Segregation analysis of blood pressure and body mass index in a rural US community
Nath, Swapan K
Abstract To assess evidence for a gene with large effect on systolic blood pressure (SBP), diastolic blood pressure (DBP), and body mass index (BMI), we conducted segregation analyses on 261 nuclear families collected from a rural Caucasian community in Michigan. The families were ascertained through a hypertensive proband. Each phenotype was adjusted for significant covariate effects (e.g., gender and age). We used class D regressive models to conduct the segregation analyses. Our analysis results support the segregation of a major gene for BMI, but not for SBP or DBP A recessive locus effect provided the best explanation for BMI where -43% of the variance of BMI was due to this gene.
KEY WORDS: SEGREGATION ANALYSIS, COMMINGLING ANALYSIS, BLOOD PRESSURE, BODY MASS INDEX
Elevated or high blood pressure (HBP) and obesity (OB) are significant risk factors for cardiovascular disease, diabetes, and essential hypertension. They are also the leading causes of morbidity and mortality in Western societies (McGill 1996). An estimated 55% of adults are classified as overweight, and obesity is associated with increased morbidity and mortality from several chronic diseases (Dom et al. 1997; Taubes 1998). Blood pressure and obesity level are complex quantitative traits, thought to be influenced by a variety of physiological, behavioral, and sociocultural factors (Ward 1983; Ward 1990). As a result, the identification of susceptibility genes that contribute to HBP and OB will be difficult. We have undertaken segregation analyses to determine the likelihood that a gene with large effect influences blood pressure and obesity level. The analyses can be seen as a precursor to the actual identification of genes by means of linkage and/or association analyses. It is also important to assess the degree to which any single factor may be contributing to a complex trait such as blood pressure and obesity for risk assessment and clinical epidemiology practices.
Numerous studies have been conducted in an attempt to understand the genetic architecture and epidemiology of HBP and OB. However, many of these studies have been inconclusive, and even contradictory. For example, Rice et. al. (1990) used segregation analysis in French Canadian nuclear families and did not find evidence of a major gene effect for systolic blood pressure (SBP) or diastolic blood pressure (DBP). Weissbecker (1993) could not find evidence of either a major gene or polygenic effect on DBP level in a large family. Majumder et al. (1995) compared hypertension prevalence in two populations from India, and found that the observed familial resemblance of blood pressure levels in the two populations was primarily due to cultural rather than genetic inheritance. Perusse et al. (1991) found evidence for a major gene effect for SBP. Schork et al. (1994) studied the contribution of pleiotropy to variation in blood pressure and body mass index (BMI) in a large population-based sample and reported evidence that blood pressure and BMI seem to be influenced to some degree by a few common genes. Cheng et al. (1998) studied an Israeli population and found evidence for a major gene segregation for SBP level, with a possible pleiotropic effect BMI as well. Livshits et al. (1999) conducted segregation analyses on two ethnically distinct populations from Chuvasha and Turkmenia of the former Soviet Union and found evidence for a major gene influence on SBP and DBP after “sub-sampling” of pedigrees. Ginsburg et al. (1998) also found evidence for a major gene segregating in BMI across five ethnically different populations. Finally, Colilla et al. (2000) found evidence for a major gene for BMI in African American families.
These conflicting findings are difficult to reconcile. We, therefore, undertook a large-scale study to assess evidence for major gene influences on blood pressure and obesity level in a very well characterized Caucasian population within the United States.
Proband Selection. Probands for this study were identified by two basic strategies. First, probands were selected from families who participated in Rounds 1-3 of the Tecumseh Community Health Study (TCHS) from 1958 to 1968. If at least one parent’s blood pressure was >140 mmHg and/or >90 mmHg in any of the first three rounds, an offspring from such a family would be eligible for consideration as a proband. In order to qualify as a proband, however, subjects had to be between the ages of 18-45 years old, untreated for high blood pressure, and in the upper 15% percentile for age- and gender-specific blood pressure. Probands recruited in this way were considered as being from a “high blood pressure” family. Further, if potential probands had participated in Rounds 1, 2, or 3 of the TCHS as youths and had a blood pressure in the upper 25th percentile for their age/gender distribution at that time, they would also be eligible to participate as probands. Once a proband was identified, all first-degree relatives (parents and siblings) were recruited without regard to age, blood pressure status, or treatment for hypertension.
