Familial aggregation of blood pressure and anthropometric variables in patrilocal households.код для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 79305-311 (1989) Familial Aggregation of Blood Pressure and Anthropometric Variables in Patrilocal Households P.J. BYARD, B.N. MUKHERJEE, S.K. BHATTACHARYA, J.M. RUSSELL, AND D.C. RAO Departments of Pediatrics and Epidemiology and Biostatistics, Case Western Reserve University, Cleweland, Ohio 44106 (P.J.B.);Anthropometry & Human Genetics Unit, Indian Statistical Institute, Calcutta-35,India (B.N.M.,S.K.B.);Division of Biostatistics (J.M.R.,D.C.R.),and Departments of Psychiatry and Genetics (0.C.R.), Washington University School of Medicine, St. Louis, Missouri 63110 KEY WORDS folds Anthropometrics, Familial correlations, India, Skin- ABSTRACT Fifty-three households in a small Indian fishing community were surveyed for blood pressure, pulse rate, and anthropometric measurements (height, weight, and three skinfolds). I n addition to nuclear family relationships, correlations for extended family members and in-laws living within a common household were estimated by maximum likelihood. Based on likelihood ratio tests, the hypothesis that correlations among genetically unrelated pairs from the same household are zero is rejected for systolic blood pressure. Among genetically related individuals, the degree of relationship does not affect the magnitude of the blood pressure correlations. For the anthropometric measurements, family resemblance is significant only for first-degree relatives, except that the correlation for uncle-child pairs is significant for subscapular skinfold, and brother-in-law-sister-in-lawpairs resemble each other for height and weight. The results suggest that common household environment is a significant determinant of blood pressure but not fatness in this population. Although familial aggregation of hypertension and obesity is well established in western populations (Acheson and Fowler, 1967; Annest et al., 1979; Biron et al., 1976; Hartz et al., 1977; Havlik et al., 1979;Johnson et al., 1965; Miall and Oldham, 1963), the effect of household environment on normal variation in blood pressure and fatness levels has not been demonstrated systematically. In western populations, nuclear families living in a single household share both genes and environment, and with the exception of spouse pairs, twins, or adoptions, there are no family relationships which allow examination of the role of household environment independently of shared genes. I n cultures which promote alternative types of household compositions, it is possible to study the effects of shared household environment on several types of genetically unrelated individuals. Here we describe one such arrangement found in patrilocal societies, where @ 1989 ALAN R. LISS, INC. wives take up residence with their husbands and their extended families. MATERIALS AND METHODS Population Data were collected in a small fishing community located in the state of West Bengal, India, about 180 km southwest of Calcutta. The respondents comprise a nearly complete sampling of household inhabitants old enough to participate (generally, above 6 years of age). A total of 413 people in 53 households were measured for blood pressure, pulse rate, height, weight, and three skinfolds (biceps, triceps, and subscapular). Details of measurement technique have been reported previously (Mukherjee et al., 1988). Figure 1presents a typical household structure found in this group. The core of the Received March 21, 1988; revision accepted September 22, 1988. 306 P.J. BYARD ET AL. Fig. 1. Pedigree showing typical household composition. Males are shown as squares, females as circles. This pedigree shows two brothers and their wives in the middle generation, their mother and deceased father in the top generation, and their unmarried sons and daughters in the bottom generation. household consists of a group of brothers, their mother, more rarely their father, their wives, sons, and unmarried daughters. The greater representation of females in the older generation is due either to greater longevity in women or to age disparity among spouses at marriage (wives were much younger than their husbands, especially in earlier times). Since most women marry and join the households of their husbands, the adult women in a household are more often related to other household members by marriage, not by blood. Familial correlations Although many different types of relationships are found in these households, we limited analysis to relationships with significant representation. Thus, ten familial correlations were estimated for each variable: marital, parent-child, sibling, grandparentchild, cognate uncle-nephewhiece (genetically related), affinal aunt-nephewhiece (related by marriage), first cousins, parent in law-daughter in law, brother in law-sister in law, and wives of brothers. Before estimating familial correlations, measurements were adjusted for age and sex (Mukherjee et al., 1988). This was done with stepwise multiple regression of each variable on age, age2, and age3 within each of four groups: adult males (220 years), adult females (220 years), young males (<20), and young females (<20). The standardized residuals were then normalized by assigning Blom scores (Blom, 1958). Normalization was done separately within each age-sex group to approximate a univariate normal distribution. After age adjustment in this way, there is no residual correlation between any of the anthropometric or blood pressure variables and age (r ranges from -0.006 to 0.006, P > 