been collected prospectively from a large cohort with good rates of follow-up and detailed demographic data to allow adjustment of regression models for covariates. The authors plan to analyze and report data at later follow-ups for this cohort to determine whether diet quality in childhood predicts cardiometabolic risk in late adolescence.

CONCLUSION

In conclusion, child diet quality was associated with reduced markers of diabetes risk in early adolescence in India. It should be noted that the majority of children were normoglycemic. In addition, the findings support the validity and usefulness of the GDQS as a predictor of poor diet quality outcomes.

Identifying barriers and opportunities to increase diet quality among children is an important priority for national and international policymakers. Investment in young children’s diet quality is likely to bring benefits for lifelong human capital, as well as a reduction in healthcare costs and benefits to planetary health. Future research should address barriers to quality diets and thus develop interventions targeted at groups that are at moderate or high risk of poor diet quality outcomes.

Author Contributions

S.H.K., J.E.A., N.B.A., S.B., M.D., C.B., A.C.G., C.H.D.F. and G.V.K. contributed to the study conception and design; S.H.K., S.R.V., K.N.K., T.K.N., S.V.J., K.K. and G.V.K. contributed to data collection, analysis, and interpretation; S.H.K. was responsible for drafting of the manuscript and final approval of the article. All authors approved the final version of the article.

Funding

This article appears as part of the supplement, “Validation of the Global Diet Quality Score (GDQS) as a Metric of Diet Quality for Children 2 to 14 Years of Age,” sponsored by The Rockefeller Center (Grant Number: 2021 FOD 024). Additional funding for this work was provided by the Parthenon Trust, Switzerland; the Wellcome Trust, UK; and the Medical Research Council, UK.

Conflicts of Interest

None declared.

Data Availability

The study’s data are not freely available. However, the Parthenon Cohort team is open to data sharing with bona fide researchers, and subject to Government of India regulations. For further information contact the corresponding author: Sarah Kehoe at: (sk@mrc.soton.ac.uk).