Alpha-Linolenic Acid (ALA) Is Inversely Related to Development of Adiposity in School-Age Children

ORIGINAL ARTICLE Alpha-linolenic acid (ALA) is inversely related to development of adiposity in school-age children

W Perng1, E Villamor2, M Mora-Plazas3, C Marin3 and A Baylin2

BACKGROUND/OBJECTIVES: Studies in adults indicate that dietary polyunsaturated fatty acid (PUFA) composition may play a role in development of adiposity. Because adipocyte quantity is established between late childhood and early adolescence, understanding the impact of PUFAs on weight gain during the school-age years is crucial to developing effective interventions. SUBJECTS/METHODS: We quantified N-3 and N-6 PUFAs in serum samples of 668 Colombian schoolchildren aged 5–12 years at the time of recruitment into a cohort study, using gas–liquid chromatography. Serum concentrations of N-3 (alpha-linolenic acid (ALA), eicosapentaenoic acid, docosahexaenoic acid) and N-6 PUFAs (linoleic acid, gamma-linolenic acid, dihomo-gamma-linolenic acid, arachidonic acid) were determined as percentage total fatty acids. Children’s anthropometry was measured annually for a median of 30 months. We used mixed-effects models with restricted cubic splines to construct population body mass index-for-age z-score (BAZ) growth curves for age- and sex-specific quartiles of each PUFA. RESULTS: N-3 ALA was inversely related to BAZ gain after adjustment for sex, baseline age and weight status, as well as household socioeconomic level. Estimated BAZ change between 6 and 14 years among children in the highest quartile of ALA compared with those in the lowest quartile was 0.45 (95% confidence interval: 0.07, 0.83) lower (P-trend = 0.006). CONCLUSIONS: N-3 ALA may be protective against weight gain in school-age children. Whether improvement in PUFA status reduces adiposity in pediatric populations deserves evaluation in randomized trials. European Journal of Clinical Nutrition (2015) 69, 167–172; doi:10.1038/ejcn.2014.210; published online 1 October 2014

INTRODUCTION actions. Specifically, PUFAs in the N-6 pathway may stimulate 3 Childhood obesity poses one of the most serious public health adipogenesis, whereas the N-3 series can reduce fat mass 4 challenges. Many countries, including those in Latin America,1 have through amelioration of inflammation and adipocyte experienced a marked rise in pediatric overweight and obesity rates hypotrophia.5 in the past three decades, and the relative increase in prevalence Although there is no clear link between N-6 PUFA and human was sharper in developing as compared to developed countries.2 obesity, studies in adults suggest that N-3 PUFA intake is inversely 6,7 Although the nutrition transition was identified as a major related to body mass index (BMI). Prebirth cohorts provide some 8 contributor to overweight and obesity in less affluent settings, evidence that maternal PUFA intake during pregnancy and 9 obesity rates continue to rise and weight-gain prevention remains lactation influences offspring body composition in early child- elusive. Disentangling the effects of specific dietary components is hood. However, findings from randomized controlled trials a critically urgent area of investigation. evaluating the effects of maternal peripartum N-3 PUFA supple- Dietary fatty acid composition may have a role in development mentation on offspring adiposity have been inconsistent,10 and of adiposity. Special attention has been given to the omega-3 current evidence in children is restricted to a handful of mixed – (N-3) and omega-6 (N-6) essential polyunsaturated fatty acids findings from small cross-sectional studies.11 13 Considering that (PUFAs). Dietary precursors of N-3 and N-6 PUFAs include 18:3 adipocyte quantity is established between late childhood and (N-3) alpha-linolenic acid (ALA) and 18:2 (N-6) linoleic acid (LA), early adolescence,14 understanding the potential impact of PUFAs respectively. As the primary 18-carbon member of the N-3 series, on weight gain during the school-age years is critical to ALA can be desaturated and elongated into eicosapentaenoic acid developing effective interventions. Furthermore, there is need to (EPA) and docosahexaenoic acid (DHA). Through a shared evaluate these associations in developing countries where enzymatic pathway, LA (the main 18-carbon N-6 PUFA) is essential N-3 PUFA deficiency is widely prevalent and recent converted to gamma-linolenic acid (GLA), dihomo-gamma- shifts toward increased consumption of N-6 PUFAs coincide with linolenic acid (DGLA) and arachidonic acid (AA). These long- the increase in obesity rates. chain PUFAs can be oxidized to produce eicosanoids, which are In this study, we prospectively examined the association of N-3 hormone-like signaling molecules that are involved in important and N-6 PUFA biomarkers at the time of recruitment into a cohort biological processes including inflammation and cell differentia- study with changes in adiposity in low- and middle-income tion. Because the N-3 and N-6 PUFA families compete for the same schoolchildren from Bogotá, Colombia, a country at the early enzymes in their biosynthesis, they have interactive and opposing stages of the nutrition transition.

1Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA, USA; 2Department of Epidemiology, University of Michigan School of Public Health, Ann Arbor, MI, USA and 3Fundación para Investigación en Nutrición y Salud, FINUSAD, Bogotá, Colombia. Correspondence: Dr W Perng, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, 133 Brookline Avenue, Boston, MA 02215, USA. E-mail: [email protected] Received 8 July 2014; revised 26 August 2014; accepted 29 August 2014; published online 1 October 2014 PUFA and weight gain in children W Perng et al 168 SUBJECTS AND METHODS measured height and weight in 49% of the mothers and from self-reported Study population data otherwise. We classified maternal weight status according to the standard adult BMI categories. Household socioeconomic status corre- This study included a subsample of the Bogotá School Children Cohort, an sponded to the government’s classification (1—lowest through 6— ongoing investigation of health and nutrition conducted among children highest; maximum of 4 in study population) assigned to each household from public schools in Bogotá, Colombia. Details on recruitment and study for tax and planning purposes. We assessed the significance of these design have been published.15 Briefly, 3202 school children aged 5–12 associations with the Wald test. For ordinal characteristics, we obtained a years were recruited from Bogotá’s public school system in February 2006. test for linear trend. The study population is representative of families from low- and middle- Next, we examined associations of the PUFAs quantified at baseline with income socioeconomic backgrounds in the city, as we used a random change in BAZ during follow-up. We estimated average BAZ growth curves sampling strategy and the public school system enrolls the majority of for sex- and age-specific quartiles of each PUFA marker using mixed-effects children from these groups. The parents or primary caregivers of all models for repeated measurements with restricted cubic splines,17 as children gave written informed consent before enrollment into the study. previously described.18 Cubic splines represent nonlinear terms for the The Ethics Committee of the National University of Colombia Medical distribution of age-at-assessment that allow smoothing of nonlinear BAZ School approved the study protocol. The Institutional Review Board at the changes over time. The cublic spline function consists of piecewise cubic University of Michigan approved the use of data and samples from polynomials that are smoothly joined at joint points, or ‘knots.’ It is the study. ‘ ’ ’ restricted because the polynomials at the tails are constrained to be At enrollment, we sent the children s families self-administered linear. We fixed the knots at 5.7, 9.2, 10.8 and 13.8 years to reflect questionnaires that inquired on maternal sociodemographic (age, marital curvilinear portions of the WHO BMI-for-age child growth reference.16 In status, education, socioeconomic status) and anthropometric character- ’ the spline models, BAZ was the outcome, and the predictors included sex- istics (height, weight), as well as the child s physical activity and sedentary and age-specific quartiles of each PUFA marker, linear and spline terms for habits (82% response). Trained research assistants visited the schools to child age-at-assessment in decimal years and interaction terms between obtain anthropometric measurements and a fasting blood sample from the quartiles of the PUFA marker and the child age terms. Random effects for children. Height was measured without shoes to the nearest 1 mm using a the intercept and the linear term for age were included to account for wall-mounted portable stadiometer (Seca, Hanover, MD, USA), and weight within-child correlations between repeated BAZ measurements in the was measured in light clothing to the nearest 0.1 kg on solar-powered estimation of the variance. These methods do not require an equal number electronic scales (Tanita, Arlington Heights, IL, USA). We obtained follow-up of measurements, or that the measurements were obtained at the same anthropometric measurements in June and November 2006, and once ’ time for participants; thus, all available measurements were included in the yearly thereafter by visiting the schools or the children s homes if they models. Using the growth curves constructed for children in each quartile were absent from school on the day of assessment. of the PUFA markers, we estimated attained BAZ at 6 and 14 years. Our primary outcome of interest was the change in BAZ between these two Laboratory methods age points. We also estimated the difference in BAZ change between We obtained an 8-h fasting blood specimen by venipuncture in 88% of children in the lowest and highest quartiles of each PUFA at 6 and 14 years. fi children at baseline and placed one aliquot in a tube without antic- In multivariable analysis, we estimated differences and 95% con dence oagulant for the separation of serum. On the same day, the samples were intervals (CI) in BAZ change from 6 to 14 years after accounting for known fi ’ transported on dry ice and protected from sunlight to the National predictors of childhood weight gain. The nal model included the child s Institute of Health in Bogotá. We carried out a complete blood count and sex, baseline age and weight status and household socioeconomic status. fi fi cryopreserved serum at − 80 °C until transportation to the United States for We evaluated for effect modi cation by sex and also conducted strati ed analyses. We extracted lipids from serum samples and prepared fatty acid analyses. There was no indication that relations of the PUFA markers with BAZ were different for boys and girls; thus, the results presented are for all methyl esters of total lipids with BF3-methanol. Methyl esters were extracted from a thin-layer chromatography plate, and the solvents were children. dried and resuspended in hexane. Approximately 2 ml of sample was All analyses were carried out with the use of SAS 9.3 (SAS Institute Inc, injected via an autosampler and analyzed on a gas–liquid chromatography Cary, NC, USA). machine using a 100 m SP-2560 column with optimum conditions for separation (Model 6890 N, Agilent, Santa Clara, CA, USA). Eluted peaks RESULTS were analyzed with the Chemstation software (Agilent). The concentration of each fatty acid was determined using a calibration curve with C17:0 Mean ± s.d. age of children was 8.8 ± 1.7 years at the time of methyl ester as the standard. Serum concentrations of N-3 ALA, EPA and recruitment; 54.6% were girls. Mean BAZ at baseline was DHA and N-6 LA, GLA, DGLA and AA were each expressed as a percentage 0.09 ± 0.98, and 18% (n = 122) of the children were overweight fi of total fatty acids. Inter-assay coef cients of variation ranged from 1.1% to or obese (BAZ41). During the median of 30.1 months of follow-up 2.3% for all PUFAs. (interquartile range: 29.8–30.8 months), each child contributed a median of four BMI measurements, for a total of 2891 Statistical analysis measurements. We selected a random sample of 687 children for fatty acid quantification At baseline, boys had 0.04 (95% CI: 0.02, 0.06)% higher GLA and among those with a stored serum specimen. Of them, 668 children who 0.09 (95% CI: 0.04, 0.15)% higher DGLA than girls. Older children had valid anthropometric measurements at baseline and at least one had higher ALA, DHA and DGLA concentrations but lower AA:EPA fi additional follow-up measurement comprised the nal study population. +DHA than younger children (Table 1). Baseline HAZ was These children did not differ from the rest of the Bogotá School Children positively related to AA (P-trend = 0.09) and AA:EPA+DHA Cohort in terms of age or baseline anthropometry, or maternal demographic characteristics or BMI status. Nevertheless, compared with (P-trend = 0.05). Higher BAZ at baseline corresponded with lower children excluded from analyses, the study sample included more girls LA (P-trend = 0.02) and higher GLA (P-trend = 0.008) and DGLA (54.6% vs 50.1%) and a higher proportion of children from families in the (P-trendo0.0001). There was a weak inverse association between upper two socioeconomic strata (66.2% vs 59.5%). maternal BMI and LA. Compared with poorer children, those of Our exposures of interest were serum concentrations of N-3 (ALA, EPA higher socioeconomic status had lower ALA, EPA, GLA and DGLA and DHA) and N-6 (LA, GLA, DGLA and AA) PUFAs. We also examined the and higher DHA, LA and AA. ratio of AA to EPA+DHA (AA:EPA+DHA), an indicator of the relative In the longitudinal analysis, we estimated average BAZ growth proportion of N-6 to N-3 intake. First, we compared the distribution of each curves for quartiles of each N-3 (Table 2) and N-6 PUFA (Table 3). PUFA by child and maternal characteristics to identify variables that may We found an inverse relation between ALA and BAZ change confound the association between PUFAs and development of adiposity. Children’s body mass index (BMI)-for-age (BAZ) and height-for-age z-scores during follow-up (P-trend = 0.008), which persisted after account- (HAZ) were calculated using the World Health Organization (WHO) sex- ing for sex, baseline age and weight status and household specific growth reference for children 5–19 years.16 Child weight status was socioeconomic status (P-trend = 0.006). Estimated adjusted BAZ categorized as thin (BAZo − 2), adequate (BAZ ⩾ − 2 and ⩽ 1), overweight change from age 6 to 14 years was 0.45 (95% CI: 0.07, 0.83) z lower (BAZ41 and ⩽ 2) and obese (BAZ42). Maternal BMI was calculated from for children in the fourth quartile of N-3 ALA than for those in the

