Increased Adiposity and Low Height-for-Age in Early Childhood Are Associated With Later Metabolic Risks in American Indian Children and Adolescents

ABSTRACT

Background

Growth abnormalities in childhood have been related to later cardiometabolic risks, but little is known about these associations in populations at high risk of type 2 diabetes.

Objectives

We examined the associations of patterns of growth, including weight and height at ages 1–59 months, with cardiometabolic risk factors at ages 5–16 years.

Methods

We linked anthropometric data collected at ages 1–59 months to cardiometabolic data obtained from a longitudinal study in a southwestern American Indian population at high risk of diabetes. Analyses included 701 children with ≥1 follow-up examination at ages 5–16 years. We derived age- and sex-specific weight-for-height z-scores (WHZ) and height-for-age z-scores (HAZ) at ages 1–59 months. We selected the highest observed WHZ and the lowest observed HAZ at ages 1–59 months and analyzed associations of z-scores and categories of WHZ and HAZ with cardiometabolic outcomes at ages 5–16 years. We used linear mixed-effects models to account for repeated measures.

Results

Overweight/obesity (WHZ >2) at ages 1–59 months was significantly associated with increased BMI, fasting and 2-hour postload plasma glucose, fasting and 2-hour insulin, triglycerides, systolic blood pressure, diastolic blood pressure, and decreased HDL cholesterol at ages 5–16 years relative to normal weight (WHZ ≤1). For example, at ages 5–9 years, 2-hour glucose was 10.4 mg/dL higher (95% CI: 5.6–15.3 mg/dL) and fasting insulin was 4.29 μU/mL higher (95% CI: 2.96–5.71 μU/mL) in those with overweight/obesity in early childhood. Associations were attenuated and no longer significant when adjusted for concurrent BMI. A low height-for-age (HAZ < −2) at ages 1–59 months was associated with 5.37 mg/dL lower HDL (95% CI: 2.57–8.17 mg/dL) and 27.5 μU/mL higher 2-hour insulin (95% CI: 3.41–57.6 μU/mL) at ages 10–16 years relative to an HAZ ≥0.

Conclusions

In this American Indian population, findings suggest a strong contribution of overweight/obesity in early childhood to cardiometabolic risks in later childhood and adolescence, mediated through persistent overweight/obesity into later ages. Findings also suggest potential adverse effects of low height-for-age, which require confirmation.

See corresponding editorial on page 1803.

Introduction

Infancy and early childhood comprise critical stages of development, and physical growth during this vulnerable period of life may impact future cardiometabolic risks (1, 2). Longitudinal studies show that children at the highest end of the distribution for weight (whether expressed as weight, BMI, or weight-for-height) or who gain weight rapidly have increased cardiometabolic risks in later life (3, 4). Excessive weight in early childhood also increases the risks of persistent obesity and its related complications in later childhood (5) and adolescence (6). However, it remains unclear how early in life these patterns of increased childhood adiposity and later cardiometabolic risks are established. This information could help determine whether obesity prevention efforts are best directed to very early childhood, before a pattern of overweight or obesity is established, or to later in childhood.

Relative to childhood adiposity, the cardiometabolic consequences of impaired growth (i.e., reduced linear growth rate) are less well understood. While some studies have shown adverse cardiometabolic risks in those with low height-for-age, others have not, and in general, evidence for associations of low height-for-age in childhood with subsequent cardiometabolic risks is inconclusive (7–10). Impaired growth often results from adverse early life environments (i.e., poor nutrition, repeated infection, and inadequate psychosocial stimulation), which have been associated with adult metabolic dysregulation in insulin, glucose, and triglycerides (11, 12).

In addition, little is known about associations of patterns of growth at ages <5 years with cardiometabolic outcomes in American Indian populations, who are at high risk of obesity (13) and type 2 diabetes (14–16), and whose patterns of growth seem to differ from the general pediatric population in the United States (17). There is considerable ethnic variability in the associations of adiposity and early life exposures with cardiometabolic risks (18); thus, studies from populations at high risk of type 2 diabetes may be informative.

In the present study, we use data obtained from American Indians from the southwestern United States who participated in a longitudinal epidemiological study of type 2 diabetes to investigate associations of patterns of growth, including weight-for-height and height-for-age at ages 1–59 months, with cardiometabolic risk factors in later childhood (ages 5 to 9 years) and adolescence (ages 10 to 16 years). This population has a high prevalence of type 2 diabetes and obesity (14).

Previous studies in this population have documented positive associations of body weight and weight gain in childhood with subsequent cardiometabolic risks (19) and type 2 diabetes (20–22). However, these studies have only included children at ages ≥5 years. The relationship of patterns of growth at ages younger than 5 years of age with subsequent cardiometabolic risks in this population of American Indians has not been previously assessed.

Methods

Study design and participants

We analyzed anthropometric data collected from well-child care examinations at ages 1–59 months as predictors of cardiometabolic risk factors measured at ages 5–16 years of age in a longitudinal cohort study. The longitudinal cohort study of type 2 diabetes and its complications was conducted from 1965 to 2007 in a southwestern American Indian community in Arizona. Residents in the community who were 5 years of age or older were invited to participate in research examinations every 2 years (23). Biennial assessments included anthropometric measures, a detailed medical history, and biochemical tests. In the well-child care examinations, anthropometric data were obtained in children <5 years at visits conducted at health-care facilities in the community between January 1990 and June 2000 (17). Children were routinely scheduled to have height and weight measurements at or near ages 1, 2, 4, 6, 9, 12, 18, 21, 24, 36, and 48 months. Anthropometric data were available for 1898 children (7398 examinations). Out of 1898 children with valid anthropometric data, 708 had at least 1 follow-up examination in the longitudinal study. The baseline characteristics were similar, and the differences were relatively small, between children who did not have any follow-up examinations (n = 1190) and those included in the analysis (n = 708; Supplemental Table 1).

Because the routine clinical examinations in children <5 years were conducted from 1990 to 2000, and the longitudinal cohort study ended in 2007, follow-up data were only available through age 16 years. Follow-up examinations were divided into 2 age categories: 5 to 9 years (childhood) and 10 to 16 years (adolescence). Participants were included in both age groups if they were examined during both age ranges. Since diabetes, or its treatments, can influence levels of cardiometabolic risk factors, we further excluded children who had developed diabetes before the first follow-up examination, regardless of the age at which it occurred, from analyses of these risk factors (n = 7). For children who were initially nondiabetic but who developed diabetes during follow-up (n = 17), only examinations prior to diabetes onset were included. Our main analytical sample is based on 701 participants: 527 of these had 746 examinations at ages 5 to 9 years, and 418 had 598 examinations at ages 10 to 16 years (Supplemental Figure 1).

The study was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases. Parents provided written informed consent for minors aged <18 years, and child participants gave assent.

Study variables

Early life measures (1–59 months)

Length/height and weight were measured by trained staff (17). Recumbent length was measured in children <2 years, and standing height was measured in children ≥2 years using stadiometers. Weight was measured using electronic scales in children with no clothing other than underpants or dry diapers.

We derived age- and sex-specific weight-for-height z-scores (WHZ) and height-for-age z-scores (HAZ) using the 2006 WHO Child Growth Standards (24) and 2000 CDC Growth Charts (25). The WHZ provides an indirect measure of child adiposity (26, 27). High values (e.g., a WHZ above 2) are used to define overweight/obesity. The HAZ is a measure of attained linear growth. Low values (e.g., an HAZ below −2), are proposed by the WHO to define stunting (24).

In the United States, the WHO Child Growth Standards are recommended to monitor growth for children <2 years and the CDC Growth Charts are recommended for children 2 years and older (28). Previous studies using the CDC growth references in this population have shown, on average, high values of WHZ after age 6 months and slightly low values of HAZ up until age 24 months (17). We plotted mean WHZ and HAZ values by age and sex using the WHO and CDC growth references, and found similar patterns with both references; however, there were systematic differences, with the WHO references giving higher WHZ and lower HAZ values than the CDC references (Supplemental Figure 2). For consistency across the early childhood examinations, we therefore used the WHO Child Growth Standards for all assessments in children <5 years. We also estimated correlations of WHZ and HAZ scores between the WHO and CDC growth references by age, and between WHZ and HAZ scores by age using both references. Correlation coefficients of WHZ and HAZ between the WHO and CDC growth references were >0.98 for all pairwise comparisons, so they largely reflect shared information (Supplemental Table 2). Overall, correlation coefficients between WHZ and HAZ scores were weak and negative (Supplemental Table 3).

Where a participant had ≥1 anthropometric examination at ages 1–59 months, we selected the highest observed WHZ score and the lowest observed HAZ score and used them as predictors of study outcomes (under the assumption that ever having a high WHZ or low HAZ is the relevant biological exposure). We also created categorical variables for the highest WHZ and lowest HAZ based on the WHO criteria (29). Three categories were created for the WHZ: 1) normal weight (WHZ ≤1); 2) intermediate weight (1 < WHZ ≤ 2, which is the WHO definition of “possible risk of overweight”); and 3) overweight/obesity (WHZ >2). Four examinations (0.30% of the data) had a WHZ < −2 (“wasted” according to the WHO criteria) and were excluded from the analysis of WHZ categories. We also created categories for HAZ, with the first being a low height-for-age (HAZ < −2, based on the WHO definition of “stunting”). The WHO does not propose cutoff points to define intermediate and high HAZ scores, so we created 2 analogous categories: an HAZ below the median of the reference population (−2 ≤ HAZ < 0) was considered intermediate, and an HAZ above the median of the reference population (HAZ ≥0) was considered high. The distribution of participants in these 2 categories was more balanced under our cutoff points than when using the ranges of −2 ≤ HAZ < −1 and HAZ ≥ −1 for intermediate and high HAZ scores, respectively. For a more extreme comparison with an HAZ < −2, we used an HAZ ≥0 as the reference category in statistical analyses.

