French Caucasian subjects born between 1971 and 1985 were identified from a population-based registry encompassing more than 20,000 births in the metropolitan area of the city of Haguenau, France. Only singletons were included. Gestational age was determined from the date of the mother's last menstrual period and by physical examination during pregnancy, confirmed by ultrasound measurements when available (> 80%). Case-control association tests were performed on two unrelated groups, selected on birth data derived from the local reference curves drawn for gender and gestational age: SGA (birth weight < 10th percentile; n = 627 subjects; 294 men and 333 women) and AGA (birth weight between 25th and 75th percentile; n = 764 subjects; 369 men and 395 women).

Familial association tests were performed on 3,012 individuals from 628 pedigrees, among which 744 children with SGA, and 1,128 children with AGA. All individuals from the case-control study were included in the familial study.

Birth parameters, consisting of birth weight, length, and ponderal index, were collected from the registry. At a mean age of 22 yr, all subjects underwent a medical examination to assess anthropometric parameters (weight, height, waist to hip ratio). We assessed the pancreatic beta-cell function by HOMA beta-cell (HOMA-B) index, as the product of the fasting plasma insulin level (μU/ml) and 20, divided by the fasting plasma glucose level (mmol/l) minus 3.5. To estimate insulin resistance, HOMA-IR was calculated as the product of the fasting plasma insulin level (μU/ml) and the fasting plasma glucose level (mmol/l), divided by 22.5. Fasting insulin, HOMA-B and HOMA-IR were then converted into natural logarithm (Ln) value to normalize the distribution. Total cholesterol, high-density lipoprotein (HDL)-cholesterol, and triglyceride concentrations were measured fasting.

The study protocol was reviewed and approved by the faculty ethics committee, and all subjects and parents gave signed written consent.

All individuals were French Caucasians and were recruited using a multimedia campaign run by the "Centre National de la Recherche Scientifique" (CNRS), UMR8090. All subjects were obese children (n = 845, sex ratio (m/w) = 407/438, age = 11 ± 3 years, BMI z-score = 6.1 ± 2.9, birth weight = 3,448 ± 549 g). The study protocol was approved by the local ethic committee and an informed consent was obtained from each subject before participating in the study.

High-throughput genotyping of the s7903146 variant was performed using the TaqMan^® SNP Genotyping Assays (Applied Biosystems, Foster City, Calif. USA). The PCR primers and TaqMan probes were designed by Primer Express and optimized according to the manufacturer's protocol. There was a 98% genotyping success rate and the genotyping error rate was assessed by sequencing 384 control and 384 T2D individuals and by re-genotyping a random 10% sample. No difference was found with the first genotyping results, thus the genotyping error rate was 0%.

Tests for deviation from Hardy-Weinberg equilibrium (HWE) and for association were performed with the De Finetti program 17. Considering that the model of inheritance of the rs7903146 T allele has been found to be additive in the French population 14, we only analyzed allelic odds ratios. Given a 30% allelic frequency for the rs7903146 T allele and a frequency of 10% of SGA, the power of the case-control and family studies was then > 75% to observe an odds ratio = 1.2. Family-based association tests were performed by TDT methods implemented in the FBAT software. For the mother-child pairs, the effect of the T allele on birth weight was tested by linear regression model adjusted for the child's gender and gestational age. For that analysis, the power was > 80% to observe a birth weight variation of 80 g and 60 g in the AGA and SGA groups, respectively. A Student's T-test was applied to compare quantitative traits between SGA and AGA, without taking into account the rs7903146 genotype. Quantitative traits were then measured by linear regression models. These models took known covariables into account: gender and gestational age when testing for birth weight, birth length and ponderal index; age at examination and gender when testing for adult weight, adult height, body mass index (BMI), and waist to hip ratio; age at examination, gender, BMI, when testing for total plasma cholesterol, HDL, triglyceride, fasting glucose, fasting insulin, HOMA-B, HOMA-IR. The power was 80% to observe a variation in fasting glucose concentration of 0.01 g/l and to observe a variation in Ln(HOMA-B) of 0.09 units in the SGA and AGA groups. All tests were two-tailed, and p < 0.05 was considered to be statistically significant. No Bonferroni corrections were applied due to insufficient power. All p values were two-tailed. SPSS 14.1 software (SPSS Inc., Chicago, IL, USA) was used for general statistical analyses.

Comparison of the rs7903146 T allele frequencies in AGA (30.7%; 368 C/C, 323 C/T and 73 T/T) and SGA (29.0%; 319 C/C, 252 C/T and 56 T/T) are presented in Figure 1. The genotypic distributions between the two groups were in Hardy-Weinberg equilibrium (p = 0.86 for AGA and p= 0.56 for SGA) but were not statistically different (allelic OR = 0.92 [0.78–1.09], p = 0.34). Then the respective effects of mother's and child's genotypes on birth weight were tested by linear regression models (Table 2). No differences in birth weight were found within the 361 mother-child pairs of the SGA group (p = 0.93) and within the 215 mother-child pairs of the AGA group (p= 0.59). Family association-based studies did not show a distortion of T allele transmission in SGA subjects (48.5% for the T allele vs 51.5% for the C allele, p = 0.52). In 845 obese children (BMI = 97^th percentile), we tested the contribution to BW of the rs7903146 genotype by linear regression and after adjustments for gestational age and gender. In this sample, 404 children were C/C (BW = 3,466 ± 506 g), 356 were C/T (BW = 3,425 ± 567 g) and 85 were T/T (BW = 3,464 ± 668 g). The genotypic distribution was in Hardy-Weinberg equilibrium and no difference was found between the 3 genotypic groups (p = 0.61)

We first analyzed birth and adult (at around 22 years of age) metabolic parameters in AGA and SGA, without taking into account the rs7903146 genotype (Table 1). Insulin resistance (assessed by HOMA-IR index) and lipids profile were significantly different between SGA and AGA individuals. Then, we analyzed the effect of the rs7903146 T allele on quantitative traits by birth status and by genotype (Table 1). No significant effect was detected on any of the metabolic parameters in the SGA group. In the AGA group, however, trends towards a lower Ln(HOMA-B) and towards a higher fasting glucose were detected within the carriers of the T allele. No interaction between the birth status (AGA/SGA) and the rs7903146 variant was found in any of the quantitative traits. The absence of interaction between the genotype and the birth status for fasting glucose and Ln(HOMA-B) may be due to a lack of power.