Table 1. N-3 and N-6 polyunsaturated fatty acids by child sociodemographic and maternal characteristics in 668 school-age children from Bogotá, Colombiaa

Nb N-3 N-6 N-6:N-3

ALA 18:3 (N-3) EPA 20:5 (N-3) DHA 22:6 (N-3) LA 18:2 (N-6) GLA 18:3 (N-6) DGLA 20:3 (N-6) AA 20:4 (N-6) AA:EPA+DHA

Overall 0.49 (0.15) 0.22 (0.15) 2.30 (0.90) 30.28 (3.10) 0.30 (0.15) 1.61 (0.38) 5.96 (1.16) 2.62 (0.93)

Child sex F 365 0.49 (0.15) 0.21 (0.13) 2.28 (0.86) 30.30 (3.07) 0.28 (0.14) 1.57 (0.35) 5.96 (1.11) 2.61 (0.87) M 303 0.48 (0.15) 0.22 (0.16) 2.32 (0.93) 30.25 (3.13) 0.32 (0.17) 1.66 (0.39) 5.96 (1.20) 2.62 (1.00) Pc 0.76 0.68 0.62 0.86 0.002 0.002 0.98 0.94

Child’s age (years) 5–6 115 0.47 (0.15) 0.22 (0.13) 2.18 (0.77) 29.61 (3.21) 0.30 (0.16) 1.56 (0.30) 6.03 (1.08) 2.72 (0.84) 7–8 215 0.48 (0.14) 0.22 (0.13) 2.20 (0.87) 30.83 (2.89) 0.30 (0.16) 1.55 (0.36) 5.95 (1.13) 2.71 (1.02) 9–10 286 0.49 (0.15) 0.22 (0.16) 2.40 (0.90) 30.28 (3.00) 0.30 (0.15) 1.66 (0.40) 6.00 (1.19) 2.54 (0.89) 11–12 52 0.53 (0.19) 0.20 (0.13) 2.42 (1.17) 29.45 (3.75) 0.30 (0.16) 1.67 (0.38) 5.64 (1.24) 2.44 (0.91) P-trendd 0.02 0.33 0.007 0.91 0.68 0.0003 0.23 0.008

