Genetic and Environmental Influences on Left Ventricular Mass
A Family Study
Abstract—The relations between left ventricular mass (LVM) with age, gender, body size, and blood pressure were investigated in healthy adults and children from 149 nuclear families. LVM was strongly correlated with overall weight, especially in the children. Genetic analysis indicated that the segregation of LVM was compatible with an additive polygenic model, with a heritability estimate of 34% before adjustment for weight and 28% after adjustment for weight. Genetic and/or familial environmental factors played a strong role in the correlation of LVM and weight; they accounted for all of the correlation between the 2 traits in adults and 59% of the correlation in children. Spouses exhibited a strong correlation in their weight, which suggested that common family environment may contribute to the family correlations and to the observed heritability of the trait. LVM was strongly correlated with blood pressure before adjustment for weight, but this correlation could be attributed to nonfamilial environment rather than genetic factors. After adjustment for weight, the intertrait correlations between LVM and blood pressure were nonsignificant. Thus, adjustment for weight accounts for all common determinants of LVM and blood pressure.
Left ventricular mass (LVM) is an important clinical measure because of its association with hypertension and its significance as a risk factor for cardiovascular disease. In both healthy individuals and patients with disease, LVM is likely to be determined by a combination of genetic factors and adaptive responses to actions of environmental and mechanical factors. During embryogenesis, heart development is under the control of a series of transcription and growth factors acting on cell differentiation and on heart morphogenesis.1
Two general and complementary approaches are possible to investigate the factors that are important in determination of LVM. The first consists of molecular approaches that assess the roles of single genes, for example, by targeted disruption of the gene in mice. The role in heart development of the nuclear factor of activated T cells (NFAT) was identified by such an approach.2 3 The human homologue of genes identified by such approaches can also be examined for variation and studied for their relation with LVM in normal individual or patients with disease. The second type of approach consists of epidemiological studies in well-characterized populations and genetic investigations in families designed to assess relations between LVM and other morphological and hemodynamic parameters. Such studies can reveal important information about genetic and environmental relations between such parameters, and ultimately they will be combined with molecular approaches to determine the full spectrum of gene and environmental factors that are responsible for variation in LVM. Here, we report a genetic epidemiology study designed to investigate the relation of LVM measured by echocardiography with weight and blood pressure in adults and offspring from 149 nuclear families. Using this approach, we show that LVM and weight are strongly correlated within individuals and between individuals within families, providing evidence for the importance of both genetic and nonfamilial environmental factors that act jointly on the 2 traits. In contrast, the relation between LVM and blood pressure appears to be largely due to nonfamilial environmental factors and not genetic factors, within the normal range of trait variation examined here.
Families and Anthropometric Measurements
One hundred forty-nine nuclear families were recruited from January 1995 to September 1995 as part of a larger cohort of 1000 white European families collected in the Center of Preventive Medicine in Vandoeuvre-Les-Nancy (Stanislas Cohort).4 All families were composed of both natural parents (33 to 60 years of age) and at least 2 offspring (8 to 30 years of age) and volunteered to have a free health check-up examination. All members of a family were examined on the same day and completed a self-administered questionnaire that provided information on medical history. All subjects gave informed consent to participate in the study, which was approved by the local ethics committee.
Blood pressure was measured by an automatic device (Dynamap) every 2 minutes in the supine position during 12 minutes, after a 5-minute rest, with an appropriate sized cuff. The mean of the last 5 measurements was used in the statistical analyses. Weight (to the nearest 100 g) and height (to the nearest 1 mm) were measured without shoes or clothes.
