AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 81:563-572 (1990)
Genetic Epidemiological Study of Blood Pressure in a Sedentary Rural Agricultural Population of West Bengal, India PARTHA P. MAJUMDER, S.K. BHAlTACHARYA, B.N. MUKHERJEE, AND D.C. RAO Human Genetics Division, Department of Biostatistics, University of Pittsburgh, Pittsburgh, (P.P.M.); Anthropornetry and Human Genetics Unit, Indian Statistical Institute, Calcutta, India (P.P.M., S.K.B., B.N.M.); Division of Biostatistics, Washington University School of Medicine, St. Louis, (D.C.R.).
KEY WORDS
Family data, Path Analysis, Hypertension
ABSTRACT
To study the genetic epidemiology of blood pressure (BP), data on 78 families were collected from a sedentary agricultural population of eastern India. The general levels of both systolic (SBP) and diastolic (DBP) blood pressures are found to be low (mean SBP = 106.41 mm Hg; mean DBP = 63.94 mm Hg). Trends of blood pressures with age are similar to those reported earlier (e.g., in the Framingham study). Environmental variables - e . g . , occupation and tobacco u s e 4 0 not have any direct significant effect on blood pressure variability in this population. Path analysis of family data shows a highly significant familial aggregation and yields a genetic heritability (maximum) estimate of 0.3 for both SBP and DBP. Sib-sib and mother-child correlation estimates are, respectively, 0.3 and 0.25. Father-child correlation estimates are 0.13 for SBP and near zero for DBP. A pseudopolygenic model yields the best fit to the data on SBP, while for DBP a proper resolution of various models considered could not be obtained.
It is well-established that essential hypertension is a major public health burden. It is an important contributory factor to coronary heart disease, stroke, and kidney failure (Siervogel, 1983). Although risk factors for hypertension seem to cut across cultural patterns and geographic regions (Marmot, 1979), in contrast to many westernized societies, the majority of traditional societies have low mean values of blood pressure and are characterized by an insignificant increase of blood pressure with age (Epstein and Eckhoff, 1967). Many studies have also shown that the etiologic factors affecting the distributions of blood pressure in traditional societies are different from those in westernizedlurbanized societies (Siervogel, 1983; Ward, 1983). However, even among traditional societies, the factors identified to be significant determinants of blood pressure distributions are not the same (Ward, 1983), and there are exceptions to the general finding that hypertension is rare in traditional societies (Neilson and Williams, 1978; Marmot, 1979). While the etiologic
@ 1990 WILEY-I,ISS, I N C
factors contributing to a rise in the general level of blood pressure in a population may be many, the most important are salt intake and de ees of obesity, physical activity, and psycho ogical stress (Oliver, 1980; Ward, 1983). It may be pointed out here that salt intake and obesity are primarily related to diet and physical activity. In spite of large environmental effects on blood pressure, the within-population percentage of variability in blood pressure attributable to familial aggregation is fair1 constant: 2040% (Siervogel, 1983; W a r l 1983). Doubtless, therefore, genetic factors-major genes or polygenes-play a considerable role in determining blood pressure variability within populations. The present investigation is motivated by the fact that there are no systematic genetic epidemiological studies conducted on sedentary rural populations of India that provide a contrast to the westernized populations in
!?.
Received April 12,1989; revision accepted June 22,1989
564
P.P. MAJUMDER ET AL
women spend most of their days performing household chores, but sometimes they help the men in the fields. The standard of living is low: The majority live in mud-walled, thatched houses. They eat large quantities of rice and rice products with small amounts of lentils and leafy vegetables cooked in oil, but eat little or no animal protein. Although we have no quantitative data, the salt intake of individuals in this population is moderate. Rice is cooked without salt; salt is used while cooking lentils and vegetables. Major meals are eaten at home. Eating out, including snacks, is not popular mainly because of lack of money. Drinking water does not contain an excessive amount of salt. Three consecutive blood pressure (systolic and diastolic) and pulse rate readings were taken on each individual above 15 years of age in the sitting position with an electronic digital instrument. Data were gathered on tobacco use (smoking in the case of males and chewing in the case of females), alcohol consumption, use of antihypertensive drugs, and whether suffering from any major tension. (Major tension was a personal assessment of the respondant of immediate psychological stress and pertained to the fortnight preceding the date of survey. Death, birth, and marriage of immediate family members, school and college examinations, and so forth, were cited by the investigator as events that could lead to major tension.) MATERIALS AND METHODS There is no use of contraceptive drugs by A total of 78 randomly selected nuclear women. The following anthropometric meafamilies, comprising 325 individuals, were surements were taken on each individual: screened during June to August 1986. There height, weight, and skinfold thicknesses (biwas no instance of multiple marriage in ceps, triceps, and subscapular). these families. Thus no stepchildren are inFor purposes of uniformity, we have concluded in the data, and all sibs are full sibs. sidered data of only those individuals for All households belong to a set of six villages whom observations on both primary (systolic located in the southern deltaic coastal region and diastolic blood pressures) and secondary of the state of West Bengal, about 160 km (tobacco use, height, weight, and so forth) southwest of Calcutta, India. The villages variables were completely available. These are small, with about 60 households on aver- were available on 281 individuals (164 males age in each village. All households belong to and 117 females) distributed among 77 nuthe Hindu middle caste. From a social point clear families. It is pertinent to point out that of view, therefore, the present sample is all male offspring (irrespective of age) and fairly homogeneous. There was no evidence unmarried female offspring (age at marriage of consanguinity up to the second-cousin usually between 17 and 22 years) were living level. at home with their parents; married female Agriculture is the primary occupation of offspring belonged to separate households, the inhabitants of these villages. Most of the but such offspring included in the present agriculture is nonmechanized and tradi- data set lived in close geographical proximity tional; agricultural labor is primarily per- to the parental home and represent about formed by men. Some men are associated 20% of all female offspring included in this with the transportation business. The data set.
respect to diet, which is a prime determinant of blood pressure. While the diet in westernized societies is primarily high in protein and fat content, the rural populations of India thrive on a high carbohydrate diet with little fat and protein. Rural India is also primarily agricultural, and since agriculture is mostly nonmechanized, farmers are regularly engaged in much farm-related physical activity. Psychological stress, which is another prime determinant of blood pressure, is more difficult to quantify, but perhaps traditional societies are under lower stress than highly industrialized societies. Thus the rural sedentary population of India perhaps provides an extreme contrast for studies of blood pressure. It is, therefore, of interest to examine the role of various factors in the determination of blood pressure within a population characterized by a high carbohydrate diet, high physical activity, low intake of fat, moderate to severe protein malnutrition, and perhaps also low psychological stress. With these aims in view, a project was undertaken in 1986 by the Indian Statistical Institute, Calcutta, India, to study the epidemiology of blood pressure under contrasting ecosystems. The present paper reports some findings of the first phase of the project undertaken among a sedentary agricultural population from the coastal region of West Bengal, India.
BLOOD PRESSURE IN EASTERN INDIA
565
variables, a subset of significant variables was identified by using a stepwise regression algorithm. Having identified a subset of significant predictor variables, standardized phenotypes were computed as Ps = (& - X,)/s,, where X, is the observed value of blood pressure (systolic or diastolic) for an individual, X, is the corresponding predicted value (predicted using the significant subset of predictors identified at the previous step of analysis), and sRis the standard deviation of the residuals. These standardized phenotypes were then normalized (PN)by ranking the Ps values and taking the inverse normal transformation of the ranks (Rao et al., 1984). Analyses of variance were then performed on the PNvalues to detect effects of categorical secondary variables. Although the original blood pressure measurements showed significant skewness and kurtosis, the normalized PNvalues had distributions close to the Gaussian (normal) distribution. The normalized phenotypes (PN) were then analyzed to investigate the extent of
Since blood pressure and pulse rate measurements were taken three times on each individual, we ignored the first reading and took the average of the second and third readings as representative. This was done because the variance among individuals was the highest for the first reading and also because the correlation coefficient between second and third readings (0.92-0.96) was higher than between first and second readings (0.86) or between first and third readings (0.82). To study the effects of categorical secondary variables (e.g., education and occupation) on blood pressures, we first removed the effects of those quantitative secondary variables (e.g., height and weight) that were found to have significant effects on the primary variables. Assessment of the significance of the quantitative secondary variables was done by treating this set of variables as a set of predictor variables for predicting either systolic or diastolic blood pressure. From among the set of predictor P
Fig. 1. Path model showing familial aggregation of blood pressure in nuclear families. P, T, and R denote, respectively, phenotype, transmissible factor, and resid-
ual environment. Subscripts F, M, C and C, denote, respectively,father, mother, and two cihdren. Residuals are correlated ( s ) in sibships.
