Page 80 - The prevalence of the Val66Met polymorphism in musicians: Possible evidence for compensatory neuroplasticity from a pilot study

Validation of genes in childhood obesity | 3

written consent, and the studies were approved by histograms, we ran Kruskal–Wallis test when there
the Ethics Committee of the Federal University of were three genotypes or Mann–Whitney test when
Health Sciences of Porto Alegre. Further details for two genotypes were present.
both cohorts can be found elsewhere (13,17). For longitudinal analyses, we calculated the percent-
age change in BMI between 1 and 3.5years of age as
Molecular analyses [{BMI 3 BMI 1 }/BMI 1 ]×100 for each individual and com-
pared their means among genotypes using ANOVA or
Ten single nucleotide polymorphisms (SNPs) were ge- ORIGINALRESEARCH
T-tests. This variable was also adjusted for the effects
notyped: MC4R rs17782313, TMEM18 rs6548238,
of gender, cohort and intervention group using multiple
BDNF rs6265 and rs10767664, KCTD15
linear regressions before the comparisons between ge-
rs11084753, NEGR1 rs2815752, SH2B1 rs7498665,
notypes, as described before for continuous variables.
SEC16B rs10913469, OLFM4 rs9568856 and
Control for multiple testing was performed by con-
HOXB5 rs9299. For the São Leopoldo cohort, DNA
trolling the false discovery rate (FDR) at a 0.10 level
was extracted from whole blood, and in the Porto
as described by Benjamini and Hochberg in 1995
Alegre cohort, DNA was extracted from epithelial
(18). P-values were corrected by the number of
buccal cells, both using methods of precipitation with
phenotypes analyzed for each gene variant, as there
a high salt concentration. The 10 gene variants were
was a correlation between the phenotypes measured.
genotyped through real-time polymerase chain reac-
Analyses for each gene were considered as indepen-
tion using TaqMan® SNP genotyping assays (Applied
dent hypotheses. Therefore, all corrected P-values in
Biosystems, Carlsbad, CA, USA).
Table 3 and Supporting Information Tables S2, S3
and S4 should be evaluated using 0.10 as critical
Statistical analysis
value. All statistical analyses were performed using
Categorical variables are presented as relative frequen- the Statistical Package for the Social Sciences (SPSS,
cies (%) and continuous variables as means and stan- for windows version 20.0).
dard deviations. Genotype and allele frequencies were
calculated by gene counting. The chi-square test was Results
used to verify if these frequencies were in agreement
with those expected under Hardy–Weinberg equilib- There were 745 children followed-up since birth until
rium and to compare allele frequencies between the 3.2 (Porto Alegre) and 3.9 years old (São Leopoldo)
two localities and genotype/allele frequencies within and were genotyped for the 10 gene polymorphisms.
ethnic groups. The average age of children at the second evaluation
Normality for each continuous variable was was 1year old, and at the last one was 3.5 years old.
assessed by visual inspection using histograms of the At 1 year of age, 41.1% of children were overweight
respective distributions. Mean variables with normal and 13.7% were obese. These figures diminished a
distribution were compared among genotypes for little at 3.5 years old, when 30.7% of children were
each SNP using analysis of variance (ANOVA). When overweight and 10.5% were obese. Descriptive data
significant results were found, the post hoc Tukey test are presented in Table 1.
was used to identify which genotypes had different Minor allele frequencies (MAFs) for the polymor-
means from the others. All continuous variables were phisms studied are shown in Table 2, together with
adjusted for the effects of gender, cohort and interven- the National Center for Biotechnology Information
tion group using multiple linear regressions without dbSNP Database for European population (HapMap
genotypes and then adding the respective residues CEU: Utah residents with ancestry from northern
to the mean of the variable before the comparisons and western Europe), Sub-Saharan African popula-
between genotypes by ANOVA. For those genotypes tion (HapMap YRI: Yoruba in Ibadan, Nigeria) and
with fewer than 20 individuals, the mean values for the Asian populations frequencies (HapMap CHB:
the rare genotype were combined with heterozygotes Han Chinese in Beijing, China; and HapMap JPT:
for analyses, and mean variables were compared with Japanese in Tokyo, Japan). There were only minor
T-tests for independent variables. differences between allele frequencies in the two
Sum of skin-folds were transformed into their natu- cohorts (Table 2). The frequencies for the majority of var-
ral logarithms prior to analysis in order to achieve a iants were closer to the CEU data (NEGR1 rs2815752,
normal distribution. When the continuous variables TMEM18 rs6548238, BDNF rs6265 and rs10767664,
were not normally distributed even after logarithmic SH2B1 rs7498665, KCTD15 rs11084753) or interme-
transformation [sugar-dense foods {SDF} and lipid- diate between CEU and YRI (OLFM4 rs9568856,
dense foods {LDF}], after visual inspection at HOXB5 rs9299). For SEC16B rs10913469 and MC4R

© 2016 World Obesity. Pediatric Obesity ••, ••–••