Serum Trans Fatty Acids Are Not Associated with Weight Gain or Linear Growth in School-Age Children^,,

Abstract

Background: Animal and human adult studies indicate that long-term intake of trans fatty acids (TFAs) may be associated with weight gain. High intake of fast foods and snacks, which are rich in TFAs, is linked to overweight status among school-age children. However, the specific effects of TFAs in this population are unknown.

Objective: We examined whether serum TFAs, used as biomarkers of intake, are associated with faster weight gain and linear growth during school years.

Methods: We quantified TFAs by GLC in serum samples of 668 children aged 5–12 y at the time of recruitment into an ongoing cohort study performed in Bogota (Colombia) since 2006. Serum proportions of trans palmitoleic acid (16:1t), trans oleic acid (18:1t), trans linoleic acid (18:2t), and total TFAs were used as biomarkers of intake. Anthropometric characteristics were measured periodically for a median of 30 mo. Body mass index-for-age z scores (BAZs) and height-for-age z scores (HAZs) were calculated with the use of the WHO reference. We estimated mean changes in BAZs and HAZs over follow-up according to quartiles of each TFA at baseline by using mixed-effects regression models with restricted cubic splines.

Results: Proportions of trans palmitoleic acid, trans oleic acid, trans linoleic acid, and total TFAs (mean ± SD, % of total serum FAs), were 0.22 ± 0.06, 0.91 ± 0.37, 0.96 ± 0.27, and 2.10 ± 0.59, respectively. Serum TFAs were not associated with changes in BAZs and HAZs after adjusting for sex, baseline age, and socioeconomic status. In a subgroup analysis by sex, serum trans palmitoleic acid was positively associated with the estimated change in HAZs from ages 6 to 14 y in boys (with use of the first quartile as the reference, differences in HAZs for trans palmitoleic acid quartiles were 0.73, 0.53, and 0.70, P-trend = 0.03).

Conclusions: Proportions of serum TFAs, used as biomarkers of TFA intake, were not associated with weight gain in children aged 6–14 y in low- and middle-income populations in Bogota. The proportion of trans palmitoleic acid was positively associated with linear growth in boys. Longer follow-up and studies in diverse cohorts with wider ranges of TFA intake are warranted.

Introduction

Obesity is a global pandemic that affects adults and children worldwide (1). Childhood overweight is an important predictor of adult overweight (2), cardiovascular disease in adulthood, and other chronic diseases (3). Yet, the specific causal biological processes leading to the rapid excessive weight gain currently taking place among children are not well understood. Evidence suggests that long-term intake of trans FAs (TFAs)⁷ may be associated with weight gain in animal and human studies (4, 5). Although high intake of fast foods and snacks, usually with high content of TFAs, is linked to overweight among children (6–8), to our knowledge, there are no studies of the effects of dietary TFAs on childhood overweight and linear growth.

TFAs are exogenous FAs for humans because we cannot synthesize them, and human enzyme systems are designed to handle double bonds in FAs in the cis position. Therefore, TFAs are believed to interfere with metabolic processes that could affect weight gain and linear growth. TFAs are linked to insulin resistance and obesity by inducing alterations in cell membranes (9) and through increasing concentrations of proinflammatory factors, including IL-6, TNF, and PGs (10). Furthermore, TFAs interfere with desaturation in the n–3 (ω-3) and n–6 (ω-6) FA pathways (11, 12) and may affect adiposity and linear growth, because n–3 FAs are linked to those outcomes in some studies (13). Because TFA intake is difficult to measure with dietary questionnaires, given the rapid change in the food supply, and because TFAs cannot be synthesized by humans, we propose to use serum TFAs as biomarkers of intake (14, 15).

