Associations of childhood BMI traits with blood pressure and glycated hemoglobin in 6–9-year-old Samoan children

Abstract

Introduction

Prevalence and risk factors for elevated glycated hemoglobin (HbA1c) and blood pressure (BP) are poorly understood among Pacific children. We examined associations of HbA1c and BP in 6–9 year-olds with body mass index (BMI) at ages 2, 5, and BMI velocity between 2–9 years in Samoa.

Methods

HbA1c (capillary blood) and blood pressure were measured in n=410 Samoan children who were part of an ongoing cohort study. Multilevel models predicted BMI trajectory characteristics. Generalized linear regressions assessed associations of childhood characteristics and BMI trajectories with HbA1c and BP treated as both continuous and categorical outcomes. Primary caregiver-reported childhood characteristics were used as covariates.

Results

Overall, 12.90% (n=53) of children had high HbA1c (≥5.7%) and 33.17% (n=136) had elevated BP. BMI at 5-years and BMI velocity were positively associated with high HbA1c prevalence in males. A 1 kg/m² per year higher velocity was associated with a 1.71 (95% CI: 1.07, 2.75) times higher prevalence of high HbA1c. In females, higher BMI at 5-years and greater BMI velocity were associated with higher BP at 6–9 years (95% CI: 1.12, 1.40, and 1.42, 2.74, respectively).

Conclusion

Monitoring childhood BMI trajectories may inform cardiometabolic disease screening and prevention efforts in this at-risk population.

Introduction

Cardiometabolic diseases are the leading cause of preventable death in Pacific Island nations and represent an increasing burden on health systems.^1,2 Among their modifiable risk factors, obesity, high HbA1c, and elevated BP may be established in childhood and confer elevated risk later in life.^3–5 HbA1c ≥5.7% at age 10–19 years has been shown to predict diabetes risk in adulthood.⁶ Similarly, elevated BP in children, which is often related to obesity, can progress into hypertension in adulthood, resulting in increased cardiometabolic risk.⁷

While high BMI and rapid gain in BMI have been associated with obesity development and elevated BP in childhood,^8,9 variations in the tempo and timing of BMI gain may differentially influence how elevated cardiometabolic risk is established during childhood. Previous research studies have demonstrated that higher infant BMI and weight gain may protect children with obesity from truncal adiposity, insulin resistance, and adverse health effects in adulthood^10–12, while later BMI gain is associated with increased cardiometabolic risk¹³. The timing and patterns of BMI change (the shape of the BMI trajectory) are known to vary within and between populations. Pacific people particularly are poorly represented in the existing childhood literature and there is little knowledge of how BMI at particular ages and changes in BMI over the course of childhood relate to cardiometabolic risk. The one existing study on this topic, from the Pacific Island Families birth cohort study in Auckland, New Zealand found that both higher weight at 2.5 years of age and relatively rapid weight gain between then and 14 years of age were associated with several risk factors for metabolic disease, including insulin resistance, triglycerides, and blood pressure^14.

In the context of substantial population-level adiposity, the Ola Tuputupua’e “Growing Up” cohort in Samoa is an ideal study population to pursue scientific questions about the relationships between BMI trajectories and cardiometabolic risk in childhood. We have previously reported high and increasing obesity prevalence based on World Health Organization child growth standards/references, with 16.5% of the cohort having obesity at age 2–4 years,¹⁵ 25% at age 3.8–6 years,¹⁶ and 36% at age 5.5–11 years.¹⁷ Here we assess the extent to which BMI traits in early- and mid-childhood are associated with HbA1c and BP among Samoan children.

Methods

Study Design, Setting, Participants

The study used longitudinal data from the ongoing Ola Tuputupua’e cohort, which was designed to investigate child growth, development, and health in Samoa. Assessments took place in 2015 (children aged 2 to 4 years), 2017–2018 (3.5 to 7 years old), and 2019–2020 (5.5–11 years old). As previously reported,¹⁵ a convenience sample of 319 eligible biological mother-child pairs was initially recruited across the three census regions on the island of ‘Upolu (the most populated of the Samoan islands): Rest of Upolu, Northwest Upolu, and the Apia Urban Area. In 2017–2018, follow-up assessments with children in the original cohort and recruitment of new eligible participants expanded the cohort to 501 pairs across 11 ‘Upolu villages.