566
P.P. MAJUMDER ET AL.
familial a gregation for each of the two BP variables BP and DBP using a sim le path model shown in Figure 1. The mode postulates that a phenotype (P) is determined by independent linear additive effects of a transmissible factor (T) and a residual (R): P = tT + rR, where t' represents the proportion of blood pressure variability explained by the transmissible factor. Since T includes both genetic and familial environmental factors, t' may be referred to as generalized heritability. The residual variance caused b nonfamilial factors is simply r2 = 1 - t H, since the variance of the standardized P, is unity. Vertical transmission of the transmissible factor (TI, from parents to children, is represented by the parameters 7F and TM, distinguishing paternal (7F)and maternal (7M) effects. Thus specific maternal effects can be detected by testing the inequality O f TF and T ~ In. addition to transmissible effects, a common sibship residual correlation (s) is also incorporated so that the phenotype of a child (Pc, subscript C denoting a child) is given by Pc = tTc + rcRc, where the residuals (Re) are correlated within sibships (s). Again, re2 = 1 - t2.Finally, marital resemblance is represented by a simple correlational path (a "copath," see Cloninger, 1980), p. Since the residuals (R) are assumed to be uncorrelated among other family members, they are not shown in Figure 1. This model , involves five parameters in all; p, t, T ~ TM, and s. Not all five parameters can be estimated from nuclear families, as the data basically reduce to four correlations (spouse-spouse, father-child, mother-child, and sib-sib). At most four parameters can be estimated uniquely, by fixing any one of 7F, 7111, or s. However, the analysis can still provide useful information on the magnitude and nature of familial aggregation. Under the model, the four familial correlations have the expectations given in Table 7. The log likelihood function for a given nuclear family is taken as the logarithm of an appropriate multivariate normal density function, defining the joint distribution of the phenotypes in the family. The corresponding covariance matrix is defined in terms of the four expected correlations given above. The overall log likelihood for the entire random sample of families is given by the sum of all individual functions, one for each family. Standard maximum likelihood methods and likelihood ratio tests are used to fit the model t o the data.
5
P
567
BLOOD PRESSURE IN EASTERN INDIA
and 21 female) who reported suffering from major tension. With respect to these three categorical variables, we decided t o perform analyses 1)by ignoring the variables and 2) after excluding the “positive”cases (i.e., alcohol and antihypertensive drug users and those suffering from major tension). The frequency distributions of systolic (SBP) and diastolic (DBP) blood pressures are presented in Tables 2 and 3, separately for each sex. Sex differences are not striking. The skewness and kurtosis values, except for SBP among males (skewness = 1.2, kurtosis = 3.61, are small. By using the World Health Organization’s definition of hypertension (SBP > 160 mm Hg and/or DBP > 95 mm Hg), only four individuals (one male and three female) are found to be hypertensive; this yields a prevalence of 1.42%of hypertension. It may be pointed out that all four individuals belong to separate households. The upper 95%cut-off points for the distributions of SBP among males, females, and in the total sample are (approximately) 130, 142, and 140 mm Hg, respectively; the corresponding values for DBP are 78, 88, and 85 mm Hg. Trends of mean blood pressures are presented in Table4 separately for
RESULTS
Table 1gives the distribution of the sample subjects, by sex, with respect to age groups and three categorical variables-ducation, occupation, and tobacco use. It should be pointed out that although at the time of data collection the information was collected in finer categories, to avoid vagaries of small cell frequencies certain categories had to be merged. For example, the occupation category “manual labor” comprised separate categories, e.g., fishing, agriculture, that were pooled; the category “other” comprised separate categories, e.g., white-collar work, household work, that were pooled. Similarly, tobacco users were also initially classified as medium, heavy, and light, but had to be pooled for reasons of small sample size. From the sample distribution (see Table 11,we decided to ignore occupational differences for females. With respect to the other categorical variables, there were in all 13 consumers of alcohol (12 male and 1female), of whom 10 were irregular or casual consumers. There were, in all, six users (three male and three female) of antihypertensive drugs and 36 people (15 male TABLE 2. Frequency distribution SRP (mm Hg)
580 81-90 91-100 101-110 111-120 121-130 131-140 141-150 151-160 2160
Total
Male Frequency
Percent
2 15 36 58 33 11 4 4 0 1 164
1.2 9.2 22.0 35.4 20.1 6.7 2.4 2.4 0.0 0.6 100.0
of systolic blood pressure
Female Frequency
1 9 33 37 17 8 5 6 1 0 117
(SBP) Total
Percent
Frequency
0.9 7.7 28.2 31.6 14.5 6.8 4.3 5.1 0.9 0.0 100.0
3 24 69 95 50 19 9 10
Percent
1.1 8.5 24.5 33.8 17.8 6.8 3.2
1
3.5 0.4
1 28 1
0.4 100.0
TABLE 3. Frequency distribution of diastolic blood pressure IDBP) DBP (mm Hg) 540 4 1-50 5 1-60 61-70 71-80 81-90 91-100 2100 Totaf
Male Frequency 6
20 47 61 24 5 0 1 164
Percent
3.7 12.2 28.7 37.2 14.6 3.0 0.0 0.6 100.0
Female Frequency
1 6 20 41 34 11 3 1
117
Total Percent
Frequency
0.9 5.1 17.1 35.0 29.1 9.4 2.6 0.9 100.0
7 26 67 10’2 58 16 3 2 281
Percent
2.5 9.3 23.8 36.3 20.6 5.7 1.1
0.7 100.0
P.P. W U M D E R ET AL
a
m
-m c c
a
m v)
P.
m L?
3
7 m W
a
m
v)
PI
m L? W d
m n a
3 a
m
m d
C
' 9
"a
ffi (I:
m a CI
Lc
7
m W
a
3
males and females. Analysis of variance was performed to test age and sex differences in mean values. For SBP it was found that although the main effect of sex is nonsignificant at the 5% level (F ratio = 0.529, d.f. = l,269), there is a significant sex x age interaction (F ratio = 2.545, d.f. = 5,269). For DBP, the main effect of sex is significant (F ratio = 29.233; d.f. = 1,269),but the sex x age interaction is absent (F ratio = 0.448, d.f. = 5,269). There is a pronounced effect of age for both SBP (F ratio = 6.689, d.f. = 5,269) and DBP (F ratio = 21.495, d.f. = 5,269). In all further analyses, we have, therefore, retained both the sex and age group classifications. However, to avoid the vagaries of small sample sizes, certain age groups were pooled. We examined the scatter diagrams of age vs. SBP/DBP and found a difference in trend (for both males and females) above and below 45 years of age. This is also clearly seen from the correlation coefficients presented in Table 5. For SBP, the correlation between age and SBP is much lower for individuals (both male and female) below the age of 45 years in comparison with those above the age of 45 years. For DBP, there is reversal in trend; the correlation is positive, large, and significant (among both males and females) for individuals below 45 years of age, while it is negative, smaller, and nonsignificant for individuals above 45 years of age. We have, therefore, taken 45 years as a cut-off point and have analyzed data on individuals 5 4 5 years and >45 years of age separately. Since blood pressures are correlated with age, anthropometric measurements, and pulse rate, we decided to remove the effects of these quantitative variables before studying the effects of categorical variables such as education and occupation. To allow for nonlinear trends of blood pressures with age, we have also included age2 and age3 in the list of independent variables along with age, height, weight, skinfold thicknesses, and pulse rate. As mentioned earlier, since the relationships of blood pressure and age are different for different sex and age subgroups, we have obtained separate regression equations for males and females above and below 45 years of age. It was found that of the nine predictor variables (age, age2, age3, pulse rate, height, weight, skinfold thicknesses [biceps, triceps, and subscapularl), in most subgroups only two or three variables were significant. The results of the stepwise re-
569
BLOOD PRESSURE IN EASTERN INDIA
T A B L E 5 Correlation coefficients between age and blood uressure in individuals helongma to different age clashes Variables'
All
Age, SBP Age, RDP
0.215* 0.483*
Male 5 4 5 vears
>45 vears
All
Female 5 4 5 vears
>46 vears
0.087 0.486*
0.226 -0.180
0.418* 0.485*
0.074 0.359*
0.280 -0.1 62
I SBP, systnhc blood pressure; DBP, diastolic hloud pressure *Significant a t the 5'Xt level.