Childhood and adolescent measures (5–16 years)

In participants >5 years, anthropometric measures were taken by trained research personnel at each visit, with participants dressed in light clothing and no shoes. Height and weight were measured using a stadiometer and a calibrated scale, respectively. We computed age- and sex-specific BMI z-scores using the 2000 CDC Growth Charts (25). Serum total cholesterol, HDL cholesterol, and triglycerides were measured in venous blood obtained after an overnight fast. Plasma glucose and venous serum insulin concentrations were measured in venous blood obtained while fasting and 2 hours after a 75 g oral-glucose-tolerance test, interpreted by the WHO criteria (30). Venous plasma glucose concentrations were determined using the hexokinase method (Ciba Corning Express 550 analyzer) or the glucose oxidase method (Vital Diagnostics Envoy). HDL cholesterol was measured by precipitation (Ciba Corning Express) or enzymatic methods (Vital Diagnostics Envoy), while total cholesterol and triglycerides were measured by enzymatic methods (Ciba Corning Express or Vital Diagnostics Envoy). Serum insulin levels were determined by radioimmunoassay, using a Concept 4 analyzer (ICN Pharmaceuticals, Inc.). Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were obtained at the first and fourth Korotkoff sounds, respectively. We considered an adverse cardiometabolic profile to consist of higher BMI, fasting and 2-hour glucose, fasting and 2-hour insulin, cholesterol, triglycerides, SBP, and DBP values and a lower HDL cholesterol value.

Potentially confounding variables

We accounted for the participant's sex, age at follow-up examination, and exposure to maternal diabetes in utero, as previous studies show maternal diabetes is associated with increased offspring risks of obesity and cardiometabolic disease (31, 32). Participants were classified as being exposed to diabetes in utero if their mother had a diagnosis of diabetes observed in the longitudinal study before the child's birth date. We did not have access to routine pregnancy glucose tolerance tests in these mothers to assess gestational diabetes, and these clinical diagnoses generally reflect diabetes that preceded the pregnancy. Out of those that were exposed to diabetes in utero (n = 88), the proportion of mothers who were diagnosed with type 2 diabetes before conception (>9 months before the child's birth) was 93.2%. We additionally accounted for BMI z-scores at follow-up. We did not consider dietary, socioeconomic, and physical activity factors due to data unavailability.

Statistical analysis

All analyses were conducted using the SAS 9.4 (SAS Institute). Because some variables were nonnormally distributed (i.e., fasting and 2-hour insulin and triglycerides), we summarized participant's characteristics using medians (50th percentile) and IQRs (25th–75th percentile) across age groups and categories of WHZ and HAZ scores.

We examined associations of child adiposity and linear growth in early childhood with cardiometabolic outcomes using linear mixed models. Fitted models included categories of WHZ or HAZ (with normal weight and height-for-age above the median as reference categories, respectively), along with age, sex, and intrauterine exposure to maternal diabetes as covariates. We also fitted models with WHZ or HAZ in a continuous form and additionally estimated semipartial correlations between WHZ or HAZ and cardiometabolic outcomes from the parameters of the mixed model. Fasting insulin, 2-hour insulin, and triglyceride data were log-transformed to reduce skewness. To account for multiple follow-up examinations in the same individual, the mixed models were fit with an autoregressive covariance matrix that allowed for greater correlations between measures that were in closer proximity in time with each other. To test whether concurrent BMI mediated the association, an additional model was fit that included concurrent BMI as an additional covariate, except where height or BMI was an outcome.

Taking the maximum observed WHZ or minimum observed HAZ over ages 1–59 months is potentially influenced by anomalous values. Thus, to examine the robustness of the findings, we also conducted analyses using the mean WHZ and HAZ scores (rather than the highest observed WHZ and lowest observed HAZ values) as predictors of study outcomes, adjusting for the same set of covariates previously described. To examine the influence of WHZ and HAZ at younger ages, we additionally ran models examining associations with cardiometabolic outcomes, restricting early childhood examinations to ages 1–24 months (which reduced the available sample size to 572).

In addition, we characterized individuals based on HAZ categories at ages 1–59 months and BMI categories at ages 5–9 and 10–16 years to examine whether cardiometabolic risks are influenced by “double exposure” to these risk factors (e.g., a low HAZ in early childhood and a high BMI at follow-up). We examined associations between these categories (with an HAZ ≥ −2 and a BMI <85th percentile as the reference categories) and cardiometabolic risk factors using linear mixed models, adjusting for age, sex, and intrauterine exposure to maternal diabetes as covariates.

WHZ and HAZ trajectories in children <5 years

To examine associations of weight and height gain in early childhood with later cardiometabolic outcomes, we also derived WHZ and HAZ trajectories using a latent class trajectory analysis (LCTA) (33, 34). An LCTA is a special case of growth mixture modeling that allows for the identification of distinct groups with similar underlying growth trajectories (35, 36), whose degrees and directions of change can vary freely. For model stability, the identification of trajectories was restricted to participants with ≥3 height and weight measures collected at ages 1–59 months: there were 510 participants who had ≥3 weight and height measures. The number of groups in the model and each group's functional relationship with time (linear, quadratic, or cubic) were varied until the optimal number of trajectory groups was determined by model fit statistics. The solution that best fitted the data was taken as the solution with the lowest Bayesian Information Criterion values, the highest posterior probability (>0.70 for each class), and a minimum of 10% of participants in each latent class. Once trajectories were determined, individuals were assigned to the class with the highest posterior probability (35). The derived latent classes were then used in linear mixed models, where trajectory membership was treated as an independent variable.

The sample size for these models was based on 510 participants: 388 of these had 555 examinations at ages 5 to 9, and 284 had 375 examinations at ages 10 to 16 years. Fasting insulin, 2-hour insulin, and triglyceride models were log-transformed to reduce skewness. All models were adjusted for age, sex, and intrauterine exposure to maternal diabetes. We also fitted models additionally adjusting for concurrent BMI, except where height or BMI was an outcome.

Results

Out of 7 children excluded from the analysis of cardiometabolic risk factors for presenting diabetes, 6 were exposed to diabetes in utero, 3 had overweight/obesity, and 3 had a low height-for-age. Similarly, out of the additional 17 children who developed diabetes during follow-up, 12 were exposed to diabetes in utero, 5 had overweight/obesity, and 2 had low a height-for-age.

The 701 children included in this study had 1344 examinations distributed across 2 age categories at follow-up. The mean ± SD follow-up time and number of examinations in childhood were 6.31 ± 1.95 years and 1.34 ± 0.56 examinations, respectively. In adolescence, the mean ± SD follow-up time was 10.1 ± 1.82 years, and the mean number of examinations was 1.35 ± 0.57. The mean ages ± SDs at which the maximum WHZ and minimum HAZ scores occurred in early childhood were 25.0 ± 17.2 and 20.7 ± 17.1 months, respectively.

During early childhood, overweight/obesity occurred in 30.1% of children. The proportion of participants exposed to maternal diabetes in utero was higher among those with overweight/obesity and a low height-for-age (15.6% and 18.5%, respectively), relative to other WHZ and HAZ categories (Table 1). Those in higher HAZ categories tended to continue to have a higher height and BMI at follow-up (Supplemental Figure 3A and B), except at ages >13 years, where sample sizes were small (8, 20, and 3 examinations for an HAZ above the median, an HAZ below the median, and a low HAZ, respectively). Characteristics of individuals according to categories of HAZ and WHZ are shown in Table 1.

TABLE 1.

Cardiometabolic characteristics of children and adolescents by categories of WHZ and HAZ at ages 1 to 59 months¹

