INS VNTR Class Genotype and Indexes of Body Size and Obesity : Population-Based Studies of 7,999 Middle-Aged Men and Women

We used data from the Norfolk cohort of the European Prospective Investigation into Cancer (EPIC) and the Ely Study. EPIC is an ongoing prospective cohort study of ∼25,000 people aged 45–75, resident in Norfolk, and recruited from general practice registers between 1993 and 1997 (9). It is an ethnically homogeneous Caucasian population. These participants completed a health examination. In January 1998, we invited the cohort for a second health examination, and 15,786 people had attended by October 2000. All participants gave informed consent. Details of recruitment, anthropometric measurements, and health examinations following standardized protocols have been published (9). All participants completed a detailed health and lifestyle questionnaire, which included a question on birth weight. A d-loop nonstretch fiberglass tape was used for the head circumference measures. DNA was extracted from the 15,786 whole blood/EDTA samples (Whatman Biosciences, Ely, U.K.) taken at the second health examination. BMI was calculated from the weight and height measures.

For the purposes of this analysis, we examined data from two subcohorts of the 15,786 EPIC study participants who attended the second health examination (9). EPIC1 comprises 3,744 participants with fully arrayed DNA available for genotyping. EPIC2 is a random sample of 5,000 participants who were free of baseline disease (cancer, coronary heart disease, and diabetes) and had fully arrayed DNA available, completed food frequency questionnaire data, and had blood HbA_1c (A1C) and BMI measured at the two clinical assessments. We excluded 759 participants from EPIC2 as they were also present in the EPIC1 sample set. However, we used this duplicate set to assess concordance between genotyping methods, as described below.

We also used data from the Medical Research Council Ely Study, a prospective population-based cohort study of the etiology and pathogenesis of type 2 diabetes and related metabolic disorders (10). Again, the study population consists of an ethnically homogeneous Caucasian population in which phenotypic data have been recorded at clinical assessments using standardized protocols for anthropometric measurements. All participants were between 40 and 65 years of age at baseline and completed a health and lifestyle questionnaire, which included a question on birth weight. This cohort was recruited from a population-sampling frame with a high response rate (74%), making it representative of the general population for this area of eastern England.

The INS VNTR allele length falls into two broad classes in European populations: class I (26–63 repeats) and class III (141–209 repeats). We genotyped two subcohorts of the EPIC-Norfolk study (EPIC1 [n = 3,446] and EPIC2 [n = 3,720]) and 833 participants in the Ely study for the −23Hph1 polymorphism. There is 99.6% concordance between the A to T polymorphism at the −23Hph1 site CCACT at nucleotide 2,401 of the insulin gene (accession #V00565 and rs689) and the adjacent INS VNTR class I and III alleles in European populations (11). Hence −23/Hph1 genotype infers INS VNTR class I/III genotype. Genotype data for the EPIC1 study set were obtained from Invader assays (Third Wave Technologies, Madison, WI) and for the EPIC2 and the Ely study sets with TaqMan chemistry (Applied Biosystems, Warrington, U.K.), as previously reported (12). We assessed concordance between methods by typing a subset of 759 samples from EPIC1 and EPIC2 using both methods.

We tested for Hardy–Weinberg equilibrium using the χ² test. We calculated the effect of genotype on continuous anthropometric and metabolic characteristics using linear regression. In the primary analysis, we used an additive genetic model, assuming a linear relationship between the number of T (class III) alleles and the trait of interest. In a secondary analysis, we used a recessive model, comparing class III homozygotes with class I carriers, as demonstrated previously (8). Possible interactions between covariates and genotype were assessed with log-likelihood ratio tests. Stata version 8.2 was used for all analyses (Stata Corp, College Station, TX).

In the three cohort studies, the −23Hph1 gene variant was in Hardy–Weinberg equilibrium (EPIC1 P = 0.71, EPIC2 P = 0.24, and Ely P = 0.36). In a subgroup of 759 duplicate samples from EPIC1 and EPIC2, concordance between the two genotyping methods was 99%. There was no material difference in the association between genotype and indexes of body size and obesity in sex-stratified analyses; therefore, all data are presented for men and women combined. Table 1 shows demographic and anthropometric characteristics according to −23Hph1 genotype for the three cohorts. We found that the proportion of women varied by genotype (P = 0.02) in the EPIC1 study set; however, this observation was not evident in the other cohorts and is likely to be a chance finding. There was no association between the −23Hph1 genotype and adult head circumference, BMI, or body fat within the individual cohorts (Table 1). Similarly, in the pooled analysis adjusted for study, we found no substantial association between −23Hph1 genotype and indexes of growth and obesity (Table 2). In additional analyses, we found no consistent association between the −23Hph1 genotype and weight change (data not shown). However, we found marginal evidence for association between the −23Hph1 genotype and percentage body fat in the pooled analysis (P = 0.04). Nevertheless, based on the number of statistical tests, this association may be a false-positive (13). Exclusion of outliers (greater than three SDs for any variables) did not materially alter these findings (data not shown).