Height-for-age z-scoree o − 2.0 62 0.50 (0.16) 0.23 (0.13) 2.24 (0.81) 30.46 (3.35) 0.33 (0.16) 1.64 (0.44) 5.70 (1.20) 2.50 (0.82) − 2.0 to o − 1.0 212 0.48 (0.14) 0.21 (0.13) 2.34 (0.81) 30.61 (2.80) 0.28 (0.15) 1.56 (0.38) 5.94 (1.21) 2.53 (0.92) − 1.0 to o1.0 370 0.48 (0.15) 0.22 (0.16) 2.29 (0.97) 30.06 (3.18) 0.31 (0.15) 1.63 (0.36) 6.02 (1.13) 2.68 (0.94) ⩾ 1.0 24 0.56 (0.17) 0.19 (0.11) 2.27 (0.76) 30.22 (3.54) 0.32 (0.19) 1.62 (0.26) 5.92 (0.92) 2.66 (1.00) P-trendd 0.70 0.83 0.93 0.11 0.62 0.40 0.09 0.05

BMI-for-age z-scoree o − 2.0 12 0.52 (0.14) 0.25 (0.17) 2.21 (0.57) 30.32 (3.35) 0.27 (0.14) 1.74 (0.48) 5.91 (1.48) 2.42 (0.34) ⩾ − 2.0 to ⩽ 1.0 534 0.49 (0.15) 0.21 (0.13) 2.27 (0.87) 30.41 (3.04) 0.29 (0.15) 1.57 (0.36) 5.96 (1.14) 2.63 (0.93) 41.0 to ⩽ 2.0 106 0.49 (0.14) 0.23 (0.21) 2.43 (1.05) 29.81 (3.22) 0.34 (0.17) 1.74 (0.36) 5.98 (1.22) 2.56 (0.95) 42.0 16 0.49 (0.14) 0.23 (0.14) 2.29 (0.78) 28.81 (3.47) 0.33 (0.14) 1.89 (0.46) 5.69 (1.09) 2.52 (1.09) P-trendd 0.92 0.52 0.18 0.02 0.008 o0.0001 0.72 0.61

Maternal BMI o18.5 kg/m2 16 0.43 (0.12) 0.15 (0.11) 2.38 (0.72) 29.00 (2.19) 0.31 (0.14) 1.65 (0.26) 5.58 (0.89) 2.39 (0.82) 18.5–24.9 kg/m2 341 0.49 (0.15) 0.22 (0.16) 2.30 (0.89) 30.57 (3.12) 0.30 (0.15) 1.58 (0.35) 6.02 (1.18) 2.64 (1.00) 25.0–29.9 kg/m2 166 0.49 (0.15) 0.20 (0.13) 2.31 (0.96) 29.67 (3.27) 0.31 (0.16) 1.61 (0.36) 5.92 (1.16) 2.60 (0.83) ⩾ 30 kg/m2 43 0.49 (0.13) 0.24 (0.14) 2.31 (0.90) 30.19 (2.62) 0.28 (0.17) 1.68 (0.49) 5.88 (1.29) 2.55 (0.93) P-trendd 0.74 0.69 0.98 0.09 0.80 0.26 0.63 0.73

Household SESf 1 (lowest) 49 0.49 (0.13) 0.28 (0.12) 1.88 (0.68) 30.37 (2.43) 0.35 (0.14) 1.75 (0.35) 5.92 (1.13) 3.03 (1.39) 2 177 0.52 (0.16) 0.22 (0.18) 2.26 (0.86) 29.82 (3.15) 0.31 (0.15) 1.64 (0.36) 5.80 (1.18) 2.56 (0.86) 3 353 0.47 (0.14) 0.21 (0.13) 2.37 (0.92) 30.32 (3.14) 0.29 (0.16) 1.58 (0.38) 6.01 (1.14) 2.57 (0.89) 4 (highest) 89 0.46 (0.13) 0.20 (0.13) 2.30 (0.92) 30.97 (3.06) 0.31 (0.16) 1.60 (0.40) 6.11 (1.18) 2.69 (0.86) P-trendd 0.0004 0.006 0.006 0.02 0.07 0.006 0.05 0.27 Abbreviations: AA, arachidonic acid; ALA, alpha-linolenic acid; BMI, body mass index; DGLA, dihomo-gamma-linolenic acid; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; F, female; GLA, gamma-linolenic acid; LA, linoleic acid; M, male; SES, socioeconomic status. aValues are means (s.d.). bTotals may be o668 because of missing values. cWald test from bivariate linear regression models. dFrom bivariate regression models in which a variable representing the ordinal predictor was introduced as continuous. eAccording to the World Health Organization (WHO) 2007 Child Growth Reference.16 fAccording to the government's classiﬁcation for tax and planning purposes.