Volunteers underwent standard M-mode echocardiography with a commercially available ultrasonograph (Ultramark 4 ultrasound system) equipped with a 3.5-MHz transducer and strip-chart recorder. Measurements were made according to the recommendations of the American Society of Echocardiography with a leading edge–to–leading edge convention with the ultrasound beam at or just below the tips of the mitral valve leaflets and were averaged over 3 cycles.5 LVM for parents and their children was calculated according to published formula, which has been validated anatomically for children with normal hearts and for adults.6 7 LV measurements obtained by the two observers were previously shown to be highly correlated, and interreader variability was <6% for all parameters.8
Individuals were assigned to 1 of 4 classes, based on their gender and generation (ie, their status as parents or offspring). The natural logarithm was applied to all variables as a variance-stabilizing transformation. The effects of age, body mass index (BMI), height and weight on LVM, systolic blood pressure (SBP), and diastolic blood pressure (DBP) were assessed separately for each gender/generation class. Parameters for linear regression were estimated for variables showing significant effects on the trait (P<0.10) in a stepwise regression, and the adjusted variables were standardized to have mean 0 and variance 1 within each class. Linear regression analyses were carried out with the general linear model procedure of the SAS (SAS Institute, Inc) or the regression modules of BMDP (BMDP Statistical Software, Inc). The offspring were divided into age classes of 7 to 11, 12 to 13, 14 to 15, 16 to 17, and ≥18 years to test if the within-individual or parent-offspring correlations differed significantly by age class. The tests were undertaken separately for male and female offspring and for the different combinations of parent and offspring gender. The within-individual correlations of LVM with weight, SBP, and DBP and the parent-offspring correlations of LVM did not differ significantly by age class (P>0.05). Although the within-individual correlation between SBP and DBP exhibited nominally significant differences by age class in male offspring (P=0.02) and female offspring (P=0.03), as did the within-individual correlation between weight and SBP in female offspring (P=0.02), these results were nonsignificant after adjustment for the number of statistical comparisons that were made. Thus, the offspring were grouped without respect to age classes for further within-individual and familial correlation analyses. (The possibility of differences in the sib-sib correlations as a function of the age classes was not examined because of the small number of pairs for each combination of classes).
Familial correlation models were applied to estimate intratrait correlations and test various variance component models. The most general model allowed for different correlations between father-son, father-daughter, mother-son, mother-daughter, brother-brother, brother-sister, and sister-sister, with a nonzero spouse-spouse correlation. The general model of familial correlation was compared with nested models, with constraints on the parameters to test (1) absence of spouse-spouse correlation, (2) equality of sib-sib correlation irrespective of the gender of the sibling, and (3) equality of parent-offspring correlation irrespective of the gender of the parent or child. We also tested equality of parent-offspring and sib-sib correlations while imposing (2) and (3) simultaneously. This model can be described in terms of a familial component of variance, VF, and a nonfamilial environmental component or error variance, Ve, where the total trait variance is VP=VF+VE. If VF is assumed to be due to genetic factors, then this is equivalent to an additive genetic model in which the intratrait correlation for first-degree relatives is 1/2VF. The trait heritability was estimated as VF/VP.
Bivariate analysis was performed under the additive genetic model to evaluate genetic and environmental correlation between pairs of traits. We assume that the within-individual intertrait covariance (ie, the covariance between 2 different traits measured within the same individual) can be partitioned as CP=CF+CE, where CF is the familial component and CE is a nonfamilial environmental or error component. If the familial component is due to additive genetic factors, the intertrait covariance for first-degree relatives is 1/2 CF. Parameter estimates were obtained by the method of maximum likelihood. The correlation between familial factors affecting the traits has been estimated as CF/VF11/2VF21/2. Similarly, correlation between environment factors affecting the trait is CE/VE11/2VE21/2. In these equations, VF1 and VF2 stand for the familial components of variance, and VE1 and VE2 stand for the nonfamilial environmental components of variance for traits 1 and 2, respectively. In some analyses, CE was allowed to vary according the generation of the individual.
All familial correlation and segregation analyses were carried out with the computer program PAP or with BMDP (unbalanced repeated-measures models with structured covariance matrixes). Tests of nested models were made with the likelihood ratio statistic, that is, twice the difference in the natural logarithm of the likelihood of the data under the two models. Significance of the likelihood ratio statistic was evaluated by comparison with a χ2 distribution with degrees-of-freedom equal to the difference in the number of parameters estimated in the two models. The null hypothesis (model with the smallest number of parameters) was rejected at a value of P<0.05. Major gene analysis was performed by classic methods.9
LVM and Blood Pressure in Healthy Adults and Their Children
LVM, blood pressure, weight, and height were measured on 624 individuals from 149 nuclear families (Table 1⇓). Mean LVM differed significantly by gender and generation grouping. Overall, weight was the best predictor variable, that is, the most strongly correlated, of LVM in each generation/gender grouping (correlation from 0.30 in women to 0.81 in male offspring, P<10−4). This was followed by height, which is itself strongly correlated to weight. In the parental generation, adjustment for weight accounted for 17% of the variance of LVM in fathers and 12% of its variance in mothers, whereas it accounted for 70% and 34% of LVM variance in male and female offspring, respectively. The correlation of LVM with weight is significantly higher in male compared with female offspring and in the offspring generation compared with the parental generation.