Adipogenesis occurs throughout life, but it is highest during the first year of life and before puberty (16). Therefore, those life periods are critical for the prevention of overweight through dietary interventions. Understanding the impact of dietary TFAs on overweight risk and linear growth has substantial implications in both developed and developing countries where intake of TFAs is still above recommendations. Because dietary risk factors are known to track from childhood into adulthood (17), it is important to identify those factors that can be modified early to prevent future cardiovascular disease risk. With the use of data from an ongoing cohort study of school-age children in Bogota, Colombia, we examined whether TFAs from the diet, as measured by serum biomarkers, are associated with faster weight gain in childhood [i.e., changes in BMI-for-age z scores (BAZs)] and linear growth [i.e., changes in height-for-age z scores (HAZs)].

Methods

The study population is a subsample of the Bogota School Children Cohort, a prospective study of nutrition and health among children from public schools in Bogota, Colombia, ongoing since 2006. Details of the study design were previously reported (18). Briefly, 3202 schoolchildren aged 5–12 y were recruited from the Bogota public school system in February 2006. The study population is representative of low- and middle-income socioeconomic backgrounds in the city because we used a random sampling strategy and the public school system enrolls most of the children from these groups (19).

We collected baseline information from the children’s families with the use of a self-administered questionnaire that was completed by the mother or primary care provider and returned by 82% of households. This questionnaire inquired about maternal sociodemographic characteristics (age, marital status, education, and socioeconomic level), anthropometry (self-reported height and weight), and information about physical activity and sedentary habits of the child. Trained research assistants visited the schools to obtain anthropometric measurements and a fasting blood sample from the children. Height was measured without shoes to the nearest 1 mm with the use of a wall-mounted portable Seca 202 stadiometer (Seca), and weight was measured in light clothing to the nearest 0.1 kg with the use of Tanita HS301 solar-powered electronic scales (Tanita). Follow-up anthropometric measurements were obtained in June and November 2006 and once yearly thereafter by visiting the schools or homes of the children when they were absent from school on the day of assessment.

The study protocol was approved by the Ethics Committee of the National University of Colombia Medical School. The Institutional Review Board at the University of Michigan approved the use of data and samples from the study.

FA analysis.

We used TFAs in serum as a biomarker of dietary intake (15). We selected a random sample of 687 children that had enough serum volume (200 μL) for FA analysis. We analyzed 32 FAs, including the following TFAs: trans palmitoleic acid (16:1n–7t), petroselaidic acid (18:1n–12t), elaidic acid (18:1n–9t), vaccenic acid (18:1n–7t), 18:2n–6ct, 18:2n–6tc, and 18:2n–6tt. Results for other FAs were described elsewhere (20). Fasting blood specimens were obtained by venipuncture in 2816 cohort participants (88%) and transported on the same day to the National Institute of Health in Bogota where we performed a complete blood count and separated serum after centrifugation at 1500 × g for 15 min. Samples were then stored at −80°. These analyses were performed at the University of Michigan Metabolomics and Obesity Center Lipidomics Laboratory. With the use of methods described by Bligh and Dyer (21), we extracted lipids from serum samples and prepared FA methyl esters of total lipids with BF₃-methanol (22). Methyl esters were extracted from a TLC plate, and the solvents were dried and resuspended in a volume of hexane that was appropriate for the estimated sample concentration. A sample (1–2 mL) was injected via an auto sampler and analyzed on a GLC machine (Agilent, Model 6890N) with use of a 100-m SP-2560 column with optimum conditions for separation. Eluted peaks were analyzed with Chemstation software (Agilent). The proportion of each FA was determined with a calibration curve with C17:0 methyl ester as the standard. CVs for TFAs were between 1.1% and 2.3%.

Statistical analysis.

Of the 687 children selected for FA analysis, 668 with valid anthropometric measurements at baseline and ≥1 additional follow-up measurement comprised the final study population. These children did not differ from the rest of the cohort in terms of age, baseline anthropometric characteristics, maternal demographic characteristics, or weight. Nevertheless, compared with children excluded from analyses, the study sample included more girls (54.6% vs. 50.1%; P = 0.04) and a higher proportion of children from families in the upper 2 socioeconomic strata (66.2% vs. 59.5%; P = 0.002).