first quartile. Among the N-6 PUFAs, higher GLA levels at baseline To date, only a few small cross-sectional investigations were also related to lower BAZ gain, although the trend only examined the relation of specific PUFAs with weight status in approached significance (P-trend = 0.05). In comparison with children, and the findings regarding ALA have been inconsistent. children in the lowest GLA quartile, estimated adjusted change In a study of 60 overweight and 60 normal-weight French in BAZ for those in the highest quartile was 0.42 (95% CI: 0.00, adolescents, overweight youth had lower PUFA-to-saturated fatty 0.84) z lower (Table 3). BAZ change was not associated with the acids ratio than normal-weight participants, although ALA other PUFA markers or with AA:EPA+DHA. concentrations were similar between the two groups.12 In another study of 10 obese and 15 lean adolescents, lower PUFA concentrations were observed in obese participants, mostly driven DISCUSSION by DHA, with no difference in ALA.13 In recent investigations of In this longitudinal investigation of school-age children in Bogotá, Australian schoolchildren11 and pre-pubertal Spanish children,19 Colombia—a setting where obesity is becoming a serious public erythrocyte and plasma ALA levels were positively related to health problem—we explored the relationship of PUFA markers obesity status. The challenge of reconciling these discrepancies is with BMI trajectories. We found that lower serum concentrations further compounded by the cross-sectional nature of the designs, of N-3 ALA were related to greater gains in adiposity during the as the associations could be explained by an effect of adiposity on follow-up period. fatty acid profiles.20

Table 2. Estimated change in BMI-for-age z among 668 school-age children from Bogotá, Colombia according to quartiles of N-3 polyunsaturated fatty acids and AA:EPA+DHAa

N BMI-for-age z-scoreb Changeb Difference in changec Pd

6 years 14 years 14–6 years β (95% CI)

ALA 18:3 (N-3) 0.006 Q1 168 − 0.14 ± 0.12 0.22 ± 0.10 0.36 ± 0.15 Reference Q2 166 − 0.23 ± 0.14 0.32 ± 0.14 0.55 ± 0.20 0.17 (−0.29, 0.62) Q3 169 − 0.01 ± 0.12 0.07 ± 0.11 0.08 ± 0.15 − 0.28 (−0.69, 0.12) Q4 165 0.06 ± 0.11 − 0.01 ± 0.12 − 0.08 ± 0.14 − 0.45 (−0.83, − 0.07) Difference Q4 − Q1 0.20 (−0.12, 0.53) − 0.24 (−0.54, 0.07)

EPA 20:5 (N-3) 0.24 Q1 170 − 0.04 ± 0.13 0.15 ± 0.13 0.19 ± 0.18 Reference Q2 164 − 0.24 ± 0.13 0.25 ± 0.13 0.50 ± 0.17 0.32 (−0.15, 0.79) Q3 166 − 0.18 ± 0.12 0.09 ± 0.11 0.27 ± 0.15 0.08 (−0.36, 0.52) Q4 168 0.14 ± 0.11 0.10 ± 0.11 − 0.04 ± 0.14 − 0.19 (−0.62, 0.23) Difference Q4 − Q1 0.17 (−0.17, 0.51) − 0.06 (−0.39, 0.28)

DHA 22:6 (N-3) 0.09 Q1 166 0.01 ± 0.12 0.05 ± 0.11 0.04 ± 0.15 Reference Q2 168 − 0.03 ± 0.11 0.14 ± 0.11 0.17 ± 0.14 0.16 (−0.22, 0.55) Q3 169 − 0.27 ± 0.14 0.08 ± 0.12 0.35 ± 0.19 0.34 (−0.11, 0.79) Q4 165 − 0.04 ± 0.12 0.31 ± 0.14 0.35 ± 0.17 0.35 (−0.09, 0.78) Difference Q4 − Q1 − 0.05 (−0.38, 0.29) 0.26 (−0.10, 0.62)

AA:EPA+DHA Q1 167 0.09 ± 0.12 0.36 ± 0.13 0.27 ± 0.16 Reference 0.50 Q2 165 − 0.13 ± 0.13 0.15 ± 0.13 0.29 ± 0.17 0.03 (−0.42, 0.47) Q3 170 − 0.08 ± 0.13 0.02 ± 0.11 0.10 ± 0.16 − 0.18 (−0.62, 0.25) Q4 166 − 0.20 ± 0.12 0.04 ± 0.11 0.23 ± 0.15 − 0.09 (−0.51, 0.34) Difference Q4 − Q1 − 0.29 (−0.62, 0.05) − 0.32 (−0.67, 0.02) Abbreviations: AA, arachidonic acid; ALA, alpha-linolenic acid; BMI, body mass index; CI, conﬁdence interval; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid. aQuartiles for indicators are sex- and age-speciﬁc according to their distributions in the study population. bValues are mean ± s.e. Estimates are from growth curves built using mixed-effects models with restricted cubic splines that accounted for within-child repeated BMI measurements. cDifferences in change are adjusted for sex, baseline age, baseline weight status and household socioeconomic stratum. dTest for linear trend from a linear regression model where an ordinal variable that represented quartiles of each fatty acid indicator was entered as a continuous predictor.

