American Journal of Epidemiology
Copyright © 2003 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved
Vol. 157, No. 4
Printed in U.S.A.
DOI: 10.1093/aje/kwf208
Genetic and Environmental Contributions to Cardiovascular Disease Risk in
American Indians
The Strong Heart Family Study
Kari E. North1, Barbara V. Howard2, Thomas K. Welty3, Lyle G. Best4, Elisa T. Lee5, J. L. Yeh6,
Richard R. Fabsitz6, Mary J. Roman7, and Jean W. MacCluer1
1
Department of Genetics, Southwest Foundation for Biomedical Research, San Antonio, TX.
Medstar Research Institute, Washington, DC.
3 Aberdeen Area Tribal Chairmen’s Health Board, Rapid City, SD.
4 Missouri Breaks Industries Research, Inc., Timber Lake, SD.
5 Center for American Indian Health Research, School of Public Health, University of Oklahoma Health Sciences Center,
Oklahoma City, OK.
6 Epidemiology and Biometry Program, National Heart, Lung, and Blood Institute, Bethesda, MD.
7 Cornell University Medical College, New York, NY.
2
Received for publication November 12, 2001; accepted for publication September 16, 2002.
The aims of the Strong Heart Family Study are to clarify the genetic determinants of cardiovascular disease (CVD)
risk in American Indians and to map and identify genes for CVD susceptibility. The authors describe the design of
the Strong Heart Family Study (conducted between 1998 and 1999) and evaluate the heritabilities of CVD risk
factors in American Indians from this study. In the first phase of the study, approximately 950 individuals, aged 18
years or more, in 32 extended families, were examined. The examination consisted of a personal interview, physical
examination, laboratory tests, and an ultrasound examination of the carotid arteries. The phenotypes measured
during the physical examination included anthropometry, lipoproteins, blood pressure, glycemic status, and clotting
factors. Heritabilities for CVD risk factor phenotypes were estimated using a variance component approach and the
program SOLAR. After accounting for the effects of covariates, the authors detected significant heritabilities for
many CVD risk factor phenotypes (e.g., high density lipoprotein cholesterol (heritability = 0.50) and diastolic blood
pressure (heritability = 0.34)). These results suggest that heredity explains a substantial proportion of the variability
of CVD risk factors and that these heritabilities are large enough to warrant a search for major risk factor genes.
cardiovascular diseases; environment; genetic predisposition to disease; Indians, North American; risk factors
Abbreviation: CVD, cardiovascular disease.
Cardiovascular disease (CVD) is the leading cause of
death in the United States, but until recently, it was thought
to be rare among American Indians (1, 2). The Strong Heart
Study, a population-based study of risk factors for cardiovascular disease and related disorders in American Indians,
changed this notion by demonstrating a high prevalence of
CVD and CVD risk factors among 13 Indian communities in
three geographic areas (3).
Although CVD mortality in the United States has been
steadily declining in recent decades, the American Indians of
the Strong Heart Study display increasing mortality rates and
incidence rates (4, 5). Indeed, CVD mortality rates among
American Indians aged 45–74 years were as high or higher
than corresponding state rates in Arizona, Oklahoma, and
North and South Dakota between 1984 and 1988 (5).
Furthermore, Welty et al. (6) have reported increases in
levels of many CVD risk factors across time.
The Strong Heart Family Study was initiated to investigate
genetic effects on CVD risk factors. Risk factors for CVD
have been identified both within and between populations,
Reprint requests to Dr. Kari E. North, Department of Epidemiology, Bank of America Center, 137 E. Franklin St., Suite 306, Chapel Hill, NC
27514-3628 (e-mail: [email protected]).
303
Am J Epidemiol 2003;157:303–314
304 North et al.
including diabetes, elevated blood pressure, adverse lipid
profile, smoking, a family history of CVD, male gender,
obesity, insulin resistance, a sedentary lifestyle, and a diet
high in saturated fat and cholesterol (7, 8). Although many
possible candidate genes for CVD have been identified, for
only a few of these genes is their relation to CVD understood, and their contributions to variation in CVD risk
factors tend to be small. Little is known about genes that
contribute to population variability in precursors of CVD.
Because metabolic and epidemiologic studies have identified many quantitative CVD risk factors, there is great potential to identify the effects of specific genes on the
development of CVD. The American Indian populations of
the Strong Heart Study have high prevalences of diabetes,
obesity, insulin resistance, and several other CVD risk
factors, and therefore alleles that contribute to disease risk
are most likely overrepresented. The identification of genes
that increase disease susceptibility in American Indians will
provide an opportunity to understand variability in response
to environmental risk factors. This report describes the
Strong Heart Family Study and summarizes the genetic and
environmental contributions to CVD risk factors in American Indians from the first phase of the Strong Heart Family
Study, as a first step in our search for CVD risk factor genes.
MATERIALS AND METHODS
Strong Heart Study
The Strong Heart Study was initiated in 1988 as a longitudinal study of CVD and its risk factors in 13 American
Indian communities in three geographic regions in Arizona,
Oklahoma, and North and South Dakota.
The Strong Heart Study has included three clinical examinations and mortality and morbidity surveillance of the original cohort, resident tribal members aged 45–74 years. In
1998–1999, a pilot family study, the Strong Heart Family
Study, was added in which 8–12 extended families (more
than 300 family members at least 18 years of age) were
recruited and examined at each center. An extension of the
pilot family study is currently in progress and will involve
recruitment of approximately 90 additional families, 30 from
each center, comprising 2,700 family members. The Strong
Heart Study protocol was approved by the Indian Health
Service Institutional Review Board, by the institutional
review boards of the participating institutions, and by the 13
participating tribes.
Strong Heart Family Study
The study populations. The Strong Heart Family Study
population includes 13 communities in three geographic
areas in Arizona, Oklahoma, and North and South Dakota.
Details of the three field centers, published previously (9),
are briefly described here.