Eligible children were of Samoan ethnicity based on maternal report of four Samoan grandparents and were either age 2–4 years old if enrolled in 2015 or age 4–7 years old if enrolled in 2017. Eligible mothers were at least 18 years old, not pregnant at initial enrollment, had no severe physical or cognitive impairments, and were willing and able to complete the study. Children who were determined to be outside of the eligible age range upon verification of their birth date or had missing data for any characteristic were excluded. Data from n=485 eligible children were used to develop BMI trajectory models; n=410 children with data on BP/HbA1c at 6–9 years were included in outcome analyses.

Yale and Brown University Institutional Review Boards and the Health Research Committee of the Samoan Ministry of Health approved all study procedures (IRB# 2000020519 and IAA#18–41 959). Written informed consent was obtained from parents and assent from children (once they reached 7 years of age) at each assessment.

Measurement of BMI

Duplicate measures of weight (HD 351 weight scale; Tanita Corporation of America, Arlington Heights, IL) and height (Stadiometer Pfister Imports, New York, NY) were collected at each research assessment using standardized procedures and the two measures averaged for analyses. BMI was calculated as weight in kilograms divided by height in meters squared.

Cardiometabolic risk marker assessment

The outcomes were measurements of HbA1c and BP collected at age 6 to 9 years. We collected a finger-prick, capillary blood sample to assess HbA1c using a point-of-care A1cNOW test system (PTS Diagnostics, IN).¹⁸ Considering that there are no standard cut-off values to define elevated HbA1c in children, we categorized children with HbA1c ≥ 5.7% as having high levels because HbA1c above this threshold (commonly indicative of prediabetes in adults) in childhood has been prospectively associated with higher diabetes incidence in adulthood in a Pima Indian community.⁶

Systolic and diastolic BP were measured while the child was sitting, three times on the left arm using an Omron HEM-907XL Automatic BP Monitor (Omron Healthcare, IL, USA). The average of the last two measurements was calculated. For six children (n=2 females) only two measurements were completed and the average of these was used. Average BP was then adjusted for age, sex, and height at the time of measurement and the residualized values were used in analyses. Based on these values, children were assigned to BP percentiles using 2017 AAP references¹⁹ and classified as having elevated BP if either systolic or diastolic BP were equal to or exceeded the 90^th percentile.

Child, maternal, and household characteristics

Based either on associations previously observed among the cohort, or those noted in prior studies, we considered several covariates in our analyses. Data were derived from a questionnaire administered to mothers at the time of their initial enrollment into the cohort (2015 or 2017). Child age was calculated in years by subtracting the reported date of birth from the date of any survey and physical assessment. We asked mothers to recall whether they ever breastfed or gave their children their pumped milk (yes/no). We calculated the daily total energy intake of children using a food frequency questionnaire with a thirty-day reference period that has been validated and adapted from previous adult studies in Samoa.^20–22 Children were categorized as having either ‘high’ or ‘low’ adherence to an identified modern dietary pattern using factor analysis of the food frequency questionnaire data.^16,23 Physical activity levels were estimated using 3-items (15-point score) from the Netherlands Physical Activity Questionnaire for young children that provided the greatest internal consistency with accelerometer data in this cohort.²⁴ Children were categorized into tertiles based on physical activity score distribution to distinguish low, medium, and high physical activity levels. Total daily screen time was estimated by the sum of reported minutes per day spent watching television, using a computer, or phone.

To maintain the same categories that were previously reported in the cohort¹⁵ maternal age was categorized as either “18–24.9 years”, “25–39.9 years”, or “≥40 years”. Comparable to Samoan census data^25,26 mothers reported years of education received and marital status (married or cohabitating or not). Maternal weight and height were measured using the same protocols as those described for children to calculate BMI. Maternal hypertension was defined as measured systolic BP ≥140 mmHg or diastolic BP ≥90 mmHg.²⁷ Total annual household income was dichotomized as ≤10,000 tala versus > 10,000 tala based on the response distribution. An 18-item household asset score, which measures ownership of consumer durables such as television or refrigerator and was categorized into quartiles. Urbanicity was based on the census region of the village where the child resided at enrollment, with rural Rest of Upolu, peri-urban Northwest Upolu, and Apia Urban Area.¹⁵

Statistical analyses

To identify additional variables for inclusion in our multilevel models, we first examined individual, maternal, and household characteristics in the sample and by sex among children at initial enrollment. We also examined associations among these participant characteristics and the CMD outcomes (HbA1c and BP; Supplementary Tables 1 and 2). For this analysis, CMD outcomes were treated as continuous variables, and Wilcoxon two-sample tests were used for continuous participant characteristics that were not normally distributed; Spearman rank’s tests were used to assess the correlations between continuous variables with skewed distributions.