Our longitudinal study is less susceptible to reverse-causation suggesting that GLA is involved in the suppression of fat bias because we examined prospective changes in BAZ, the accretion.28 This finding warrants additional investigation in other optimal measure of adiposity development in children that pediatric populations. reflects fat growth rather than accrual of lean mass.21 Our results Our findings point toward potential nutritional interventions indicate that higher ALA is related to lower gains in adiposity. that could be implemented in populations with fatty acid status There are a several mechanisms that could explain this associa- comparable to that of our study population. A previous study that tion. First, ALA reduces fat deposition by promoting the investigated the composition and usage of cooking fats in Bogotá, expression of genes involved in hepatic fat oxidation22 and Colombia found that mixed vegetable oils (composed of soybean, thermogenesis.23 Second, ALA can displace LA from the shared corn and palm oil) and sunflower oil were most commonly used in delta6-desaturase enzyme, and prevent its conversion to AA, this population.29 Biochemical analyses revealed that the mixed which is a trigger for adipogenesis through prostacyclin synthesis vegetable oils used by those of lower socioeconomic status had (an effector of adipocyte differentiation), and activation of cAMP the highest ALA content. On the other hand, sunflower oil, which pathways that favor pre-adipocyte maturation.3 Third, greater was preferred by the more affluent, contained negligible amounts intake of fatty acids prone to oxidation is less likely to result in of PUFAs and higher trans fat content.29 Because Colombia has weight gain than intake of fatty acids prone to storage. ALA is the one of the lowest intakes of marine fish worldwide,30 home- most highly oxidized fatty acid in humans,24 with 420% of cooking oils represent the main source of dietary fatty acids in this ingested ALA catabolized for energy.25 Furthermore, greater ALA setting. Thus, a feasible way to increase healthy PUFA intake could intake appears to increase oxidation rates rather than accumulat- be to raise the proportion of soybean oil, which is rich in ALA, in ing in tissues.26 the widely used mixed vegetable oils or to promote replacement We also noted a marginally significant inverse relation between of cooking fats with soybean oil. Nutrition education campaigns N-6 GLA and BMIZ change, which was unexpected as GLA is an may also be helpful for consumers to understand the importance intermediate in the biosynthesis of AA from LA—a pathway that of choosing healthy fats, and could be an effective way to foster ultimately leads to the production of inflammatory and adipo- selection of healthier cooking oils. genic AA-derived eicosanoids.3 However, a study in Zucker rats Our study has some limitations. First, we relied on one found that GLA administration reduced weight only among obese measurement of the PUFA markers at the time of enrollment as animals, implicating impaired LA to GLA desaturation in the the primary exposure. However, serum fatty acid concentrations etiology of obesity.27 Similarly, in a randomized controlled trial of can be adequate biomarkers of long-term intake.31 Second, 24 formerly obese adults, supplementation with 890 mg/day of because PUFA concentrations are expressed as a relative GLA markedly reduced weight regain after 1 year (2.2 vs 8.8 kg), percentage of total fatty acids, higher levels of one PUFA may

Table 3. Estimated change in BMI-for-age z among 668 school-age children from Bogotá, Colombia according to quartiles of N-6 polyunsaturated fatty acidsa

N BMI-for-age z-scoreb Changeb Difference in changec Pd

6 years 14 years 14–6 years β (95% CI)

LA 18:2 (N-6) 0.46 Q1 168 0.01 ± 0.14 0.34 ± 0.12 0.33 ± 0.17 Reference Q2 166 − 0.08 ± 0.12 0.14 ± 0.13 0.22 ± 0.17 − 0.09 (−0.53, 0.36) Q3 169 − 0.10 ± 0.11 0.06 ± 0.13 0.16 ± 0.16 − 0.17 (−0.61, 0.27) Q4 165 − 0.15 ± 0.12 0.03 ± 0.12 0.18 ± 0.16 − 0.14 (−0.58, 0.29) Difference Q4 − Q1 − 0.17 (−0.53, 0.20) − 0.32 (−0.64, 0.01)

GLA 18:3 (n-6) 0.05 Q1 167 − 0.46 ± 0.13 0.11 ± 0.12 0.56 ± 0.16 Reference Q2 167 − 0.09 ± 0.12 0.05 ± 0.12 0.14 ± 0.16 − 0.38 (−0.81, 0.05) Q3 165 0.09 ± 0.12 0.17 ± 0.12 0.08 ± 0.16 − 0.48 (−0.91, − 0.05) Q4 169 0.13 ± 0.12 0.26 ± 0.12 0.13 ± 0.16 − 0.42 (−0.84, 0.00) Difference Q4 − Q1 0.58 (0.24, 0.92) 0.15 (−0.18, 0.49)