Although the Dakota Center has examined participants
from three Sioux Indian tribes (figure 1), the majority of
Dakota Strong Heart Family Study participants were
recruited from one tribe, the Cheyenne River Sioux tribe of
the Cheyenne River Reservation, South Dakota. Overall, the
mean and median levels of percentage of Indian heritage
were 63 percent and 75 percent, respectively, with 121
participants reporting 100 percent American Indian heritage.
The Oklahoma center encompasses seven different tribes,
primarily Plains and Southeastern American Indians
(Apache, Caddo, Comanche, Delaware, Fort Sill Apache,
Kiowa, and Wichita tribes in the original study) living in
the Lawton-Anadarko area in southwestern Oklahoma.
Unlike the tribes in Arizona and the Dakotas, members of
these tribes do not live on reservations (there are no
reservations in Oklahoma) (10). The majority of the 310
Strong Heart Family Study participants from the Oklahoma
Center are members of three tribes, the Kiowa (n = 186),
Comanche (n = 80), and Delaware (n = 20). The Oklahoma
Center participants had varying degrees of self-identified
American Indian heritage. A total of 224 participants
reported 100 percent American Indian heritage, and the
mean and median levels were 87 percent and 100 percent,
respectively.
The Arizona Center is located in Phoenix. This center has
enrolled primarily Pima Indians, but there are also representatives of the closely related Maricopa and Tohono O’odham
Indian tribes (9). The participants live in three communities
near Phoenix: the Gila River and Salt River reservations
(Pima/Maricopa) and the Ak-Chin Reservation (Tohono
O’oodham/Pima) (3). The Gila River community is the
largest of the three (figure 1). The Ak-Chin Indian community, located on the border of the Gila River Indian community (figure 1), has the smallest population, numbering in
total 500 (8). The majority of the 345 family participants
from the Arizona Center are members of two tribes, the Gila
River (n = 194) and Salt River (n = 135) Pima/Maricopa. The
majority of Arizona Center participants (319 of 345)
reported 100 percent American Indian heritage, with mean
and median levels of 98 percent and 100 percent,
respectively.
Family recruitment. In preparation for this large-scale
family study, we expanded upon data collected for the original Strong Heart Study cohort, using family history forms
completed by each participant in phase I. These forms
included the names and years of birth and death of each
participant’s parents, full and half siblings, and offspring.
They were computerized and linked into pedigrees using a
name-matching algorithm (11) and other demographic information. From the 575 sibships identified in this manner
(Dakota Center (n = 202), Oklahoma Center (n = 210), and
Arizona Center (n = 163)), preliminary family trees were
generated and sent to the field centers for potential
recruitment.
To ensure that families were sufficiently large, we
required that each have a “core sibship” with at least five
living members, of whom three were original Strong Heart
Study cohort members. The cohort members were required
to have at least 12 living offspring who were at least 18 years
of age (see prototype family tree in figure 2). The core
sibship formed the basis for recruitment of extended families, which included members of the core sibship and their
parents (if alive), spouses, offspring, spouses of offspring,
and grandchildren who were 18 years of age or more.
Am J Epidemiol 2003;157:303–314
Cardiovascular Disease in American Indians 305
FIGURE 1.
Approximate locations of Strong Heart Family Study communities, 1998–1999. Adapted with permission from Lee et al. (10).
Older members of each family were interviewed to determine the probable interest of the families in participating.
From this information, the list at each center was narrowed
to 9–12 families. In total, 32 extended families were
recruited (nine in the Dakota Center, 11 in the Oklahoma
Center, and 12 in the Arizona Center) and, as of February
2001, data were available for a total of 1,011 individuals (of
whom 30 were nonbiologic relatives who were given courtesy examinations).
Family relationships were confirmed in interviews with
the Strong Heart cohort members identified in the initial
family trees. The family trees were continuously updated
using Pedigree/Draw (12), with notations as to which family
members should be the focus of the recruitment effort. To
the extent possible, the validity of stated family relationships
was verified during participant interviews. Every attempt
was made to recruit both parents of every descendant of the
core sibships, including any spouses who are non-Indian.
Phenotypic, demographic, and lifestyle data. The longterm goal of the Strong Heart Family Study is to map and
identify genes that contribute to the risk of cardiovascular
disease in American Indians. Genes that influence CVD risk
factors are likely to interact with each other and with environmental factors. This study is using a strategy specifically
designed to deal with this complexity by selecting phenotypes that are readily measurable in large numbers of individuals, that show reasonable repeatability over time, and
that show promise of providing useful data for genetic
analysis.
FIGURE 2. Strong Heart Family Study recruitment criteria, 1998–1999. A prototype family demonstrating the recruitment criteria: a sibship with
at least three Strong Heart Study participants, five or more siblings, and at least 12 offspring of the Strong Heart Study participants. In this figure,
black shading indicates Strong Heart Study cohort participants, diagonal lines indicate additional siblings, and dotted shading indicates the 12
children. Squares designate males and circles designate females.
Am J Epidemiol 2003;157:303–314
306 North et al.
TABLE 1. Phenotypes and their abbreviations used for analysis of cardiovascular disease risk
factors in American Indians from the Strong Heart Family Study, 1998–1999
Category of phenotype
Obesity
Lipid
Phenotype
Abbreviation
Weight (kg)
Body mass index (kg/m2)
BMI
Waist/hip ratio (cm)
WHR
Body fat mass (%)
FM
Fat-free mass (%)
FFM
Low density lipoprotein cholesterol (mg/dl)
LDL-C
High density lipoprotein cholesterol (mg/dl)
HDL-C
Triglyceride (mg/dl)
Blood pressure
Diabetes
Apolipoprotein A-1 (mg/ml)
ApoA-1
Apolipoprotein B (mg/ml)
ApoB
Lipoprotein a (mg/dl)
Lp(a)
Very low density lipoprotein cholesterol (mg/dl)
VLDLC
Very low density lipoprotein triglycerides (mg/dl)
VLDTG
Systolic blood pressure (mmHg)
SBP
Diastolic blood pressure (mmHg)
DBP
Fasting glucose (mg/dl)
Insulin (pmol/liter)
Clotting factors
Fibrinogen (mg/dl)
Plasminogen activator inhibitor 1 (ng/ml)
The Strong Heart Family Study examination consisted of a
personal interview, physical examination, laboratory tests,
and carotid ultrasonography. Several categories of phenotypes were assessed, such as body mass index and composition, lipoproteins, blood pressure, glycemic status, and
clotting measures. Standard protocols, used for the collection of all data, are described in detail in previous publications (3, 9).