	Age category	Normal weight	Intermediate weight	Overweight and obesity		Low height-for-age	HAZ below the median	HAZ above the median
Variables	years	WHZ ≤1	1 < WHZ ≤ 2	WHZ >2	P-trend2	HAZ < −2	−2 ≤ HAZ < 0	HAZ ≥0	P-trend2
Baseline characteristics
n, %	<5	213 (30.4)	277 (39.5)	211 (30.1)	—	124 (17.7)	436 (62.2)	141 (20.1)	—
Female, n (%)	<5	102 (47.9)	145 (52.3)	116 (55.0)	0.143	52 (41.9)	234 (53.7)	77 (54.6)	0.046
Exposure to diabetes in utero, n (%)	<5	23 (10.8)	32 (11.5)	33 (15.6)	0.132	23 (18.55)	55 (12.61)	10 (7.09)	0.005
Height, cm	<5	84 (71–97)	78 (67–92)	82 (68–100)	0.888	62 (56–73)	73 (61–91)	89 (76–100)	<0.001
HAZ	<5	−0.67 (−1.42 to 0.10)	−0.57 (−1.3 to 0.11)	−0.12 (−1.21 to 0.57)	—	−2.65 (−3.17 to −2.27)	−0.98 (−1.43 to −0.54)	0.42 (0.16–0.75)	—
WHZ	<5	0.44 (0.00–0.75)	1.48 (1.24–1.69)	2.68 (2.26–3.31)	—	0.99 (0.26–1.76)	0.90 (0.14–1.52)	1.24 (0.42–1.84)	—
Follow-up characteristics
Examinations,3n (%)	5–9	208 (27.9)	321 (43.0)	217 (29.1)	—	151 (20.3)	456 (61.1)	139 (18.6)	—
	10–16	188 (31.5)	240 (40.1)	170 (28.4)	—	79 (13.2)	372 (62.2)	147 (24.6)	—
Height, cm	5–9	128 (119–135)	128 (120–136)	132 (124–140)	<0.001	125 (118–132)	129 (121–136)	135 (127–141)	<0.001
	10–16	154 (146–160)	155 (149–161)	157 (150–162)	0.028	150 (144–155)	154 (147–159)	161 (155–167)	<0.001
BMI z4	5–9	0.42 (−0.23 to 1.22)	0.96 (0.21–2.02)	2.50 (1.11–3.87)	<0.001	0.76 (0.02–2.18)	0.96 (0.29–2.22)	2.14 (0.78–3.32)	<0.001
	10–16	1.11 (0.34–1.83)	1.63 (0.67–2.51)	2.49 (1.31–3.74)	<0.001	1.40 (0.56–2.50)	1.44 (0.58–2.51)	2.27 (1.04–3.28)	<0.001
Plasma fasting glucose, mg/dL	5–9	84 (81–88)	85 (80–90)	87 (82–91)	0.004	85 (80–90)	85 (81–90)	86 (82–90)	0.465
	10–16	89 (86–94)	89 (86–96)	91 (87–96)	0.093	90 (86–96)	90 (86–95)	90 (86–96)	0.735
Plasma 2-hour glucose, mg/dL	5–9	90 (79–103)	92 (80–109)	99 (85–115)	<0.001	90 (79–110)	92 (81–108)	97 (84–108)	0.253
	10–16	99 (87–114)	102 (89–117)	108 (92–125)	0.057	99 (89–115)	103 (89–117)	104 (91–120)	0.550
Serum fasting insulin, μU/mL	5–9	4.6 (2.6–7.8)	4.7 (2.7–8.6)	8.4 (4.6–16.2)	<0.001	4.1 (2.3–9.4)	5.2 (3.0–9.6)	7.8 (4.2–11.4)	<0.001
	10–16	10.0 (5.2–18.2)	12.0 (7.1–19.7)	15.6 (9.3–27.9)	<0.001	10.2 (5.7–19.6)	11.4 (6.3–20.0)	15.4 (9.4–28.4)	<0.001
Serum 2-hour insulin, μU/mL	5–9	33.3 (16.2–50.7)	31.6 (18.4–62.8)	58.8 (26.1–123.4)	<0.001	31.9 (20.0–68.1)	37.4 (20.5–71.2)	49.2 (24.9–94.4)	0.163
	10–16	56.6 (26.4–105.7)	68.9 (37.3–157.4)	83.7 (46.5–196.3)	0.001	61.8 (30.8–167.4)	65.1 (35.8–131.4)	75.7 (39.7–173.4)	0.283
Serum total cholesterol, mg/dL	5–9	147 (131–165)	148 (133–167)	151 (135–171)	0.537	147 (131–166)	148 (131–167)	154 (136–170)	0.417
	10–16	150 (132–167)	142 (126–161)	149 (128–168)	0.014	144 (127–161)	145 (126–165)	150 (135–166)	0.313
Serum HDL cholesterol, mg/dL	5–9	47 (40–56)	47 (40–56)	42 (36–51)	<0.001	47 (40–55)	46 (39–55)	44 (38–53)	0.343
	10–16	43 (38–51)	42 (36–52)	41 (35–47)	0.019	42 (37–51)	42 (36–49)	44 (37–52)	0.174
Serum triglycerides, mg/dL	5–9	50 (36–69)	61 (42–89)	78 (50–108)	<0.001	58 (39–94)	59 (41–88)	69 (51–98)	0.027
	10–16	79 (56–113)	84 (62–119)	90 (62–129)	0.144	89 (66–118)	80 (58–120)	89 (66–125)	0.519
SBP, mmHg	5–9	96 (86–104)	96 (90–106)	100 (90–112)	<0.001	96 (84–104)	98 (90–108)	100 (90–110)	0.054
	10–16	109 (100–118)	110 (100–118)	110 (104–120)	0.044	106 (98–112)	110 (100–118)	112 (104–120)	<0.001
DBP, mmHg	5–9	52 (50–60)	54 (50–60)	58 (50–64)	<0.001	50 (48–60)	56 (50–60)	56 (50–60)	0.002
	10–16	60 (52–67)	60 (52–68)	61 (54–70)	0.026	58 (50–68)	60 (52–68)	62 (56–70)	0.003

¹Values are medians (IQRs) unless otherwise indicated. WHZ and HAZ scores were derived using the 2006 WHO Child Growth Standards (24). Abbreviations: DBP, diastolic blood pressure; HAZ, height-for-age z-score; SBP, systolic blood pressure; WHZ, weight-for-height z-score.

² P-trend values were calculated using linear mixed-effects models (from inclusion of an ordered numeric variable).

³Values are frequencies and row percentages.

⁴Derived using the 2000 CDC Growth Charts (25).

Table 2 shows associations of WHZ categories with cardiometabolic outcomes. In models that adjusted for age, sex, and maternal diabetes, overweight/obesity (WHZ >2) at ages 1–59 months was significantly associated with a higher height and a greater BMI relative to normal weight (WHZ ≤1) at ages 5–9 and 10–16 years. It was also associated with increased fasting and 2-hour glucose, fasting and 2-hour insulin, SBP, and DBP values and decreased HDL cholesterol at ages 5–9 and 10–16 years. We also found positive and significant associations between overweight/obesity at ages 1–59 months (relative to normal weight) and serum triglycerides at ages 5–9 years. With an additional adjustment for concurrent BMI, associations were attenuated and no longer significant. Associations between intermediate weight (1 < WHZ ≤ 2) and cardiometabolic outcomes were also positive but of lesser magnitude, except HDL cholesterol, for which associations were negative.

TABLE 2.

Associations of categories of WHZ at ages 1 to 59 months with cardiometabolic outcomes in childhood and adolescence¹