0.8). Correlations among household members were estimated by maximum likelihood, using a general purpose algorithm called MLECOR (Rao et al., 1987) that, in turn, uses the numerical optimization subroutine GEMINI (Lalouel, 1979).A FORTRAN program called PEDCOR was used to input pedigree structure. The PEDCOR program accepts any arbitrary matrix of familial or other relationships, thus allowing estimation of correlations for any type of genetic or non-genetic relationship. Hypothesis testing For each variable, we first estimated all 10 correlations listed above, calling this the “general” model. In subsequent runs, we set certain correlations equal to each other or fixed them at zero in order to test hypotheses about household resemblance. Statistical significance was evaluated by likelihood ratio criterion. The log-likelihood value obtained under the reduced model is subtracted from the value obtained under the general model, and doubled to yield a x2 statistic. The test statistic has been shown to be distributed asymptotically as a x2 with 10-k degrees of freedom (df), where k is the number of correlations being estimated in the reduced set (McGue et al., 1984). Several simple hypotheses about the effects of household environment can be tested in this way. If common household environment is the primary determinant of resemblance, then we would expect all members of the household to be similar, and all 10 correlations could be set equal. Thus, we estimate a single correlation coefficient for all relationships, and test the hypothesis using a x2 with 10 - 1 = 9 df. Rejection of this hypothesis need not, however, implicate a genetic effect, since the household environment may have different effects on different classes of relatives. As a further test of the hypothesis of no genetic effect, we set all the first-, second-, and third-degree genetic relationships equal, BLOOD PRESSURE AND ANTHROPOMETRICS 307 giving rise to a single parameter, all the nongenetic relationships equal to a second parameter, and the marital correlation to a third parameter, and tested the fit of this model against that obtained with the general model. This hypothesis is tested using a x2 with 10 3 = 7 df. Finally, if the genetic effect is the only determinant of family resemblance, we could fit a model in which all the nongenetic correlations (except marital, which may be affected by phenotypic assortment as well a s by common household environment) are fixed a t zero. This is tested using a x2 with 10 - 6 = 4 df. It should be pointed out that this is not a specific test for no household environmental effect, as the genetic relationships also share such effects. While rejection of this hypothesis implies significant household environmental effect, failure to reject should not be interpreted as a demonstration of genetic effect. Although a number of more complex models could be tested with this method, they would be unlikely to yield much more information than could be obtained through examination of the familial correlations and their standard errors (Table 1). Some testing of specific hypothesis, however, guards against the problem of multiple comparisons when a large number of familial correlations is presented. RESULTS As shown in Figure 2, this sample is considerably leaner and has lower median systolic blood pressure than NCHS standards for white Americans. In addition, fatness and blood pressure levels are less variable in this sample than in western groups, and the effect of fatness level on blood pressure is minimal (Mukherjee et al., 1988). Estimates of correlations among household members, with their standard errors and estimated sample sizes, are presented in Table 1. Using twice the standard error as the 95%confidence limit, marital correlations are significantly different from zero only for height and diastolic blood pressure. Correlations for first-degree relatives (parent-child and sibling pairs) are significant for all variables. The grandparent-child correlations are significant only for diastolic blood pressure. Cognate uncle-child correlations are significant for subscapular skinfold and both blood pressure measurements, whereas cousins are significantly correlated for the blood pressure variables only. $- 1 -73 fia .3 P.J. BYARL) ET AL 308 14- lS01 140- h E E 12- m 3 10- h 5 130- V *- .v) 120- 0 C Y E110- 8- a 0 .- i 64- 25- 2; 20 10 50 40 30 m* --NcHsdard. 90 80 4 10 20 30 50 40 60 70 1507 A Sample U.dbn (Fomalu) A bmpb u.dbn (Femak.) x v) 57 2 10 I . I I 20 30 Age in Years . . 40 I I 8 0 - r , , , 50 10 20 , , 30 I , I 40 . 1 50 3 , 60 < 70 Age in Years Fig. 2. Median values of triceps skinfold and systolic blood pressure compared to NCHS standards. Among the nongenetic in-law relationships, there are few significant correlations. None of the affine aunt-child and or parent-in-lawdaughter-in-lawcorrelations is significantly different from zero. Brother-in-law-sister-inlaw pairs are significantly correlated for height and weight, with borderline statistical significance. Wives of brothers have a significant correlation coefficient for diastolic blood pressure only. Table 2 presents the results of likelihood ratio tests for the reduced hypotheses. The hypothesis that all ten household correlations are equal (hypothesis 1, the household environment model) is rejected for every variable except diastolic blood pressure. A model allowing one correlation for all genetically related individuals, one for spouse pairs, and a third for all genetically unrelated pairs (hypothesis 2, the three-correlation model) is rejected for height, weight, biceps, and triceps