The exposures of interest were serum proportions of trans palmitoleic acid, trans oleic acid (18:1t), trans linoleic acid (18:2t), and total TFAs. First, we compared the distribution of each FA by child and maternal characteristics to identify variables that may confound the associations between FAs and anthropometric change. Children’s BAZs and HAZs were calculated with use of the sex-specific growth reference for children 5–19 y from the WHO (23). Household socioeconomic status (SES) corresponded to the government’s classification [1 (lowest) through 6 (highest); maximum of 4 in the study population) assigned to each household for tax and planning purposes (24). We assessed the statistical significance of these associations with the Wald chi-square test from univariate linear regression models. For ordinal characteristics, we obtained a test for linear trend by entering the variable of interest as a continuous predictor.

Next, we examined associations of TFAs quantified at baseline with change in BAZs and HAZs during follow-up. Follow-up measures had a median of 4, ranging from 2 to 5. We estimated mean BAZ and HAZ growth curves for sex- and age-specific quartiles of each FA marker by using mixed-effect models for repeated measurements with restricted cubic splines that represented nonlinear terms for the distribution of age at assessment that allow smoothing of nonlinear changes in anthropometry over time (20, 25, 26). In these analyses, knots were placed at 5.7, 9.2, 10.8, and 13.8, because these ages reflect curvilinear portions of the BMI-for-age and height-for-age growth reference for children 5–19 y (23). In the spline models, the outcomes were BAZ and HAZ, and the predictors included sex- and age-specific quartiles of each TFA biomarker, linear and spline terms for child age at assessment in decimal years, and interaction terms between FA quartiles and the child age terms. Random effects for the intercept and the linear term for age were included to account for the within-child correlation of repeated anthropometric measurements in the estimation of the variance (27). Because these methods do not require an equal number of measurements, or that measurements be obtained at the same time for all participants, all available measurements were included in the models. With the use of the growth curves constructed for children in each quartile of the TFA biomarkers, we estimated attained BAZs and HAZs at 6 and 14 y. Our primary outcome of interest was the change in each anthropometric measure between these 2 age points. Next, we estimated the difference in BAZ and HAZ change between children in the lowest and highest quartiles of TFAs at 6 and 14 y.

In multivariable analysis, we estimated adjusted differences and 95% CIs in anthropometric change from 6 to 14 y by introducing into the mixed models covariates that were associated with the FA markers in bivariate analyses or were known predictors of child growth. The final model included the child’s sex, baseline age, and household SES. In all models, we used empirical estimates of variance to overcome deviations from the multivariate normality assumption. Because the age distribution of children at baseline ranged from 5 to 12 y, we tested for possible birth cohort effects in the construction of the growth curves by including an interaction term between baseline age and age-adjusted quartiles of each TFA indicator; the interaction terms provided no evidence of birth cohort effects. We also examined associations of the FAs with BAZs and HAZs separately for boys and girls.

Values in the text are means ± SDs. P < 0.05 was considered significant for all statistical tests. All analyses were performed with the use of the Statistical Analyses System software 9.3 (SAS Institute Inc.).

Results

The age of children was 8.8 ± 1.7 y at the time of recruitment; 45.4% were boys. BAZ and HAZ at the time of recruitment were 0.09 ± 0.98 and −0.84 ± 0.97, respectively. During the median of 30.1 mo of follow-up (IQR: 29.8–30.8 mo), each child contributed a median of 4 measurements for weight and height, yielding a total of 2891 measurements.

Values for serum trans palmitoleic acid, trans oleic acid, trans linoleic acid, and total TFAs, expressed as % of total serum FAs, can be found in Table 1. Other FAs in serum were distributed as follows: saturated FAs: 30.5 ± 1.9; MUFAs: 22.2 ± 2.4; and PUFAs: 45.2 ± 3.5 (of those n–3 FAs represent 3.5 ± 1.0). BAZs or HAZs at baseline were not associated with TFAs. 18:2 TFAs were inversely associated with being born in Bogota and household SES. None of the other sociodemographic variables listed in Table 1 were associated with trans oleic acid, trans palmitoleic acid, or total TFAs.

TABLE 1.