Our study does not support previous reports suggesting that INS VNTR class is associated with indexes of body size and obesity or related metabolic traits. By contrast, these data suggest that INS VNTR class genotype is not associated with indexes of body size and adult obesity in middle-aged men and women.

Variation in insulin expression influences growth, development, and metabolism. Phenotypes of animal models of insulin gene deletion exhibit reduced viability, growth, and metabolic control (1). Circulating maternal and fetal insulin is also a positive regulator of fetal growth and birth size in humans (14). However, the mechanistic processes underlying any association between INS VNTR class and indexes of growth or body size and metabolic regulation are speculative. In vitro studies have suggested that class I allele carriers may have higher insulin expression (15), but in vivo studies have found inconsistent associations between INS VNTR class and indexes of circulating levels of insulin (4,6,16). Moreover, whereas higher cord blood IGF-II levels, a regulator of fetal growth and development, have been associated with class III homozygote children (4), in vitro data have found no association (17) or have suggested the class III allele is associated with lower IGF-II transcription (5). In the present study, we found no association between INS VNTR class and circulating levels of insulin or IGF-II.

The INS VNTR class I allele may be overrepresented in obese children (18). Investigations of obese subpopulations also suggest that children carrying a paternal class I allele may have a greater rate of weight gain and risk of obesity (18,19). However, this imprinting effect on body weight regulation has not been replicated in studies of children from unselected populations (7) nor in studies of overweight and obese Caucasian and Pima adults (20,21). Nevertheless in unrelated individuals, one study of children and adolescents reported that class III homozygotes may have lower BMI compared with class I carriers (7). Conversely, it has also been shown that class III homozygote children and young adults may have a greater BMI than class I carriers (4,6). By contrast, we found no association between INS VNTR class as inferred by the −23Hph1 genotype and BMI in 7,999 middle-aged men and women, confirming previous reports based on middle-aged study participants (7,22).

One contemporary birth cohort has reported that INS VNTR class is associated with head circumference at birth and in early childhood (23), which has been corroborated in a larger subgroup of this cohort (4). These studies suggest that class III homozygotes may have larger head circumferences at birth and at 7 years. However, a subsequent study of 5,091 participants failed to find any association between the INS VNTR class and head circumference at 1 year (8). Similarly, in our analysis based on 7,149 study participants, we found no evidence for an association between adult head circumference and INS VNTR class.

We did not find a main genotype effect on birth weight in 3,473 study participants, which is consistent with several earlier studies (6,8,23). However, a study of Pima Indians reported that class III carriers may have lower birth weights (21). This observation is at variance with early results from the Avon Longitudinal Study of Parents and Children cohort, which suggested that, compared with class I homozygotes, class III carriers may have relatively higher birth weights in a subgroup of 348 “nonchangers” (23). However, this observation was not replicated in an additional subcohort of the study (4). Subsequent studies of a U.K. population cohort series and a Finnish birth cohort comprising a total of 2,795 participants have also failed to replicate the association between genotype and birth weight among nonchangers (6,8). Collectively, these observations suggest that the INS VNTR class genotype is unlikely to be a principal regulator of birth weight.

One limitation of this investigation was reliance on self-reported birth weight. Random misclassification of birth weight may lead to attenuation of the association among genotype and indexes of early growth. However, self-reports correlate well with birth records and mothers’ reports of birth weight (24). Birth weight in this cohort also showed the expected correlations with adult anthropometry. Moreover, our self-reported birth weight data showed the well-established inverse associations with measures of blood pressure and glucose tolerance (M.S.S., N.J.W., unpublished observations). The high concordance between genotyping methods suggests that it is unlikely that genotyping error could account for the lack of association between INS VNTR class genotype and indexes of body size and adult obesity.

We did not have data relating to other birth characteristics such as length of gestation, birth order, or postnatal growth realignment that might obscure or modify any genotype–growth/obesity index relation. However, to confound the association between INS VNTR and indexes of body size and obesity, these birth characteristics would need to be associated with both the INS VNTR class genotype and the specific outcome, for which there is no evidence (4,8). Associations between INS VNTR and indexes of body size and obesity have been found largely in contemporary birth cohorts. It is possible that these associations characterize the effect of INS VNTR class and changes in intergenerational environmental factors on growth, body size, and weight (4,6). Yet, effect-modification between INS VNTR and indexes of body size and obesity has not been reliably reproduced, and the biological basis for these possible interactions is unclear (6). Moreover, given the size of the present study, any association between INS VNTR class and indexes of body size and adult obesity in specific subgroups is unlikely to be very strong in this population.

In conclusion, in this population cohort series of 7,999 study participants, we found no evidence to suggest that INS VNTR class genotype is an important determinant of body size and adult obesity in middle-aged men and women.