skinfold, for which genetic effects are there- fore implicated, but not for subscapular skinfold or blood pressure, for which household environment appears to be the major determinant. The hypothesis that all the nongenetic correlations (except marital) are zero TABLE 2. ,yz values for likelihood ratio tests of hypothesis' Hypothesis Nos. Variable Height Weight Biceps skinfold Triceps skinfold Subscapular skinfold Systolic blood pressure Diastolic blood pressure 1 2 3 26.17* 26.31* 23.50* 29.16' 25.01* 27.35* 13.66 23.22* 17.44' 25.11* 14.51* 9.78 12.03 4.33 6.24 6.73 1.85 7.56 6.00 10.47* 3.52 'No. 1: Household environment only (a11 ten Correlations equal). No. 2. Different household environment effects (three correlations: marital, genetic, and nongenetic). No. 3 Genetic effect only (nongenetic correlationsfixed at zero). * P C 0.05. RL001) PRESSURE AND ANTHROPOMETRICS (hypothesis 3, the genetic model) is rejected only for systolic blood pressure. DISCUSSION In this population, shared household environment appears to play a major role in family resemblance for systolic and diastolic blood pressure. This is reflected in the likelihood ratio tests, in that the household environment model (1) cannot be rejected for diastolic blood pressure, and the genetic model (3) can be rejected for systolic blood pressure. There is some suggestion of a household effect for subscapular skinfold, in that the three-correlation model (2), with all genetic relationships constrained to have equal correlations, cannot be rejected for that variable. This results primarily from the fact that resemblance for cognate uncle-child pairs (r = .29) is as high as the resemblance for first-degree relatives. However, caution must be used when interpreting such results, because failure to reject a hypothesis may, in part, be due to lack of sufficient power. This is especially true when sample size is small. Nonetheless, the effect of household environment on body build has been documented for several western samples of parents and their adopted children (Garn et al., 1979; Hartz et al., 1977), and it may be that differences between the results for subscapular skinfold and the other indicators of body composition in this sample occur because there is too little variation in fat deposition at sites other than subscapular in this lean population. The pattern of familial resemblance exhibited for blood pressure in this population does differ considerably from that observed in western populations, however. Chazan and Winklestein (1964) found household aggregation of hypertension in a group of American households, both in genetically related and unrelated individuals, but most studies have found no resemblance for normal blood pressure variation in genetically unrelated household members. Adopted children and their adoptive parents have nearzero correlations, both in the Framingham offspring study (Feinleib, 1979) and in the Quebec study of Biron et al. (1976). Since adopted children were born elsewhere and probably have not yet reached adulthood, the length of time spent with their adoptive parents is probably less than the length of cohabitation for some of the genetically unre- 309 lated or distantly related pairs used in the present analysis. In the Framingham offspring study, correlations between spouses of the propositi and their unrelated in laws are near zero (Feinleib, 1979). Of course, these in-laws live in completely separate households. Similar results in the present analysis suggest that either length of cohabitation has not been sufficient to have a n effect on in-law pairs, or that in-laws do not share enough common environment at the proper stage of the life cycle to have a significant effect, even when they occupy the same household. The marital correlation for diastolic blood pressure is comparable to the correlations for first-degree relatives in this populations. A similar result is reported by Tseng (1967)in a fishing community from Taiwan. Reports of marital correlations from western populations are conflicting. Johnson et al. (1965) found no significant correlation among spouse pairs for blood pressure at all. Others conclude that spouse resemblance for blood pressure remains even when several indices of household environment (and spouse similarity for them), such a s education, occupation, and dietary sodium intake, are controlled (Speers et al., 1986). Length of cohabitation for spouse pairs h a s been suggested as a n intervening factor, but length of cohabitation does not affect the marital correlation for blood pressure in the Framingham study (Feinleib, 1979; Sackett et al., 1975),or in Taiwanese agricultural or fishing communities (Tseng, 1967). In the present study, no accurate information on length of cohabitation is available, but the length of cohabitation for spouse pairs and in-law pairs should be comparable. If cohabitation is responsible for the spouse resemblance, one would expect in-law pairs to have significant correlations for blood pressure as well. On the other hand, assortative mating for blood pressure seems unlikely, since blood pressure is not detectable in normal social interactions. Although there is some marital resemblance for height in this group, there is none for weight or skinfolds, and phenotypic assortment for height alone cannot account for the marital resemblance for blood pressure. One explanation for the discrepancy in marital and in-law resemblance is that household environment is shared unequally by different types of relatives. This hypothesis would require detailed study of household 310 P.J. BYARU ET AL