Serum TFAs by child sociodemographic characteristics in 668 school-age children from Bogota, Colombia¹

Characteristic	n2	16:1t, % of total FAs	18:1t, % of total FAs	18:2t, % of total FAs	All TFAs, % of total FAs
Overall		0.22 ± 0.06	0.91 ± 0.37	0.96 ± 0.27	2.10 ± 0.59
Child’s sex
Female	365	0.22 ± 0.05	0.90 ± 0.27	0.96 ± 0.26	2.08 ± 0.51
Male	303	0.22 ± 0.06	0.92 ± 0.46	0.97 ± 0.29	2.12 ± 0.68
P3		0.51	0.35	0.40	0.37
Child’s age, y
5–6	115	0.23 ± 0.06	0.95 ± 0.28	0.96 ± 0.25	2.14 ± 0.52
7–8	215	0.22 ± 0.05	0.88 ± 0.24	0.96 ± 0.27	2.06 ± 0.49
9–10	286	0.22 ± 0.06	0.88 ± 0.30	0.96 ± 0.28	2.07 ± 0.55
11–12	52	0.22 ± 0.07	1.06 ± 0.90	1.01 ± 0.30	2.29 ± 1.11
P-trend4		0.49	0.74	0.48	0.66
Child born in Bogota
Yes	556	0.22 ± 0.06	0.92 ± 0.39	0.96 ± 0.27	2.10 ± 0.61
No	58	0.21 ± 0.04	0.89 ± 0.24	1.05 ± 0.30	2.16 ± 0.51
P3		0.38	0.49	0.04	0.45
Height-for-age z score5
< −2.0	62	0.23 ± 0.07	0.90 ± 0.29	0.96 ± 0.29	2.09 ± 0.56
−2.0 to < −1.0	212	0.22 ± 0.06	0.90 ± 0.29	0.98 ± 0.29	2.11 ± 0.58
−1.0 to < 1.0	370	0.22 ± 0.06	0.92 ± 0.43	0.96 ± 0.26	2.10 ± 0.62
≥1.0	24	0.21 ± 0.04	0.84 ± 0.22	0.89 ± 0.20	1.95 ± 0.36
P-trend4		0.76	0.89	0.28	0.63
BMI-for-age z score5
< −2.0	12	0.22 ± 0.07	0.82 ± 0.15	0.87 ± 0.13	1.91 ± 0.28
−2.0 to < 1.0	534	0.22 ± 0.06	0.92 ± 0.39	0.97 ± 0.28	2.11 ± 0.62
1.0 to < 2.0	106	0.22 ± 0.06	0.87 ± 0.25	0.95 ± 0.26	2.05 ± 0.49
≥2.0	16	0.23 ± 0.06	0.81 ± 0.12	0.87 ± 0.18	1.91 ± 0.30
P-trend4		0.77	0.05	0.30	0.12
Household SES6
1 lowest	49	0.22 ± 0.06	0.86 ± 0.22	1.03 ± 0.28	2.10 ± 0.52
2	177	0.23 ± 0.05	0.94 ± 0.51	0.99 ± 0.31	2.16 ± 0.75
3	353	0.22 ± 0.06	0.91 ± 0.32	0.94 ± 0.26	2.07 ± 0.55
4 highest	89	0.22 ± 0.06	0.88 ± 0.24	0.96 ± 0.22	2.05 ± 0.44
P-trend4		0.56	0.64	0.03	0.17

¹Values are means ± SDs unless otherwise indicated. SES, socioeconomic status; TFA, trans FA.

²Totals may be <668 because of missing values.

³Wald test from univariate linear regression models.

⁴From univariate regression models in which a variable that represented the ordinal predictor was introduced as continuous.

⁵According to the WHO growth standard for children 5–19 y of age (23).

⁶According to the government’s classification for tax and planning purposes (24).

None of the serum TFAs were significantly associated with BAZ change during follow-up after adjusting for sex, baseline age, and SES (Table 2). Analyses stratified by sex yielded comparable results (data not shown).

TABLE 2.