Briefly, fasting blood samples were obtained during the
physical examination for the measurement of lipids, lipoproteins, apolipoproteins, insulin, glucose, plasma creatinine,
plasma fibrinogen, and plasminogen activator inhibitor 1.
All variables were assayed at MedStar Research Institute,
Washington, DC, and the University of Vermont using standard laboratory methods as previously described (3, 9). Body
fat mass was measured using an RJL bioelectric impedance
meter (RJL Systems, Detroit, Michigan). The percentage of
body fat was estimated by the RJL formula based on total
body water (13). Blood pressure was measured three times,
and the average of the last two measurements was used for
analysis. Diabetes status was determined using World Health
Organization criteria (14). Carotid ultrasonography was
performed using previously described methods (15).
Intimal-medial thickness and minimum and maximum diameters of the common carotid arteries were measured from
digitized M-mode images. The presence of focal plaque was
assessed using B-mode scanning with Doppler quantification of significant stenosis. Vascular stiffness of the common
carotid artery was estimated using methods that incorporate
arterial diameters and the central arterial pressure waveform
back-calculated from radial artery tonometry by means of
validated transfer functions (16).
PAI-1
Information was also collected on demographic characteristics, lifestyle variables, medical history, and reproductive
history. During the personal interview, information on the
following demographic characteristics was obtained by
questionnaire: income, residence, marital status, number of
household members, tribal enrollment, degree of Indian heritage, education, and other cultural factors. Participants were
asked lifestyle questions, with a focus on smoking, alcohol
intake, physical activity, and diet (24-hour recall). For questions on alcohol intake, current and ever drinking were
defined as having had at least one alcoholic beverage in the
last year and/or lifetime, respectively. Smoking was defined
as having had at least 100 cigarettes. A medical history was
taken, including the Rose questionnaire for angina pectoris
and intermittent claudication. A reproductive history was
taken including questions concerning parity, gravidity,
menopausal status, and estrogen use.
Phenotypic data were transmitted by the Oklahoma Coordinating Center to the Southwest Foundation, where they
were transferred to our pedigree data management system,
PEDSYS (17).
Analytic techniques
To identify and evaluate the genetic and environmental
contributions to CVD risk factors, we used a variance
component approach implemented in SOLAR (18). The
quantitative phenotype for an individual (y) is modeled as
(19, 20)
y = µ + ∑ βj v ij + g i + e i
(1)
Am J Epidemiol 2003;157:303–314
Cardiovascular Disease in American Indians 307
TABLE 2. Strong Heart Family Study numbers of examined relative pairs, 1998–1999
Relationship
North and South
Dakota
(n = 326)
Oklahoma
(n = 310)
Arizona
(n = 345)
Strong Heart
Family Study
(n = 981)
Parent-offspring
336
245
319
900
Siblings
330
319
350
999
86
77
44
207
Avuncular
954
802
721
2,477
Grand avuncular
395
139
261
795
Half siblings
Grandparent-grandchild
129
35
127
291
First cousins
1,142
843
866
2,851
First cousins once removed
1,091
409
909
2,409
Second cousins
487
57
233
777
Other
621
322
145
1,088
5,571
3,248
3,975
12,794
Total relative pairs
where µ is the mean of the trait in males, βj is the regression
coefficient for the covariate j, vij is the value of covariate j in
individual i, and gi and ei represent the deviations from µ for
the individual i that are attributable to additive genetic
effects and unmeasured environmental effects, respectively.
gi and ei are assumed to be uncorrelated with one another and
normally distributed with mean 0 and variances σ g2 and σ e2 .
In the simplest model, the covariance between a set of relative pairs (Ω) is a function of the additive genetic variance
and the random environmental variance. To permit the analysis of arbitrary pedigree structures, each of the variances is
multiplied by a structuring matrix. The structuring matrix for
the additive genetic variance is two times the matrix of
kinship coefficients ( Φ ), while the structuring matrix for the
environmental variance is an identity matrix (m) (diagonal
elements are ones and the rest of the elements are zeroes),
which permits unique environments for each individual (18).
Thus,
Ω = 2 Φσ A2 + I σ e2 .
Once the expected mean and the covariance matrix for
each pedigree are defined, the likelihood of a pedigree is
evaluated using the multivariate normal density function and
summed over all pedigrees. The likelihood of the phenotypes
of the family members is assumed to follow a multivariate
normal distribution, but the method is robust to violations of
this assumption (18). The p values for the heritability calculations are obtained by likelihood ratio tests, where the likelihood of a model in which heritability is estimated is
compared with the likelihood of the model in which the heritability is constrained to zero. Twice the difference in the
natural logarithmic likelihoods is asymptotically distributed
as a 1/2:1/2 mixture of a chi-squared variable with 1 df and a
point mass at zero (21).
Data
Five classes of CVD risk factors were examined including
obesity, lipoprotein, blood pressure, diabetes related, and
Am J Epidemiol 2003;157:303–314
clotting phenotypes. The specific CVD risk factors and their
abbreviations are reported in table 1. The variables used for
the estimation of covariate effects include basic demographic (age, sex, age × sex interaction, age2, age2 × sex
interaction, self-reported American Indian heritage, and
diabetes status), lifestyle (current and/or ever-smoking status
and current and/or ever consumer of alcoholic beverages),
and reproductive history (current and/or ever estrogen use)
measures.
Because the sample size is small within any one center, our
analysis is based upon the combined data from all three
centers. To obtain an estimate of between-center differences,
we included each center as a covariate in the analyses. When
larger sample sizes become available in 2003, independent
analyses will be conducted in each center.
