Family resemblance for anthropometric and blood pressure measurements in black Caribs and Creoles from St. Vincent Island

AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 73:33-39 (1987)

Family Resemblance for Anthropometric and Blood Pressure Measurements in Black Caribs and Creoles From St. Vincent Island

JANIS HUTCHINSON AND PAMELA J. BYARD Department ofdnthropology, University of Houston, Houston, Texas (J.H.); Department of Pediatrics, Case Western Reserve University, Cleveland, Ohio 44106 (P J.B.)

KEY WORDS Path analysis, Caribbean

ABSTRACT This paper examines family resemblance for five anthropometric measurements (height, weight, triceps skinfold, upper arm circumference relaxed [UACR] and flexed [UACF] and for systolic and diastolic blood pressure in a group of adult Caribbean islanders of primarily African ancestry. Six hypotheses about family resemblance are tested by using path analysis and likelihood ratios. Significant intergenerational transmission is found only for height and UACR. For weight, UACF, and diastolic blood pressure, nontransmissible sibling resemblance is the primary component of family resemblance, although significant marital resemblance exists for diastolic blood pressure. Triceps skinfold and systolic blood pressure show no evidence of any family resemblance. Although results for highly heritable traits such as height are comparable to reports from other populations, measurements with a large contribution from common family environment or residual environmental effects, such as triceps skinfold or blood pressure, have much lower family resemblance in this population than in other populations. We hypothesize that this difference is due to the fact that adult children and their parents do not share a common household in this culture and to the presence of major nonfam- ilia1 environmental factors contributing to obesity and hypertension in this population.

Both obesity and hypertension have been implicated as important modifiable risk factors for chronic disease (Kannel et al., 1967; Marks, 1966; Stamler, 1967). Research has indicated that these risk factors have multifactorial etiologies (Miall and Oldham, 1963; Miall et al., 1967; Johnson et al., 1965; Tyro- ler, 1977). In addition to numerous environmental factors such as diet, stress, and activity levels, most evidence points to a genetic predisposition for hypertension and obesity (Biron et al., 1976; Harburg et al., 1973; 1977). Although much research into the etiology of these traits has been conducted on U.S. populations, data from populations outside the United States may provide etiological clues not available elsewhere.

In this paper, we examine family resemblance for anthropometric measurements and blood pressure in a group with a high preva-

lence of obesity and of hypertension, the Black Caribs and Creoles of St. Vincent Is- land. Familial correlations are reported, and path analysis is used to test hypotheses about marital, sibling, and parent-child resemblance.

MATERIALS AND METHODS Data collection

St. Vincent Island is located in the Carib- bean 21 miles southwest of St. Lucia (Fig. 1). Demographic, anthropometric, and blood pressure data were collected from the communities of Sandy Bay, Owia, and Fancy on St. Vincent during the summer of 1979 and fall of 1982. Adults over the age of 18 years were interviewed and measured in their

Received August 4, 1966; revision accepted November 14, 1966.

0 1987 ALAN R. LISS, INC

34 J. HUTCHINSON AND P.J. BYARD

Puerto Virgin Is. a 5 Anguilla Rico

0 Barbuda cb OAntigua

N 0

Guadeloupe v o 0 Dominica

mi 0 5 10 MartiniqueO 0.9. Lucia

ST. VINCENT +9 Bar:ados

p” Grenada

kms

I 0 200 400 mi

Fig. 1. Map showing location of study communities on St. Vincent island.

homes, at health clinics, and in a schoolhouse in Fancy. Although other ethnicities were present in the population, this study concen- trates on those with primarily African ancestry-Black Caribs and Creoles. The Black Carib population of St. Vincent is a mixture of African, Carib-Arawak, and European, whereas Creoles are primarily African and European. Both groups are 70-90% African, according to blood admixture estimates (Hutchinson and Crawford, 1981). The sample used for study of family resemblance is approximately 46% Black Carib and 44% Creole, with the remaining 10% consisting of Carib-Creole mixed and people of other ethnicities who had married into the population. The history and culture of the St. Vincent population has been described in detail elsewhere (Crawford, 1984; Hutchinson, 1986).