Estimated change in BMI-for-age z score among 668 school-age children according to quartiles of serum TFAs¹

			BMI-for-age z score2		Change2
Serum TFAs	n	Median, % total FAs	6 y	14 y	14 y–6 y	Difference in change3β (95% CI)	P-trend4
16:1t							0.95
Q1	167	0.16	−0.26 ± 0.12	0.10 ± 0.13	0.35 ± 0.16	Reference
Q2	166	0.20	−0.03 ± 0.13	0.07 ± 0.13	0.11 ± 0.17	−0.25 (−0.71, 0.21)
Q3	168	0.23	0.08 ± 0.12	0.15 ± 0.12	0.07 ± 0.15	−0.28 (−0.71, 0.14)
Q4	167	0.28	−0.11 ± 0.13	0.26 ± 0.11	0.37 ± 0.16	0.03 (−0.42, 0.47)
Difference Q4–Q15			0.14 (−0.20, 0.49)	0.16 (−0.16, 0.49)
Total 18:1t							0.58
Q1	167	0.64	−0.10 ± 0.12	0.09 ± 0.12	0.19 ± 0.16	Reference
Q2	167	0.80	−0.15 ± 0.14	0.02 ± 0.12	0.16 ± 0.18	−0.02 (−0.49, 0.46)
Q3	166	0.92	−0.06 ± 0.12	0.19 ± 0.12	0.25 ± 0.15	0.07 (−0.37, 0.51)
Q4	168	1.18	−0.01 ± 0.12	0.27 ± 0.11	0.28 ± 0.15	0.10 (−0.34, 0.54)
Difference Q4–Q15			0.09 (−0.25, 0.42)	0.18 (−0.15, 0.51)
Total 18:2t							0.73
Q1	168	0.69	−0.09 ± 0.14	0.13 ± 0.14	0.23 ± 0.19	Reference
Q2	166	0.85	−0.06 ± 0.12	0.11 ± 0.12	0.17 ± 0.16	−0.05 (−0.53, 0.43)
Q3	167	0.98	−0.11 ± 0.14	0.05 ± 0.12	0.17 ± 0.17	−0.06 (−0.56, 0.43)
Q4	167	1.29	−0.05 ± 0.10	0.26 ± 0.10	0.31 ± 0.13	0.09 (−0.37, 0.54)
Difference Q4–Q15			0.04 (−0.29, 0.38)	0.13 (−0.21, 0.47)
Total TFAs
Q1	168	1.57	−0.06 ± 0.12	0.06 ± 0.13	0.12 ± 0.17	Reference	0.64
Q2	166	1.86	−0.31 ± 0.14	0.03 ± 0.12	0.33 ± 0.18	0.22 (−0.26, 0.71)
Q3	166	2.13	0.04 ± 0.13	0.21 ± 0.12	0.17 ± 0.16	0.06 (−0.40, 0.52)
Q4	168	2.69	0.01 ± 0.10	0.28 ± 0.10	0.27 ± 0.14	0.16 (−0.27, 0.59)
Difference Q4–Q15			0.07 (−0.24, 0.37)	0.22 (−0.12, 0.55)

¹Quartiles for indicators are sex and age specific according to their distributions in the study population. Q, quartile; TFA, trans FA.

²Values are means ± SEs. Estimates are from growth curves built with use of mixed-effects models with restricted cubic splines that accounted for within-child repeated BMI measurements.

³Differences in change are adjusted for sex, baseline age, and socioeconomic status.

⁴Test for linear trend from a linear regression model where an ordinal variable representing quartiles of the FA indicator was entered as a continuous predictor.

⁵Values are mean (95% CI).

We next examined the associations of serum TFAs and change in HAZs (Supplemental Table 1). None of the serum TFAs were significantly associated to HAZ change during follow-up after adjusting for sex, baseline age, and SES (Supplemental Table 1). Analyses stratified by sex showed a significant positive association among boys for serum trans palmitoleic acid (Table 3) but not for girls (Table 4). Other serum TFAs were not associated with HAZ change overall or by sex.

TABLE 3.