Data Description. A total of 261 nuclear families were collected in the manner described above. Only full siblings and parents were considered in the analyses. Half-sibs and multiple spouses were discarded from our analyses. Data were collected on several variables: anthropometric, hemodynamic, family history, medication use, and demographics. BMI was derived by dividing the body weight in kilograms by the square of height in meters.
Blood Pressure Protocol. Technicians measuring BP underwent audiometric testing and then were trained in auscultatory blood pressure technique by means of a videotape instruction guide and a video test. Criteria for a passing score on the test included a minimum of 80% of correct readings ( 3 mmHg); readings could not be systematically high or low, had to be reproducible, and could not show digit preference. After successfully passing the video test, each technician was individually evaluated by a trained, certified blood pressure technician using a double-headed stethoscope. In order to pass, a trainee had to have blood pressure readings within 3 mmHg of the supervisor’s measurements. Finally, the technique of blood pressure measurement with the Dinamap device was compared to that with the auscultatory technique and readings had to agree to within +8 mmHg.
This study included two participating field centers. At both centers a clinical examination took place on participants who were in a nonfasting state. At the time of the measurement, the participant emptied his/her bladder and remained seated quietly for 5 minutes. A cuff of appropriate size was chosen by measuring the circumference of the participant’s right arm at the midpoint between the acromion and the olecranon. A 30-sec radial pulse was taken, and with the antecubital fossa at heart level, an initial systolic pulse disappearance level (i.e., occlusion pressure) was measured. Two manual measurements were then taken with a standard mercury manometer, followed by two measurements with each of two automated BP devices (Omron and Dinamap) (Cooper et. al. 1997) in fixed order. Recertification through use of the double-headed stethoscope and an electronic device was conducted at regular intervals using videotapes.
Data Adjustment. Sampled individuals were divided into four groups of relatives: fathers, mothers, sons, and daughters. Phenotypic values were then adjusted for the effects of age and relative type through a stepwise multiple regression procedure using SAS (SAS Institute 1997) prior to performing segregation analyses. Phenotype values were regressed on age, age ^sup 2^, and relative type, as well as their interaction terms. Only statistically significant (i.e., at the 5% level) terms were retained. These regressions accounted for 30%, 17%, and 7% of the variance of SBP, DBP, and BMI, respectively. The ‘residuals’ from these regressions were standardized to a mean of 0 and variance of 1 and used for subsequent segregation analysis. In addition, to reduce the possibility that the evidence for a major gene could be due to the presence of the extreme outliers, all isolated and outlying residual values greater than 4 standard deviations from the mean were excluded from analysis. Five individuals were excluded from the analyses.
Ascertainment Correction. Since the families were ascertained through hypertensive probands, an ascertainment correction was applied in the segregation analysis. This involved conditioning each family’s likelihood on the blood pressure phenotype of the proband (Thomson and Cannings 1980). While the underlying genetic relationship between blood pressure and BMI is still unclear, the phenotypes have been found to be associated with each other (Dyer and Elliot 1989; Schork et al. 1994). In our data, the correlation between SBP and BMI was 0.24 (p
Commingling Analysis. Commingling analyses were performed prior to segregation analyses to test if a mixture of two or three age- and gender-adjusted normal distributions fit the standardized SBP, DBP, and BMI values better than a single normal distribution (Day 1969; Schork et al. 1996). To accomplish this, we fit to the data three models: (1) a single normal distribution, consistent with no major gene effect on the trait; (2) a mixture of two normal distributions, consistent with dominant or recessive alleles at a two-allele locus; and (3) a mixture of three normal distributions, consistent with a codominant allele at a two-allele locus. These models assume no genetic transmission from one generation to the next, as well as complete homogeneity across generations. Further, we also tested for skewness in the distributions, as evidence for multiple distributions in any variable could reflect skewness in a single distribution, and not result from a mixture (Schork et al. 1994). The parameters estimated in the various models included three genotype-specific means, (mu)^sub AA^, (mu)^sub Aa^, and (mu)^sub aa^, and a common variance, sigma^sup 2^ term, and residual familial correlations for father-mother pairs (pFM), parent-offspring pairs (ppo), and sibling pairs (pss).