Anthropometric measurements were collected by using standard techniques. Weight was measured by bounds by using a Counse- lor portable scale. For blood pressure readings, subjects were seated with their left arm resting on a table or bench. A baumanometer with a 14-cm-wide compression cuff was po- sitioned on the left arm at heart level and systolic and diastolic (phase V) pressures were then measured and recorded. Blood pressure readings on individuals who stated that they were on hypertension medication or whose urinalysis test indicated kidney malfunction or diabetes were excluded from

the sample. A total of 123 males and 206 females were used in this study, but only 117 males and 192 females have usable blood pressure measurements because of these in- clusion criteria.

Age adjustment Although only adults were included in the

study, significant age trends were found for several variables. The effects of age were re- moved by stepwise multiple regression of age, age’, and age3 separately in males and females. The regression equations used are shown in Table 1. All of the age-adjusted variables (except for stature in females) ex- hibit significant skewness and kurtosis. Therefore, the standardized residuals were normalized by assigning Blom scores (Blom, 1958). Normalization was done separately for mothers, fathers, sons, and daughters, in or- der to ensure univariate normality for each group used in the analysis of family resemblance.

Familial correlations After age adjustment and normalization,

familial correlations were estimated by maximum likelihood, using the MLECOR com- puter program (Rao et al., 1982). In this application, we have four different types of nuclear family members: fathers, mothers, sons, and daughters. In addition to the four means and variances, these relatives generate eight different familial correlations:

ANTHROPOMETRICS AND BLOOD PRESSURE 35

Fig. 2. Path diagram for the TAU model applied in this analysis. The diagram depicts phenotypes and transmissible factors for a nuclear family composed of a father, mother (subscripts 1 and 2 in the top half of the diagram), a son and a daughter (subscripts 1 and 2 in the bottom half of the diagram).

father-mother, father-son, father-daughter7 mother-son, mother-daughter, son-son, son- daughter, and daughter-daughter. In prac- tice, the four means are fixed at sample values, along with the variances for fathers and for mothers. Variances for sons and daughters, eight familial correlations, and sample sizes are then estimated. The estimates of sample size are useful because they more nearly reflect the information available in multiple pairs taken from the same families than when all possible pairs are used in prod- uct-moment correlations.

Path analysis Path analysis of familial resemblance ex-

amines the transmissibility of traits by par- titioning the covariance among family members into components described by standardized partial regression coefficients (Wright, 1921). These regression Coefficients, or path coefficients, are estimated by solving systems of simultaneous equations written in terms of the path coefficients. The model of family resemblance used to generate these equations may be more easily understood by use of a path diagram (Fig. 2). In this diagram, observed phenotypes of family mem-


bers (shown as squares labeled “P”), are related through inferred sources of variation (shown as circles, transmissible factors, labeled “T”, or residual environment, labeled “R’). Expected correlations between the phenotypes of relatives are derived by tracing the paths between them (shown as lines, la-

7 , “s”, or “e”) according to certain rules (see Li, 1975). All components of the diagram may be subscripted according to sex, where 1 =male and 2 =female.

The arrows on the paths denote inferred causality. Single-headed arrows indicate that one element of the diagram is thought to cause another, while paths with arrows at each end denote correlation between two variables, neither of which is necessarily the cause of the other. A headless line, as shown for the path p, signifies a copath, with special tracing rules (Cloninger, 1980).

For phenotype data in nuclear families, the TAU model of transmission is appropriate (Rice et al., 1978). This model does not at- tempt to separate genetic transmission from the effects of common family environment, but it does not assume that all transmission is genetic, as traditional estimates of heritability do. Instead, a parameter called 7 is used to represent the path between the transmissible factors of parents and children. This parameter may be estimated separately according to the sex of the parent and the sex of the child (rll =father-son, r12 =father- daughter, 721 =mother-son, and 722 =mother- daughter), estimated without sex differences (711 = 712 = 721 = 7221, or may be fixed at one- half to conform to polygenic transmission

The path t represents the effects of a per- son’s transmissible factor on his or her phenotype. When squared, this parameter is known as transmissibility, or the proportion of phenotypic variability which is determined by transmissible factors. Transmissi- bility may be subscripted according to sex so that tf ,= transmissibility in males and ti = transmissibility in females. However, in this application, a single transmissibility estimate for both sexes gives an adequate fit to the data.