We used a two-stage approach to obtain estimates of conditional associations between BMI trajectory characteristics and cardiometabolic risk marker outcomes (examined as both continuous and categorical outcomes) by sex. In the first stage, BMI trajectories were estimated in two-level multilevel models, with random intercepts and slopes. Among n=485 children in the cohort with data to develop BMI trajectories (Supplementary Figure 1), the majority (n=437, 90.1%) had at least two measurements (Supplementary Table 3). BMI trajectories were estimated using sex-specific multilevel models with measurement occasion at level 1 and individuals at level 2. The models’ random effects represent 1) BMI at age 2 years (early childhood), 2) BMI at age 5 years (mid-childhood), and 3) BMI velocity between 2 and 9 years (kg/m²/year). Based on the best model fit, the trajectory shape was specified as a restricted cubic spline with 3 knots using the ‘mkspline’ function in STATA 16 (StataCorp LLC, TX, USA) to model the non-linear changes in BMI during childhood. The first spline term was equivalent to decimal age in years. The intercept and first spline term were allowed to have random effects at level 2, with an unstructured variance-covariance matrix. Assessment of model fit was based on the comparison of observed vs. fitted BMI, the standardized residuals, and the Bayesian information criteria. The best linear unbiased predictions of the random effects were estimated using the ‘reffects’ function in STATA 16. The random effect estimates of the intercept indicated the individual-level deviation from mean BMI at age 2 years and at age 5 years, with a positive intercept random effect implying relatively high (above the sample mean) BMI at the respective ages. The random effect estimates of the slope indicated the individual-level deviation from the sample mean BMI velocity, with high (above the sample mean) velocity indicated by a positive slope random effect.

With these random effect estimates, we proceeded to the second stage with multivariable general linear regression models and stratified by sex. Since HbA1c values for the sample were right-skewed, we transformed the data by using the natural logarithm and multiplied by 100. The 100log_e scale of HbA1c was used as the outcome to estimate the symmetric percentage differences in the outcome across BMI trajectory characteristics without any back-transformation of the regression coefficients.²⁸ For BP outcomes, we used the residuals on their original scales (mmHg) to estimate the mean differences. We then fitted modified Poisson regression models (using proc genmod in SAS 9.4, SAS Institute NC, USA) to estimate the PR for elevated HbA1c and elevated BP at ages 6 to 9 years old according to BMI trajectory characteristics.²⁹ Participant characteristics associated with cardiometabolic outcomes, as well as those that could be used to minimize potential confounding (ever breastfed status, maternal hypertension, household asset quartile, and household total annual income) or selection bias (child age and household urbanicity) were included in the models.

Results

Sample characteristics at age 2 to 5 years

When we compared various child, maternal, and household characteristics of the n=410 included in our outcome analysis to the n=485 eligible children in the wider cohort, they had lower median age, estimated daily energy intake, weight, and height at study enrollment compared to those excluded (Supplementary Table 4). HbA1c at 6–9 years ranged from 4.0 to 7.8%, with a median of 5.2% (Q1-Q3:5.0–5.4) in females and 5.3% (Q1-Q3:5.1–5.6) in males. Nearly one in ten females (n=20, 9.7%) had a HbA1c ≥5.7%; prevalence was higher in males (n=33, 16.3%). After adjusting for child age, sex, and height, mean systolic and diastolic BP were 104 ± 9 and 64 ± 9 mmHg respectively. Based on the 2017 AAP guidelines¹⁹, the prevalence of elevated BP was 32.4% (n=67) in females and 34.0% (n=69) in males. Associations between sample characteristics and cardiometabolic risk markers are described in Supplementary Tables 1 and 2.