Segregation Analysis. To investigate evidence for a major gene influencing phenotypic variation in SBP, DBP, and BMI, we performed complex segregation analysis (Elston and Stewart 1971). The basic segregation analysis model postulates that the observed distribution of a quantitative trait can be expressed as the sum of the following components: (1) a single genetic or nontransmitted environmental factor with a major effect on the respective trait; (2) the small additive allelic effects of a large number of independent polygenic loci; and (3) a random, nontransmissible effect. The polygenic and the random environmental effects were assumed to be distributed normally. The major gene effect was modeled as segregation of two alternate alleles, `’A’ and ‘a’, with the ‘A’ allele inducing elevated trait values and, the ‘a’ allele low trait values, respectively. There are thus three postulated corresponding genotypes (AA, Aa, and aa). The model permits the estimation of the following parameters: qA, the relative frequency of high trait value (A) allele frequency; three transmission probabilities reflecting the probability that a person with one of the three genotypes will transmit allele A (T^sub AA^, T^sub Aa^, and tau,); three genotype means (AA, AA, and aa); a common variance sigma^sup 2^; and residual familial correlations for father-mother pairs (pFM), parent-offspring pairs (ppo), and sibling pairs (pss).
Parameter estimation and tests of competing hypotheses were performed by maximizing the likelihood of the data under an unrestricted general model and also under several submodels specified by imposing certain constraints on the general model. A model with no transmission of a major gene fixes the transmission parameters equal to the (A) allele frequency (,AA= tAp = tauaa = qA). A ‘mixed’ Mendelian model with residual familial correlations fixes the transmission parameters to Mendelian expectations (rAA = 1, TAG = 0.5, and ra = 0).
For assessing the mode of inheritance of a major gene, certain parameters were constrained as well, thus, a constraint of uAA = pA. was made for the dominant model; uAa = utd for the recessive model; and uAa = 1/2(uAA + Uaa.) for a purely additive model. These models were compared with the more general codominant model in which all parameters are unconstrained. The phenotypic variance attributed to the major gene effect was calculated by pi(l – Tc)(IAA _ [taG)2′ where n is the proportion of individuals in one of the distributions. Maximum likelihood estimates were obtained to calculate these quantities, with the substitution n = (1 – qA) (Heiba et al. 1994).
Our segregation analysis took advantage of the class D regressive models (Bonney 1984), which have been shown to be robust against false detection of a major gene (Demenais and Bonney 1989), using the REGC computer module implemented in the S.A.G.E. (2000) computer program.
We also tested the null hypothesis that familial correlations without a major gene effect explain trait variation against a model that includes a major gene and residual familial correlations. Two additional criteria concerning the parent-offspring transmission of the major effect are required to confirm the presence of a major gene: (1) failure to reject Mendelian transmission, and (2) rejection of the no parent-offspring transmission model when compared with the general transmission model. Using these criteria safeguards against spurious evidence for a major gene segregation (Demenais et. al. 1993).
Tests of hypotheses were carried out using the likelihood ratio test constructed as twice the difference between log likelihoods for nested models, and is asymptotically distributed as X2. The number of degrees of freedom for this xz statistic is equal to the difference in the number of independently estimated parameters between the two models.