In addition to transmissibility, parameters of the TAU model include p, the correlation between phenotypes of mates due to phenotypic assortative mating, and s, nontransmissible correlation among the environments of siblings. The s parameter may differ according to the sex of each member of the sib pair

beled < C p 9 9 , < < t 7 7 , (( ?)

7ij = 1/21.

(so that s11 = brother-brother, s12 7 brother- sister, and s22 = sister-sister relationships). Thus, all sib pairs will have some resemblance due to transmissible factors from the parents, but there may be sex differences in sibling resemblance due to the effects of nontransmissible factors, such as different degrees of common sibling environment according to sex.

The path coefficients are estimated by maximum likelihood, using a method 3 ap- proach, described by Rao et al. (1984). In this technique, path parameters are fit directly to the family phenotype data instead of to the familial correlations, thus eliminating a number of statistical assumptions concern- ing the independence and multivariate normality of the correlation coefficients. Path parameter estimates that most closely fit the observed family data within the constraint of the path model are derived. The model used must be determined-that is, the family data must contain enough information to derive the path coefficients. In this case, we have eight familial relationships and can estimate at most eight path coefficients. The original

s12, and s22) can be reduced by assuming a polygenic mode of transmission (rij = l/2) and a single transmissibility estimate for both sexes (tl =t2). With these assumptions, a general model with five estimated parameters (p, t, sll, 512, and s22) can be used. The goodness of fit for the general model is assessed by comparing the log of the likelihood value obtained from the path analysis (given as -21n[L] plus a constant) to the log likelihood value for the familial correlations. Sub- traction yields a value which is asymp- totically distributed as a x2 with n-k degrees of freedom, where n is the number of familial correlations and k is the number of parameters estimated in the path model. In this case the degrees of freedom will be three (eight familial correlations minus five estimated path parameters).

One of the more useful and informative features of path analysis is the ability to test hypotheses about family resemblance through the use of likelihood ratio x2 tests. To do this, we fix certain of the parameters at predetermined values (e.g., at zero or equal to another parameter) and compare the likelihood for the reduced model to the likelihood for the general model. Again, this is done by subtracting the second value of -21n[L] +c from the first. The result is distributed

ten parameters (P, tl, t2, 711, 712, 721, 722, Sl l ,

freedom, where n is the number of parameters estimated in the first (general) model and k is the number of estimated path parameters in the second (reduced) model. In this way, we may test several null hypotheses: no marital resemblance (p=O), no sex effects on sibling-shared environment (sll= 512 = s22), no sibling-shared environmental effects at all (sll=s12=~22=0), and no transmission from parent to child (t=r=O). Finally, combinations of these hypotheses can be tested to arrive at a parsimonious model of family resemblance for each variable.

RESULTS -.\ d Familial correlation and sample size esti- $ matesareshownforeachtypeofrelationship '5

in Table 2. None of the marital correlations -$

37

O m a c D s z % % z F % a V

+ %

.i M

, i * g 8 (xi g m h l + & 2 n azdgzsS" 2

I I I

e r n o (xi w t- - m t- a * m m m e r i ( x i N a U .* e j+: m m m m a % g I: s

2 2 8 g 8 8 8 z I I I

g g g 9 2 2 g E

u3 9 II t!

child correlations are significant, except for stature and the father-daughter correlation for upper arm circumference relaxed (UACR). Highly significant sib pair correlations exist for height, weight, UACR, and upper arm circumference flexed (UACF), although this

2 2

3

is primarily true for same-sex pairs. The sister-sister correlation for diastolic blood pressure is significant a t the .05 level.