Associations of BMI traits with HbA1c and high HbA1c (≥5.7%)

Relative to the sample mean, one kg/m² higher BMI at age 2 years was associated with a 1.0% lower HbA1c in 6- to 9-year-old males after adjusting for child age and characteristics at enrollment: ever breastfeeding status, household asset quartile, urbanicity, household total annual income; however, the corresponding confidence interval was imprecise (adjusted 95% CI: −2.38 – 0.26, Table 2). BMI velocity between 2 and 9 years was associated with a higher prevalence of high HbA1c in males (adjusted PR: 1.71, adjusted 95% CI: 1.07–2.75) but not in females. Consequently, an association was also observed between BMI at 5 years and high HbA1c (adjusted PR: 1.20; adjusted 95% CI: 1.02–1.40) in males.

Table 2.

Unadjusted and adjusted differences in HbA1c and blood pressure among children at age 6–9 years (n=410) according to BMI trajectory characteristics, 2015–2020

	Female (n=207)				Male (n=203)
	β	95% CI		P	β	95% CI		P
Percentage difference in HbA1c (%)
BMI (kg/m2) at age 2 years
Model 1†	0.22	−1.30	1.75	0.774	−0.72	−2.07	0.63	0.295
Model 2‡	0.49	−1.02	2.00	0.525	−1.06	−2.38	0.26	0.115
BMI (kg/m2) at age 5 years
Model 3†	0.19	−0.56	0.95	0.616	0.60	−0.22	1.41	0.150
Model 4§	0.18	−0.66	1.02	0.680	0.68	−0.12	1.49	0.097
BMI velocity (kg/m2/year)
Model 5**	0.53	−2.03	3.09	0.683	1.73	−0.28	3.74	0.092
Model 6§	0.53	−1.99	3.05	0.680	2.05	−0.37	4.47	0.097
Mean difference in systolic blood pressure (mmHg)
BMI (kg/m2) at age 2 years
Model 1†	0.48	−1.20	2.16	0.573	−0.02	−1.48	1.44	0.978
Model 2‡‡	0.81	−0.87	2.49	0.343	−0.16	−1.67	1.36	0.840
BMI (kg/m2) at age 5 years
Model 3†	0.83	−0.01	1.68	0.054	0.56	−0.35	1.48	0.227
Model 4§§	1.11	0.15	2.07	0.023	0.74	−0.21	1.68	0.128
BMI velocity (kg/m2/year)
Model 5**	2.78	−0.10	5.66	0.059	1.11	−1.11	3.34	0.326
Model 6§§	3.33	0.46	6.20	0.023	2.21	−0.64	5.05	0.128
Mean difference in diastolic blood pressure (mmHg) ††
BMI (kg/m2) at age 2 years
Model 7†	0.44	−1.23	2.11	0.604	−1.51	−2.87	−0.15	0.030
Model 8‡‡	0.64	−1.02	2.31	0.448	−1.24	−2.64	0.15	0.081
BMI (kg/m2) at age 5 years
Model 9†	1.34	0.49	2.20	0.002	0.28	−0.51	1.08	0.486
Model 10§§	1.73	0.73	2.72	<0.001	0.10	−0.74	0.93	0.821
BMI velocity (kg/m2/year)
Model 11**	4.91	1.97	7.86	0.001	1.60	−0.42	3.62	0.121
Model 12§§	5.18	2.20	8.16	<0.001	0.29	−2.22	2.80	0.821

^*Results from generalized regression models, stratified by sex. The percentage of natural log-transformed HbA1c was the main outcome of interest and the model coefficient estimates of the BMI trajectory were interpreted as a symmetric percentage difference in HbA1c.

^†Models included only the predicted random effect estimates of the intercept (either BMI at age 2 or 5 years)

^‡Adjusted for child age at the time of HbA1c measurement, 4 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala).

^§Adjusted for child age at the time of HbA1c measurement, BMI at age 2 years (intercept random effect), 4 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala).

^**Models included only the predicted random effect estimate of the slope

^††Results from generalized regression models. Residuals for systolic blood pressure or diastolic blood pressure were the main outcomes of interest.

^‡‡Adjusted for child age and height at blood pressure measurement, 5 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), maternal hypertension (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala).

^§§Adjusted for child age and height at blood pressure measurement, BMI at age 2, (intercept random effect), 5 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), maternal hypertension (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala).