The analysis of each phenotype was restricted to those
individuals for whom all covariate data were complete. In
addition, lipoprotein phenotypes were not analyzed for individuals currently taking lipid-lowering medications (n = 13),
and blood pressure measures were not analyzed for 143 individuals currently taking antihypertensive medications.
Insulin and glucose values were not analyzed for individuals
currently taking antidiabetic medications (n = 153). For all
phenotypes, any individual value greater than 4 standard
deviations from the mean was removed from analysis.
During preliminary analyses, the effects of tribal membership and household, independent of additive genetic effects,
were estimated (data not shown). Because no tribal membership or household effects were identified, tribal membership
and household effects were removed from further analyses.
However, linear center effects were identified (multiple
tribes at each center), and the center was included as a covariate in all further analyses.
The initial analysis screened for the following classes of
covariates: sex, age, age × sex interaction, diabetes status,
percentage of Indian heritage, center, education, estrogen
use, alcohol consumption, and smoking status. Any covariates whose effects were significant at the p ≤ 0.10 level in the
initial analysis were retained in subsequent analyses, even if
the significance levels decreased after inclusion of other
308 North et al.
TABLE 3. A summary by mean values of descriptive statistics of cardiovascular disease risk
factors in males, Strong Heart Family Study, 1998–1999
Category of phenotype and
phenotype
North and South Dakota
Oklahoma
Arizona
Obesity
Weight (kg)
91.36 (17.09)*
92.28 (21.88)
98.31 (21.36)
BMI† (kg/m2)
29.54 (5.40)
30.04 (6.12)
33.16 (7.10)
WHR†
0.94 (0.08)
0.94 (0.07)
0.99 (0.06)
FM† (%)
27.19 (6.44)
27.54 (6.70)
33.16 (6.88)
FFM† (%)
65.94 (8.86)
65.74 (11.67)
66.58 (11.53)
LDL-C† (mg/dl)
129.45 (34.32)
120.13 (31.08)
113.09 (29.71)
HDL-C† (mg/dl)
40.56 (12.68)
40.72 (12.83)
39.76 (12.36)
Total cholesterol (mg/dl)
196.01 (39.28)
184.27 (36.09)
175.63 (33.82)
Triglyceride (mg/dl)
Lipid‡
122.12 (57.44)
133.51 (64.77)
126.28 (65.96)
ApoA-1† (mg/ml)
1.27 (0.22)
1.30 (0.25)
1.27 (0.26)
ApoB† (mg/ml)
1.07 (0.29)
1.02 (0.27)
1.00 (0.30)
Lp(a)† (mg/dl)
3.48 (2.79)
4.02 (3.06)
3.16 (3.16)
VLDLC† (mg/dl)
20.85 (9.09)
20.38 (9.99)
18.80 (11.07)
VLDTG† (mg/dl)
70.49 (41.13)
81.66 (56.62)
69.75 (54.10)
SBP† (mmHg)
125.48 (11.92)
124.63 (12.99)
127.70 (14.87)
DBP† (mmHg)
81.46 (9.63)
76.24 (9.85)
79.28 (10.81)
Fasting glucose (mg/dl)
97.13 (11.49)
95.88 (13.94)
102.77 (15.44)
Insulin (pmol/liter)
16.33 (10.53)
19.47 (13.48)
22.57 (13.83)
304.00 (63.33)
306.73 (66.35)
330.45 (75.98)
48.62 (37.76)
61.86 (37.13)
57.07 (38.64)
Blood pressure§
Diabetes¶
Clotting factors
Fibrinogen (mg/dl)
PAI-1† (ng/ml)
* Numbers in parentheses, standard deviation.
† BMI, body mass index; WHR, waist/hip ratio; FM, body fat mass; FFM, fat-free mass; LDL-C, low
density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol; ApoA-1, apolipoprotein A-1;
ApoB, apolipoprotein B; Lp(a), lipoprotein a; VLDLC, very low density lipoprotein cholesterol; VLDTG,
very low density lipoprotein triglycerides; SBP, systolic blood pressure; DBP, diastolic blood pressure;
PAI-1, plasminogen activator inhibitor 1.
‡ Individuals taking cholesterol-lowering medications were not included in these analyses.
§ Individuals taking antihypertensive medications were not included in these analyses.
¶ Individuals taking diabetes-related medications were not included in these analyses. For all
phenotypes, any individual value greater than 4 standard deviations from the mean was removed from
analysis.
covariates. After the initial covariate screening, maximum
likelihood methods were used to estimate the effects of covariates and additive effects of genes.
RESULTS
Descriptive statistics
Because we recruited extended families, the sample of
examined individuals included a large number of relative
pair types (table 2). The sample included information on
12,800 relative pairs: nearly 1,900 pairs of first-degree relatives (i.e., 900 parent-offspring pairs and 999 sibling pairs),
3,025 pairs of second-degree relatives (i.e., 2,477 avuncular
pairs (aunt/uncle-niece/nephew) and 291 grandparent-grandchild pairs), 4,012 pairs of third-degree relatives (i.e., 795
great-avuncular pairs and 2,851 first cousin pairs), and 4,863
pairs of relatives of fourth degree or greater (i.e., 2,409 pairs
of first cousins once removed and 777 pairs of second
cousins).