DGLA 20:3 (N-6) 0.38 Q1 168 − 0.39 ± 0.12 − 0.03 ± 0.11 0.36 ± 0.15 Reference Q2 165 − 0.13 ± 0.12 0.16 ± 0.13 0.28 ± 0.17 − 0.10 (−0.52, 0.32) Q3 164 0.02 ± 0.12 0.04 ± 0.13 0.02 ± 0.17 − 0.34 (−0.77, 0.09) Q4 171 0.17 ± 0.13 0.41 ± 0.11 0.24 ± 0.16 − 0.12 (−0.54, 0.29) Difference Q4 − Q1 0.56 (0.22, 0.91) 0.44 (0.14, 0.75)

AA 20:4 (N-6) 0.27 Q1 166 0.05 ± 0.12 0.16 ± 0.12 0.11 ± 0.15 Reference Q2 167 0.01 ± 0.13 0.20 ± 0.12 0.19 ± 0.16 0.09 (−0.34, 0.52) Q3 167 − 0.13 ± 0.13 0.07 ± 0.10 0.21 ± 0.16 0.10 (−0.31, 0.52) Q4 168 − 0.23 ± 0.12 0.15 ± 0.15 0.38 ± 0.18 0.25 (−0.20, 0.70) Difference Q4 − Q1 − 0.28 (−0.60, 0.05) −0.01 (−0.38, 0.36) Abbreviations: AA, arachidonic acid; BMI, body mass index; CI, conﬁdece iterval; DGLA, dihomo-gamma-linolenic acid; GLA, gamma-linolenic acid; LA, linoleic acid. aQuartiles for indicators are sex- and age-speciﬁc according to their distributions in the study population. bValues are mean ± s.e. Estimates are from growth curves built using mixed-effects models with restricted cubic splines that accounted for within-child repeated BMI measurements. cDifferences in change are adjusted for sex, baseline age, baseline weight status and household socioeconomic stratum. dTest for linear trend from a linear regression model where an ordinal variable that represented quartiles of each fatty acid indicator was entered as a continuous predictor.

correspond to a lower relative amount for others; caution is 3 Massiera F, Saint-Marc P, Seydoux J, Murata T, Kobayashi T, Narumiya S et al. appropriate when interpreting markers of fatty acid intake. Third, Arachidonic acid and prostacyclin signaling promote adipose tissue development: generalizability of these findings to populations with different a human health concern? J Lipid Res 2003; 44:271–279. proportions of PUFA intake maybe limited. 4 Cintra DE, Ropelle ER, Moraes JC, Pauli JR, Morari J, Souza CT et al. Unsaturated fatty acids revert diet-induced hypothalamic inflammation in obesity. PLoS One In conclusion, higher N-3 ALA and, possibly, higher N-6 GLA are 7 inversely associated with weight gain in school-age children. 2012; : e30571. 5 Ruzickova J, Rossmeisl M, Prazak T, Flachs P, Sponarova J, Vecka M et al. Omega-3 Despite the strengths of our longitudinal investigation, it is not PUFA of marine origin limit diet-induced obesity in mice by reducing cellularity of possible to conclude that fatty acid status is causally related to adipose tissue. Lipids 2004; 39: 1177–1185. development of adiposity in children. This question deserves 6 Buckley JD, Howe PR. Long-chain omega-3 polyunsaturated fatty acids may be evaluation in intervention trials—perhaps, for example, through a beneficial for reducing obesity-a review. Nutrients 2010; 2:1212–1230. randomized intervention of cooking oils in this particular 7 Martinez-Victoria E, Yago MD. Omega 3 polyunsaturated fatty acids and population. body weight. Br J Nutr 2012; 107(Suppl 2): S107–S116. 8 Donahue SM, Rifas-Shiman SL, Gold DR, Jouni ZE, Gillman MW, Oken E. Prenatal fatty acid status and child adiposity at age 3 y: results from a US CONFLICT OF INTEREST pregnancy cohort. Am J Clin Nutr 2011; 93: 780–788. 9 Pedersen L, Lauritzen L, Brasholt M, Buhl T, Bisgaard H. Polyunsaturated fatty acid The authors declare no conflict of interest. content of mother's milk is associated with childhood body composition. Pediatr Res 2012; 72:631–636. ACKNOWLEDGEMENTS 10 Muhlhausler BS, Gibson RA, Makrides M. Effect of long-chain polyunsaturated fatty acid supplementation during pregnancy or lactation on infant and child This study was supported by the University of Michigan Nutrition Obesity Research body composition: a systematic review. Am J Clin Nutr 2010; 92: 857–863. Center (NORC) pilot grant (P30 DK089503) and the ASISA Research Fund at the 11 Burrows T, Collins CE, Garg ML. Omega-3 index, obesity and insulin resistance in University of Michigan. children. Int J Pediatr Obes 2011; 6: e532–e539. 12 Klein-Platat C, Drai J, Oujaa M, Schlienger JL, Simon C. Plasma fatty acid composition is associated with the metabolic syndrome and low-grade inflammation REFERENCES in overweight adolescents. Am J Clin Nutr 2005; 82:1178–1184. 1 Rivera JA, de Cossio TG, Pedraza LS, Aburto TC, Sanchez TG, Martorell R. 13 Karlsson M, Marild S, Brandberg J, Lonn L, Friberg P, Strandvik B. Serum phospho- Childhood and adolescent overweight and obesity in Latin America: lipid fatty acids, adipose tissue, and metabolic markers in obese adolescents. a systematic review. Lancet Diabetes Endocrinol 2014; 2: 321–332. Obesity 2006; 14: 1931–1939. 2 de Onis M, Blossner M, Borghi E. Global prevalence and trends of overweight and 14 Spalding KL, Arner E, Westermark PO, Bernard S, Buchholz BA, Bergmann O et al. obesity among preschool children. Am J Clin Nutr 2010; 92: 1257–1264. Dynamics of fat cell turnover in humans. Nature 2008; 453: 783–787.

