Study population and sample design
Between December 2005 and January 2006, 119 families (1,712 individuals) were selected in Baependi, a city in a rural area (752 Km2, 18,072 inhabitants) located in Minas Gerais State, Brazil. Probands were identified from the community at large in several stages. First, eleven census districts (from a total of twelve) were selected for study. Second, residential addresses within each district were randomly selected (first by randomly selecting a street, second a household). Finally, eligibility criteria (any individual living in the selected household who was 18 years old or above) within each household were established.
Once a proband was enrolled, all his/her first-degree (eg, parents, siblings, and offspring), second-degree (eg, half-siblings, grandparents/grandchildren, aunts/uncles, nieces/nephews, and double cousins), and third-degree (eg, first cousins, great-uncles/great-aunts, and great-nephews/great-nieces) relatives and his/her respective spouse's relatives, who were at least 18 years old, were invited to participate. After the proband's first contact, first degree relatives were invited to participate by phone; these included all living relatives in the city of Baependi (urban and rural area) and surrounding cities. To recruit the participants, the project was advertised through provincial, religious, and municipal authorities, in local television, newspaper, and radio messages, through physicians, and by phone calls. For physical examination, a clinic was established in a quiet but easily accessible sector of Baependi. Only individuals age 18 and older were eligible to participate in the study.
Trait description
A questionnaire was administered to each participant to obtain information on family relationships, demographic characteristics, medical history, and environmental risk factors such as smoking habit, alcohol use, physical activity, and prescription drug use (e.g., anti-hypertensive, for diabetes, for dyslipidemia). Questionnaire was administered and filled out by research assistants specially trained for this task. The questionnaire was based on the WHO-MONICA epidemiological instrument and was previously used by our group in other epidemiological projects [9, 10].
Anthropometric measures such as weight, body height and waist circumference were measured following standardized procedures. Height was measured in centimeters and weight in kilograms using a calibrated digital balance. Body mass index (BMI) (weight in Kg/height in meters2) was calculated and overweight defined as BMI ≥ 25 Kg/m2 or obesity, BMI ≥ 30 Kg/m2. Waist circumference (WC) was measured half way between the lowest rib and the iliac crest while the subject was at minimal respiration. An individual was considered sedentary who worked seated or did not walk during work and who did not have any physical activity during leisure time. Smoking status was defined as "current smoker" when smoking has occurred during the last six months. Blood pressure was measured using a standard digital sphygmomanometer (OMRON, Brazil) on the left arm after 5 minutes rest, in the sitting position. Systolic and diastolic blood pressures were calculated from three readings (mean value of all measurements), with a minimal interval of 3 minutes. Physical examination and electrocardiogram were performed concurrently by trained medical students.
Fasting blood glucose, total cholesterol, lipoprotein fractions, and triglycerides were assayed by standard techniques in 12 hour fasting blood samples. Serum samples were stored at -80°C and genomic DNA was extracted following a standard salting-out procedure.
The study protocol was approved by the ethics committee of the Hospital das Clinicas, University of Sao Paulo, Brazil, and each subject provided informed written consent before participation.
Statistical Analysis
Descriptive statistics were calculated to describe familial structure of the Baependi data set. In addition, descriptive measures, such as mean, median, standard deviation, skewness and kurtosis, were calculated for all traits considering the total sample and gender stratification. When the normality assumption did not hold for a specific trait, natural log-transformation was applied followed by a new data assessment. After the trait transformations, when the residual kurtosis remained too high we tried to prevent biased heritability estimates using robust estimation implemented in SOLAR through the tdist procedure.
Polygenic heritability estimates were calculated, adjusted for age, gender and medication use, for each cardiovascular related trait, using the variance-components approach implemented in the SOLAR package [11]. Heritability was calculated as the proportion of the total phenotypic variance explained by additive genetic effects after accounting for covariates.
The overall aim of these analyses was to determine the extent to which unmeasured genetic factors, and measured environmental and lifestyle factors, contributed to variation in a large panel of cardiovascular-related traits. Information on the covariance among relatives was used to estimate the polygenic (or additive genetic) component of variance. The variance component model is a well-known tool for heritability estimates in family studies and is only briefly described here [11, 12]. Under this model, the level of the trait for individual I (denoted by y
i
) is described as follows:
where μ is the general mean of the trait, and β
j
is the regression coefficient for covariate j, which assumes the value X
ij
for individual i. Measures of the qualitative covariates (eg, female gender, diabetes, medication use, etc.) were scaled so that the regression coefficient represents the effect of having the covariate present compared with having it absent. The remaining parameters, g
i
and e
i
are the residual genetic effect due the polygenic term, and random error component, respectively. The random effects, g
i
and e
i
are assumed to be uncorrelated and normally distributed with mean zero and variance and respectively. As usual, the error component is unique to each individual, whereas the polygenic component is shared between individuals in proportion to their kinship coefficient. Thus, the covariance between traits for individuals i and i' is given by:
Parameter 2φ
ii
is the coefficient of relationship between individuals i and i
'. The likelihood of the traits of family members is assumed to follow a multivariate normal distribution. Estimates of the mean and variance components are obtained using maximum likelihood methods.
Firstly the variance component model was fitted without any covariate effects, denoted by unadjusted model (Model 1). Secondly two sets of covariate effects on each traits were considered: under Model 2 the covariates were sex, age, age2, and sex × age interaction; under Model 3, besides the covariates from Model 2, current medication use was also considered. The likelihood ratio test (LRT) was applied to test whether the additive polygenic effect in each analysis accounted for a significant component of the variation for the trait under study, after adjusting for the covariates. This test compares the likelihood of a full model (covariates, and additive polygenic and residual random components) with that of a nested model (covariates, and the residual component only). The LRT is asymptotically distributed as a mixture of , and .
The inclusion of covariates in the variance component model might influence the proportion of the phenotypic variance associated with the polygenic effect. Increased precision might be obtained if the heritability estimates increase with the covariate adjustments in comparison to the reduced or unadjusted model. But, when the heritability estimates decrease with the covariate inclusion, it means that the covariate explains a part of the familial relationship associated to the polygenic effect.
Risk factors were analyzed both as continuous and categorical variables. Dichotomized traits were defined according to the ATPIII criteria [13], as follows: truncal obesity was diagnosed when waist was more than 102 cm in men and more than 88 cm in women. High fasting glucose was defined when glucose levels were ≥110 mg/dL or with current use of anti-diabetic medication. The definition of low HDL cholesterol was considered when levels were < 40 mg/dL in men and < 50 mg/dL in women, high triglycerides ≥ 150 mg/dL, high LDL-c ≥ 130 mg/dL, high total cholesterol ≥ 200 mg/dL or use of lipid lowering medication. Hypertension was defined as mean systolic blood pressure of ≥ 140 mmHg and/or diastolic blood pressure of ≥ 90 mmHg or the current use of anti-hypertensive medication. Heritabilities for dichotomized traits were also estimated under the variance component model using SOLAR. Because the current medication information was absorbed into the trait's dichotomization criterion, Model 3 was not fitted to those traits.