Height and weight were sufficiently correlated so that in most instances adjustment for weight accounted for nearly all the variance attributed to height. Only in male offspring did a stepwise regression procedure lead to the inclusion of both weight and height as significant predictors of LVM, but the latter accounted for only an additional 2% of the variance once weight was taken into account. Weight was essentially equivalent to body surface area as a predictor of the traits, whereas height and weight were always better than BMI as a predictor of LVM Age was not significantly correlated with LVM after adjustment for weight and/or height, but adjustment for age rendered the gender differences in correlation nonsignificant in each generation (although the tendency of higher correlations in male subjects remained). In the following, attention is focused on weight as the simplest overall predictor of LVM, and LVM is adjusted for age as well as gender and generation before genetic analysis.
SBP and DBP varied significantly with age in all gender/generation combinations, with the highest correlation coefficients being observed in the offspring (from 0.28 for DBP to 0.55 for SBP in female and male offspring, respectively, P<0.004). With the exception of DBP in female subjects, weight was also significantly correlated with blood pressure in the parental generation. Weight and height showed similar correlation coefficients for both SBP and DBP in the offspring generation in which the correlation coefficients ranged from 0.69 for SBP in male subjects to 0.27 for DBP in female subjects (P<10−3). The combined effects of age and weight accounted for 12% to 14% of the total variance of blood pressure in the parental generation except for DBP in female subjects, in which only 3% of the total variance was explained by these variables. In the offspring generation, these predictors accounted for 8% to 9% of the total variance in DBP in male and female subjects. SBP showed a stronger relation with the predictor variables in the offspring, in which these accounted for 16% of the variance in female offspring and 49% in male offspring.
The relation between LVM and SBP was significant, without taking into account weight, particularly in the offspring generation. DBP exhibited a similar but less significant pattern of correlation with LVM. Once weight was taken into account, LVM was largely independent of blood pressure, whereas adjustment for the predictor variables, including weight, did not substantially modify the correlation between SBP and DBP. Correlations between the SBP and DBP ranged from 0.45 to 0.67 after adjustment for age and weight (P<10−4).
Familial Resemblance of LVM and Body Weight
Traits were regressed for age and then standardized to have mean 0 and variance 1 in each gender/generation category. Family analysis showed significant correlation between LVM measured in first-degree relatives. The data were compatible with an additive genetic model (equal parent-offspring and sibling correlations) and gave a heritability estimate of 34%.
As discussed in the previous section, LVM and weight exhibited a strong within-individual correlation (ie, correlation between the traits measured within the same individual). Thus, we also decided to examine familial correlations for weight. Parent-offspring and sibling pairs exhibited similar correlation for weight, which was again compatible with an additive genetic model; however, the spouse correlation for weight was also significant and of similar value to the correlation in first-degree relatives (0.32 for spouse pairs versus 0.31 for parent-offspring and sibling pairs). Although the significant spouse correlation suggests that nongenetic familial effects could have an important influence on weight, we did not observe any further indications of family environmental effects. For example, we found no significant differences in father-offspring, mother-offspring, or sib-sib correlations; nor did we find any significant differences in the correlations according to the gender of the members of these pairs (results not shown). The spouse correlation could also be the consequence in total or in part of assortative mating.