The descriptive statistics for cardiovascular disease risk
factors and the sex-specific demographic characteristics of
Strong Heart Family Study participants are reported in tables
3, 4, 5, and 6. There was little variance in sample size across
phenotypes and center. Additionally, there was a maximum
difference of 1.0 and 3.4 years among the mean ages of
Am J Epidemiol 2003;157:303–314
Cardiovascular Disease in American Indians 309
TABLE 4. A summary by mean values of descriptive statistics of cardiovascular disease risk
factors in females, Strong Heart Family Study, 1998–1999
Category of phenotype and
phenotype
North and South Dakota
Oklahoma
Arizona
Obesity
Weight (kg)
82.66 (19.50)*
86.37 (16.71)
90.12 (21.53)
BMI† (kg/m2)
30.64 (6.25)
30.61 (6.23)
34.90 (7.21)
WHR†
0.90 (0.09)
0.90 (0.07)
0.94 (0.06)
FM† (%)
39.17 (7.83)
39.64 (7.45)
45.04 (7.94)
FFM† (%)
49.24 (7.61)
47.81 (6.90)
49.19 (7.27)
LDL-C† (mg/dl)
116.89 (28.59)
111.06 (25.52)
108.09 (26.36)
HDL-C† (mg/dl)
45.12 (11.90)
47.82 (13.31)
40.76 (10.19)
Total cholesterol (mg/dl)
184.51 (34.17)
181.66 (30.72)
169.65 (32.74)
Triglyceride (mg/dl)
123.32 (60.86)
116.71 (53.71)
123.17 (56.84)
Lipid‡
ApoA-1† (mg/ml)
1.37 (0.25)
1.45 (0.28)
1.29 (0.25)
ApoB† (mg/ml)
0.97 (0.27)
0.95 (0.23)
0.97 (0.26)
Lp(a)† (mg/dl)
4.13 (3.24)
3.82 (3.28)
2.98 (2.34)
VLDLC† (mg/dl)
18.73 (8.52)
17.28 (8.23)
17.46 (9.57)
VLDTG† (mg/dl)
67.63 (50.18)
56.10 (44.18)
62.04 (46.81)
SBP† (mmHg)
119.76 (14.10)
121.42 (15.67)
118.77 (15.49)
DBP† (mmHg)
74.51 (9.27)
74.17 (8.88)
73.75 (8.98)
Fasting glucose (mg/dl)
94.15 (12.19)
93.34 (15.77)
98.89 (17.16)
Insulin (pmol/liter)
19.53 (13.65)
17.80 (12.01)
24.20 (13.35)
315.28 (70.95)
325.43 (64.93)
359.73 (67.27)
45.51 (38.71)
55.73 (38.37)
59.03 (38.13)
Blood pressure§
Diabetes¶
Clotting factors (mg/dl)
Fibrinogen (mg/dl)
PAI-1† (ng/ml)
* Numbers in parentheses, standard deviation.
† BMI, body mass index; WHR, waist/hip ratio; FM, body fat mass; FFM, fat-free mass; LDL-C, low
density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol; ApoA-1, apolipoprotein A-1;
ApoB, apolipoprotein B; Lp(a), lipoprotein a; VLDLC, very low density lipoprotein cholesterol; VLDTG,
very low density lipoprotein triglycerides; SBP, systolic blood pressure; DBP, diastolic blood pressure;
PAI-1, plasminogen activator inhibitor 1.
‡ Individuals taking cholesterol-lowering medications were not included in these analyses.
§ Individuals taking antihypertensive medications were not included in these analyses.
¶ Individuals taking diabetes-related medications were not included in these analyses. For all
phenotypes, any individual value greater than 4 standard deviations from the mean was removed from
analysis.
males and females, respectively, at the three centers. The
prevalence of diabetes was 30 percent, with a higher proportion of females than males (tables 5 and 6). Approximately
57 percent of the participants reported current alcohol
consumption, and approximately 37 percent of men and 32
percent of women reported that they currently smoked
(tables 5 and 6). Approximately 19 percent of women
reported the ever use of estrogen.
Covariate effects
In the initial variance component analysis that included all
of the covariates previously described, several CVD risk
factors were only moderately heritable (data not shown).
Am J Epidemiol 2003;157:303–314
Recognizing that American Indian populations have among
the highest rates of diabetes worldwide, we reasoned that
genes influencing diabetes status also may affect the pattern
of these CVD risk factors, and that by “factoring out” the
effects of diabetes status, we may have removed some of the
genetic component in CVD risk. Similarly, inclusion of the
percentage of Indian heritage as a covariate in essence
factored out the effects of genes common to American
Indian populations, genes that may be important for CVD
risk. We therefore removed the covariates percentage of
Indian heritage and diabetes status from the analysis to determine if these covariates were dampening the genetic signal.
Excluding these covariates did increase the heritability esti-
310 North et al.
TABLE 5. Demographic characteristics of male American Indians participating in the Strong
Heart Family Study, 1998–1999
North and South
Dakota estimate
Oklahoma
estimate
Arizona
estimate
Age (years)
41.36
41.93
40.96
Self-identified Indian heritage (%)
59
85
98
Diabetic (%)
19
18
43
Ever smoking (%)*
58
63
59
Current smoking (%)
40
42
31
Ever drinking (%)†
82
82
89
Current drinking (%)
70
56
75
Category of variables and variable
Demographic
Lifestyle
* Current and ever smoking are defined as having had at least 100 cigarettes in the last year or
lifetime, respectively.
† Current and ever drinking are defined as having had at least one alcoholic beverage in the last year
or lifetime, respectively.
mates for most variables examined (data not shown), and
therefore we removed them from further analyses.
Table 7 displays the significant covariates for each CVD
risk factor and the total proportion of variance accounted for
by the significant covariates. Significant center effects were
identified for 14 of 20 phenotypes examined. However, it is
important to note that there is likely to be a confounding of
this covariate with two others, tribal membership and
percentage of Indian heritage. Once sample sizes are
increased, analyses will be done separately for each center.
The proportion of the total phenotypic variance accounted
for by the measured covariates ranged from 1 percent (very
low density lipoprotein cholesterol) to 50 percent (lean body
mass) (table 7) and varied within category; for example, for
the obesity phenotypes, the covariate effects ranged from 10
percent for body mass index to 50 percent for lean body
mass.