	Childhood (5–9 years)2				Adolescence (10–16 years)3
	Intermediate weight (1 < WHZ ≤ 2)	Overweight or obese (WHZ >2)	P value	P-trend	Intermediate weight (1 < WHZ ≤ 2)	Overweight or obese (WHZ >2)	P value	P-trend
Height, cm
Model 1	0.85 (−0.14 to 1.85)	3.98 (2.88–5.07)	<0.001	<0.001	2.38 (1.09–3.67)	4.09 (2.68–5.49)	<0.001	<0.001
Model 2	0.85 (−0.15 to 1.85)	3.92 (2.83–5.02)	<0.001	<0.001	2.37 (1.08–3.66)	4.07 (2.67–5.48)	<0.001	<0.001
BMI-z
Model 1	0.57 (0.30–0.84)	2.04 (1.74–2.34)	<0.001	<0.001	0.51 (0.25–0.77)	1.46 (1.18–1.74)	<0.001	<0.001
Model 2	0.56 (0.30–0.82)	1.95 (1.67–2.24)	<0.001	<0.001	0.50 (0.25–0.76)	1.44 (1.17–1.72)	<0.001	<0.001
Plasma fasting glucose, mg/dL
Model 1	0.80 (−0.41 to 2.01)	2.26 (0.93–3.59)	0.003	0.002	0.84 (−0.48 to 2.17)	1.94 (0.50–3.38)	0.030	0.009
Model 2	0.76 (−0.44 to 1.96)	2.04 (0.71–3.36)	0.009	0.004	0.83 (−0.48 to 2.16)	1.92 (0.48–3.36)	0.032	0.010
Model 3	0.14 (−1.02 to 1.30)	−0.27 (−1.69 to 1.13)	0.796	0.583	0.37 (−0.94 to 1.69)	0.58 (−0.95 to 2.12)	0.746	0.475
Plasma 2-hour glucose, mg/dL
Model 1	3.96 (−0.55 to 8.47)	10.4 (5.56–15.2)	<0.001	<0.001	1.34 (−3.52 to 6.22)	6.15 (0.78–11.5)	0.062	0.026
Model 2	3.82 (−0.62 to 8.26)	9.47 (4.66–14.2)	<0.001	<0.001	1.40 (−3.44 to 6.25)	6.11 (0.76–11.4)	0.064	0.026
Model 3	1.05 (−3.10 to 5.20)	−0.27 (−5.24 to 4.70)	0.792	0.938	–2.39 (−6.95 to 2.17)	−4.25 (−9.66 to 1.15)	0.294	0.120
Serum fasting insulin,4 μU/mL
Model 1	0.75 (−0.06 to 1.64)	4.29 (2.96–5.71)	<0.001	<0.001	1.62 (−0.43 to 3.52)	5.96 (3.23–8.79)	<0.001	<0.001
Model 2	0.74 (−0.08 to 1.62)	4.11 (2.85–5.50)	<0.001	<0.001	1.65 (−0.31 to 3.50)	5.78 (3.14–8.55)	<0.001	<0.001
Model 3	−0.19 (−0.95 to 0.60)	−0.25 (−1.16 to 0.65)	0.851	0.483	−0.49 (−2.10 to 1.00)	−1.04 (−2.85 to 0.63)	0.523	0.248
Serum 2-hour insulin,4 μU/mL
Model 1	5.67 (−2.45 to 13.3)	30.3 (17.6–44.7)	<0.001	<0.001	14.0 (−1.71 to 30.5)	29.7 (8.84–53.1)	0.001	<0.001
Model 2	5.42 (−2.76 to 13.2)	28.20 (16.0–41.5)	<0.001	<0.001	14.7 (−1.03 to 31.0)	29.9 (9.40–52.6)	0.001	<0.001
Model 3	−0.71 (−9.26 to 7.71)	−0.18 (−10.2 to 9.70)	0.886	0.992	2.08 (−13.5 to 18.4)	−9.46 (−26.2 to 8.20)	0.400	0.305
Serum total cholesterol, mg/dL
Model 1	0.97 (−4.02 to 5.97)	2.97 (−2.47 to 8.42)	0.545	0.292	−8.23 (−13.9 to −2.53)	−3.60 (−9.78 to 2.57)	0.017	0.213
Model 2	0.95 (−4.04 to 5.95)	2.79 (−2.67 to 8.27)	0.590	0.326	−8.26 (−13.9 to −2.57)	−3.66 (−9.84 to 2.51)	0.016	0.206
Model 3	0.43 (−4.62 to 5.48)	1.01 (−5.06 to 7.10)	0.947	0.762	−9.42 (−15.1 to −3.68)	−7.03 (−13.7 to −0.35)	0.005	0.025
Serum HDL cholesterol, mg/dL
Model 1	−0.12 (−2.30 to 2.05)	−4.54 (−6.91 to −2.16)	<0.001	<0.001	−0.01 (−2.07 to 2.05)	−2.79 (−5.04 to −0.55)	0.017	0.013
Model 2	−0.10 (−2.27 to 2.07)	−4.35 (−6.72 to −1.97)	<0.001	<0.001	0.02 (−2.03 to 2.08)	−2.72 (−4.96 to −0.49)	0.019	0.015
Model 3	1.39 (−0.67 to 3.45)	0.78 (−1.69 to 3.27)	0.413	0.467	1.39 (−0.56 to 3.35)	1.26 (−1.02 to 3.54)	0.345	0.288
Serum triglycerides,4 mg/dL
Model 1	8.93 (3.15–14.9)	19.8 (12.5–27.7)	<0.001	<0.001	2.86 (−6.79 to 12.3)	9.66 (−0.30 to 19.8)	0.135	0.054
Model 2	8.86 (3.12–14.7)	19.3 (11.9–27.1)	<0.001	<0.001	2.84 (−6.67 to 12.2)	9.43 (−0.61 to 19.6)	0.152	0.062
Model 3	4.80 (−1.37 to 11.1)	3.04 (−4.76 to 10.9)	0.252	0.334	−2.59 (−11.9 to 7.19)	−5.54 (−17.2 to 5.35)	0.539	0.276
SBP, mmHg
Model 1	3.03 (0.81–5.25)	6.86 (4.44–9.27)	<0.001	<0.001	1.77 (−0.49 to 4.04)	4.47 (2.00–6.94)	0.001	<0.001
Model 2	3.02 (0.80–5.24)	6.78 (4.36–9.21)	<0.001	<0.001	1.73 (−0.52 to 3.98)	4.40 (1.94–6.85)	0.002	<0.001
Model 3	1.46 (−0.65 to 3.58)	1.34 (−1.20 to 3.89)	0.378	0.303	0.71 (−1.51 to 2.93)	1.47 (−1.12 to 4.06)	0.538	0.278
DBP, mmHg
Model 1	1.77 (0.09–3.45)	3.84 (2.00–5.67)	<0.001	<0.001	0.97 (−0.88 to 2.83)	3.28 (1.25–5.30)	0.005	0.002
Model 2	1.76 (0.08–3.44)	3.72 (1.89–5.56)	<0.001	<0.001	0.97 (−0.88 to 2.83)	3.28 (1.25–5.30)	0.005	0.002
Model 3	0.92 (−0.72 to 2.57)	0.78 (−1.19 to 2.77)	0.534	0.379	0.33 (−1.51 to 2.18)	1.43 (−0.73 to 3.60)	0.398	0.198

¹The reference category is normal weight, defined as a WHZ ≤1. Values are β coefficients (95% CI) from linear mixed-effects models. The P values for trends were calculated using linear mixed-effects models (from inclusion of an ordered numeric variable). Model 1 was adjusted for sex and age. Model 2 was additionally adjusted for exposure to diabetes in utero. Model 3 was additionally adjusted for concurrent BMI. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure; WHZ, weight-for-height z-score.

²Sample size is 744 for height and BMI, 629 for fasting glucose, 416 for 2-hr glucose, 534 for fasting insulin, 353 for 2-hr insulin, 630 for total cholesterol, 627 for HDL cholesterol, 600 for triglycerides, 742 for SBP, and 737 for DBP.

³Sample size is 596 for height and BMI, 573 for fasting glucose, 537 for 2-hr glucose, 520 for fasting insulin, 466 for 2-hr insulin, 580 for total cholesterol, 579 for HDL cholesterol, 561 for triglycerides, 597 for SBP, and 593 for DBP.

⁴Values are presented as differences in adjusted geometric means, derived from the β coefficients of the linear mixed models, with 95% CI.

Associations between WHZ score, analyzed as a continuous variable, and cardiometabolic outcomes showed similar results to those of WHZ categories (Table 3). The WHZ at ages 1–59 months was positively and significantly correlated with fasting and 2-hour glucose, fasting and 2-hour insulin, serum triglyceride, SBP, and DBP values and was negatively correlated with HDL cholesterol at ages 5–9 and 10–16 years. Again, most associations were attenuated and no longer statistically significant with a further adjustment for concurrent BMI. This suggests the associations may reflect obesity that develops in early childhood and persists at older ages; indeed, individuals in the high WHZ categories in early childhood tended to continue to have high BMIs at follow-up (Supplemental Figure 3C).

TABLE 3.

Associations of weight-for-height (WHZ) and height-for-age (HAZ) z-scores at ages 1 to 59 months with cardiometabolic outcomes in childhood and adolescence¹