Estimated change in height-for-age z score among 303 school-age boys according to quartiles of serum TFAs¹

			Height-for-age z score2		Change2
Serum TFAs	n	Median, % total FAs	6 y	14 y	14 y–6 y	Difference in change3β (95% CI)	P-trend4
16:1t							0.03
Q1	76	0.16	−0.85 ± 0.16	−0.89 ± 0.18	−0.04 ± 0.21	Reference
Q2	75	0.20	−1.10 ± 0.16	−0.42 ± 0.20	0.68 ± 0.20	0.73 (0.16, 1.29)
Q3	77	0.23	−0.98 ± 0.13	−0.50 ± 0.17	0.48 ± 0.18	0.53 (0.00, 1.06)
Q4	75	0.28	−1.38 ± 0.15	−0.72 ± 0.16	0.66 ± 0.18	0.70 (0.17, 1.23)
Difference Q4–Q15			−0.52 (−0.96, −0.09)	0.17 (−0.29, 0.64)
Total 18:1t							0.80
Q1	75	0.62	−1.11 ± 0.16	−0.66 ± 0.23	0.45 ± 0.22	Reference
Q2	76	0.79	−0.99 ± 0.16	−0.55 ± 0.17	0.45 ± 0.21	0.00 (−0.60, 0.60)
Q3	77	0.91	−1.03 ± 0.15	−0.73 ± 0.16	0.30 ± 0.18	−0.14 (−0.69, 0.42)
Q4	77	1.20	−1.16 ± 0.15	−0.61 ± 0.15	0.54 ± 0.19	0.10 (−0.46, 0.66)
Difference Q4–Q15			−0.05 (−0.48, 0.39)	0.05 (−0.48, 0.58)
Total 18:2t							0.50
Q1	76	0.68	−1.31 ± 0.16	−0.63 ± 0.19	0.68 ± 0.20	Reference
Q2	76	0.85	−0.88 ± 0.16	−0.60 ± 0.20	0.28 ± 0.21	−0.40 (−0.97, 0.16)
Q3	76	1.01	−1.00 ± 0.14	−0.71 ± 0.14	0.29 ± 0.17	−0.39 (−0.91, 0.13)
Q4	75	1.33	−1.09 ± 0.15	−0.62 ± 0.18	0.47 ± 0.20	−0.21 (−0.77, 0.35)
Difference Q4–Q15			0.22 (−0.21, 0.64)	0.01 (−0.50, 0.52)
Total TFAs							0.82
Q1	77	1.56	−1.14 ± 0.16	−0.62 ± 0.23	0.52 ± 0.22	Reference
Q2	75	1.86	−1.04 ± 0.19	−0.70 ± 0.14	0.34 ± 0.23	−0.18 (−0.80, 0.44)
Q3	75	2.13	−0.93 ± 0.13	−0.66 ± 0.17	0.27 ± 0.17	−0.24 (−0.79, 0.31)
Q4	76	2.81	−1.18 ± 0.15	−0.60 ± 0.16	0.58 ± 0.19	0.06 (−0.51, 0.63)
Difference Q4–Q15			−0.04 (−0.48, 0.39)	0.02 (−0.53, 0.57)

¹Quartiles for indicators are sex and age specific according to their distributions in the study population. Q, quartile; TFA, trans FA.

²Values are means ± SEs. Estimates are from growth curves built with use of mixed-effects models with restricted cubic splines that accounted for within-child repeated height measurements.

³Differences in change are adjusted for baseline age and socioeconomic status.

⁴Test for linear trend from a linear regression model in which an ordinal variable that represented quartiles of the FA indicator was entered as a continuous predictor.

⁵Values are mean (95% CI).

TABLE 4.