Heritability of CVD risk factors
Table 8 presents the heritability estimates, measured as the
proportion of residual phenotypic variance due to the additive effect of genes, after the effects of covariates have been
accounted for, for selected CVD risk factors. The proportion
of residual phenotypic variance due to the additive effects of
genes ranges from 23 percent (systolic blood pressure and
natural log-transformed fibrinogen) to 54 percent (waist/hip
ratio). All of the obesity phenotypes have heritabilities
greater than 40 percent. The heritabilities of all lipid phenotypes are 34 percent or greater. For example, 8 percent of the
phenotypic variance in high density lipoprotein cholesterol
is attributable to covariate effects, and 50 percent of the
remaining phenotypic variance is due to the additive effects
of genes. Both blood pressure phenotypes display significant
components of genetic variation (p ≤ 0.0001). In addition,
TABLE 6. Demographic characteristics of female American Indians participating in the Strong
Heart Family Study, 1998–1999
Category of variables and variable
North and South
Dakota estimate
Oklahoma
estimate
Arizona
estimate
Demographic
Age (years)
41.19
44.60
42.24
Self-identified Indian heritage (%)
66
88
98
Diabetic (%)
24
23
49
Ever smoking (%)*
60
46
39
Current smoking (%)
48
27
20
Ever drinking (%)†
73
68
71
Current drinking (%)
61
36
49
Ever estrogen use (%)
18
24
14
Current estrogen use (%)
10
17
7
Lifestyle
* Current and ever smoking are defined as having had at least 100 cigarettes in the last year or
lifetime, respectively.
† Current and ever drinking are defined as having had at least one alcoholic beverage in the last year
or lifetime, respectively.
Am J Epidemiol 2003;157:303–314
Cardiovascular Disease in American Indians 311
TABLE 7. Significant covariate effects (p ¯ 0.10) in American Indians of the Strong Heart Family Study from maximum likelihood
decomposition of variance using sequential oligogenic linkage analysis routines, 1998–1999
Phenotype and trait
Sex
Age
X
X
Age × sex
Estrogen
Center
Drinking
Smoking
% of variance due
to covariates
Obesity
Weight
BMI*
X
WHR*
X
X
FM*
X
X
X
FFM*
X
X
X
LDL-C*
X
X
X
HDL-C*
X
Total cholesterol
X
X
ApoA-1*
X
X
ApoB*
X
X
X
X
X
13
X
X
X
X
10
X
X
X
X
25
X
X
46
X
X
50
Lipids†
ln* triglyceride
X
X
X
X
X
X
X
X
X
X
X
8
8
11
X
2
X
16
X
Lp(a)*
X
VLDLC*
X
VLDTG*
X
7
1
1
X
3
Blood pressure‡
SBP*
X
X
DBP*
X
X
X
X
X
16
X
11
Diabetes§
Fasting glucose
ln insulin
X
X
X
X
X
X
X
X
X
13
X
5
Clotting factors
ln fibrinogen
PAI-1*
X
X
X
X
X
X
17
6
* BMI, body mass index; WHR, waist/hip ratio; FM, body fat mass; FFM, fat-free mass; LDL-C, low density lipoprotein cholesterol; HDL-C,
high density lipoprotein cholesterol; ln, natural log transformed; ApoA-1, apolipoprotein A-1; ApoB, apolipoprotein B; Lp(a), lipoprotein a;
VLDLC, very low density lipoprotein cholesterol; VLDTG, very low density lipoprotein triglycerides; SBP, systolic blood pressure; DBP, diastolic
blood pressure; PAI-1, plasminogen activator inhibitor 1.
† Individuals taking cholesterol-lowering medications were not included in these analyses.
‡ Individuals taking antihypertensive medications were not included in these analyses.
§ Individuals taking diabetes-related medications were not included in these analyses. For all phenotypes, any individual value greater than 4
standard deviations from the mean was removed from analysis.
the heritability of the two diabetes phenotypes is 29 percent
or greater, and the heritability of the two clotting phenotypes
is 23 percent or greater. Because substantial heritability estimates for obesity measures were found, we additionally
adjusted CVD risk factors for body mass index (table 8). Not
unexpectedly, adjustment for body mass index slightly
reduced the genetic signal for many of the CVD risk factors
examined.
DISCUSSION
The Strong Heart Family Study provides an exceptional
opportunity to characterize the genetic contributions to risk
of cardiovascular disease in American Indians. In many US
populations, the identification and recruitment of extended
families can be difficult because of the dispersal of family
Am J Epidemiol 2003;157:303–314
members over large geographic areas. However, many of the
participants in the Strong Heart Family Study have spent
their entire lives as residents of the same communities. For
the genetic analysis of common complex traits such as CVD
risk factors, the extended family sampling strategy is more
powerful than other approaches (22–24), and the availability
of extended families offers an advantage for gene localization. The relative biologic homogeneity of the American
Indian communities in the Strong Heart Family Study also
increases the likelihood of identifying genetic determinants
of disease risk. Finally, the American Indian populations of
the Strong Heart Family Study have high prevalences of
diabetes, obesity, and other CVD risk factors and, to the
extent that genetic factors contribute to susceptibility to
these diseases, American Indians also may have high
frequencies of alleles that contribute to disease risk. We have
312 North et al.
TABLE 8. Proportion of variance due to genes for selected cardiovascular disease risk factors
from analysis of American Indians from the Strong Heart Family Study, 1998–1999
Category of phenotype and
phenotype
Proportion of variance
due to genes*
Proportion of remaining variance due to
genes when additionally adjusted for
BMI*,†
%
SE†
%
SE
Weight
51
0.07
NA†
BMI
44
0.07
NA
WHR†
54
0.07
44
0.07
FM†
52
0.06
53
0.06
FFM†
53
0.07
46
0.07
LDL-C†
39
0.06
39
0.06
HDL-C†
50
0.07
44
0.07
Total cholesterol
39
0.06
39
0.06
ln† triglyceride
40
0.07
36
0.07
ApoA-1†
39
0.07
34
0.07
ApoB†
34
0.07
33
0.07
Lp(a)†
51
0.09
51
0.09
VLDLC†
45
0.09
48
0.09
VLDTG†
41
0.10
43
0.10
SBP†
23
0.06
20
0.06
DBP†
34
0.07
30
0.07
Fasting glucose
29
0.08
17
0.08
ln† insulin
44
0.08
36
0.08
ln fibrinogen
23
0.07
16
0.07
PAI-1†
26
0.06
25
0.06
Obesity
Lipids‡
Blood pressure§
Diabetes¶
Clotting factors
* All heritabilities significant (p ≤ 0.0001), except for fibrinogen and fasting glucose when additionally
adjusted for body mass index (p ≤ 0.001).
