VOLUME 19 NUMBER 10 | OCTOBER 2011 | www.obesityjournal.org
nature publishing group
Obesity is an increasing health problem worldwide. In the
United States, there has been a threefold increase in prevalence
of obesity over the past 5 decades, with most of the rise occur-
ring in more recent years (1,2). Both lifestyle and genetic fac-
tors contribute to obesity.
Numerous genome-wide association and candidate gene
studies have led to the identification of variants that are associ-
ated with BMI, a widely used measure of obesity (3–6). Most of
these studies have been conducted in white populations, while
the genetic architecture of obesity may differ across popula-
tions. American Indians, in particular, have a high prevalence of
obesity (7), but only a few linkage studies have been performed
in this population (8,9). No genome-wide association study
(GWAS) for BMI in American Indians has been reported.
We, therefore, performed a GWAS of BMI in Pima Indians.
In this population, BMI is highly heritable (7,10). In addition,
longitudinal data (collected from 1965 to 2004) are available,
and analysis of the repeated measures can provide additional
information in identifying variants associated with BMI.
ReseaRch Design anD PRoceDuRes
study participants and phenotypes
Full-heritage American Indians (n = 1,120; age ≥15 years) were selected
for the GWAS from participants in a longitudinal study (Table 1).
Criteria for inclusion in the GWAS included: (i) membership in a
family potentially informative for association with young-onset type
2 diabetes (11) or (ii) participation in detailed studies of metabolic and
anthropometric phenotypes in nondiabetic healthy individuals (12).
Although 436 individuals were singletons, the remainder were derived
from 247 sibships. Size of sibships ranged from two to eight indi-
viduals. As individuals were selected in sibships, this was taken into
account in the analysis, but we did not attempt to account for other
Of the 1,120 participants, 508 were male and 612 were female. Informed
consent was obtained for all participants and this study was approved by
the institutional review board of the National Institute of Diabetes and
Digestive and Kidney Diseases.
Data included BMI at multiple examinations (range: 1–17 examina-
tions for a given individual) collected from 1965 to 2004. Exams were
excluded if an individual was pregnant. Average BMI from all 5,904
exams in these individuals was 33.5 kg/m2. Table 1 provides details of
BMI measurements for both longitudinal BMI (i.e., all examinations)
and maximum BMI measurements.
genotyping and quality control
Individuals were genotyped according to the manufacturer’s protocol
with the Genome-Wide Human SNP Array 6.0 (Affymetrix, Santa
Clara, CA), which contains 909,508 single-nucleotide polymorphisms
(SNPs) with known positions in the human genome. Genotypes were
generated using the BIRDSEED algorithm. Details on the quality control
procedures are given in the Supplementary Methods and Procedures
online. A total of 454,194 SNPs passed these quality control procedures
A Genome-Wide Association Study of BMI
in American Indians
Alka Malhotra1, Sayuko Kobes1, William C. Knowler1, Leslie J. Baier1, Clifton Bogardus1
and Robert L. Hanson1
Numerous studies have been done to understand genetic contributors to BMI, but only a limited number of studies
have been done in nonwhite groups such as American Indians. A genome-wide association study (GWAS) for BMI was
therefore performed in Pima Indians. BMI measurements from a longitudinal study of 1,120 Pima Indians and 454,194
single-nucleotide polymorphisms (SNPs) from the 1 million Affymetrix SNP panel were used (35% of SNPs were
excluded due to minor allele frequency <0.05). Data included BMI measured at multiple examinations collected from
1965 to 2004, as well as the maximum BMI at one of these visits. General and within-family tests were performed using
a maximum-likelihood based mixed model procedure. No SNP reached a genome-wide significance level (estimated at
P < 4.94 × 10−7). For repeated measures analyses, the strongest associations for general and within-family tests mapped
to two different regions on chromosome 6 (rs9342220 (P = 1.39 × 10−6) and rs7758764 (P = 2.51 × 10−6), respectively). For
maximum BMI, the strongest association for the general tests mapped to chromosome 4 (rs17612333; P = 1.98 × 10−6)
and to chromosome 3 (rs11127958; P = 1.53 × 10−6) for the within-family tests. Further analysis is important because only
a few of these regions have been previously implicated in a GWAS and genetic susceptibility may differ by ethnicity.
Obesity (2011) 19, 2102–2106. doi:10.1038/oby.2011.178
1Phoenix Epidemiology and Clinical Research Branch, National Institute of Diabetes and Digestive and Kidney Diseases, Phoenix, Arizona, USA.
Correspondence: Alka Malhotra (email@example.com)
Received 5 October 2010; accepted 12 May 2011; published online 23 June 2011. doi:10.1038/oby.2011.178