Descriptive statistics for the study subjects are given in Table 1. Marked age differences exist across generations, and genders, as do differences in SBP, DBP, and BMI. Table 2 provides the results of the stepwise multiple regression analysis. It is evident from Table I that a significant effect of relative type (fathers, mothers, sons, and daughters), and age exists on all three traits. Furthermore, a quadratic age effect was also observed for all the traits.
The results of commingling analysis are presented in Table 3. A mixture of two distributions fit significantly better than a single normal distribution for both SBP and BMI (p
In the segregation analysis, we fit four models, including: (1) a general model with arbitrary transmission; (2) a completely homogeneous environmental model with no assumed major gene transmission (,=AA _ FAQ = TAa =aa = qA); (3) a polygenic model assuming no major gene; and (4) a mixed Mendelian model. HardyWeinberg equilibrium of the trait-influencing allele was assumed in all the models. The results of a segregation analysis for SBP, DBP, and BMI are shown in Tables 4-6, respectively. For SBP and DBP, all three nested submodels were rejected when compared to the general full model of arbitrary transmission. For BMI, while both environmental and polygenic models were rejected, the mixed Mendelian major gene model was not rejected when compared to the unrestricted general model. A model assessing pure Mendelian transmission with no familial correlations, however, was significantly different from the mixed Mendelian transmission model (p
Given evidence for a major gene effect on BMI, we tested the single major locus Mendelian model with different mode of inheritance. We compared the mixed Mendelian model with three nested submodels (i.e., models assuming a dominant, recessive and additive gene effect). Results are presented in Table 7. The recessive model was the best-fitting and most parsimonious model compared to dominant and additive models. The estimated allele frequency for the high
BMI allele (qA) was estimated to be about 63%. Our analysis has provided evidence for the presence of a major gene for BMI, and the variance due to this effect is about 43%. It is also evident from our results that there is evidence for a large spouse correlation under each model, which might be due to the effect of common familial environment or to assortative mating.
We used segregation analysis to examine evidence for major gene segregation in SBP, DBP, and BMI based on nuclear family data gathered from a Caucasian US population. We did not find evidence of a major gene effect for either SBP or DBP in our data. However, the limitations of the models used in segregation analysis do not necessarily preclude the possibility that there is an underlying major gene with large effect influencing SBP or DBP. Alternatively, there may be many genes contributing to SBP and DBP variation, each with relatively moderate effects (oligogeny) and interacting with other genes and environmental factors. Moreover, potential genetic heterogeneity due to marry-in founders (spouses) of different racial backgrounds may also confound the detection of major gene effect.
Our results, from both the commingling and segregation analyses, do offer evidence for major gene effect on BMI, with a significant amount of residual familial correlation due to other factors. The most likely mode of inheritance of this gene is recessive, and the variance attributable to this gene is about 43% of the total BMI variance. The frequency of the high BMI-associated allele at this locus was estimated to be about 63%.
Segregation analyses are not unproblematic. For example, the study by Perusse et al. (1991) showed that model misspecification could significantly affect the result of a segregation analysis. In addition, lack of power to detect genes with only moderate effects on a trait is also often a problem. Borecki et. al. (1995) found, by way of a simulation study, that it was not always possible to resolve transmission patterns in about 20%-70% of the cases they examined when the assumed model was recessive, and about 3%-15% of the time when the assumed model was dominant. Therefore, while tests on transmission probabilities can often reduce rates of false positive inference of a major gene, they may also lead to a failure to detect the presence of a major gene when one indeed exists, particularly under a recessive model. A more complex type of analysis such as multivariate modeling, e.g., multilocus segregation analysis, incorporating detailed information about common environmental covariates, which includes gene-environmental interactions with more family data, could be required (Blangero and Konigsberg 1991; Allison et al. 1998).