To further explore the relation between LVM and weight, we undertook a bivariate analysis of the two traits within families. The bivariate analysis allowed the relative contributions of familial (genetic and/or environmental) factors and nonfamilial environmental factors acting on both traits to be assessed (see Methods). For LVM and weight, the familial component of the correlation was estimated to be 0.34±0.04. A marked difference between correlations attributed to nonfamilial environmental factors was found in the parent and offspring generations. Familial factors appeared to explain all the intertrait correlation in the adult generation, whereas the intertrait correlation caused by nonfamilial environmental factors was estimated to be 0.25±0.05 in the offspring generation. The differences between the adult and offspring are consistent with the results of the correlation analysis, in which the largest correlation between LVM and weight was found in the offspring generation. We estimated that the correlation between familial factors affecting both LVM and weight was 66%, whereas the correlation between nonfamilial environmental factors affecting the two traits was 55% in offspring and nonsignificant in adults. The spouse intertrait correlation was also significant, indicating the possible importance of correlated familial environmental factors common to LVM and weight.
Table 2⇓ shows the results of the familial correlation analysis for LVM after adjustment for weight to remove effects that are common to the two traits. The parent-offspring and sibling correlations for LVM after this adjustment were 0.13±0.04 and 0.15±0.07, respectively, and the spouse correlation was nonsignificant. The heritability of LVM adjusted for weight was estimated to be 28%, reduced from 34% before adjustment. No evidence of major genes with predominant effects on the traits was found in segregation analyses of LVM and body weight in univariate or bivariate analyses (results not shown).
Familial Resemblance of Blood Pressure
Familial correlation of blood pressure was analyzed after adjustment for age and standardizing in generation/gender categories, without and with correction for weight (Table 2⇑). The pattern of correlations in first-degree relatives were consistent with an additive genetic model, with heritability estimates of 50% for SBP and 22% for DBP before adjustment for weight and 50% and 24%, respectively, after adjustment (Table 2⇑). The spouse correlations were significant (except for SBP after adjustment for weight) and of magnitude similar to the correlation in first-degree relatives. Using the bivariate model (Table 3⇓), we estimated correlations of 72% (75%) between the familial factors and 56% (57%) between the environmental factors influencing SBP and DBP before (after) adjustment for weight. Segregation analysis under the mixed model did not support the presence of major gene effects.
We also undertook bivariate analysis of SBP with LVM and weight and of DBP with weight to partition the covariance of these traits into familial and nonfamilial components (Table 4⇓). LVM showed no significant overall correlation with DBP or with SBP or DBP after adjustment for weight, and partitions of the covariance were not estimated for these instances. The bivariate analysis showed that all the covariance of the blood pressure traits with either LVM or weight could be attributed solely to nonfamilial factors (with the exception of significant spouse correlation). Thus, common genetic or other common familial factors do not appear to be an important source of correlation between blood pressure and LVM in first-degree relatives from these families.
We have investigated the relation of LVM with body weight, blood pressure, and other variables in male and female adults and their offspring. Overall, the strongest predictor of LVM was found to be weight. The correlation between the two traits varied with generation and gender, with the strongest relation observed in the offspring generation, particularly male offspring. In the parental generation, the traits showed a less marked but still strongly significant relation, with the largest correlation also in male subjects. A similar relation between LVM and body weight and height in male compared with female subjects has been described previously.10 The relation between the traits in the offspring generation in our data are stronger than that reported by Verhaaren et al.11 They found a correlation between 0.30 and 0.40 in a study of monozygotic and dizygotic twins, with mean age of 11.2±0.2, somewhat younger than the offspring in our families. Interestingly, our results are in agreement with those of Daniels et al12 in a cross-sectional study of LVM in children and adolescents. In this group of subjects, they found that lean body mass explained 75% of the variance of LVM, whereas fat mass explained only 1.5% of this variance. Variation of LVM with age in adults has been reported in a previous study13 but was not replicated here, possibly because of the narrow range of age of the parents in our study. Indeed, after taking gender, generation, and weight into account, LVM showed essentially no significant relation with other variables including age, height, and blood pressure in our data. However, after adjustment of LVM for age, the relation between LVM and body weight did not differ significantly by gender (although it was still larger in male subjects). Therefore, we decided to adjust LVM for gender, generation, and age before genetic analysis. Family analysis of LVM and weight showed that transmission of these traits was compatible with an additive genetic model. Using a bivariate analysis, we found that the intertrait correlation between LVM and body weight caused by familial components (CF/CP) was 0.28. The nonfamilial environmental component, CE, differed markedly in the offspring and parental generations. For the former, CE/CP was 0.25, whereas in the latter it did not differ significantly from 0. On the basis of the individual trait heritability and the familial intertrait correlation estimates, we could also evaluate the degree of relation between familial factors affecting both traits. This estimate was 0.66, which shows that the familial factors influencing LVM and weight are strongly correlated. Such a correlation means that at least some of the familial factors that affect weight also affect LVM. Thus, if the familial factors are genetic, this would suggest that at least some of the genes acting on the traits are pleiotropic, that is, influence both traits. Thus, LVM and weight are determined in large part by common genetic and/or familial environmental factors in both adults and children, whereas the impact of correlated nonfamilial environmental factors appears to be much larger in children compared with adults.