Goodness of fit for the general model is presented in Table 3, along with likelihood ratio tests for several reduced models. None ofthe goodnessoffit x2values is significant, indicating that the general model gives a

variable.

2

9 6

2 -

.I

2 E

3

good fit to the observed data for every

Hypotheses tested by likelihood ratio are #

8 2 z z ? g ? 0

I

' w 4 s a % 2 3 i% G

2 Q) p- % rn cr) rl ~ o m m

+I 8 2 8 0 0 , ? z z + O

0 I I

fi Q, ?-4 hl (xi m h l m m w a l . ~

3 ~ c.l " O W

zzz?zzg I 2 I I

~ ?-4

tion among sibs is independent of sex

sibling resemblance at all (s11= s12 = s22 = 0)

offspring (t = r = 0) is rejected for height and UACR, but not for weight, UACF, or blood pressure.

parsimonious model for each variable shown

mates and their standard errors. For height,

(sll =s12 = s22) is rejected for weight, UACR, and UACF. The hypothesis of no residual

is also rejected for these variables. The hypothesis of no transmission from parents to

Combinations of these hypotheses yield the

in Table 4. Also given are parameter esti-

the only significant parameter is t2, which is

0 I

~ ( x i r i ? - 4 0 0 " h l t - * * w t - h l ( x i o o

.e * * * * f X 3 g d s g : 2 2 2 :

8 2 2 g 2 2 g g

& & ;%& & % & Z & & +I

$ 2 3 .zz 5z G C 3 5 2 5 ; k;

I

r 0 $ k

~ E Z Z Z 4 2 gs2 ek :.e,z..;; c r ~ c r % % m m m


TABLE 3. x2 values for goodness-of-fit for the general path model and likelihood ratio tests for reduced models

General s11=s12 s11=s12 model p=o = s22 =szz=O t=7=0 (df = 3) (df = 1) (df=2) (df = 3) (df=I)

Height Weight Triceps UACR UACF

0.483 0.096 0.398 0.439 23.469* 1.704 0.431 7.576* 12.443* 3.220 0.632 0.002 3.047 3.266 1.205 3.065 1.235 9.679* 9.726* 3.901* 2.566 0.836 9.479* 10.823* 2.618 ~~

Systolic BP 3.907 0.983 0.138 0.942 0.000 Diastolic BP 0.482 4.922* 4.910* 5.234 0.310

*Significant at a= .05.

TABLE 4. Parameter estimates and standard errors under the most parsimonious model

df

Height 0 0.789 f ,091 0 0 0 0.994 4 Weight 0 0 0.690 f .015 0.453 k ,110 0.236 k ,121 3.634 2 Triceps Skf 0 0 0 0 0 3.938 5 UACR 0 0.361 i ,167 0.137 +_ ,461 0.464 i ,160 0.318 +_ ,288 1.235 1 UACF 0 0 0.582 f ,157 0.489 ,105 0.041 i ,136 3.554 2 Systolic BP 0 0 0 0 0 1.925 5 Diastolic BP 0.280 i ,116 0 0 0 0 6.508 4

x2 ~

P t 2 s11 512 522

one-half and p and sij are fixed at zero. This heritability level is comparable to other published studies. As expected from examination of the familial correlations, parent-child transmission for weight is not significant, leading to a parsimonious model with t2=7=0. The highly significant sibling resemblance for weight is incorporated entirely into the nontransmissible component, sij. Sibling resemblance is not independent of sex for weight; brother-brother pairs are more similar than opposite sex or sister-sister pairs.

The parsimonious model for triceps skinfold is extremely parsimonious: no family resemblance of any kind. This conclusion no doubt reflects the changeable nature of skinfold measurements and the smallish sample sizes in this study. Upper arm circumference is transmissible only when taken relaxed, not flexed. In either case, sex differences play a role in sibling resemblance.

Systolic blood pressure displays no family resemblance, but diastolic blood pressure ex- hibits significant marital resemblance. This would imply that nonfamilial environmental factors influence systolic pressure, whereas nontransmissible family environmental factors affect diastolic pressure.