† BMI, body mass index; SE, standard error; NA, not applicable; WHR, waist/hip ratio; FM, body fat
mass; FFM, fat-free mass; LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein
cholesterol; ln, natural log transformed; ApoA-1, apolipoprotein A-1; ApoB, apolipoprotein B; Lp(a),
lipoprotein a; VLDLC, very low density lipoprotein cholesterol; VLDTG, very low density lipoprotein
triglycerides; SBP, systolic blood pressure; DBP, diastolic blood pressure; PAI-1, plasminogen activator
inhibitor 1.
‡ Individuals taking cholesterol-lowering medications were not included in these analyses.
§ Individuals taking antihypertensive medications were not included in these analyses.
¶ Individuals taking diabetes-related medications were not included in these analyses. For all
phenotypes, any individual value greater than 4 standard deviations from the mean was removed from
analysis.
both an opportunity and an obligation to try to identify these
susceptibility alleles.
The extended family structures within our sample enabled
the estimation of common environmental effects, independent of additive genetic effects. Household and tribal
membership effects are unmeasured nongenetic factors that
are shared more closely by members of the same household
or tribe than by individuals in different households or tribes
and may represent unmeasured dietary or lifestyle factors
(25). In this study, shared household environment and shared
tribal membership had no significant effect on CVD risk
factors (data not shown). Thus, neither measure of common
environment (household and tribal membership) improved
the fit of the variance component models. Given these findings, it may be that our measure of shared environment,
while appealing in its simplicity, is imprecise. In future analAm J Epidemiol 2003;157:303–314
Cardiovascular Disease in American Indians 313
yses, more sophisticated models of shared environment can
be evaluated, such as shared sibship environment and
common parent-offspring environment (26).
Variation attributable to measured covariates was substantial for many of the traits (ranging from 1 percent to 50
percent) (table 7). Unfortunately, covariate data were incomplete for three possibly important covariates: nutrition, physical activity, and amount of smoking per day or year.
Nonetheless, several of the probable concomitants of CVD
risk factors were identified in this population (i.e., age, sex,
center, the use of reproductive hormones, current and/or ever
smoking status, and current and/or ever alcohol consumption).
Heritabilities were estimated after accounting for covariate
effects (table 8). The estimation of the genetic component of
variance was limited to that attributable to additive effects
and may result from actions of more than one gene. If other
nonadditive sources of genetic variation exist, such as dominance or epistasis, then these observed heritabilities will
represent lower bounds. Therefore, these estimates were
conservative. In any case, the heritability estimates reported
in table 8 demonstrate that genetic effects explain a substantial proportion of the variability for many CVD risk factors
and related phenotypes and that these heritabilities are large
enough to warrant further research. Moreover, the magnitude of these heritabilities was demonstrated by the minimal
reduction of the genetic signal when additionally adjusted
for body mass index.
Direct comparison of the heritability estimates from this
study with those obtained from other studies is problematic.
Different study designs, ascertainment schemes, methods of
parameter estimation, and population-specific environmental contributions to the phenotypic variance can affect
heritability estimates, resulting in different heritabilities
even when the genetic variance estimates in the different
populations are similar (25). Additionally, similar heritability estimates for a phenotype in different populations do
not constitute evidence for the same genes in the expression
of a trait, nor do dissimilar heritability estimates constitute
evidence for the exclusion of the same genes in the expression of a trait (27). Despite these caveats, it is worth noting
that the heritability estimates from this study are in the range
of commonly reported heritability estimates from other
populations (25, 28–30). It is not surprising that for some of
the phenotypes (e.g., lipoprotein a and glucose) the proportion of variance that is explained by genetic variation is
reduced, since American Indian populations may be more
genetically homogeneous than are larger population groups.
Nonetheless, our results are comparable with those reported
among the Pima Indians; for example, Sakul et al. (31)
reported similar heritabilities for obesity measures (e.g.,
body mass index = 49 percent), and Hanson et al. (32)
reported a similar heritability for fasting insulin (36 percent).
We are now genotyping each family member for nearly
400 anonymous markers distributed throughout the genome.
A variance component linkage analysis of full pedigree data
will be applied to determine the chromosomal locations of
genes that influence disease risk factors (e.g., quantitative
trait loci). The identification and localization of quantitative
trait loci in American Indians will enable the examination of
several specific questions. For example, are there quantitaAm J Epidemiol 2003;157:303–314
tive trait loci that have large effects in explaining plasma
cholesterol levels in American Indians? Can these genes be
mapped to specific chromosomal regions? Such questions
are essential for the decomposition of the risk of CVD and/or
diabetes in the general population.
ACKNOWLEDGMENTS
Funding for this study was provided by a cooperative
agreement that includes grants U01 HL65520, U01
HL41642, U01 HL41652, U01 HL41654, and U01 HL65521
from the National Heart, Lung, and Blood Institute.
The authors would first like to thank the American Indian
participants in the Strong Heart Family Study. Without their
participation, this project would not have been possible. In
addition, the cooperation of the Indian Health Service hospitals and the directors of the Strong Heart Study clinics, Betty
Jarvis, Marcia O’Leary, and Dr. Tauqeer Ali, and the many
collaborators and staff of the Strong Heart Study have made
this project possible. The authors would also like to thank
Drs. John Blangero, Tony Comuzzie, and Jeff Williams for
assistance with analytic approaches; Michael Crawford,
David Frayer, Jeff Gilger, and Jim Mielke for their helpful
suggestions on portions of these analyses; and Lisa Martin
for her editorial assistance.
REFERENCES
1. Nelson RG, Sievers ML, Knowler WC, et al. Low incidence of
fatal coronary heart disease in Pima Indians despite high prevalence of non-insulin-dependent diabetes. Circulation 1990;81:
987–95.