	WHZ				HAZ
	Childhood (5–9 years)2		Adolescence (10–16 years)3		Childhood (5–9 years)2		Adolescence (10–16 years)3
	β (95% CI)	Semi-partial r (95% CI)	β (95% CI)	Semi-partial r (95% CI)	β (95% CI)	Semi-partial r (95% CI)	β (95% CI)	Semi-partial r (95% CI)
Height, cm
Model 1	1.47 (1.14–1.80)	0.30 (0.22–0.38)	1.23 (0.79–1.67)	0.21 (0.12–0.31)	2.55 (2.26–2.83)	0.54 (0.42–0.64)	2.66 (2.23–3.08)	0.45 (0.35–0.55)
Model 2	1.45 (1.12–1.79)	0.30 (0.21–0.38)	1.22 (0.77–1.66)	0.21 (0.12–0.31)	2.60 (2.32–2.89)	0.55 (0.43–0.65)	2.68 (2.26–3.10)	0.46 (0.35–0.56)
BMI-z
Model 1	0.80 (0.71–0.88)	0.56 (0.48–0.63)	0.54 (0.45–0.62)	0.45 (0.36–0.53)	0.24 (0.14–0.34)	0.17 (0.07–0.27)	0.21 (0.11–0.31)	0.16 (0.07–0.26)
Model 2	0.75 (0.67–0.84)	0.55 (0.47–0.62)	0.52 (0.43–0.60)	0.44 (0.35–0.52)	0.28 (0.18–0.37)	0.21 (0.10–0.31)	0.22 (0.13–0.32)	0.18 (0.08–0.28)
Plasma fasting glucose, mg/dL
Model 1	0.76 (0.37–1.17)	0.15 (0.06–0.23)	0.69 (0.23–1.15)	0.12 (0.04–0.20)	0.25 (−0.14 to 0.65)	0.05 (−0.04 to 0.14)	−0.36 (−0.85 to 0.13)	−0.05 (−0.15 to 0.03)
Model 2	0.64 (0.23–1.04)	0.12 (0.03–0.21)	0.65 (0.19–1.12)	0.11 (0.03–0.20)	0.33 (−0.06 to 0.72)	0.06 (−0.02 to 0.15)	−0.33 (−0.82 to 0.15)	−0.05 (−0.15 to 0.03)
Model 3	−0.28 (−0.74 to 0.18)	−0.05 (−0.16 to 0.05)	0.17 (−0.34 to 0.67)	0.03 (−0.06 to 0.12)	−0.01 (−0.39 to 0.38)	−0.01 (−0.10 to 0.10)	−0.61 (−1.10 to −0.12)	−0.10 (−0.19 to −0.01)
Plasma 2-hour glucose, mg/dL
Model 1	4.53 (3.05–6.00)	0.28 (0.19–0.37)	2.35 (0.64–4.07)	0.11 (0.03–0.20)	0.49 (−1.11 to 2.10)	0.03 (−0.06 to 0.13)	−0.15 (−1.96 to 1.66)	−0.006 (−0.10 to 0.09)
Model 2	4.01 (2.50–5.51)	0.25 (0.15–0.35)	2.15 (0.43–3.86)	0.10 (0.02–0.19)	0.77 (−0.80 to 2.36)	0.05 (−0.05 to 0.15)	0.04 (−1.76 to 1.85)	0.002 (−0.09 to 0.10)
Model 3	0.08 (−1.65 to 1.81)	0.01 (−0.13 to 0.14)	−1.64 (−3.42 to 0.13)	−0.08 (−0.18 to 0.01)	−0.29 (−1.74 to 1.16)	−0.02 (−0.11 to 0.08)	−1.52 (−3.22 to 0.17)	−0.07 (−0.17 to 0.02)
Log of serum fasting insulin4
Model 1	1.28 (1.21–1.35)	0.36 (0.26–0.46)	1.22 (1.16–1.29)	0.31 (0.20–0.41)	1.11 (1.05–1.18)	0.16 (0.05–0.26)	1.09 (1.03–1.17)	0.13 (0.03–0.24)
Model 2	1.26 (1.19–1.33)	0.35 (0.24–0.45)	1.20 (1.14–1.27)	0.29 (0.18–0.40)	1.12 (1.06–1.19)	0.18 (0.07–0.29)	1.10 (1.04–1.17)	0.15 (0.04–0.26)
Model 3	0.96 (0.91–1.00)	−0.08 (−0.22 to 0.04)	0.98 (0.94–1.02)	−0.04 (−0.16 to 0.07)	1.02 (0.98–1.07)	0.06 (−0.05 to 0.19)	0.99 (0.95–1.04)	−0.002 (−0.09 to 0.09)
Log of serum 2-hour insulin4
Model 1	1.33 (1.23–1.42)	0.36 (0.26–0.46)	1.19 (1.11–1.29)	0.20 (0.09–0.30)	1.04 (0.96–1.13)	0.05 (−0.07 to 0.18)	1.02 (0.93–1.11)	0.02 (−0.09 to 0.13)
Model 2	1.30 (1.20–1.41)	0.34 (0.24–0.44)	1.18 (1.09–1.28)	0.19 (0.08–0.29)	1.04 (0.96–1.14)	0.06 (−0.05 to 0.20)	1.02 (0.94–1.12)	0.03 (0.08–0.14)
Model 3	0.99 (0.92–1.08)	−0.007 (−0.14 to 0.12)	0.95 (0.89–1.03)	−0.05 (−0.17 to 0.05)	0.97 (0.90–1.03)	−0.04 (−0.14 to 0.06)	0.92 (0.86–0.99)	−0.10 (−0.20 to 0.01)
Serum total cholesterol, mg/dL
Model 1	0.79 (−0.84 to 2.44)	0.04 (−0.05 to 0.13)	−0.11 (−2.09 to 1.87)	−0.004 (−0.10 to 0.09)	1.38 (−0.26 to 3.02)	0.06 (−0.03 to 0.16)	2.10 (−0.01 to 4.22)	0.08 (−0.01 to 0.17)
Model 2	0.69 (−0.98 to 2.37)	0.03 (−0.05 to 0.12)	−0.22 (−2.22 to 1.78)	−0.009 (−0.10 to 0.09)	1.46 (−0.19 to 3.11)	0.06 (−0.02 to 0.16)	2.17 (0.05–4.29)	0.08 (−0.01 to 0.17)
Model 3	−0.08 (−2.07 to 1.91)	−0.004 (−0.11 to 0.10)	−1.41 (−3.65 to 0.81)	−0.059 (−0.16 to 0.05)	1.23 (−0.45 to 2.92)	0.05 (−0.03 to 0.15)	1.82 (−0.33 to 3.98)	0.07 (−0.02 to 0.16)
Serum HDL cholesterol, mg/dL
Model 1	−1.76 (−2.47 to −1.04)	−0.19 (−0.26 to −0.11)	−1.11 (−1.83 to −0.39)	−0.12 (−0.21 to −0.03)	0.02 (−0.71 to 0.75)	0.002 (−0.08 to 0.08)	0.85 (0.08–1.62)	0.08 (0.001–0.18)
Model 2	−1.66 (−2.39 to −0.94)	−0.18 (−0.25 to −0.10)	−1.00 (−1.72 to −0.27)	−0.11 (−0.20 to −0.02)	−0.06 (−0.78 to 0.67)	−0.005 (−0.08 to 0.07)	0.79 (0.02–1.55)	0.08 (0.007–0.17)
Model 3	0.36 (−0.45 to 1.17)	0.04 (−0.04 to 0.13)	0.52 (−0.24 to 1.28)	0.06 (−0.04 to 0.16)	0.68 (−0.00 to 1.35)	0.08 (0.002–0.16)	1.50 (0.77–2.22)	0.17 (0.07–0.26)
Log of serum triglycerides4
Model 1	1.11 (1.08–1.15)	0.27 (0.19–0.34)	1.04 (1.00–1.07)	0.09 (−0.004 to 0.17)	1.01 (0.98–1.05)	0.03 (−0.06 to 0.12)	0.98 (0.95–1.02)	−0.03 (−0.12 to 0.05)
Model 2	1.11 (1.07–1.14)	0.26 (0.18–0.33)	1.03 (0.99–1.07)	0.07 (−0.01 to 0.16)	1.02 (0.98–1.05)	0.04 (−0.05 to 0.13)	0.98 (0.95–1.02)	−0.03 (−0.11 to 0.05)
Model 3	1.01 (0.97–1.05)	0.03 (−0.07 to 0.13)	0.96 (0.93–1.00)	−0.08 (−0.19 to 0.02)	0.98 (0.94–1.00)	−0.05 (−0.14 to 0.02)	0.95 (0.92–0.99)	−0.10 (−0.18 to −0.02)
SBP, mmHg
Model 1	2.21 (1.47–2.95)	0.21 (0.11–0.29)	1.57 (0.80–2.34)	0.16 (0.07–0.24)	1.22 (0.49–1.95)	0.11 (0.03–0.20)	1.29 (0.48–2.10)	0.13 (0.02–0.23)
Model 2	2.20 (1.45–2.95)	0.21 (0.11–0.30)	1.46 (0.69–2.24)	0.15 (0.06–0.23)	1.28 (0.54–2.01)	0.12 (0.03–0.21)	1.37 (0.57–2.18)	0.14 (0.03–0.23)
Model 3	0.04 (−0.80 to 0.88)	0.004 (−0.08 to 0.10)	0.42 (−0.42 to 1.26)	0.04 (−0.04 to 0.13)	0.47 (−0.22 to 1.18)	0.05 (−0.03 to 0.12)	0.92 (0.12–1.71)	0.09 (−0.003 to 0.19)
DBP, mmHg
Model 1	1.49 (0.93–2.04)	0.19 (0.10–0.27)	0.91 (0.27–1.55)	0.11 (0.02–0.20)	0.76 (0.20–1.32)	0.09 (0.03–0.16)	0.82 (0.15–1.48)	0.09 (0.01–0.18)
Model 2	1.44 (0.87–2.01)	0.18 (0.10–0.26)	0.92 (0.28–1.56)	0.11 (0.02–0.20)	0.82 (0.26–1.37)	0.10 (0.04–0.17)	0.82 (0.15–1.49)	0.10 (0.04–0.17)
Model 3	0.37 (−0.28 to 1.03)	0.05 (−0.04 to 0.15)	0.23 (−0.47 to 0.93)	0.03 (−0.06 to 0.12)	0.39 (−0.15 to 0.94)	0.05 (−0.01 to 0.11)	0.52 (−0.14 to 1.19)	0.05 (−0.01 to 0.11)

¹Values are β coefficients (95% CIs) from linear mixed-effects models or r coefficients (95% CI) from semi partial correlations, as appropriate. Model 1 was adjusted for sex and age. Model 2 was additionally adjusted for exposure to diabetes in utero. Model 3 was additionally adjusted for concurrent BMI. Abbreviations: DBP, diastolic blood pressure; HAZ, height-for-age z-score; SBP, systolic blood pressure; WHZ, weight-for-height z-score.

²Sample sizes in childhood are 746 for height and BMI, 631 for fasting glucose, 416 for 2-hour glucose, 535 for fasting insulin, 353 for 2-hour insulin, 632 for total cholesterol, 629 for HDL cholesterol, 602 for triglycerides, 743 for SBP, and 738 for DBP.

³Sample sizes in adolescence are 598 for height and BMI, 574 for fasting glucose, 538 for 2-hour glucose, 521 for fasting insulin, 467 for 2-hour insulin, 581 for total cholesterol, 580 for HDL cholesterol, 562 for triglycerides, 597 for SBP, and 595 for DBP.

⁴β coefficient is expressed as a multiplier with 95% CI.

Results from analyses using the mean WHZ score were generally similar to those using the highest observed WHZ value (Supplemental Table 4).

Table 4 shows associations of HAZ categories at ages 1–59 months with cardiometabolic outcomes. In models that adjusted for age, sex, maternal diabetes, and concurrent BMI, a low height-for-age (HA < −2) was associated with 5.37 mg/dL lower HDL cholesterol (95% CI: 2.57–8.17 mg/dL), 27.5 μU/mL higher 2-hour insulin (95% CI: 3.41–57.6 μU/mL), 5.04 mmHg lower SBP (95% CI: 1.85–8.22 mmHg) and 2.85 mmHg lower DBP (95% CI: 0.19–5.52 mmHg) at ages 10–16 years. With an additional adjustment for height at follow-up, the differences in HDL cholesterol and 2-hour insulin were similar and remained statistically significant, while those for blood pressure were attenuated and, except for SBP in adolescence, no longer significant (Supplemental Table 5). When the intermediate and high HAZ categories were combined as the reference category, similar conclusions were obtained (Supplemental Table 6).

TABLE 4.