structure and day-to-day functioning for and genetic and cultural heritabilities are useful primarily for examining the possible verification. The fact that the wives-of-brothers correla- effects of shared environment on family tion for systolic blood pressure is significant, resemblance for modifiable disease risk facwhereas the other in-law correlations are not, tors, not for establishing rigid standards for requires some comment. The possibility of the effect of household environment on a trait genetic relatedness among women marrying worldwide. into a particular sibship was examined, and ACKNOWLEDGMENTS two pairs of related wives were excluded from This research was supported in part by PHS analysis. While the possibility of remote, unknown relatedness exists, it would be grants GM28719, HL33973, and MH31302. unlikely to have a detectable effect on the LITERATURE CITED correlation coefficient. Instead, the possibility that individuals who share the same gen- Acheson RM, Fowler GB (1967) On the inheritance of stature and blood pressure. J . Chronic Dis. 20:731-745. eration, status, and sex role expectations JL, Sing CF, Biron P, and Mongeau J G (1979) may react physiologically in similar ways Annest Familial aggregation of blood pressure and weight in must be entertained. It should be noted that adoptive families. I. Estimation of the relative contributions of genetic and common environmental factors to the similarity among wives of brothers for blood pressure correlations among family members. blood pressure does not appear to be mediated Am. J . Epidemiol. 110:492-503. by any similarity in weight or adiposity, Biron P, Mongeau J, Bertand D (1976) Familial aggregasince their correlations are near zero for skintion of blood pressure in 558 adopted children. Can. Med. folds and negative for weight. In fact, there is Assoc. J . 115773-774. no consistent relationship between family Blom G (1958) Statistical Estimates and Transformed Beta Variables. New York John Wiley and Sons. resemblance for body build and family resemJA, Winklestein W, J r (1964) Household aggregablance for blood pressure in this population, Chazan tion of hypertension: Report of a preliminary study. J. suggesting that the resemblance for blood Chronic Dis. 17:9-18. pressure is not mediated by genetic or envi- Feinleib M (1979) Genetics and familial aggregation of blood pressure. I n G Onesti and C Klimt (eds.): New ronmental control of body build. The correlaYork: Grume and Stratton, pp. 35-48. tion in individuals of blood pressure and SM, Cole PE, and Bailey SM 91979) Living together body build is low in this population, espe- Garn a s a factor in family-line resemblance. Hum. Biol. cially in men (Mukhejee et al., 1988). 51565-587. In conclusion, studies on nuclear families Hartz A, Giefer E, and Rimm AA (1977) Relative importance of the effect of family environment and heredity and adopted children in western populations on obesity. Ann. Hum. Genet. 41:185-193. have provided conflicting information regardHavlik RJ, Garrison RJ, Feinleib M, Padgett S, Castelli ing the effect of household environment on WP, and McNamara P (1979) Blood pressure aggregavariation in blood pressure and fatness levtion in families. Am. J. Epidemiol. 110:304-312. els. The influence of household environment Johnson BC, Epstein FH, and Kjelsberg MO (1965) Distributions and familial studies of blood pressure and on these traits may be understood through serum cholesterol levels in a total community-Tecumseh. the study of multigeneration pedigrees with Mich. J. Chronic Dis. 18:147-160. several types on nongenetic relationships. Lalouel J-M (1979) GEMINI-A computer program for The cultures where these types of houseoptimization of general nonlinear functions. Departholds are found may sponsor more cohesive ment of Medical Biophysics and Computing Technical Report No. 14, Salt Lake City, Utah. extended families, in which members of households live in a more homogeneous McGue M Wette R, and Rao DC (1984) A Monte Carlo evaluation of three statistical methods used in path common environment. However, the effects analysis. Genet. Epidemiol. 34:129-155. of cohabitation in the present study are Miall WE, and Oldham PD (1963) The hereditary factor in detectable primarily in spouses or in pairs of arterial blood-pressure. Br. Med. J. [Clin. Res.] i:75-80. relatives that are genetically related to some Mukherjee BN, Byard PJ, Bhattacharya SK, and Rao DC (1988) Blood pressure in a rural West Bengal fishing degree, such a s first cousins. Although wives community: An epidemiologic profile. Hum. Biol. 60: of brothers are significantly correlated for 69-79. systolic blood pressure, none of the other in- Rao DC, Vogler GP, McGue M, and Russell J M (1987) law relationships have significant correlaMaximum likelihood estimation of familial correlations from multivariate quantitative data on pedigrees: A tions. The suggestion of a n effect for housegeneral model and examples. Am. J . Hum. Genet. hold environment is found through testing of 41:1104-1116. simple hypotheses using likelihood ratio tests Sackett DL, Anderson GD, Milner R, and Feinleib M on the familial correlation estimates. Cross(1975) Concordance for coronary risk factors among spouses. Circulation 52:589-595. cultural comparisons of familial correlations BLOOD PRESSURE AND ANTHROPOMETRICS Speers M, Kasl SV, Freeman DH, Jr, and Ostfeld AM (1986) Blood pressure concordance between spouses. Am. J. Epidemiol. 123:818-829. 31 1 Tseng W-P (1967) Blood pressure and hypertension in a agricultural and a fishing population in Taiwan. Am. J. Epidemiol. 86:513-525.