T A B L E 6. Results of stepwise regression analysis: Significant predictor variables and regression coefficients Sex and aee subprouo Sex Age (years)
Regression equation SBP DBP SBP DBP
= 74.34 + 0.63 (weight) = 9.29 0.52 (age) 0.42 (weight) 0.19 (pulse rate) = 1.22 0.73 (age) 1.37 (weight) = 29.80 1.03 (weight) 0.85 (subscapular skinfold thickness)
545
SRP DRP
>45
SBP DBP
= = = =
Male
545
Male
>45
Female Female
+ + + + + 73.58 t 0.81 (subscapular skinfold thickness) + 0.18 (pulse rate) 25.86 + 0.42 (age) + 0.69 (subscapular skinfold thickness) + 0.18 (pulse rate) 70.60 + 1.13 (weight) 42.39 t 0.62 (subscapular skinfold thickness) + 0.28 (pulse rate) +
-
T A B L E 7. M a x i m u m likelihood estimatesof familial correlations for systolic (SHP)and diastolic (DBP) bloodpressure Maximum likelihood estimate Correlation Spouse-spouse Father-child Mother-child Sib-sih
SBP
DBP
-0.01 t 0.15 0.1'3 i 0.14 0.24 0.12 0.31 i 0.13
-0.04 i 0.15 -0.04 i 0.15
+
gression analyses are presented in Table 6. It is seen that age and weight are the two important predictors of blood pressure. Nonlinear trends in age were found to be absent. For further analyses, standardized Z scores of the phenotypic data were computed that were then normalized using the method described in the previous section. The effects of the categorical variables were studied using analysis of variance techniques on the normalized blood pressure values. For males we have considered the effects of education, occupation, and tobacco use separately, while for females we have ignored the occupation classification for reasons mentioned earlier. Apart from performing analysis of variance separately for sex and age subgroups, we have also performed (within each sex and age subgroup) analyses separately by considering 1) all individuals ignoring alcohol use, antihypertensive drug use, and tension suffering; and 2) only individuals who are nonusers of either alcohol or antihypertensive drugs and not suffering from ma-
*
0.26 0.12 0.32 i 0.13
Expected correlation under the path model P t 2 ( 7 y + p TM)
ti s
( 7+ ~ p
rF)
+ tz ( Tp + T M ' ) + 2 7v 7M)
jor tension. Analysis of variance results showed that, except for SBP among males aged 1 4 5 years (on which occupation has a significant effect), the categorical variables do not have any significant effect on either SBP or DBP for all the other sex and age categories. The significant effect of occupation on SBP for males 5 4 5 years vanished when users of alcohol and antihypertensive drugs and tension sufferers were removed from the analysis. Prior to path analysis, we estimated the four familial correlations from the data using maximum likelihood methods. These, presented in Table 7, are remarkably similar for SBP and DBP. The lack of a significant marital correlation suggests that cohabitation effects per se are absent, at least in adults. The mother-child and sib-sib correlations are of similar magnitudes for both SBP and DBP. In contrast, however, the father-child correlations are much smaller and are virtually nonexistent for DBP. The results of the path analysis are sum-
570
P.P. W U M D E R ET AL.
marized in Table 8, which presents estimates of parameters under each of several hypotheses, with the associated values of -21nL+C, where 1nL is the log likelihood and C is a constant. For each BP variable, we fitted five hypotheses: a general hypothesis with ‘TF = 0.5;a pseudopolygenic hypothesis given by TF = T~ = 0.5; the hypotheses of no paternal transmission (TF = 0 ) and no sibship residual correlation (s = 0 ) ; and the hypothesis of no family resemblance (all parameters set t o zero). Note that not all four hypotheses are hierarchical, and, therefore, standard likelihood ratio tests cannot be performed €or each hypothesis. However, several salient features emerge for each of the two phenotypes. First, familial aggregation is highly significant. Second, common sibship residual correlation is not significant. Third, specific maternal effects are not present, as indicated by similar estimates of ‘TF and TM. Finally, the generalized heritability (t2),which provides an upper-bound for genetic heritability, is only moderatearound 0.3. Although there is clear evidence of familial aggregation of both SBP and DBP in these families, the determination of a model to account for the familial aggregation turned out to be difficult, especially for DBP. Since the models are not hierarchical, and hence likelihood ratio tests cannot be applied, we decided to use Akaike’s information criterion (AIC) (1985) for model selection. AIC is defined as AIC = -2 In L + 2 x (number of parameters in the model), and the principle for model selection is to select the one with the smallest value of AIC (Akaike, 1985). As is seen from Table 8, for SBP, the model with the smallest value of AIC is the pseudopolygenic model with TF = ‘TM = 0.5. Under this model, the estimatesoftheparametersaret2 = 0.44 k 0.13, s = 0.15 0.16, and p = 0.01 2 0.15. For
*
DBP, however, the model with TF = 0, which is the model postulating no paternal transmission, seems to be the model of choice, and the difference in AIC value between this model and the pseudopolygenic model (that is, ‘TF = ‘TM = 0.5,which is the selected model for SBP) is 2.81. In fact, when TF is treated as a parameter, its estimate is 0.00 (Table 8) when s = 0. Under this model, the estimates ofthe other parameters are t2 = 0.46 ? 0.36, TM = 0.61 +- 0.40, and p = -0.08 0.16.