DISCUSSION

Although transmissibility for height and UACR in this study is similar to that reported for other populations (Byard et al., 1984; Devor et al., 1986), parent-child resemblance for weight, triceps skinfold, UACF, and blood pressure is lower in this study

than in most published reports, leading to a parsimonious path model with no intergenerational transmission for these variables. There are several possible reasons for this difference. Compared to other traits such as stature, reported familial correlations for blood pressure are generally low (Acheson and Fowler, 1967). However, many studies do report heritability levels significantly above zero. It should be noted that many of these studies have very large sample sizes (i.e., Krieger et al., 1965, and Morton et al., 1980, had over 1,000 pairs for each type of relation- ship). The parent-child correlations reported here are depressed even in comparison to low reported values, however. Transmissibility on St. Vincent may be especially low because most of the “children” in this study are adults living in their own households. Thus, parents and their children have not shared common household environment for some time. This residence pattern has been found in other Black Carib communities undergoing modernization, as migration to obtain jobs in the cash economy has become a way of life during young and middle adulthood (Gonzalez, 1984). A decrease in familial resemblance for blood pressure with increasing modernization was also found in Samoans (McGarvey et al., 1980). Marital resemblance, however, increases slightly with modernization in Samoans (James et al., 1983), to a level (.27 for systolic and .24 for diastolic) very similar to that reported in Ta- ble 2.

Another difference between this study and many of those in the literature is that data

ANTHROPOMETRICS AND BLOOD PRESSURE 39

collection did not proceed by family group. Individuals were measured only occasionally at the same time and in the same location as the rest of their families. This would have obvious effects on the more changeable measurements such as blood pressure.

Finally, there may be major environmental factors acting on the fatness levels and blood pressures of this population, especially in women, which are not seen in other groups. The mean value for triceps skinfold among the adult women sampled is 19 mm, but some women have measurements over 50 mm and weights over 200 kg. Likewise, ten women (5%) were detected in this survey with diastolic blood pressures over 95, not counting the 14 whose blood pressure readings were excluded in the original screening test (most of whom were undergoing treatment for hypertension). The present analysis has shown that the factors involved in these extreme phenotypes are not familial in nature. A more detailed study of potential risk factors, such as individual dietary intake of fats and sodium, is needed to ascertain the etiology of obesity and hypertension in this population.

ACKNOWLEDGMENTS

This research was supported in part by Biomedical Science Research grant 4932- X706-4 from the University of Kansas and by National Institutes of Health grant KO4 DE028-05. We thank the people of St. Vin- cent for their cooperation in the fieldwork, and the Division of Biostatistics, Washington University School of Medicine, where part of this analysis was conducted while the junior author was a postdoctoral trainee and visit- ing assistant professor.

LITERATURE CITED

Acheson, RM, and Fowler, GB (1967) On the inheritance of stature and blood pressure. J. Chronic Dis. 20:731- 745.

Barnicot, N, Bennett, F, Woodburn, J, Pilkington, T, and Antonis, A (1972) Blood pressure and serum cholesterol in the Hadza of Tanzania. Hum. Biol. 44:87-116.

Biron, P, Mongeau, J, and Bertand, D (1976) Familial aggregation of blood pressure in 558 adopted children. Can. Med. Assoc. J. 115:773-774.

Blom, G (1958) Statistical Estimates and Transformed Beta Variables. New York: John Wiley and Sons.

Byard, PJ, Sharma, K, Russell, JM, and Rao, DC (1984) A family study of anthropometric traits in a Punjabi community: 11. An investigation of familial transmission. Am. J. Phys. Anthropol. 64:97-104.

Cloninger, CR (1980) Interpretation of intrinsic and ex- trinsic structural relations by path analysis: Theory and applications to assortative mating. Genet. Res.

Crawford, MH (ed) (1984) Black Caribs: A Case Study in 36:133-145.

Biocultural Adaptation. New York Plenum Press.