Despite the evidence for and against a major gene for BMI, SBP, and DBP, we should acknowledge two additional concerns about our analysis. First, it is well accepted that large pedigree analysis is the only available way to ascertain the mode of inheritance of most of human traits. Our study was based on small nuclear families, which might not show the mode of inheritance of a trait clearly. Second, the ascertainment correction for BMI may not have taken into account ascertainment properly. As mentioned earlier, our families were originally ascertained through a hypertensive proband. Although BMI is significantly correlated with SBP and DBP in our data, the underlying genetic relationship between BMI and blood pressure is still unclear.
In summary, we found strong evidence for a major gene effect on the transmission pattern of BMI among the families from a Caucasian population. Actually localizing this gene would be very important, especially because it could have strong predictive value, and could relate to cardiovascular disease risk. Our study should provide motivation to find this gene by means of genomewide linkage and association studies.
Acknowledgments S.K.N thanks Drs. R. C. Elston and C. E. Aston for helpful discussions. The data for this study was generated through the GenNet Network of the Family Blood Pressure Program (FBPP), sponsored by the National Heart, Lung and Blood Institute, through grant HL54998. We also thank two anonymous reviewers for their insightful comments.
Received 20 November 2000; revision received 3 August 2001.
1Arthritis and Immunology Research Program, Oklahoma Medical Research Foundation, Oklahoma City, OK.
2McKusick-Nathans Institute of Genetic Medicine, Johns Hopkins University School of Medicine, Baltimore, MD.
3Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, OH.
4Department of Preventive Medicine, Loyola University, Chicago, IL.
5Department of Medicine, University of Michigan Medical School, Ann Arbor, MI. Human Biology, February 2002, v. 74, no. 1, pp. 11-23.
Copyright 2002 Wayne State University Press, Detroit, Michigan 48201-1309
Allison, D.B., B. Thiel, P. St. Jean et al. 1998. Multiple phenotype modeling in gene-mapping studies of quantitative traits: Power advantage. Am. J. Hum. Genet. 63:1190-1201.
Blangero, J., and L.W. Konigsberg. 1991. Multivariate segregation analysis using the mixed model. Genet. Epidemiol. 8:299-316.
Bonney, G.E. 1984. On the statistical determination of the major gene mechanisms in continuous human traits: Regressive models. Am. J. Hum. Genet. 18:731-49.
Borecki, LB, M.A. Province, and D.C. Rao. 1995. Inferring a major gene for quantitative traits by using segregation analysis with tests on transmission probabilities: How often do we miss? Am. J. Hum. Genet. 56:319-326.
Cheng, L.S.-C., D. Carmelli, S.C. Hunt et al. 1995. Evidence for a major gene influencing 7-year increases in diastolic blood pressure with age. Am. J. Hum. Genet. 57:1169-1177.
Cheng, L.S.-C., G. Livshits, D. Carmelli et al. 1998. Segregation analysis reveals a major gene effect controlling systolic blood pressure and BMI in an Israeli population. Hum. Biol. 70:59-75.
Cooper, R.S., A. Puras, J. Tracy et al. 1997. Evaluation of an electronic blood pressure device for epidemiologic studies. Blood Pressure Monitoring 2:35-40.
Colilla, S., C. Totmi, R. Cooper et al. 2000. Genetic inheritance of body mass index in African-American and African families. Genet. Epidemiol. 18:360-376.
Day, N.E. 1969. Estimating the components of a mixture of normal distributions. Biometrika 56:463-474.
Demenais, F., and G.E. Bonney. 1989. Equivalence of the mixed and the regressive models for genetic analysis. I. Continuous traits. Genet. Epidemiol. 6:597-617.
Demenais, F., M. Martinez, and N. Andrieu. 1993. The transmission probability model is useful to prevent false inference. Am. J. Hum. Genet. 52:441-442.
Dom, J.M., E.F. Schisternman, W. Winkelstein et al. 1997. Body mass index and mortality in a general population sample of men and women. Am. J. Epidemiol. 146:919-931.