*
DISCUSSION
The present study shows that the general level of blood pressure is low in this population and, in fact, is similar to an unacculturated tribal population-the Yanomama Indians (Oliver, 1980). In India, the rural populations generally have a lower mean blood pressure than the urban populations (Padmavati and Gupta, 1959; Dalal, 1978). However, unlike most other lowblood-pressure populations in which blood pressure does not increase with age (Oliver, 1980), in the present study age is found to have a significant effect on blood pressure. The effect is more pronounced on DBP than on SBP. While systolic BP tends to rise more sharply with age after 45 years in both males and females, diastolic BP is seen to rise until 45 years, after which there is a slight decline with advancing age. These trends are similar to those found in the Framingham study (Kannel, 1980). It may, however, be pointed out that the reports claiming that unacculturated populations show little or no blood pressure increase with age have recently been criticized (Beilin, 1988).Thus our finding of a significant effect of age on blood pressure may not really indicate that the population we have studied is significantly acculturated. Indeed our experience with and studies on this population do not point to
TABLE 8. Results o f p a t h analysrs o f blood pressure Phenotype
T2
S
Systolic BP
0.38 0.44 0.46 0.30 (0) 0.25 0.27 0.44 0.46
0.09 0.15 0.35 (0) (0) -0.09 0.10 -0.01 (0)
(0)
(0)
Diastolic BP
‘Fixed values are shown in parentheses
Tp
TM
(0.5)’ (0.5)
0.63 (0.5) 0.52 0.78 (0)
(0)
0.62
(0) (0.5) (0.5) (0) 0.00
(0)
0.88
(0.5) 0.64 0.61 (0)
P
0.00 0.01 -0.00 0.00 (0) -0.11 -0.08 -0.08 -0.08 (0)
-2lnLfC
0.00 0.13 2.83 0.00 22.09 2.35 4.81 0.00 0.00 12.70
AIC+C 8.00 6.13 10.83 8.00 22.09
10.35 10.81 8.00 8.00 12.70
BLOOD PRESSURE IN EASTERN INDIA
recent acculturation to any significant degree. Further, Beilin (1988)has pointed out that chronic infection has potential blood-pressure-lowering effects, which is significant in the present context. Our investigations in this population have shown that there is a very high prevalence of chronic infections of the gastrointestinal tract (Bhattacharya et al., manuscript in preparation). Thus the low prevalence of hypertension and low mean blood pressure may actually be a result of chronic infections. It is, therefore, possible that this population may quickly catch up to the levels of hypertension observed in urbanized societies as soon as the infection rate is brought under control. Indeed, this may have been one of the confounding factors in previous studies of migrant populations (Beilin, 1988). The environmental determinants of blood pressure, e.g., occupation, tobacco use, and so forth, do not have a significant effect on this population. While the effects of environmental variables on blood pressure have been found to vary from one study to another, it is highly probable that the effects of many environmental variables are only transient and not sustained (Oliver, 1980). Moreover, many environmental variables implicated to have effects on blood pressure actually play indirect roles. For example, in the Framingham study it was found that cigarette smokers had lower blood pressures than nonsmokers, and when smokers quit smoking they experienced a rise in blood pressure (Kannel, 1980). The reason for this was found to be that smokers weighed less than nonsmokers, and, since weight or obesity positively influences blood pressure (Chiang et al., 19691, smokers had lower blood pressure values. When smokers quit they experience weight gain, resulting in an increase in blood pressure. In the present study, since we regressed out the effects of concomitant variables such as weight and skinfold thicknesses from blood pressure before studying the effects of environmental variables, such indirect effects were not discernible. The low blood pressure level in this population may be a reflection of the fact that these individuals, by and large, have low body weight and little body fat because of the combined effects of undernutrition, regular physical activity, and perhaps also chronic infection of the gastrointestinal tract. Further, although the long-term effect of mental stress on blood pressure is debated, the fact remains that
571
the individuals in this population suffer from little psychological stress. The prevalence of hypertension (about 1.5%)is certainly lower than that found in most populations. In fact, it may be interesting to investigate whether the World Health Organization’s definition of hypertension should be applied to such low-blood-pressure populations, because it may well be that clinical problems associated with hypertension may actually arise at lower blood pressure levels among individuals in such populations. Finally, the nature of familial aggregation of blood pressure in this population appears to be similar, in many ways, to what is known in other populations. The maximum genetic heritability estimate of 0.3 in this study is comparable to that in many other populations (e.g., see Morton et al., 1980; Krieger et al., 1980; Iselius et al., 1983).The higher sib-sib correlation compared with parent-ffspring correlation observed in this study is in concordance with previous studies (Biron and Mongeau, 1978; Annest et al., 1979; Rice et al., 1989). In general, the values of correlation coefficientsbetween different relative pairs are smaller in this population than those reported in some earlier studies (Annest et al., 1979). It may also be noted that the spouse-spouse correlation coefficients for both SBP and DBP are smaller in this population than in many other populations. Annest et al. (1979)have also noted a higher mother-child correlation than father-child correlation. Although this trend has been found in the present study, father-child correlation in this population is virtually zero both for SBP and DBP; the reason for this remains unclear. While for SBP the pseudopolygenic model with TF = TM = 0.5 seems to be the favored model, for DBP there seems to be some conflicting evidence. Although the generalized heritability for DBP is between 0.25 and 0.46 (Table 8), the father-child correlation is near zero (Table 7) and the chosen model favors the hypothesis of no paternal transmission. Thus the role of genetic factors in the familial aggregation of DBP in this population is unclear and may actually be minimal. We plan to collect further family data to resolve this issue. ACKNOWLEDGMENTS
Thanks are due to the Director, Indian Statistical Institute, Calcutta, for providing logistic help during field work. This work was partially supported by grants GM28719, HL33973, and MH31302 to D.C.R.
572
P.P. W D E R ET AL. LITERATURE CITED
Akaike H (1985)Prediction and entropy. In AC Atkinson and SE Feinberg (eds.): A Celebration of Statistics: The IS1 Centenary Volume. New York: Springer-Verlag, pp. 1-24. Annest JL, Sing CF, Biron P, and Mongeau J-G (1979) Familial aggregation of blood pressure and weight in adoptive families. 11. Estimation of the relative contributions of genetic and common environmental factors to blood pressure correlations between family members. Am. J . Epidemiol. 110:492-503. Beilin U (1988) Epitaph to essential hypertension-A preventable disorder of known aetiology? J. Hypertens. 6:85-94. Biron P, and Mongeau J-G (1978) Familial aggregation of blood pressure and its components. Pediatr. Clin. North Am. 25:29-33. Chiang BN, Perlman LV, and Epstein FH (1969) Overweight and hypertension: A review. Circulation 39:403-421. Cloninger CR (1980) Interpretation of intrinsic and extrinsic structural relations by path analysis: Theory and applications to assortative mating. Genet. Res. 35:733-756. Dalal PM (1978) Community survey of hypertension in “ O l d Bombay. In G Thurn and P Phipps (eds.):Essential Hypertension Symposium. Miami: Symposia Specialists, pp. 139-148. Epstein FH, and Eckhoff RD (1967)The epidemiology of high blood pressure-Geographic distribution and etiological factors. In J Stamler, R Stamler, and TM Pullman (eds.): The Epidemiology of Hypertension. New York: Grune & Stratton, pp. 155-166. Iselius L, Morton NE, and Rao DC (1983) Family resemblance of blood pressure. Hum. Hered. 33:277-286. Kannel WB (1980) Host and environmental determinants of hypertension: Perspective from the Framingham study. In H Kesteloot and JV Joossens (eds.):
Epidemiology of Arterial Blood Pressure. The Hague: Martinus Nijhoff Publishers, pp. 265-295. Krieger H, Morton NE, Rao DC, and Azevado E (1980) Familial determinants of blood pressure in Northeastern Brazil. Hum. Genet. 53:415-418. Marmot MG (1979)Epidemiological basis for the prevention of coronary heart disease. Bull. WHO 57:331-347. Morton NE, Gulbrandsen CL, Rao DC, Rhoads GG, and Kagan A (1980) Determinants of blood pressure in Japanese-American families. Hum. Genet. 53: 261-266. Neilson G, and Williams G (1978) Blood pressure and valvular and congenital heart disease in Torres Strait islanders. Med. J. Aust. [Suppl.] Ir12-16. Oliver WJ (1980) Sodium homeostasis and low blood pressure populations. In H Kesteloot and JV Joossens (eds.): Epidemiology of Arterial Blood Pressure. The Hague: Martinus Nijhoff Publishers, pp. 229-241. Padmavati S, and Gupta S (1959)Blood pressure studies in rural and urban groups in Delhi. Circulation 19:395405. Rao DC, McGue M, 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 Co. Inc., pp. 35-81. Rice T, Vogler GP, Perusse L, Bouchard C, and Rao DC (1989) Cardiovascular risk factors in a French Canadian population: Resolution of genetic and familial environmental effects on blood pressure using twins, adoptees, and extensive information on environmental correlates. Genet. Epidemiol. (Manuscript in submission.) Siervogel RM (1983) Genetic and familial factors in essential hypertension and related traits. Yearbk Phys. Anthropol. 26:37-63. Ward RH (1983) Genetic and sociocultural components of high blood pressure. Am. J. Phys. Anthropol. 62r91-105.