Associations of BMI traits with systolic and diastolic BP and elevated BP

There was no evidence of any association between BMI at 2 years and 6- to 9-year-old blood pressure in either males or females. At 6- to-9-years, females had between 1 and 5 mmHg greater systolic and diastolic BPs for every 1 kg/m² higher BMI at age 5 years and 1 kg/m² per year increase in BMI velocity relative to the sample mean (Table 2). There was more than a 10% higher prevalence of elevated BP associated with relatively higher BMI at age 5 (adjusted PR: 1.25, adjusted 95% CI: 1.12–1.40) and BMI velocity (adjusted PR: 1.97, adjusted 95% CI: 1.42–2.74, Figure 1). Similar associations were observed in males, but the effects were not as strong and the estimates imprecise.

Figure 1. Prevalence ratio of high glycated hemoglobin (HbA1c) and elevated blood pressure in Samoan children by characteristics of BMI trajectories.

High HbA1c was defined as ≥5.7% for both females and males. Elevated blood pressure was based on 2017 American Academy of Pediatric references based on age, sex, and height at time of blood pressure measurement. Estimated prevalence ratio and their corresponding 95% confidence intervals are shown for high HbA1c and elevated blood pressure in black for females (n=207) and grey for males (n=203) at age 6–9 years old. Unadjusted model estimates are shown as circles and these models included only the predicted random effect estimate of the intercept (BMI at age 2 or 5 years), or the slope (BMI velocity). Adjusted model estimates are shown as triangles. For BMI at age 2, high HbA1c outcome models were adjusted for child age at the time of HbA1c measurement, and 4 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala). Elevated blood pressure models were adjusted for child age and height at blood pressure measurement, 5 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), maternal hypertension (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala). For BMI at age 5 and BMI velocity, high HbA1c outcome models were adjusted for child age at the time of HbA1c measurement, BMI at age 2 years (intercept random effect), and 4 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala). Elevated blood pressure models were adjusted for child age and height at blood pressure measurement, BMI at age 2, (intercept random effect), 5 additional baseline characteristics at enrollment: ever breastfeeding status (yes or no), maternal hypertension (yes or no), household asset quartile (referent: low), urbanicity (referent: rural Rest of Upolu region), and household total annual income (referent: <10,000 tala).

Discussion

Our results represent the first attempt to examine associations among childhood BMI traits and mid-childhood CMD risk markers in Samoa. The key findings include a positive association between BMI at 5 years, BMI velocity, and prevalence of high HbA1c in 6- to 9-year-old males. Stronger positive associations between BMI at age 5 years and BMI velocity with elevated BP prevalence were observed in females compared to males.

The prevalence of high HbA1c and elevated BP among the Samoan children included in our sample highlight the urgent need for preventive interventions in this setting and further research to characterize high-risk groups. There were higher levels of HbA1c observed in the Samoan cohort compared to a healthy US population of 5–9-year-olds,³⁰ 6–10-year-old children in China,³¹ and 0.5–18-year-old children in Germany.³² They were also higher in mid-childhood than at 14 years of age in the Pacific Island Families Study¹⁴. While we expected that average BP in children would have increased as economic development and the nutrition transition progresses in Samoa,^33–35 the range of BPs in the Samoan cohort fell within or above the 2017 AAP recommended values¹⁹ and previous summaries of BP data for children measured in Samoa in 1979–1993 and 2003^35,36 In the Samoan population characterized by high adiposity levels, further evaluation and early intervention should be recommended for all children with elevated CMD risk markers.

Several other studies have described associations between rapid childhood BMI gain and dysglycemia, although most have used diabetes in adulthood as their outcome. In American Indians, for example, higher childhood weight and accelerated weight gain between birth and 18 years were associated with an increased risk of type 2 diabetes in adulthood³⁷. Interestingly, in that study, it was rapid weight gain pre-adolescence (birth to 8 years) that conferred the greatest risk (compared to weight gain in early or late adolescence)³⁷. In our study, there was already evidence of an association between rapid childhood BMI gain and potential dysglycemia at only 6–9 years in males, likely as a result of the chronic inflammatory state and early insulin resistance that accompanies excess adiposity³⁸. Although it did not remain significant after the final model adjustment, we did observe an unexpected negative association between BMI at age 2 years and HbA1c; again, only in males. A similar association was observed in a Danish cohort, where increased fetal growth and early weight gain up to age 3 years were thought to induce favorable adipose tissue metabolism with an increased production of insulin sensitivity-increasing adiponectin. Conversely, when weight gain occurred predominantly after 4 years in the Danish cohort, further fat accumulation favored insulin resistance with decreasing adiponectin.¹⁰ Risk and protective factors associated with high HbA1c in Pacific children require additional longitudinal investigation including studies beginning at an earlier age and an expanded set of glucose-insulin metabolism biomarkers.