Estimated change in height-for-age z score among 365 school-age girls according to quartiles of serum TFA¹

			Height-for-age z score2		Change2
Serum TFAs	n	Median, % total FAs	6 y	14 y	14 y–6 y	Difference in change3 β (95% CI)	P-trend4
16:1t							0.46
Q1	91	0.16	−0.82 ± 0.16	−0.34 ± 0.14	0.48 ± 0.20	Reference
Q2	91	0.20	−0.77 ± 0.15	−0.51 ± 0.15	0.27 ± 0.18	−0.22 (−0.74, 0.31)
Q3	91	0.24	−0.85 ± 0.14	−0.65 ± 0.15	0.21 ± 0.21	−0.28 (−0.84, 0.28)
Q4	92	0.29	−0.68 ± 0.13	−0.38 ± 0.15	0.30 ± 0.17	−0.19 (−0.70, 0.33)
Difference Q4–Q15			0.14 (−0.27, 0.56)	−0.04 (−0.44, 0.36)
Total 18:1t							0.58
Q1	92	0.65	−0.84 ± 0.16	−0.39 ± 0.12	0.45 ± 0.18	Reference
Q2	91	0.81	−0.72 ± 0.15	−0.47 ± 0.17	0.25 ± 0.21	−0.20 (−0.74, 0.34)
Q3	91	0.94	−0.79 ± 0.15	−0.57 ± 0.18	0.22 ± 0.22	−0.22 (−0.78, 0.33)
Q4	91	1.17	−0.77 ± 0.14	−0.46 ± 0.12	0.31 ± 0.17	−0.14 (−0.63, 0.35)
Difference Q4–Q15			0.07 (−0.34, 0.48)	−0.07 (−0.41, 0.27)
Total 18:2t							0.37
Q1	92	0.68	−0.82 ± 0.15	−0.34 ± 0.14	0.48 ± 0.19	Reference
Q2	90	0.85	−0.70 ± 0.15	−0.49 ± 0.17	0.21 ± 0.22	−0.27 (−0.83, 0.30)
Q3	91	0.97	−0.70 ± 0.16	−0.31 ± 0.15	0.39 ± 0.21	−0.09 (−0.64, 0.46)
Q4	92	1.28	−0.92 ± 0.12	−0.74 ± 0.14	0.18 ± 0.15	−0.30 (−0.77, 0.16)
Difference Q4–Q15			−0.10 (−0.48, 0.28)	−0.40 (−0.78, −0.02)
All TFAS							0.26
Q1	91	1.60	−0.84 ± 0.17	−0.42 ± 0.13	0.41 ± 0.20	Reference
Q2	91	1.86	−0.77 ± 0.14	−0.27 ± 0.17	0.49 ± 0.20	0.08 (−0.47, 0.64)
Q3	91	2.13	−0.78 ± 0.14	−0.69 ± 0.15	0.09 ± 0.19	−0.32 (−0.86, 0.22)
Q4	92	2.65	−0.74 ± 0.14	−0.51 ± 0.14	0.24 ± 0.17	−0.18 (−0.68, 0.33)
Difference Q4–Q15			0.10 (−0.33, 0.52)	−0.08 (−0.46, 0.29)

¹Quartiles for indicators are sex and age specific according to their distributions in the study population. Q, quartile; TFA, trans FA.

²Values are means ± SEs. Estimates are from growth curves built with use of mixed-effects models with restricted cubic splines that accounted for within-child repeated height measurements.

³Differences in change are adjusted for baseline age and socioeconomic status.

⁴Test for linear trend from a linear regression model in which an ordinal variable that represented quartiles of the FA indicator was entered as a continuous predictor.

⁵Values are mean (95% CI).

Discussion

In this longitudinal study of school-age children from Bogota, Colombia, serum TFA proportions were not associated with weight gain or linear growth during a median of 2.5 y of follow-up. However, in subgroup analysis by sex, higher proportions of serum trans palmitoleic acid were associated with faster linear growth among boys.

A literature review found limited but consistent evidence from both prospective cohort studies of adult humans and randomized controlled trials in monkeys that TFA intake is correlated with both weight gain and weight retention (28). For example, monkeys fed for 6 y with a Western-style diet that contained 8% of TFAs had a 7.2% increase in body weight, compared with a 1.8% increase in monkeys that ate a similar isocaloric diet with the same percentage of fat but with no trans fat (4). Similarly, the isocaloric substitution of a 2% increment in energy intake from TFAs with PUFAs or carbohydrates was associated with a 0.77-cm waist circumference increase among 16,587 US men aged 40–75 y who were followed for 10 y (5). Considering that these studies spanned a markedly longer time frame than the length of follow-up in the present study, it is possible that the effect of TFAs on weight be observed after long exposure periods. It is also possible that the magnitude of the TFA exposure of our study population was not large enough to show positive results. For example, the overall proportions of TFAs in the Bogota children were 2.10 ± 0.59, whereas obese and nonobese children from Spain had 2.43 ± 0.26 and 2.29 ± 0.24, respectively (29).