Associations of categories of height-for-age z-scores (HAZ) at ages 1 to 59 months with cardiometabolic outcomes in childhood and adolescence¹

	Childhood (5–9 years)2				Adolescence (10–16 years)3
	Low HAZ (HAZ < −2)	HAZ below the median (−2 ≤ HAZ < 0)	P value	P-trend	Low HAZ (HAZ < −2)	HAZ below the median (−2 ≤ HAZ < 0)	P value	P-trend
Height, cm
Model 1	−10.37 (−11.53 to −9.21)	−5.62 (−6.58 to −4.67)	<0.001	<0.001	−9.55 (−11.25 to −7.84)	−5.72 (−6.91 to −4.53)	<0.001	<0.001
Model 2	−10.56 (−11.72 to −9.41)	−5.74 (−6.68 to −4.79)	<0.001	<0.001	−9.69 (−11.39 to −7.98)	−5.78 (−6.97 to −4.60)	<0.001	<0.001
BMI-z
Model 1	−1.08 (−1.48 to −0.68)	−0.72 (−1.04 to −0.39)	<0.001	0.001	−0.77 (−1.17 to −0.38)	−0.66 (−0.93 to −0.39)	<0.001	<0.001
Model 2	−1.22 (−1.60 to −0.84)	−0.79 (−1.11 to −0.48)	<0.001	0.001	−0.86 (−1.24 to −0.47)	−0.69 (−0.96 to −0.42)	<0.001	<0.001
Plasma fasting glucose, mg/dL
Model 1	−1.32 (−2.97 to 0.31)	−1.09 (−2.45 to 0.26)	0.215	0.121	0.14 (−1.76 to 2.04)	−0.22 (−1.56 to 1.11)	0.885	0.969
Model 2	−1.60 (−3.23 to 0.02)	−1.23 (−2.57 to 0.10)	0.117	0.058	0.03 (−1.87 to 1.93)	−0.26 (−1.60 to 1.07)	0.891	0.940
Model 3	−0.13 (−1.73 to 1.47)	−0.20 (−1.51 to 1.09)	0.950	0.886	0.97 (−0.92 to 2.87)	0.51 (−0.83 to 1.85)	0.581	0.298
Plasma 2-hour glucose, mg/dL
Model 1	−5.02 (−11.42 to 1.37)	−3.27 (−8.46 to 1.91)	0.290	0.125	−1.06 (−8.18 to 6.05)	−2.61 (−7.47 to 2.25)	0.557	0.571
Model 2	−6.11 (−12.39 to 0.15)	−3.93 (−9.01 to 1.15)	0.150	0.058	−1.92 (−9.03 to 5.17)	−2.91 (−7.75 to 1.92)	0.496	0.422
Model 3	−0.65 (−6.49 to 5.18)	0.31 (−4.40 to 5.02)	0.911	0.809	3.27 (−3.38 to 9.93)	1.14 (−3.41 to 5.69)	0.626	0.348
Serum fasting insulin,4 μU/mL
Model 1	−2.51 (−4.11 to −1.07)	−2.05 (−3.42 to −0.77)	<0.001	<0.001	−4.05 (−7.44 to −0.83)	−3.82 (−6.54 to −1.24)	0.002	0.002
Model 2	−2.94 (−4.52 to −1.46)	−2.31 (−3.76 to −0.98)	<0.001	<0.001	−4.52 (−7.65 to −1.47)	−4.22 (−6.91 to −1.66)	<0.001	<0.001
Model 3	−0.41 (−1.41 to 0.60)	−0.48 (−1.38 to 0.31)	0.565	0.451	0.68 (−1.55 to 2.87)	−0.06 (−1.67 to 1.44)	0.808	0.564
Serum 2-hour insulin,4 μU/mL
Model 1	−8.36 (−23.9 to 6.51)	−8.54 (−21.6 to 3.16)	0.280	0.261	1.64 (−28.2 to 35.5)	−12.8 (−34.0 to 6.47)	0.236	0.561
Model 2	−10.9 (−27.8 to 4.69)	−10.4 (−24.5 to 1.55)	0.158	0.155	−1.13 (−30.6 to 30.8)	−14.3 (−36.1 to 4.83)	0.213	0.415
Model 3	5.02 (−7.05 to 17.1)	−0.50 (−10.6 to 8.77)	0.485	0.323	27.5 (3.41–57.6)	7.54 (−7.24 to 21.5)	0.086	0.033
Serum total cholesterol, mg/dL
Model 1	−4.55 (−11.31 to 2.20)	−3.59 (−9.09 to 1.90)	0.354	0.191	−7.36 (−15.57 to 0.85)	−3.80 (−9.54 to 1.92)	0.190	0.068
Model 2	−4.84 (−11.62 to 1.94)	−3.72 (−9.22 to 1.78)	0.319	0.166	−7.68 (−15.9 to 0.55)	−3.93 (−9.66 to 1.80)	0.166	0.058
Model 3	−3.81 (−10.76 to 3.12)	−3.00 (−8.60 to 2.59)	0.501	0.289	−6.41 (−14.77 to 1.95)	−2.96 (−8.80 to 2.87)	0.312	0.128
Serum HDL cholesterol, mg/dL
Model 1	1.68 (−1.31 to 4.68)	1.56 (−0.87 to 3.99)	0.418	0.279	−3.10 (−6.09 to −0.12)	−2.00 (−4.09 to 0.07)	0.076	0.026
Model 2	1.99 (−1.00 to 4.98)	1.69 (−0.72 to 4.12)	0.333	0.198	−2.79 (−5.77 to 0.18)	−1.88 (−3.96 to 0.18)	0.113	0.043
Model 3	−1.21 (−4.06 to 1.63)	−0.57 (−2.86 to 1.72)	0.703	0.401	−5.37 (−8.17 to −2.57)	−3.84 (−5.80 to −1.88)	<0.001	<0.001
Serum triglycerides,4 mg/dL
Model 1	−8.20 (−17.8 to 1.99)	−11.0 (−18.1 to −4.03)	0.010	0.072	3.40 (−9.82 to 16.7)	−1.64 (−11.3 to 8.00)	0.515	0.819
Model 2	−9.38 (−18.6 to 0.77)	−11.5 (−18.5 to −4.40)	0.005	0.036	2.48 (−10.5 to 15.9)	−2.19 (−11.8 to 7.22)	0.491	0.675
Model 3	1.45 (−6.91 to 10.1)	−4.34 (−10.9 to 1.72)	0.156	0.664	10.5 (−0.80 to 22.6)	4.80 (−4.01 to 13.5)	0.400	0.208
SBP, mmHg
Model 1	−4.67 (−7.64 to −1.70)	−2.19 (−4.63 to 0.24)	0.008	0.002	−6.38 (−9.63 to −3.14)	−2.18 (−4.44 to 0.08)	<0.001	<0.001
Model 2	−4.87 (−7.84 to −1.90)	−2.30 (−4.73 to 0.13)	0.005	0.001	−6.78 (−10.01 to −3.55)	−2.34 (−4.59 to −0.09)	<0.001	<0.001
Model 3	−1.39 (−4.23 to 1.45)	−0.05 (−2.36 to 2.25)	0.474	0.322	−5.04 (−8.22 to −1.85)	−0.93 (−3.16 to 1.29)	0.005	0.005
DBP, mmHg
Model 1	−3.19 (−5.42 to −0.95)	0.01 (−1.83 to 1.81)	0.002	0.004	−3.95 (−6.62 to −1.28)	−1.98 (−3.84 to −0.12)	0.011	0.002
Model 2	−3.41 (−5.65 to −1.17)	−0.11 (−1.94 to 1.70)	<0.001	0.002	−3.99 (−6.66 to −1.31)	−1.99 (−3.86 to −0.13)	0.011	0.002
Model 3	−1.52 (−3.74 to 0.69)	1.10 (−0.68 to 2.89)	0.010	0.152	−2.85 (−5.52 to −0.19)	−1.08 (−2.95 to 0.78)	0.109	0.109

¹The reference category is the HAZ above the median, defined as an HAZ ≥0. Values are β coefficients (95% CI) from linear mixed-effects models. P values for trend were calculated using linear mixed-effects models (from inclusion of an ordered numeric variable). Model 1 was adjusted for sex and age. Model 2 was additionally adjusted for exposure to diabetes in utero. Model 3 was additionally adjusted for concurrent BMI. Abbreviations: DBP, diastolic blood pressure; HAZ, height-for-age z-score; SBP, systolic blood pressure; WHZ, weight-for-height z-score.

²Sample sizes are 746 for height and BMI, 631 for fasting glucose, 416 for 2-hour glucose, 535 for fasting insulin, 353 for 2-hour insulin, 632 for total cholesterol, 629 for HDL cholesterol, 602 for triglycerides, 744 for SBP, and 739 for DBP.

³Sample sizes are 598 for height and BMI, 574 for fasting glucose, 538 for 2-hour glucose, 521 for fasting insulin, 467 for 2-hour insulin, 581 for total cholesterol, 580 for HDL cholesterol, 562 for triglycerides, 597 for SBP, and 595 for DBP.

⁴Values are presented as differences in adjusted geometric means, derived from the β coefficients of the linear mixed models, with 95% CIs.

Associations between HAZ scores, analyzed as a continuous variable, and cardiometabolic outcomes showed similar results. With adjustments for age, sex, maternal diabetes, and concurrent BMI, for every 1-unit increase in the HAZ score at ages 1–59 months, fasting glucose decreased by 0.61 mg/dL (95% CI: 0.12–1.10 mg/dL), 2-hour insulin decreased by 8% (β = 0.92; 95% CI: 0.86–0.99), HDL cholesterol increased by 1.50 mg/dL (95% CI: 0.77–2.22 mg/dL), triglycerides decreased by 5% (β = 0.95; 95% CI: 0.92–0.99), and SBP increased by 0.92 mmHg (95% CI: 0.12–1.71 mmHg) at ages 10–16 years. Most observed associations showed weak correlations (Table 3).