Our BP findings also echoed prior studies, with higher BMI and more rapid BMI gain associated with increased mid-childhood BP. In the PROBIT study, a 1 SD faster weight gain in infancy (birth to 12 months) and early childhood (12–60 months) was associated with increases in systolic blood pressure of between 0.49 and 1.32 mmHg, with the magnitude of the association increasing with age³⁹. The effects observed in our study were significantly larger, perhaps as a result of the greater weight and higher BPs observed in our Samoan sample.

Sex differences in the associations observed also warrant exploration. Given the high levels of adiposity among the sample there is the possibility, although we did not measure this directly, that the onset of pubertal maturation and accompanying hormonal and metabolic changes may have already begun occurring in females and contributed to differences observed. In the Pacific Island Families study sex differences in the associations between body size trajectories and cardiometabolic risk markers were also observed, with more rapid weight gain (based on a positive slope of Z-score trajectories) associated with insulin resistance in both sexes, BP in males, leptin in males, and adiponectin in females at age 14 years.¹⁴ Manifestations of CVD risk in response to rapid BMI gain may, therefore, differ by sex, necessitating sex-specific approaches to cardiometabolic risk screening.

Study limitations and strengths

There are several study limitations that should be considered. We recognize there is an overlap in timing between the BMI trajectories and CVD risk measurement that makes it difficult to distinguish temporality. Therefore, we present estimates of association rather than inferring causation in this study. We also note the limitations of BMI as an indicator of adiposity. Ideally, we would have had more frequent measures in childhood with which to develop trajectories that could sensitively identify key events, such as the adiposity rebound, whose timing and magnitude have been associated with CVD risk in other studies^40,41. Residual confounding may be present from in utero exposures and pubertal development which were not measured in the cohort. The categorical cutoffs used to define high HbA1c and elevated BP in childhood also require careful consideration. While a HbA1c of 5.7% has been documented as a sensitive cut-off in childhood to predict future diabetes risk in a population without diabetes⁶, further investigations are necessary to better understand the distribution of childhood HbA1c and develop risk thresholds in this setting. Among children with a HbA1c of ≥ 6.5% in the cohort, there were no additional data available to speak to timing of onset or type of diabetes. Measurements of insulin resistance (e.g., insulin, glucose, HOMA-IR) can help to understand whether high HbA1c is due to insulin resistance or Type 1 diabetes/early onset of dysglycemia prior to type 1 diabetes onset. For elevated BP, we used 2017 AAP guidelines that included US children aged ≤13 years from 11 studies conducted between 1976 and 2000 with <85^th BMI percentile but recognize the differences between that reference population and the sample studied here.¹⁹ Although the convenience sampling of the cohort may limit generalizability, children were from the rapidly modernizing (or transitioning) island of ‘Upolu and living in households with lower assets compared to the national population²⁶, who may be among those with the highest risk for cardiometabolic diseases⁴².

Conclusion

BMI trajectories offer insight into the life course processes that may confer high risk for CMDs. The moderate prevalence of high HbA1c and elevated BP among Samoans at age 6- to 9-years reinforces the need for interventions starting from early childhood to reduce CMD risk and the associations between BMI traits and CMD risk observed provide some preliminary information on which to base risk screening. Future work, including utilizing the ongoing nature of this cohort and examining the manifestation of type 2 diabetes and hypertension, should focus on continuing to identify critical periods in which to apply CMD risk screening and intervention.

Table 1.