Dyer, A.R., and P. Elliott. 1989. The INTERSALT study: Relations of body mass index to blood pressure. J. Hum. Hypertens. 3:299-308.
Elston, R.C., and J. Stewart. 1971. A general model for the genetic analysis of pedigree data. Hum. Hered. 21:523-542.
Ginsburg, E., G. Livshits, K. Yakovenko et al. 1998 Major gene control of human body height, weight and BMI in five ethnically different populations. Ann. Hum. Genet. 62:307-322.
Heiba, LM., R.C. Elston, B.E.K. Klein et al. 1994. Sibling correlations and segregation analysis of age-related maculopathy: The Beaver Dam eye study. Genet. Epidemiol. 11:51-67.
Hunt, S.C., and R.R. Willium. 1994. Genetic factors in human hypertension. In Textbook of Hypertension, J.D. Swales, ed. Oxford, UK: Blackwell Scientific, 519-538.
Livshits, G., E. Ginsburg, and E. Kobylianski. 1999. Heterogeneity of ethnic control of blood pressure in ethnically different populations. Hum. Biol. 71:685-708.
Majumder, P.P., R.N. Das, and S. Nayak. 1995 Genetic epidemiology of blood pressure in two Indian populations: Some lessons. Hum. Biol. 67:827-842.
McGill, H.C. 1996. Overview. In Atherosclerosis and Coronary Artery Disease, Vol. 5, R. Fuster, R. Ross, and E.J. Topol, eds. Philadelphia, PA: Lippincott-Raven, 25-41.
Perusse, L., PP. Moll, and C.F. Singh. 1991 Evidence that a single with gender and age dependent effects influences systolic blood pressure determination in a population-based sample. Am. J. Hum. Genet. 49:94-105.
Rice, T., C. Bouchard, and LB. Borecki, 1990. Commingling and segregation analysis of blood pressure in a French-Canadian population. Am. J. Hum. Genet. 46:37-44.
S.A.G.E. 2000. Statistical analysis for genetic epidemiology, Release 3.1. Computer program package available from the Department of Epidemiology and Biostatistics, Rammelkamp center for education and research, MetroHealth Campus, Case Western Reserve University, Cleveland, OH.
Schork, N.J., A.B. Weder, M. Trevisan et al. 1994 The contribution of pleiotropy to blood pressure and body mass index variation: The Gubbio study. Am. J. Hum. Genet. 54:361-373.
Schork, N.J., D.B. Allison, and B. Thiel. 1996. Mixture distributions in human genetics research. Stat. Methods Med. Res. 5:155-178.
Taubes, G. 1998. As obesity rates rise, experts struggle to explain why. Science 280:1367-1368
Thompson, E.A., and C. Cannings. 1880. Sampling schemes and ascertainment. In The Genetic Analysis of Common Diseases: Applications to Predictive Factors in Coronary Heart Diseases, C.F. Sing and M. Skolnick, eds. New York, NY: Alan Liss.
Ward, R.H. 1983. Genetic and sociocultural component of high blood pressure. Am. J. Phys. Anthropol. 62:91-105
Ward, R.H. 1990. Familial aggregation and genetic epidemiology of blood pressure. In Hypertension: Pathophysiology, Diagnosis, and Management, J.H. Laragh and B.M. Brenner, eds. New York, N.Y: Raven, 81-100.
Weissbecker, K.A. 1993. Segregation analysis of diastolic blood pressure in a large pedigree. Genet. Epidemiol. 10:659-664.
SWAPAN K. NATH,’ ARAVINDA CHAKRAVARTI, CHIEN-HSIUN CHEN,’ RICHARD COOPER,4 ALAN WEDER,’ AND NICHOLAS J. SCHORK’
Copyright Wayne State University Press Feb 2002
Provided by ProQuest Information and Learning Company. All rights Reserved