Results from analyses using the mean HAZ score showed generally similar associations compared to those using the lowest observed HAZ value in models that accounted for concurrent BMI. However, values were attenuated for fasting glucose in adolescence (Supplemental Table 4).

We also found that 5.6% and 7.0% of subjects at ages 5–9 and 10–16 years, respectively, had a low HAZ in early childhood and a high BMI at follow-up (Supplemental Table 7). These doubly exposed individuals tended to have significantly worse cardiometabolic profiles (higher values for most outcomes, lower HDL cholesterol) than lean individuals with higher HAZ scores, but most differences were not significant compared with individuals with higher HAZ scores in early childhood and a high BMI at follow-up (Supplemental Table 8).

Weight and height trajectories in children <5 years

An LCTA identified 3 WHZ trajectory classes with a cubic relation with time: 25.8% of subjects were in trajectory 1, 56.9% in trajectory 2, and 17.3% in trajectory 3 (Supplemental Figure 4). Trajectory 1 was characterized by starting in a relatively lower WHZ score and remaining stable throughout ages 1–59 months (mean WHZ ± SD, 0.62 ± 0.96). Trajectory 2 started at a higher WHZ score relative to trajectory 1 and remained stable throughout ages 1–59 months (mean WHZ ± SD, 1.68 ± 0.59). Last, trajectory 3 started with a slightly higher WHZ score relative to trajectory 2, followed by a steady relative increase throughout ages 1–59 months (mean WHZ ± SD, 3.54 ± 1.07). We descriptively called these trajectories low, intermediate, and increasing high, respectively. The LCTA also identified 4 HAZ score trajectory classes with a linear relation with time: 10.6% of subjects were in trajectory 1, 30.3% in trajectory 2, 43.6% in trajectory 3, and 15.5% in trajectory 4. These 4 classes showed parallel slopes (Supplemental Figure 5), indicating height stability throughout early childhood. Because we did not identify any group of children with increased height gains, we focus on results obtained from WHZ trajectories.

In models that adjusted for age, sex, and maternal diabetes, we found positive associations of the increasing high WHZ trajectory (relative to the low trajectory) with fasting and 2-hour insulin, total cholesterol, and SBP at ages 5–9 and 10–16 years. The increasing high WHZ trajectory was positively associated with fasting glucose, serum triglycerides, and DBP and negatively associated with HDL cholesterol at ages 5–9 years (Table 5). With an additional adjustment for concurrent BMI, the associations were attenuated and were no longer significant, except for total cholesterol at ages 10–16 years. Results showing associations between HAZ trajectories and cardiometabolic outcomes are similar to those obtained using HAZ scores (Supplemental Table 9).

TABLE 5.

Associations between WHZ trajectories at ages 1 to 59 months and cardiometabolic outcomes in childhood and adolescence¹

	Childhood (5–9 years)2			Adolescence (10–16 years)3
	Intermediate (trajectory 2)	Increasing high (trajectory 3)	P value	Intermediate (trajectory 2)	Increasing high (trajectory 3)	P value
Height, cm
Model 1	−0.05 (−1.21 to 1.12)	4.86 (3.29–6.42)	<0.001	2.67 (1.09–4.24)	4.32 (2.10–6.55)	<0.001
Model 2	−0.04 (−1.21 to 1.12)	4.88 (3.29–6.47)	<0.001	2.67 (1.09–4.25)	4.31 (2.07–6.55)	<0.001
BMI-z
Model 1	0.91 (0.61–1.22)	3.14 (2.73–3.54)	<0.001	0.68 (0.35–1.01)	2.03 (1.57–2.51)	<0.001
Model 2	0.90 (0.60–1.20)	3.01 (2.60–3.42)	<0.001	0.72 (0.39–1.05)	1.94 (1.48–2.41)	<0.001
Plasma fasting glucose, mg/dL
Model 1	0.91 (−0.47 to 2.30)	4.43 (2.58–6.27)	<0.001	0.90 (−0.76 to 2.57)	1.31 (−1.04 to 3.65)	0.459
Model 2	0.85 (−0.52 to 2.24)	3.98 (2.09–5.86)	<0.001	0.94 (−0.72 to 2.61)	1.21 (−1.15 to 3.58)	0.467
Model 3	−0.27 (−1.64 to 1.08)	0.22 (−1.88 to 2.34)	0.793	0.08 (−1.57 to 1.74)	−1.21 (−3.72 to 1.29)	0.496
Plasma 2-hour glucose, mg/dL
Model 1	2.75 (−2.64 to 8.14)	16.8 (9.59–23.9)	<0.001	8.74 (2.59–14.9)	12.4 (3.75–21.1)	0.005
Model 2	2.72 (−2.57 to 8.03)	14.3 (7.09–21.5)	<0.001	9.26 (3.12–15.4)	11.7 (3.02–20.4)	0.005
Model 3	−0.86 (−5.91 to 4.17)	0.33 (−7.48 to 8.15)	0.890	3.63 (−2.21 to 9.48)	−2.08 (−10.8 to 6.68)	0.196
Serum fasting insulin,4 μU/mL
Model 1	0.67 (−0.20 to 1.49)	6.77 (4.42–9.49)	<0.001	2.18 (−0.10 to 4.54)	9.48 (4.00–16.2)	<0.001
Model 2	0.65 (−0.22 to 1.47)	6.44 (4.19–9.18)	<0.001	2.54 (0.32–4.76)	7.82 (3.04–13.3)	<0.001
Model 3	−1.11 (−1.90 to −0.38)	−1.27 (−2.40 to −0.11)	0.018	−0.61 (−2.49 to 1.21)	−2.10 (−4.66 to 0.70)	0.156
Serum 2-hour insulin,4 μU/mL
Model 1	7.87 (−2.20 to 16.6)	54.5 (28.0–85.3)	<0.001	26.7 (9.80–43.9)	34.1 (1.16–76.1)	0.007
Model 2	8.05 (−2.25 to 16.9)	47.5 (24.7–74.3)	<0.001	29.2 (12.3–46.4)	30.5 (0.36–70.2)	0.004
Model 3	−2.46 (−14.3 to 7.71)	−3.76 (−20.8 to 12.8)	0.918	11.7 (−6.60 to 28.7)	−16.2 (−37.6 to 6.91)	0.022
Serum total cholesterol, mg/dL
Model 1	5.01 (−0.96 to 10.9)	8.74 (0.73–16.8)	0.087	1.22 (−5.89 to 8.33)	13.8 (3.90–23.8)	0.012
Model 2	4.96 (−1.02 to 10.9)	8.32 (0.14–16.5)	0.112	1.44 (−5.68 to 8.56)	13.3 (3.34–23.3)	0.019
Model 3	4.10 (−2.03 to 10.2)	5.23 (−4.36 to 14.8)	0.392	1.12 (−6.17 to 8.41)	12.5 (1.59–23.3)	0.047
Serum HDL cholesterol, mg/dL
Model 1	−1.87 (−4.52 to 0.77)	−7.35 (−10.9 to −3.80)	<0.001	−2.26 (−4.77 to 0.26)	−1.92 (−5.44 to 1.60)	0.208
Model 2	−1.84 (−4.49 to 0.80)	−7.12 (−10.7 to −3.50)	<0.001	−2.44 (−4.95 to 0.06)	−1.49 (−5.01 to 2.03)	0.159
Model 3	0.36 (−2.20 to 2.94)	0.82 (−3.19 to 4.85)	0.920	−0.42 (−2.80 to 1.95)	4.02 (0.46–7.57)	0.017
Serum triglycerides,4 mg/dL
Model 1	8.23 (0.55–15.5)	27.8 (15.0–41.0)	<0.001	7.31 (−4.68 to 18.5)	16.3 (−1.96 to 34.9)	0.166
Model 2	8.27 (0.59–15.4)	28.2 (15.0–42.1)	<0.001	7.85 (−3.85 to 19.0)	14.9 (−2.91 to 34.3)	0.198
Model 3	1.69 (−6.74 to 9.78)	1.22 (−12.6 to 15.2)	0.849	0.90 (−10.5 to 11.7)	−3.48 (−21.1 to 15.7)	0.791
SBP, mmHg
Model 1	3.77 (1.12–6.42)	9.76 (6.21–13.3)	<0.001	2.40 (−0.35 to 5.17)	6.17 (2.28–10.1)	0.008
Model 2	3.79 (1.13–6.44)	10.0 (6.44–13.6)	<0.001	2.55 (−0.20 to 5.31)	5.81 (1.90–9.72)	0.013
Model 3	0.96 (−1.59 to 3.53)	0.62 (−3.34 to 4.58)	0.750	0.99 (−1.73 to 3.72)	1.58 (−2.51 to 5.68)	0.699
DBP, mmHg
Model 1	1.71 (−0.29 to 3.70)	5.42 (2.74–8.09)	0.001	1.84 (−0.53 to 4.23)	2.00 (−1.35 to 5.36)	0.281
Model 2	1.72 (−0.28 to 3.72)	5.61 (2.89–8.33)	0.001	1.83 (−0.55 to 4.22)	2.03 (−1.33 to 5.41)	0.282
Model 3	0.30 (−1.69 to 2.30)	0.80 (−2.29 to 3.90)	0.878	0.70 (−1.68 to 3.09)	−1.00 (−4.58 to 2.58)	0.508

¹WHZ trajectories were derived using latent class trajectory analysis. The reference category is low WHZ trajectory (trajectory 1). Values are β coefficients (95% CIs) from linear mixed-effects models. Model 1 was adjusted for sex and age. Model 2 was additionally adjusted for exposure to diabetes in utero. Model 3 was additionally adjusted for concurrent BMI. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure; WHZ, weight-for-height z-score.