© 2015 Macmillan Publishers Limited European Journal of Clinical Nutrition (2015) 167 – 172 PUFA and weight gain in children W Perng et al 172 15 Arsenault JE, Mora-Plazas M, Forero Y, Lopez-Arana S, Marin C, Baylin A et al. 24 DeLany JP, Windhauser MM, Champagne CM, Bray GA. Differential oxidation Provision of a school snack is associated with vitamin B-12 status, linear growth, of individual dietary fatty acids in humans. Am J Clin Nutr 2000; 72: 905–911. and morbidity in children from Bogota, Colombia. J Nutr 2009; 139: 1744–1750. 25 Burdge GC, Finnegan YE, Minihane AM, Williams CM, Wootton SA. Effect of altered 16 de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. Development dietary n-3 fatty acid intake upon plasma lipid fatty acid composition, conversion of a WHO growth reference for school-aged children and adolescents. Bull World of [13C]alpha-linolenic acid to longer-chain fatty acids and partitioning towards Health Organ 2007; 85:660–667. beta-oxidation in older men. Br J Nutr 2003; 90:311–321. 17 Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med 26 Vermunt SH, Mensink RP, Simonis MM, Hornstra G. Effects of dietary 1989; 8: 551–561. alpha-linolenic acid on the conversion and oxidation of 13C-alpha-linolenic acid. 18 Perng W, Mora-Plazas M, Marin C, Villamor E. Iron status and linear growth: Lipids 2000; 35: 137–142. a prospective study in school-age children. Eur J Clin Nutr 2013; 67:646–651. 27 Phinney SD, Tang AB, Thurmond DC, Nakamura MT, Stern JS. Abnormal 19 Gil-Campos M, del Carmen Ramirez-Tortosa M, Larque E, Linde J, Aguilera CM, polyunsaturated lipid metabolism in the obese Zucker rat, with partial metabolic Canete R et al. Metabolic syndrome affects fatty acid composition of plasma lipids correction by gamma-linolenic acid administration. Metabolism 1993; 42: in obese prepubertal children. Lipids 2008; 43:723–732. 1127–1140. 20 Schwab U, Seppanen-Laakso T, Yetukuri L, Agren J, Kolehmainen M, Laaksonen DE 28 Schirmer MA, Phinney SD. Gamma-linolenate reduces weight regain in formerly et al. Triacylglycerol fatty acid composition in diet-induced weight loss in subjects obese humans. J Nutr 2007; 137: 1430–1435. with abnormal glucose metabolism--the GENOBIN study. PLoS One 2008; 3:e2630. 29 Baylin A, Mora-Plazas M, Cobos-de Rangel O, Lopez-Arana S, Campos H, Villamor E. 21 Boeke CE, Oken E, Kleinman KP, Rifas-Shiman SL, Taveras EM, Gillman MW. Corre- Predictors of usage and fatty acid composition of cooking fats in Bogota, lations among adiposity measures in school-aged children. BMC Pediatr 2013; 13:99. Colombia. Public Health Nutr 2009; 12:531–537. 22 Ide T. Effect of dietary alpha-linolenic acid on the activity and gene expression of 30 Hibbeln JR, Nieminen LR, Blasbalg TL, Riggs JA, Lands WE. Healthy intakes of n-3 hepatic fatty acid oxidation enzymes. Biofactors 2000; 13:9–14. and n-6 fatty acids: estimations considering worldwide diversity. Am J Clin Nutr 23 Murase T, Nagasawa A, Suzuki J, Wakisaka T, Hase T, Tokimitsu I. Dietary alpha- 2006; 83: 1483s–1493s. linolenic acid-rich diacylglycerols reduce body weight gain accompanying the 31 Baylin A, Kim MK, Donovan-Palmer A, Siles X, Dougherty L, Tocco P et al. Fasting stimulation of intestinal beta-oxidation and related gene expressions in C57BL/ whole blood as a biomarker of essential fatty acid intake in epidemiologic studies: KsJ-db/db mice. J Nutr 2002; 132: 3018–3022. comparison with adipose tissue and plasma. Am J Epidemiol 2005; 162: 373–381.