Genetic Epidemiological Study of Blood Pressure in a Sedentary Rural Agricultural Population of West Bengal, India PARTHA P. MAJUMDER, S.K. BHAlTACHARYA, B.N. MUKHERJEE, AND D.C. RAO Human Genetics Division, Department of Biostatistics, University of Pittsburgh, Pittsburgh, (P.P.M.); Anthropornetry and Human Genetics Unit, Indian Statistical Institute, Calcutta, India (P.P.M., S.K.B., B.N.M.); Division of Biostatistics, Washington University School of Medicine, St. Louis, (D.C.R.).
KEY WORDS
Family data, Path Analysis, Hypertension
ABSTRACT
To study the genetic epidemiology of blood pressure (BP), data on 78 families were collected from a sedentary agricultural population of eastern India. The general levels of both systolic (SBP) and diastolic (DBP) blood pressures are found to be low (mean SBP = 106.41 mm Hg; mean DBP = 63.94 mm Hg). Trends of blood pressures with age are similar to those reported earlier (e.g., in the Framingham study). Environmental variables - e . g . , occupation and tobacco u s e 4 0 not have any direct significant effect on blood pressure variability in this population. Path analysis of family data shows a highly significant familial aggregation and yields a genetic heritability (maximum) estimate of 0.3 for both SBP and DBP. Sib-sib and mother-child correlation estimates are, respectively, 0.3 and 0.25. Father-child correlation estimates are 0.13 for SBP and near zero for DBP. A pseudopolygenic model yields the best fit to the data on SBP, while for DBP a proper resolution of various models considered could not be obtained.
It is well-established that essential hypertension is a major public health burden. It is an important contributory factor to coronary heart disease, stroke, and kidney failure (Siervogel, 1983). Although risk factors for hypertension seem to cut across cultural patterns and geographic regions (Marmot, 1979), in contrast to many westernized societies, the majority of traditional societies have low mean values of blood pressure and are characterized by an insignificant increase of blood pressure with age (Epstein and Eckhoff, 1967). Many studies have also shown that the etiologic factors affecting the distributions of blood pressure in traditional societies are different from those in westernizedlurbanized societies (Siervogel, 1983; Ward, 1983). However, even among traditional societies, the factors identified to be significant determinants of blood pressure distributions are not the same (Ward, 1983), and there are exceptions to the general finding that hypertension is rare in traditional societies (Neilson and Williams, 1978; Marmot, 1979). While the etiologic
@ 1990 WILEY-I,ISS, I N C
factors contributing to a rise in the general level of blood pressure in a population may be many, the most important are salt intake and de ees of obesity, physical activity, and psycho ogical stress (Oliver, 1980; Ward, 1983). It may be pointed out here that salt intake and obesity are primarily related to diet and physical activity. In spite of large environmental effects on blood pressure, the within-population percentage of variability in blood pressure attributable to familial aggregation is fair1 constant: 2040% (Siervogel, 1983; W a r l 1983). Doubtless, therefore, genetic factors-major genes or polygenes-play a considerable role in determining blood pressure variability within populations. The present investigation is motivated by the fact that there are no systematic genetic epidemiological studies conducted on sedentary rural populations of India that provide a contrast to the westernized populations in
!?.
Received April 12,1989; revision accepted June 22,1989
564
P.P. MAJUMDER ET AL
women spend most of their days performing household chores, but sometimes they help the men in the fields. The standard of living is low: The majority live in mud-walled, thatched houses. They eat large quantities of rice and rice products with small amounts of lentils and leafy vegetables cooked in oil, but eat little or no animal protein. Although we have no quantitative data, the salt intake of individuals in this population is moderate. Rice is cooked without salt; salt is used while cooking lentils and vegetables. Major meals are eaten at home. Eating out, including snacks, is not popular mainly because of lack of money. Drinking water does not contain an excessive amount of salt. Three consecutive blood pressure (systolic and diastolic) and pulse rate readings were taken on each individual above 15 years of age in the sitting position with an electronic digital instrument. Data were gathered on tobacco use (smoking in the case of males and chewing in the case of females), alcohol consumption, use of antihypertensive drugs, and whether suffering from any major tension. (Major tension was a personal assessment of the respondant of immediate psychological stress and pertained to the fortnight preceding the date of survey. Death, birth, and marriage of immediate family members, school and college examinations, and so forth, were cited by the investigator as events that could lead to major tension.) MATERIALS AND METHODS There is no use of contraceptive drugs by A total of 78 randomly selected nuclear women. The following anthropometric meafamilies, comprising 325 individuals, were surements were taken on each individual: screened during June to August 1986. There height, weight, and skinfold thicknesses (biwas no instance of multiple marriage in ceps, triceps, and subscapular). these families. Thus no stepchildren are inFor purposes of uniformity, we have concluded in the data, and all sibs are full sibs. sidered data of only those individuals for All households belong to a set of six villages whom observations on both primary (systolic located in the southern deltaic coastal region and diastolic blood pressures) and secondary of the state of West Bengal, about 160 km (tobacco use, height, weight, and so forth) southwest of Calcutta, India. The villages variables were completely available. These are small, with about 60 households on aver- were available on 281 individuals (164 males age in each village. All households belong to and 117 females) distributed among 77 nuthe Hindu middle caste. From a social point clear families. It is pertinent to point out that of view, therefore, the present sample is all male offspring (irrespective of age) and fairly homogeneous. There was no evidence unmarried female offspring (age at marriage of consanguinity up to the second-cousin usually between 17 and 22 years) were living level. at home with their parents; married female Agriculture is the primary occupation of offspring belonged to separate households, the inhabitants of these villages. Most of the but such offspring included in the present agriculture is nonmechanized and tradi- data set lived in close geographical proximity tional; agricultural labor is primarily per- to the parental home and represent about formed by men. Some men are associated 20% of all female offspring included in this with the transportation business. The data set.
respect to diet, which is a prime determinant of blood pressure. While the diet in westernized societies is primarily high in protein and fat content, the rural populations of India thrive on a high carbohydrate diet with little fat and protein. Rural India is also primarily agricultural, and since agriculture is mostly nonmechanized, farmers are regularly engaged in much farm-related physical activity. Psychological stress, which is another prime determinant of blood pressure, is more difficult to quantify, but perhaps traditional societies are under lower stress than highly industrialized societies. Thus the rural sedentary population of India perhaps provides an extreme contrast for studies of blood pressure. It is, therefore, of interest to examine the role of various factors in the determination of blood pressure within a population characterized by a high carbohydrate diet, high physical activity, low intake of fat, moderate to severe protein malnutrition, and perhaps also low psychological stress. With these aims in view, a project was undertaken in 1986 by the Indian Statistical Institute, Calcutta, India, to study the epidemiology of blood pressure under contrasting ecosystems. The present paper reports some findings of the first phase of the project undertaken among a sedentary agricultural population from the coastal region of West Bengal, India.
BLOOD PRESSURE IN EASTERN INDIA
565
variables, a subset of significant variables was identified by using a stepwise regression algorithm. Having identified a subset of significant predictor variables, standardized phenotypes were computed as Ps = (& - X,)/s,, where X, is the observed value of blood pressure (systolic or diastolic) for an individual, X, is the corresponding predicted value (predicted using the significant subset of predictors identified at the previous step of analysis), and sRis the standard deviation of the residuals. These standardized phenotypes were then normalized (PN)by ranking the Ps values and taking the inverse normal transformation of the ranks (Rao et al., 1984). Analyses of variance were then performed on the PNvalues to detect effects of categorical secondary variables. Although the original blood pressure measurements showed significant skewness and kurtosis, the normalized PNvalues had distributions close to the Gaussian (normal) distribution. The normalized phenotypes (PN) were then analyzed to investigate the extent of
Since blood pressure and pulse rate measurements were taken three times on each individual, we ignored the first reading and took the average of the second and third readings as representative. This was done because the variance among individuals was the highest for the first reading and also because the correlation coefficient between second and third readings (0.92-0.96) was higher than between first and second readings (0.86) or between first and third readings (0.82). To study the effects of categorical secondary variables (e.g., education and occupation) on blood pressures, we first removed the effects of those quantitative secondary variables (e.g., height and weight) that were found to have significant effects on the primary variables. Assessment of the significance of the quantitative secondary variables was done by treating this set of variables as a set of predictor variables for predicting either systolic or diastolic blood pressure. From among the set of predictor P
Fig. 1. Path model showing familial aggregation of blood pressure in nuclear families. P, T, and R denote, respectively, phenotype, transmissible factor, and resid-
ual environment. Subscripts F, M, C and C, denote, respectively,father, mother, and two cihdren. Residuals are correlated ( s ) in sibships.