Weight also showed a significant spouse intratrait correlation that is likely to be due to the influence of common familial environmental factors, which could also affect parent-offspring and sibling correlations. (Spouse correlation could also reflect assortative mating in which individuals with phenotype similarities due to either genetic or environmental causes intermarry preferentially). This implies that the familial components of the variances and covariance cited above may be at least partially of environmental origin. To test the genetic model, we examined many other hypothetical patterns of familial correlation: different intertrait or intratrait correlations for parent-offspring and siblings, for father-son, father-daughter, mother-son, and mother-daughter combinations and for brother-brother, brother-sister, and sister-sister combinations. We found no significant deviations from the additive genetic model such as might be expected if familial environmental factors were largely implicated in the intertrait covariance.
If familial environmental factors are present, the decomposition of the intertrait correlation for LVM and weight that has been obtained under the additive genetic model may overstate the importance of familial factors at the expense of nonfamilial environmental factors (ie, the intertrait covariance in first-degree relatives may be greater than 1/2VF). However, the conclusion that common nonfamilial environmental factors are a greater source of correlation between LVM and weight in the offspring compared with the adult generation would still hold. One explanation for this generation difference could be that the offspring have greater physical activity and that this affects both traits. This hypothesis is supported by the data from Daniels et al12 showing the importance of lean body mass for explaining LVM variance in young subjects.
A second analysis was conducted on LVM regressed for weight to determine the importance of genetic factors on LVM after adjusting in this way for factors that were common to the traits. The familial correlations were again compatible with an additive genetic model with a heritability estimate of 0.28 (compared with 0.34 before regression). This is somewhat less than previously reported in the study by Verhaaren et al11 of monozygote and dizygotic twins, in which genetic factors accounted for 39% and 59% of total variance in male and female subjects, respectively, for LVM adjusted for body weight. Since the twins were somewhat younger than offspring from our families (see above), this might be an indication that the heritability of the trait decreases with age in children. However, the results may also be due to differences in the design or other characteristics of the 2 studies.
LVM was also related to blood pressure, but in contrast to the results for weight, we obtained no evidence that the correlation was familial. Similarly, the correlation between weight and blood pressure also appeared to be due uniquely to nonfamilial factors. After adjustment for weight, the LVM–blood pressure correlations were only marginally significant for female offspring and nonsignificant in the other generation/gender combinations. Thus, accounting for body weight adjusts for essentially all the factors that are common to LVM and blood pressure. The heritability estimates were similar before and after further adjustment for weight: 24% for DBP and 50% for SBP. The heritability estimates for DBP are in the low end of the range found in previous studies, whereas the estimates for SBP are intermediate between values found in infants and in adults in other studies (reviewed in Reference 1414 ). Pérusse et al15 also described higher heritability for SBP than for DBP in adults. For DBP, the spouse correlation in our data were marginally significant, again raising the question of the importance of familial environmental factors on blood pressure.