TFAs have been linked to insulin resistance and obesity by inducing alterations in cell membranes (9) and through increasing concentrations of proinflammatory factors, including IL-6, TNF, and PGs (10). In one study among Spanish prepubertal children plasma TFAs were not associated with insulin resistance, but they remained higher in obese children than in control participants 3 h after a meal (29). Studies among children are scarce, and to our knowledge there are no longitudinal studies of the specific effect of dietary TFAs on childhood overweight. Although the specific biological mechanism by which TFAs may be associated with weight gain is not clear yet, TFAs may have adverse effects on growth and development through interfering with essential FA metabolism (12).

We found a positive association between serum trans palmitoleic acid and linear growth among boys. Serum trans palmitoleic acid can be found mainly in dairy products and ruminant meat; thus, serum trans palmitoleic acid may be a proxy of dairy intake that is positively associated with linear growth in other populations (30), and higher intake of protein is also associated positively with linear growth (31). However, this association was not present in girls; therefore, we cannot rule out the possibility of chance. Nevertheless, this finding deserves future investigations in larger cohorts with longer follow-up.

Our study has several strengths. First, we used serum biomarkers of TFA intake that is a preferred measured of intake over questionnaires because of objectivity and absence of recall bias (14, 15). Second, the prospective design and use of repeated anthropometric measures allowed us to explore the temporal relation between serum TFAs and anthropometric measures. We also used modern analytic methods to model nonlinear growth trajectories and adjusted the estimates of association for key potential confounders. Third, our study sample is representative of the low- and middle-income pediatric populations in Bogota, Colombia, which enhances generalizability.

Our study has also some limitations. The follow-up period was relatively short (median: 2.5 y). A single measure of TFAs in serum reflects short-term intake and may not be enough to reflect long-term intake with consequent increase in measurement error. Nevertheless, because the main source of TFAs in this population is their cooking oil and people do not change their cooking oil often, we can speculate that serum can still be used as a proxy of long-term intake for this population. In any case, we cannot rule out that our null findings may be the result of using a short-term biomarker. Anthropometric measures are also prone to measurement error. We cannot completely rule out selection bias, given that the study sample was selected from children who had enough serum volume to conduct the FA analysis. However, the study sample did not differ in BMI z score compared with the whole cohort. Finally, we did not collect information on pubertal stages that may add extraneous variation to the association.

Although we did not find an association between serum TFAs and faster weight gain, we believe it is important to study the health consequences of TFA intake in children. TFAs are linked to chronic disease in adults (32), but their effects on school-age children are unknown. In the past few years, a Pan American Health Organization/WHO task force called for the elimination of TFAs from the food supply in the Americas (Trans-Fats Free Americas initiative) and recommended that unsaturated fats should be the preferred alternative (33). The mean TFA intake in Colombia is estimated to range between 1.25% and 1.49% of total energy intake. It is higher than in other South American countries, such as Bolivia, Argentina, and Uruguay (0.75–0.99%), but lower than in Brazil (1.5–1.99%) and Peru (2.0–2.49%) and much lower than in the United States (2.5–3.49%) and Mexico (3.5–4.49%) (34). However, our previous experience in Costa Rica has taught us that TFAs in the food supply can change dramatically in a few years (35). Therefore, we believe that further research and monitoring of the TFA levels in the food supply is still needed.

In conclusion we did not find an association between serum TFAs and weight gain in a cohort of school-age children. We found a positive association with serum trans palmitoleic acid and linear growth in boys. Longer follow-up and studies in diverse cohorts with wider ranges of TFA intake are warranted.