566
P.P. MAJUMDER ET AL.
familial a gregation for each of the two BP variables BP and DBP using a sim le path model shown in Figure 1. The mode postulates that a phenotype (P) is determined by independent linear additive effects of a transmissible factor (T) and a residual (R): P = tT + rR, where t' represents the proportion of blood pressure variability explained by the transmissible factor. Since T includes both genetic and familial environmental factors, t' may be referred to as generalized heritability. The residual variance caused b nonfamilial factors is simply r2 = 1 - t H, since the variance of the standardized P, is unity. Vertical transmission of the transmissible factor (TI, from parents to children, is represented by the parameters 7F and TM, distinguishing paternal (7F)and maternal (7M) effects. Thus specific maternal effects can be detected by testing the inequality O f TF and T ~ In. addition to transmissible effects, a common sibship residual correlation (s) is also incorporated so that the phenotype of a child (Pc, subscript C denoting a child) is given by Pc = tTc + rcRc, where the residuals (Re) are correlated within sibships (s). Again, re2 = 1 - t2.Finally, marital resemblance is represented by a simple correlational path (a "copath," see Cloninger, 1980), p. Since the residuals (R) are assumed to be uncorrelated among other family members, they are not shown in Figure 1. This model , involves five parameters in all; p, t, T ~ TM, and s. Not all five parameters can be estimated from nuclear families, as the data basically reduce to four correlations (spouse-spouse, father-child, mother-child, and sib-sib). At most four parameters can be estimated uniquely, by fixing any one of 7F, 7111, or s. However, the analysis can still provide useful information on the magnitude and nature of familial aggregation. Under the model, the four familial correlations have the expectations given in Table 7. The log likelihood function for a given nuclear family is taken as the logarithm of an appropriate multivariate normal density function, defining the joint distribution of the phenotypes in the family. The corresponding covariance matrix is defined in terms of the four expected correlations given above. The overall log likelihood for the entire random sample of families is given by the sum of all individual functions, one for each family. Standard maximum likelihood methods and likelihood ratio tests are used to fit the model t o the data.
5
P
567
BLOOD PRESSURE IN EASTERN INDIA
and 21 female) who reported suffering from major tension. With respect to these three categorical variables, we decided t o perform analyses 1)by ignoring the variables and 2) after excluding the “positive”cases (i.e., alcohol and antihypertensive drug users and those suffering from major tension). The frequency distributions of systolic (SBP) and diastolic (DBP) blood pressures are presented in Tables 2 and 3, separately for each sex. Sex differences are not striking. The skewness and kurtosis values, except for SBP among males (skewness = 1.2, kurtosis = 3.61, are small. By using the World Health Organization’s definition of hypertension (SBP > 160 mm Hg and/or DBP > 95 mm Hg), only four individuals (one male and three female) are found to be hypertensive; this yields a prevalence of 1.42%of hypertension. It may be pointed out that all four individuals belong to separate households. The upper 95%cut-off points for the distributions of SBP among males, females, and in the total sample are (approximately) 130, 142, and 140 mm Hg, respectively; the corresponding values for DBP are 78, 88, and 85 mm Hg. Trends of mean blood pressures are presented in Table4 separately for
RESULTS
Table 1gives the distribution of the sample subjects, by sex, with respect to age groups and three categorical variables-ducation, occupation, and tobacco use. It should be pointed out that although at the time of data collection the information was collected in finer categories, to avoid vagaries of small cell frequencies certain categories had to be merged. For example, the occupation category “manual labor” comprised separate categories, e.g., fishing, agriculture, that were pooled; the category “other” comprised separate categories, e.g., white-collar work, household work, that were pooled. Similarly, tobacco users were also initially classified as medium, heavy, and light, but had to be pooled for reasons of small sample size. From the sample distribution (see Table 11,we decided to ignore occupational differences for females. With respect to the other categorical variables, there were in all 13 consumers of alcohol (12 male and 1female), of whom 10 were irregular or casual consumers. There were, in all, six users (three male and three female) of antihypertensive drugs and 36 people (15 male TABLE 2. Frequency distribution SRP (mm Hg)
580 81-90 91-100 101-110 111-120 121-130 131-140 141-150 151-160 2160
Total
Male Frequency
Percent
2 15 36 58 33 11 4 4 0 1 164
1.2 9.2 22.0 35.4 20.1 6.7 2.4 2.4 0.0 0.6 100.0
of systolic blood pressure
Female Frequency
1 9 33 37 17 8 5 6 1 0 117
(SBP) Total
Percent
Frequency
0.9 7.7 28.2 31.6 14.5 6.8 4.3 5.1 0.9 0.0 100.0
3 24 69 95 50 19 9 10
Percent
1.1 8.5 24.5 33.8 17.8 6.8 3.2
1
3.5 0.4
1 28 1
0.4 100.0
TABLE 3. Frequency distribution of diastolic blood pressure IDBP) DBP (mm Hg) 540 4 1-50 5 1-60 61-70 71-80 81-90 91-100 2100 Totaf
Male Frequency 6
20 47 61 24 5 0 1 164
Percent
3.7 12.2 28.7 37.2 14.6 3.0 0.0 0.6 100.0
Female Frequency
1 6 20 41 34 11 3 1
117
Total Percent
Frequency
0.9 5.1 17.1 35.0 29.1 9.4 2.6 0.9 100.0
7 26 67 10’2 58 16 3 2 281
Percent
2.5 9.3 23.8 36.3 20.6 5.7 1.1
0.7 100.0
P.P. W U M D E R ET AL
a
m
-m c c
a
m v)
P.
m L?
3
7 m W
a
m
v)
PI
m L? W d
m n a
3 a
m
m d
C
' 9
"a
ffi (I:
m a CI
Lc
7
m W
a
3
males and females. Analysis of variance was performed to test age and sex differences in mean values. For SBP it was found that although the main effect of sex is nonsignificant at the 5% level (F ratio = 0.529, d.f. = l,269), there is a significant sex x age interaction (F ratio = 2.545, d.f. = 5,269). For DBP, the main effect of sex is significant (F ratio = 29.233; d.f. = 1,269),but the sex x age interaction is absent (F ratio = 0.448, d.f. = 5,269). There is a pronounced effect of age for both SBP (F ratio = 6.689, d.f. = 5,269) and DBP (F ratio = 21.495, d.f. = 5,269). In all further analyses, we have, therefore, retained both the sex and age group classifications. However, to avoid the vagaries of small sample sizes, certain age groups were pooled. We examined the scatter diagrams of age vs. SBP/DBP and found a difference in trend (for both males and females) above and below 45 years of age. This is also clearly seen from the correlation coefficients presented in Table 5. For SBP, the correlation between age and SBP is much lower for individuals (both male and female) below the age of 45 years in comparison with those above the age of 45 years. For DBP, there is reversal in trend; the correlation is positive, large, and significant (among both males and females) for individuals below 45 years of age, while it is negative, smaller, and nonsignificant for individuals above 45 years of age. We have, therefore, taken 45 years as a cut-off point and have analyzed data on individuals 5 4 5 years and >45 years of age separately. Since blood pressures are correlated with age, anthropometric measurements, and pulse rate, we decided to remove the effects of these quantitative variables before studying the effects of categorical variables such as education and occupation. To allow for nonlinear trends of blood pressures with age, we have also included age2 and age3 in the list of independent variables along with age, height, weight, skinfold thicknesses, and pulse rate. As mentioned earlier, since the relationships of blood pressure and age are different for different sex and age subgroups, we have obtained separate regression equations for males and females above and below 45 years of age. It was found that of the nine predictor variables (age, age2, age3, pulse rate, height, weight, skinfold thicknesses [biceps, triceps, and subscapularl), in most subgroups only two or three variables were significant. The results of the stepwise re-
569
BLOOD PRESSURE IN EASTERN INDIA
T A B L E 5 Correlation coefficients between age and blood uressure in individuals helongma to different age clashes Variables'
All
Age, SBP Age, RDP
0.215* 0.483*
Male 5 4 5 vears
>45 vears
All
Female 5 4 5 vears
>46 vears
0.087 0.486*
0.226 -0.180
0.418* 0.485*
0.074 0.359*
0.280 -0.1 62
I SBP, systnhc blood pressure; DBP, diastolic hloud pressure *Significant a t the 5'Xt level.