²Sample sizes are 555 for height and BMI, 470 for fasting glucose, 307 for 2-hour glucose, 418 for fasting insulin, 267 for 2-hour insulin, 466 for total cholesterol, 464 for HDL cholesterol, 447 for triglycerides, 555 for SBP, and 550 for DBP.

³Sample sizes are 375 for height and BMI, 360 for fasting glucose, 335 for 2-hour glucose, 328 for fasting insulin, 291 for 2-hour insulin, 364 for total cholesterol, 364 for HDL cholesterol, 354 for triglycerides, 374 for SBP, and 373 for DBP.

⁴Values are presented as differences in adjusted geometric means, derived from the β coefficients of the linear mixed models, with 95% CIs.

Discussion

In this longitudinal analysis, growth abnormalities were observed early in childhood. By ages 1–59 months, 30.1% of children had at least 1 examination with measurements showing overweight/obesity, and 17.7% had at least 1 examination with measurements showing a low height-for-age. Similar percentages have been documented in a previous study in this population at ages <5 years (17) and among other American Indian groups, in whom obesity often begins early in childhood (37). The point prevalence of overweight/obesity estimated in our population across all examinations at ages 1–59 months using the WHO criteria (24) was 14.7%. When restricting data to ages 1–24 months, the point prevalence was 14.0%, almost double the prevalence of overweight/obesity reported for the general US pediatric population at ages 0–2 years between the years of 1999–2010 using the WHO criteria (8.6%) (38). Similarly, the point prevalence of obesity (the CDC criteria define obesity as ≥95th the BMI percentile) (25) at ages 5–16 was 49.9% in this population, which is more than double the 18.2% observed in the US general population at ages 6–19 years (38). The present findings are consistent with previous reports indicating that American Indian children have higher rates of overweight/obesity compared to other ethnic groups in the United States (37, 39, 40).

The point prevalence of low height-for-age estimated in this population was 7.3% (across all examinations at ages 1–59 months). The HAZ is a measure of attained linear growth. Low values (e.g., a HAZ below −2) often reflect exposure to an adverse environment early in life (27, 41). Rates of low height-for-age tend to be higher among indigenous preschool-aged children relative to children from other ethnic groups, particularly in developing countries (42, 43). Differences in growth potential among different ethnic groups may be better explained by environmental factors (e.g., socioeconomic status, health, and nutrition) rather than by genetic differences (44, 45).

Our findings also show that the proportion of children exposed to diabetes in utero were higher among the overweight/obesity and low height-for-age groups than among other weight and height categories. In our previous study in this same population, at age 1.5 years, children exposed to diabetes in utero were shorter than those not exposed to diabetes (z-score 0.12 compared with 0.24; P < 0.01), while no differences between groups were observed for weight. However, by age 7.7 years, children exposed to diabetes were heavier than (z-score 0.89 compared with −0.07; P < 0.01) but had similar heights to those not exposed to diabetes (46).

The present study shows that higher adiposity at ages 1–59 months was significantly associated with greater cardiometabolic risks in childhood and adolescence in both linear and categorical analyses in models that accounted for age, sex, and intrauterine exposure to maternal diabetes. With an additional adjustment for concurrent BMI, these associations were attenuated to near null values and were no longer statistically significant, which is consistent with complete mediation by concurrent BMI (47). This suggests that the influence of early childhood adiposity on later cardiometabolic risks may be attributable to persistent overweight/obesity into later ages. Indeed, individuals in the high-weight categories at ages 1–59 months tended to continue to have high BMIs at ages 5–9 and 10–16 years (see Supplemental Figure 3C). These results are consistent with a previous study in this population showing that children with obesity at ages 5–9 years remained with obesity throughout adolescence (13). Previous studies in this population have examined risk factors associated with obesity between childhood (age 5 years) and adolescence, including measures of body composition (48), energy expenditure, and physical activity (49). Consistent with our study findings, results indicated that the body fat percentage at age 5 years was the most significant predictor of obesity at age 10 years (48). Additional factors associated with obesity included an increased screen time and low levels of physical activity (49).

Although dietary data were not available in the present study population, several dietary practices that may contribute to overweight/obesity include the wide consumption of high-fat, high-sugar foods and sugar-sweetened beverages, which is facilitated by the proliferation of fast-food restaurants and convenience food stores on and near reservations (50). It has been documented that American Indian adolescents from some tribes consume sugar-sweetened beverages at a rate of twice the national average (51) and have a greater preference for foods that are high in fat (52).

Taken together, our findings suggest that overweight/obesity contributes substantially to adverse cardiometabolic risks in this population, and that the effects of higher adiposity that occur early in childhood (WHZ was evaluated at ages 1–59 months and, on average, at age 2 years in the present study) track from early childhood to adolescence. Analyses that were restricted to WHZ and HAZ measures made at ages 1–24 months showed that associations of WHZ with cardiometabolic outcomes were somewhat attenuated compared to those made at ages 1–59 months (Supplemental Table 10).

In an LCTA, we also found that groups defined by increased rates of weight gain at ages 1–59 months had higher levels of cardiometabolic risk factors at ages 5–16 years. However, most of these associations were also attenuated by the adjustment for concurrent BMI. Those in the highest weight gain group at ages 1–59 months did have higher cholesterol at ages 10–16 years, and this was not affected by the adjustment for BMI at follow-up, suggesting independent adverse effects of increases in adiposity in early childhood. An LCTA offers advantages over single measures of growth, allowing for capture of the influences of growth velocity at certain periods of development on outcomes of interest, while controlling for repeated measures over time. However, our analysis of height-for-age trajectories did not give much more information than simply classifying individuals based on 1 examination.

Studies in other populations have also found adverse effects of early life overweight/obesity on later cardiometabolic risks. In the Bogalusa Heart Study, childhood obesity (a BMI ≥95th percentile) was associated with adverse cardiometabolic risk factors in adulthood, but associations were weak and attributable to persistent weight statuses from childhood to adulthood (53). Moreover, rapid weight gain from 1–48 months was associated with increased adiposity, higher blood pressure, and hyperinsulinemia at ages 4 to 5 years among Mexican children, and associations were mediated by the concurrent BMI (54). Similarly, in India, weight gain in early childhood was associated with higher insulin concentrations and HOMA-IR at 9–10 years of age (55).

The current findings also suggest that a low height-for-age in early childhood can affect cardiometabolic risks in later childhood and adolescence. However, the effects are much more modest than those for higher adiposity, as reflected in the weaker correlation coefficients and less consistent associations with increased cardiometabolic risks. We found that a low height-for-age at ages 1–59 months was associated with lower HDL cholesterol and higher 2-hour insulin at ages 10–16 years, accounting for age, sex, maternal diabetes, and concurrent BMI. However, it was also associated with lower SBP and DBP values, though these associations were attenuated by an adjustment for concurrent height. Potential mechanisms that underpin these associations are unclear but may involve impaired fat oxidation (56), increased visceral fat accumulation (57), and impaired insulin sensitivity (58). In Brazil, stunted children aged 8–11 years had significantly lower fat oxidation while fasting and 30 minutes after a meal (56), as well as increased insulin sensitivity (58), compared to nonstunted control children living in the same environment. Furthermore, stunting at age 2 years was associated with visceral fat accumulation in adulthood only after adjustments for current BMI (57). Because fat that is not oxidized must be stored, low fat oxidation could lead to increased fat deposition over time, contributing to the development of metabolic disorders later in life. In addition, low fat oxidation has been found to be a strong predictor of weight gain in this study population, independent of the 24-hour metabolic rate (59).

We also found that a small proportion of subjects had a low height-for-age in early childhood followed by a high BMI at ages 5–9 or 10–16 years (∼6% to 7% at each age group). These doubly exposed individuals tended to have worse cardiometabolic risk factors, but not generally more than what would be expected given their higher BMIs. In fact, we observed no significant differences between those with low height-for-age and normal height-for-age values in early childhood who later developed overweight or obesity (see Supplemental Table 8).

A strength of this study was the use of longitudinal data collected from early childhood to adolescence, allowing for the prospective examination of the links between early life nutrition and later health outcomes. Additionally, the methods used for examining associations of childhood growth (including weight and height gains) with later cardiometabolic risks allowed us to account for repeated observations in the same individual adequately.

Our study also has limitations. We did not adjust for age at puberty, diet, physical activity, or socioeconomic status due to unavailable data on these variables. Future studies that include these variables may help identify the mechanisms underlying these associations. In addition, there is an increase in the possibility of false positive findings occurring, because no mathematical correction was made for multiple comparisons (60).

Taken together, our findings suggest that in this American Indian population, overweight/obesity at ages 1–59 months strongly predicts higher BMI, fasting and 2-hour glucose, fasting and 2-hour insulin, triglyceride, SBP, and DPB values and lower HDL cholesterol levels at ages 5–16 years. These findings highlight early life as a potentially important target for interventions to decrease risks of cardiometabolic disease. They also suggest potential adverse effects of a low height-for-age, which require further confirmation.

Notes

This work was supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases.

Author disclosures: The authors report no conflicts of interest.

Supplemental Tables 1–10 and Supplemental Figures 1–5 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: DBP, diastolic blood pressure; HAZ, height-for-age z-score; LCTA, latent class trajectory analysis; SBP, systolic blood pressure; WHZ, weight-for-height z-score.

Data Availability

Data described in the manuscript will not be made available owing to privacy concerns of the small population of volunteers.