Baseline characteristics of the sample by sex at enrollment 2015–2017

	Total (N=410)	Female (n=207)	Male (n=203)	P*
Individual
Age, years	4.05 (2.85 – 4.81)	4.10 (2.89 – 4.89)	4.00 (2.83 – 4.77)	0.522
High adherence to modern diet	201 (49.02)	98 (47.34)	103 (50.74)	0.492
Physical activity tertile†: Low	182 (44.39)	102 (49.28)	80 (39.41)	0.036
Medium	117 (28.54)	60 (28.99)	57 (28.08)
High	111 (27.07)	45 (21.74)	66 (32.51)
Total energy intake,1000 cal/day	2.54 (1.59 – 4.94)	2.61 (1.57 – 5.47)	2.49 (1.64 – 4.30)	0.429
Ever breastfed	295 (71.95)	155 (74.88)	140 (68.97)	0.183
Screentime, min/day	60 (60 –120)	60 (60 – 120)	60 (60 – 120)	0.965
Body size
Weight, kg	16.40 (13.85 – 19.25)	15.95 (13.75 – 19.20)	16.85 (14.00 – 19.40)	0.427
Height, cm	99.60 ± 0.58	99.74 ± 0.84	99.45 ± 0.80	0.807
BMI, kg/m2	16.69 (15.85 – 17.72)	16.45 (15.63 – 17.63)	16.97 (16.01 – 17.80)	0.025
Maternal				0.889
Age group: 18–24 years	69 (16.83)	33 (15.94)	36 (17.73)
25–39	231 (56.34)	118 (57.00)	113 (55.67)
≥40	110 (26.83)	56 (27.05)	54 (26.60)
Education, years	12 (12 –13)	12 (12–13)	12 (12–13)	0.921
Married or cohabitating‡	324 (79.02)	168 (81.16)	156 (76.85)	0.284
Body size
Weight, kg	86.20 (74.25 – 99.00)	86.55 (76.00 – 101.70)	85.33 (72.70 – 97.00)	0.767
Height, cm	161.39 ± 0.27	161.26 ± 0.37	161.52 ± 0.38	0.630
BMI, kg/m2	33.11 (28.95 – 37.63)	33.44 (29.75 – 38.86)	32.95 (28.28 – 37.05)	0.068
Hypertension (≥140/90 mmHg)§	40 (9.76)	21 (10.14)	19 (9.36)	0.789
Household
Asset score quartile**: Low	89 (21.71)	50 (24.15)	39 (19.21)	0.206
Medium	92 (22.44)	46 (22.22)	46 (22.66)
High	127 (30.98)	55 (26.57)	72 (35.47)
Very High	102 (24.88)	56 (27.05)	46 (22.66)
Urbanicity-Region: Rural-Rest of Upolu	127 (30.98)	58 (28.02)	69 (33.99)	0.282
Periurban-Northwest Upolu	144 (35.12)	72 (34.78)	72 (35.47)
Urban-Apia Urban Area	139 (33.90)	77 (37.20)	62 (30.54)
Annual income ≥10,000 tala	84 (20.49)	41 (19.81)	43 (21.18)	0.730

^*Exact p-values or Monte-Carlo estimation of exact p-values were based on the exact Wilcoxon two-sample tests for continuous variables that are not normally distributed to assess differences by sex. P values are also based on two-sample t tests for continuous variables that are normally distributed and chi-square test for categorical variables.

^†The sum of 3-items from the Netherlands Physical Activity Questionnaire for young children. The score is out of a maximum of 15, with higher scores indicating greater physical activity, and were classified into tertiles: low (mean: 9.46, SE: 0.17), medium (mean: 13.15, SE: 0.03), and high (mean: 15.00, SE: 0).

^‡Compared to mothers who reported being separated, divorced, or never married.

^§Includes 25 with systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg (not both)

^**The sum of consumer durables owned (fridge, freezer, stereo, portable stereo, microwave oven, rice cooker, blender, sewing machine, television, VCR/DVD, couch, washing machine, landline telephone, computer/laptop, tablet, electric fan, air conditioner, and motor vehicle). The score is out of a maximum of 18, with higher scores indicating greater assets, and was classified into quartiles: low (mean:1.11, SE: 0.09), medium (mean: 3.52, SE: 0.05), high (mean: 5.93, SE: 0.07), and very high (mean: 10.72, SE: 0.28).

Abbreviations

AAP

American Academy of Pediatrics

BMI

Body mass index

BP

Blood pressure

HbA1c

Glycated hemoglobin

PR

Prevalence Ratio

Q1

Quartile 1 – 25^th percentile

Q3

Quartile 3 – 75^th percentile

SE

Standard error