T A B L E 6. Results of stepwise regression analysis: Significant predictor variables and regression coefficients Sex and aee subprouo Sex Age (years)
Regression equation SBP DBP SBP DBP
= 74.34 + 0.63 (weight) = 9.29 0.52 (age) 0.42 (weight) 0.19 (pulse rate) = 1.22 0.73 (age) 1.37 (weight) = 29.80 1.03 (weight) 0.85 (subscapular skinfold thickness)
545
SRP DRP
>45
SBP DBP
= = = =
Male
545
Male
>45
Female Female
+ + + + + 73.58 t 0.81 (subscapular skinfold thickness) + 0.18 (pulse rate) 25.86 + 0.42 (age) + 0.69 (subscapular skinfold thickness) + 0.18 (pulse rate) 70.60 + 1.13 (weight) 42.39 t 0.62 (subscapular skinfold thickness) + 0.28 (pulse rate) +
-
T A B L E 7. M a x i m u m likelihood estimatesof familial correlations for systolic (SHP)and diastolic (DBP) bloodpressure Maximum likelihood estimate Correlation Spouse-spouse Father-child Mother-child Sib-sih
SBP
DBP
-0.01 t 0.15 0.1'3 i 0.14 0.24 0.12 0.31 i 0.13
-0.04 i 0.15 -0.04 i 0.15
+
gression analyses are presented in Table 6. It is seen that age and weight are the two important predictors of blood pressure. Nonlinear trends in age were found to be absent. For further analyses, standardized Z scores of the phenotypic data were computed that were then normalized using the method described in the previous section. The effects of the categorical variables were studied using analysis of variance techniques on the normalized blood pressure values. For males we have considered the effects of education, occupation, and tobacco use separately, while for females we have ignored the occupation classification for reasons mentioned earlier. Apart from performing analysis of variance separately for sex and age subgroups, we have also performed (within each sex and age subgroup) analyses separately by considering 1) all individuals ignoring alcohol use, antihypertensive drug use, and tension suffering; and 2) only individuals who are nonusers of either alcohol or antihypertensive drugs and not suffering from ma-
*
0.26 0.12 0.32 i 0.13
Expected correlation under the path model P t 2 ( 7 y + p TM)
ti s
( 7+ ~ p
rF)
+ tz ( Tp + T M ' ) + 2 7v 7M)
jor tension. Analysis of variance results showed that, except for SBP among males aged 1 4 5 years (on which occupation has a significant effect), the categorical variables do not have any significant effect on either SBP or DBP for all the other sex and age categories. The significant effect of occupation on SBP for males 5 4 5 years vanished when users of alcohol and antihypertensive drugs and tension sufferers were removed from the analysis. Prior to path analysis, we estimated the four familial correlations from the data using maximum likelihood methods. These, presented in Table 7, are remarkably similar for SBP and DBP. The lack of a significant marital correlation suggests that cohabitation effects per se are absent, at least in adults. The mother-child and sib-sib correlations are of similar magnitudes for both SBP and DBP. In contrast, however, the father-child correlations are much smaller and are virtually nonexistent for DBP. The results of the path analysis are sum-
570
P.P. W U M D E R ET AL.
marized in Table 8, which presents estimates of parameters under each of several hypotheses, with the associated values of -21nL+C, where 1nL is the log likelihood and C is a constant. For each BP variable, we fitted five hypotheses: a general hypothesis with ‘TF = 0.5;a pseudopolygenic hypothesis given by TF = T~ = 0.5; the hypotheses of no paternal transmission (TF = 0 ) and no sibship residual correlation (s = 0 ) ; and the hypothesis of no family resemblance (all parameters set t o zero). Note that not all four hypotheses are hierarchical, and, therefore, standard likelihood ratio tests cannot be performed €or each hypothesis. However, several salient features emerge for each of the two phenotypes. First, familial aggregation is highly significant. Second, common sibship residual correlation is not significant. Third, specific maternal effects are not present, as indicated by similar estimates of ‘TF and TM. Finally, the generalized heritability (t2),which provides an upper-bound for genetic heritability, is only moderatearound 0.3. Although there is clear evidence of familial aggregation of both SBP and DBP in these families, the determination of a model to account for the familial aggregation turned out to be difficult, especially for DBP. Since the models are not hierarchical, and hence likelihood ratio tests cannot be applied, we decided to use Akaike’s information criterion (AIC) (1985) for model selection. AIC is defined as AIC = -2 In L + 2 x (number of parameters in the model), and the principle for model selection is to select the one with the smallest value of AIC (Akaike, 1985). As is seen from Table 8, for SBP, the model with the smallest value of AIC is the pseudopolygenic model with TF = ‘TM = 0.5. Under this model, the estimatesoftheparametersaret2 = 0.44 k 0.13, s = 0.15 0.16, and p = 0.01 2 0.15. For
*
DBP, however, the model with TF = 0, which is the model postulating no paternal transmission, seems to be the model of choice, and the difference in AIC value between this model and the pseudopolygenic model (that is, ‘TF = ‘TM = 0.5,which is the selected model for SBP) is 2.81. In fact, when TF is treated as a parameter, its estimate is 0.00 (Table 8) when s = 0. Under this model, the estimates ofthe other parameters are t2 = 0.46 ? 0.36, TM = 0.61 +- 0.40, and p = -0.08 0.16.