2. Indian Health Service. Trends in Indian health, 1997. Washington, DC: Division of Program Statistics, Office of Planning,
Education, and Legislation, Indian Health Service, US Department of Health and Human Services, 1998.
3. Howard BV, Lee ET, Cowan L, et al. Coronary heart disease
prevalence and its relation to heart disease in American Indians:
the Strong Heart Study. Am J Epidemiol 1995;142:254–68.
4. Howard BV, Lee ET, Cowan LD, et al. Rising tide of cardiovascular disease in American Indians. The Strong Heart Study.
Circulation 1999;99:2389–95.
5. Lee ET, Cowan LD, Welty TK, et al. All-cause mortality and
cardiovascular disease mortality in three American Indian populations, aged 45–74 years, 1984–88: the Strong Heart Study.
Am J Epidemiol 1998;147:995–1008.
6. Welty TK, Rhoades DA, Yeh F, et al. Changes in cardiovascular disease risk factors among American Indians. The Strong
Heart Study. Ann Epidemiol 2002;12:97–106.
7. American Heart Association. 1998 heart and stroke statistical
update. Dallas, TX: American Heart Association, 1998.
8. Kromhout D, Menotti A, Kesteloot H, et al. Prevention of coronary heart disease by diet and lifestyle: evidence from prospective, cross-cultural, cohort, and intervention studies.
Circulation 2002;105:893–8.
9. Lee ET, Welty TK, Fabsitz RR, et al. The Strong Heart Study:
a study of cardiovascular disease in American Indians: design
and methods. Am J Epidemiol 1990;132:1141–55.
10. Lee ET, Howard BV, Savage PJ, et al. Diabetes and impaired
glucose tolerance in three American Indian populations aged
314 North et al.
45–74 years. Diabetes Care 1995;18:599–610.
11. Sideli RV, Friedman C. Validating patient names in an integrated clinical information system. In: Clayton PD, ed. Fifteenth annual symposium on computer applications in medical
care: November 17–20, 1991, Sheraton Washington Hotel,
Washington, DC. New York, NY: McGraw Hill, 1992:588–92.
12. Mamelka PM, Dyke B, MacCluer JW, et al. Pedigree/Draw for
the apple Macintosh. 2nd ed. San Antonio, TX: Southwest
Foundation for Biomedical Research, 1993. (Population Genetics Laboratory technical report no. 1).
13. Liedtke RJ. Principles of bioelectric impedance analysis. April
1, 1997. (www.rjlsystems.com/research/bia-principles.html).
14. Alberti KG, Zimmet PZ. Definition, diagnosis, and classification of diabetes mellitus and its complications. Part 1: diagnosis
and classification of diabetes mellitus provisional report of a
WHO consultation. Diabet Med 1998;15:539–53.
15. Roman MJ, Saba PS, Pini R, et al. Parallel cardiac and vascular
adaptation in hypertension. Circulation 1992;86:1909–18.
16. Roman MJ, Pini R, Pickering TG, et al. Non-invasive measurements of arterial compliance in hypertensive compared with
normotensive adults. J Hypertens Suppl 1992;10:S115–18.
17. Dyke B. PEDSYS: a pedigree data management software. San
Antonio, TX: Department of Genetics, Southwest Foundation
for Biomedical Research, 1994.
18. Almasy L, Blangero J. Multipoint quantitative-trait linkage
analysis in general pedigrees. Am J Hum Genet 1998;62:1198–
211.
19. Falconer DS. Introduction to quantitative genetics. 3rd ed.
Essex, United Kingdom: Longman, 1989.
20. Amos CI. Robust variance-components approach for assessing
genetic linkage in pedigrees. Am J Hum Genet 1994;54:535–
43.
21. Self SG, Liang KY. Asymptotic properties of maximum likelihood ratio tests under nonstandard conditions. J Am Stat Assoc
1987;82:605–10.
22. Botstein D, White RL, Skolnick M, et al. Construction of a
genetic linkage map in man using restriction fragment length
polymorphisms. Am J Hum Genet 1980;32:314–31.
23. Amos CI, Zhu DK, Boerwinkle E. Assessing genetic linkage
and association with robust components of variance
approaches. Ann Hum Genet 1996;60:143–60.
24. Williams JT, Duggirala R, Blangero J. Statistical properties of
a variance-components method for quantitative trait linkage
analysis in nuclear families and extended pedigrees. Genet Epidemiol 1997;14:1065–70.
25. Mitchell BD, Kammerer CM, Blangero J, et al. Genetic and
environmental contributions to cardiovascular risk factors in
Mexican Americans: the San Antonio Family Heart Study.
Circulation 1996;94:2159–70.
26. Astemborski JA, Beaty TH, Cohen BH. Variance components
analysis of forced expiration in families. Am J Med Genet
1985;21:741–53.
27. Mahaney MC, Blangero J, Comuzzie AG, et al. Plasma HDL
cholesterol, triglycerides, and adiposity: a quantitative test of
the cojoint trait hypothesis in the San Antonio Family Heart
Study. Circulation 1995;92:3240–8.
28. Duggirala R, Villapando CG, Oleary DH, et al. Genetic basis of
variation in carotid artery wall thickness. Stroke 1996;27:833–
7.
29. Hong Y, Pedersen NL, Brismar K, et al. Genetic and environmental architecture of features of insulin resistance syndrome.
Am J Hum Genet 1997;60:143–52.
30. Williams RR, Hunt SC, Hopkins PN, et al. Genetic basis of
familial dyslipidemia and hypertension: 15-year results from
Utah. Am J Hypertens 1993;6:319S–27S.
31. Sakul H, Pratley R, Cardon L, et al. Familiarity of physical and
metabolic characteristics that predict the development of noninsulin-dependent diabetes mellitus in Pima Indians. Am J Hum
Genet 1997;60:651–6.
32. Hanson RL, Imperatore G, Venkat Narayan KM, et al. Family
and genetic studies of indices of insulin sensitivity and insulin
secretion in Pima Indians. Diabetes Metab Res Rev 2001;17:
296–303.
Am J Epidemiol 2003;157:303–314