Devor, EJ, McGue M, Crawford, MH, and Lin, PM (1986) Transmissible and nontransmissible components of anthropometric variation in the Alexanderwohl Mennon- ites: 11. Resolution by path analysis. Am. J. Phys. Anthropol. 69.43-92.

Gonzalez, NL (1984) Garifuna (Black Carib) household structure. In MH Crawford (ed): Black Caribs: A Case Study in Biocultural Adaptation. New York Plenum Press, pp. 51-65.

Harburg, E, Erfurt, JC, Chape, C, Hauenstein, LS, Schull, WJ, and Schork, MA (1973) Socioecological stressor areas and black-white blood pressure: Detroit. J. Chronic. Dis. 26595-611.

Harburg, E, Schork, MA, Erfurt, JC, Schull, WJ, and Chape, C (1977) Heredity, stress and blood pressure, a family set method-11. J. Chronic Dis. 30:649-658.

Hntchinson, J, and Crawford, MH (1981) Genetic determinants of blood pressure among the Black Caribs of St. Vincent. Hum. Biol. 53:453-466.

Hutchinson, J (1986) Association between stress and blood pressure variation in a Caribbean population. Am. J. Phys. Anthropol. 71:69-79.

James, GD, McGarvey, ST, and Baker, PT (1983) The effect of modernization on spouse concordance in American Samoa. Hum. Biol. 55:643-652.

Johnson, BC, Epstein, FH, and Kjelsberg, MO (1965) Distributions and familial studies of blood pressure and serum cholesterol levels in a total community- Tecumseh, Michigan. J. Chronic Dis. 18:147-160.

Kannel, WB, LeBauer, EJ, Dawber, TR, and Mac- Namara, PM (1967) Relation of body weight to devel- opment of coronary heart disease: Framingham study. Circulation 35:734-744.

Krieger, H, Morton, NE, Mi, MP, Azevedo, A, Freire- Maia, A, and Yasuda, N (1965) Racial admixture in northeastern Brazil. Ann. Hum. Gen. 29:113-125.

Li, CC (1975) Path Analysis: A Primer. Pacific Grove, California: Boxwood Press.

Marks, MH (1966) Influence of obesity on morbidity and mortality. Bull. NY Acad. Med. 36:296-312.

McGarvey, SE, Schendel, D, and Baker, PT (1980) Mod- ernization effects on familial aggregation of Samoan blood pressure: A preliminary report. Med. Anthropol. 4r321-328.

Miall, WE, Heneage, P, Khosla, T, Lovell, HG, and Moore, F (1967) Factors influencing the degree of resemblance in arterial pressure of close relatives. Clin. Sci. 33271-283.

Miall, WE, and Oldham, PD (1963) The hereditary factor in arterial blood-pressure. Br. Med. J. 1:75-80.

Morton, NE, Gulbrandsen, CI, Rao, DC, Rhoads, GC, and Kagan, A (1980) Determinants of blood pressure in Japanese-American families. Hum. Gen. 53t261- 266.

Rao, DC, Laskarzewski, PM, Morrison, JA, Khoury, P, Kelly, K, Wette, R, Russell, J, and Glueck, CJ (1982) The Cincinnati Lipid Research Clinic family study: Cultural and biological determinants of lipids and li- poprotein concentrations. Am. J. Hum. Genet. 34:888- 903.

Rao, DC, Wette, R, and Glueck, CJ (1984) Path analysis in genetic epidemiology. In A Chakravarti (ed) Human Population Genetics: The Pittsburgh Symposium. New York Van Nostrand Reinhold Company, pp. 35-81.

Rice, J, Cloninger, CR, and Reich, T (1978) Multifactorial inheritance with culture transmission and assortative mating. I. Description and basic properties of the uni- tary models. Am. J. Hum. Genet. 30-68-143.

Stamler, J (1967) Lectures on Preventive Cardiology. New York: Grune &. Stratton, Inc.

Wright, S (1921) Correlation and causation. J. Agric. Res. 20557-585.

Documents

Family resemblance for anthropometric and blood pressure measurements in black Caribs and Creoles from St. Vincent Island