*
DISCUSSION
The present study shows that the general level of blood pressure is low in this population and, in fact, is similar to an unacculturated tribal population-the Yanomama Indians (Oliver, 1980). In India, the rural populations generally have a lower mean blood pressure than the urban populations (Padmavati and Gupta, 1959; Dalal, 1978). However, unlike most other lowblood-pressure populations in which blood pressure does not increase with age (Oliver, 1980), in the present study age is found to have a significant effect on blood pressure. The effect is more pronounced on DBP than on SBP. While systolic BP tends to rise more sharply with age after 45 years in both males and females, diastolic BP is seen to rise until 45 years, after which there is a slight decline with advancing age. These trends are similar to those found in the Framingham study (Kannel, 1980). It may, however, be pointed out that the reports claiming that unacculturated populations show little or no blood pressure increase with age have recently been criticized (Beilin, 1988).Thus our finding of a significant effect of age on blood pressure may not really indicate that the population we have studied is significantly acculturated. Indeed our experience with and studies on this population do not point to
TABLE 8. Results o f p a t h analysrs o f blood pressure Phenotype
T2
S
Systolic BP
0.38 0.44 0.46 0.30 (0) 0.25 0.27 0.44 0.46
0.09 0.15 0.35 (0) (0) -0.09 0.10 -0.01 (0)
(0)
(0)
Diastolic BP
‘Fixed values are shown in parentheses
Tp
TM
(0.5)’ (0.5)
0.63 (0.5) 0.52 0.78 (0)
(0)
0.62
(0) (0.5) (0.5) (0) 0.00
(0)
0.88
(0.5) 0.64 0.61 (0)
P
0.00 0.01 -0.00 0.00 (0) -0.11 -0.08 -0.08 -0.08 (0)
-2lnLfC
0.00 0.13 2.83 0.00 22.09 2.35 4.81 0.00 0.00 12.70
AIC+C 8.00 6.13 10.83 8.00 22.09
10.35 10.81 8.00 8.00 12.70
BLOOD PRESSURE IN EASTERN INDIA
recent acculturation to any significant degree. Further, Beilin (1988)has pointed out that chronic infection has potential blood-pressure-lowering effects, which is significant in the present context. Our investigations in this population have shown that there is a very high prevalence of chronic infections of the gastrointestinal tract (Bhattacharya et al., manuscript in preparation). Thus the low prevalence of hypertension and low mean blood pressure may actually be a result of chronic infections. It is, therefore, possible that this population may quickly catch up to the levels of hypertension observed in urbanized societies as soon as the infection rate is brought under control. Indeed, this may have been one of the confounding factors in previous studies of migrant populations (Beilin, 1988). The environmental determinants of blood pressure, e.g., occupation, tobacco use, and so forth, do not have a significant effect on this population. While the effects of environmental variables on blood pressure have been found to vary from one study to another, it is highly probable that the effects of many environmental variables are only transient and not sustained (Oliver, 1980). Moreover, many environmental variables implicated to have effects on blood pressure actually play indirect roles. For example, in the Framingham study it was found that cigarette smokers had lower blood pressures than nonsmokers, and when smokers quit smoking they experienced a rise in blood pressure (Kannel, 1980). The reason for this was found to be that smokers weighed less than nonsmokers, and, since weight or obesity positively influences blood pressure (Chiang et al., 19691, smokers had lower blood pressure values. When smokers quit they experience weight gain, resulting in an increase in blood pressure. In the present study, since we regressed out the effects of concomitant variables such as weight and skinfold thicknesses from blood pressure before studying the effects of environmental variables, such indirect effects were not discernible. The low blood pressure level in this population may be a reflection of the fact that these individuals, by and large, have low body weight and little body fat because of the combined effects of undernutrition, regular physical activity, and perhaps also chronic infection of the gastrointestinal tract. Further, although the long-term effect of mental stress on blood pressure is debated, the fact remains that
571
the individuals in this population suffer from little psychological stress. The prevalence of hypertension (about 1.5%)is certainly lower than that found in most populations. In fact, it may be interesting to investigate whether the World Health Organization’s definition of hypertension should be applied to such low-blood-pressure populations, because it may well be that clinical problems associated with hypertension may actually arise at lower blood pressure levels among individuals in such populations. Finally, the nature of familial aggregation of blood pressure in this population appears to be similar, in many ways, to what is known in other populations. The maximum genetic heritability estimate of 0.3 in this study is comparable to that in many other populations (e.g., see Morton et al., 1980; Krieger et al., 1980; Iselius et al., 1983).The higher sib-sib correlation compared with parent-ffspring correlation observed in this study is in concordance with previous studies (Biron and Mongeau, 1978; Annest et al., 1979; Rice et al., 1989). In general, the values of correlation coefficientsbetween different relative pairs are smaller in this population than those reported in some earlier studies (Annest et al., 1979). It may also be noted that the spouse-spouse correlation coefficients for both SBP and DBP are smaller in this population than in many other populations. Annest et al. (1979)have also noted a higher mother-child correlation than father-child correlation. Although this trend has been found in the present study, father-child correlation in this population is virtually zero both for SBP and DBP; the reason for this remains unclear. While for SBP the pseudopolygenic model with TF = TM = 0.5 seems to be the favored model, for DBP there seems to be some conflicting evidence. Although the generalized heritability for DBP is between 0.25 and 0.46 (Table 8), the father-child correlation is near zero (Table 7) and the chosen model favors the hypothesis of no paternal transmission. Thus the role of genetic factors in the familial aggregation of DBP in this population is unclear and may actually be minimal. We plan to collect further family data to resolve this issue. ACKNOWLEDGMENTS
Thanks are due to the Director, Indian Statistical Institute, Calcutta, for providing logistic help during field work. This work was partially supported by grants GM28719, HL33973, and MH31302 to D.C.R.
572
P.P. W D E R ET AL. LITERATURE CITED
Akaike H (1985)Prediction and entropy. In AC Atkinson and SE Feinberg (eds.): A Celebration of Statistics: The IS1 Centenary Volume. New York: Springer-Verlag, pp. 1-24. Annest JL, Sing CF, Biron P, and Mongeau J-G (1979) Familial aggregation of blood pressure and weight in adoptive families. 11. Estimation of the relative contributions of genetic and common environmental factors to blood pressure correlations between family members. Am. J . Epidemiol. 110:492-503. Beilin U (1988) Epitaph to essential hypertension-A preventable disorder of known aetiology? J. Hypertens. 6:85-94. Biron P, and Mongeau J-G (1978) Familial aggregation of blood pressure and its components. Pediatr. Clin. North Am. 25:29-33. Chiang BN, Perlman LV, and Epstein FH (1969) Overweight and hypertension: A review. Circulation 39:403-421. Cloninger CR (1980) Interpretation of intrinsic and extrinsic structural relations by path analysis: Theory and applications to assortative mating. Genet. Res. 35:733-756. Dalal PM (1978) Community survey of hypertension in “ O l d Bombay. In G Thurn and P Phipps (eds.):Essential Hypertension Symposium. Miami: Symposia Specialists, pp. 139-148. Epstein FH, and Eckhoff RD (1967)The epidemiology of high blood pressure-Geographic distribution and etiological factors. In J Stamler, R Stamler, and TM Pullman (eds.): The Epidemiology of Hypertension. New York: Grune & Stratton, pp. 155-166. Iselius L, Morton NE, and Rao DC (1983) Family resemblance of blood pressure. Hum. Hered. 33:277-286. Kannel WB (1980) Host and environmental determinants of hypertension: Perspective from the Framingham study. In H Kesteloot and JV Joossens (eds.):
Epidemiology of Arterial Blood Pressure. The Hague: Martinus Nijhoff Publishers, pp. 265-295. Krieger H, Morton NE, Rao DC, and Azevado E (1980) Familial determinants of blood pressure in Northeastern Brazil. Hum. Genet. 53:415-418. Marmot MG (1979)Epidemiological basis for the prevention of coronary heart disease. Bull. WHO 57:331-347. Morton NE, Gulbrandsen CL, Rao DC, Rhoads GG, and Kagan A (1980) Determinants of blood pressure in Japanese-American families. Hum. Genet. 53: 261-266. Neilson G, and Williams G (1978) Blood pressure and valvular and congenital heart disease in Torres Strait islanders. Med. J. Aust. [Suppl.] Ir12-16. Oliver WJ (1980) Sodium homeostasis and low blood pressure populations. In H Kesteloot and JV Joossens (eds.): Epidemiology of Arterial Blood Pressure. The Hague: Martinus Nijhoff Publishers, pp. 229-241. Padmavati S, and Gupta S (1959)Blood pressure studies in rural and urban groups in Delhi. Circulation 19:395405. Rao DC, McGue M, 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 Co. Inc., pp. 35-81. Rice T, Vogler GP, Perusse L, Bouchard C, and Rao DC (1989) Cardiovascular risk factors in a French Canadian population: Resolution of genetic and familial environmental effects on blood pressure using twins, adoptees, and extensive information on environmental correlates. Genet. Epidemiol. (Manuscript in submission.) Siervogel RM (1983) Genetic and familial factors in essential hypertension and related traits. Yearbk Phys. Anthropol. 26:37-63. Ward RH (1983) Genetic and sociocultural components of high blood pressure. Am. J. Phys. Anthropol. 62r91-105.