Some previous studies have been aimed at demonstrating a significant and independent effect of blood pressure on LVM in children.16 In one study, LVM was found to be increased in children with higher blood pressure, but children having higher blood pressure were also found to have higher weight and to be taller.17 In a survey of 645 healthy subjects, Burke et al18 did not find a significant correlation between LVM and SBP or DBP after adjustment for body surface area and ponderosity. In a study of children and adolescents with essential hypertension, as defined by blood pressure greater than the 90th percentile, 38.5% of subjects had LV hypertrophy.16 Although a significant correlation was found between LVM index and mean SBP in various conditions in the sample from Daniels et al,16 SBP did not appear as a significant independent factor in a multiple regression model of LVM. In the Virginia twin study, SBP and DBP were not significantly correlated with LVM in a multiple regression analysis.19 In adults, the prevalence of LV hypertrophy has been found to be more frequent in subjects with essential hypertension but without significant difference for SBP and DBP among hypertensive patients with and without LV hypertrophy.20 In the Framingham study, SBP did not appear to be a significant predictor for LVM in a multivariate analysis.21
Thus, our results are consistent with other studies, and taken together, they suggest that blood pressure is not a major determinant of LVM or LV hypertrophy, within a range of normal or moderately elevated blood pressure, after the trait has been adjusted for its relation to weight. This conclusion cannot be extrapolated to individuals with chronically elevated blood pressure, in which LV hypertrophy constitutes one of the major clinical hallmarks.22
On the other hand, LVM and weight are strongly correlated apparently through the effects of familial factors, probably both genetic and environmental, and nonfamilial environmental factors acting in the normal range of these traits. Common genetic factors might modulate the growth of the heart and of the whole body, ensuring the harmonious development from the embryonic to adult stage.
On the basis of the results of the segregation analyses, which showed no evidence of a major gene or genes, it is likely that the multiple genes (polygenes) affect LVM in addition to familial environmental factors. Despite the difficulties that can be anticipated as the result of multifactorial inheritance, appropriately designed genetic studies or other genomic approaches to identify the common genes underlying these traits is of great interest, as this would provide new insights of the molecular relations between body shape and LVM.
- Received January 11, 2000.
- Revision received February 1, 2000.
- Accepted May 13, 2000.
Siest G, Visvikis S, Herbeth B, Gueguen R, Vincent-Viry M, Sass C, Beaud B, Lecomte E, Steinmetz J, Locuty J, Chevrier P. Objectives, design and recruitment of a familial and longitudinal cohort for studying gene-environment interactions in the field of cardiovascular risk: the Stanislas cohort. Clin Chem Lab Med. 1998;36:35–42.
Sahn DJ, DeMaria A, Kisslo J, Weyman A. Recommendations regarding quantitation in M-mode echocardiography: results of a survey of echocardiographic measurements. Circulation. 1978;58:1072–1083.
Zureik M, Bonithon-Kopp C, Lecomte E, Siest G, Ducimetiere P. Weights at birth and in early infancy, systolic pressure, and left ventricular structure in subjects aged 8 to 24 years. Hypertension. 1996;27:339–345.
Daniels SR, Meyer RA, Liang YC, Bove KE. Echocardiographically determined left ventricular mass index in normal children, adolescents and young adults. J Am Coll Cardiol. 1988;12:703–708.
Daniels SR, Kimball TR, Morrison JA, Khoury P, Witt S, Meyer RA. Effect of lean body mass, fat mass, blood pressure, and sexual maturation on left ventricular mass in children and adolescents: statistical, biological, and clinical significance. Circulation. 1995;92:3249–3254.
Ward R. Familial aggregation and genetic epidemiology of blood pressure. In: Laragh JH, Brenner BM, eds. Hypertension: Pathophysiology, Diagnosis, and Management. New York, NY: Raven Press, Ltd; 1995:67–88.
Daniels SD, Meyer RA, Loggie JM. Determinants of cardiac involvement in children and adolescents with essential hypertension. Circulation. 1990;82:1243–1248.
Schieken RM, Clarke WR, Lauer RM. Left ventricular hypertrophy in children with blood pressures in the upper quintile of the distribution: the Muscatine Study. Hypertension. 1981;3:669–675.
Burke GL, Arcilla RA, Culpepper WS, Webber LS, Chiang YK, Berenson GS. Blood pressure and echocardiographic measures in children: the Bogalusa Heart Study. Circulation. 1987;75:106–114.
Goble MM, Mosteller M, Moskowitz WB, Schieken RM. Sex differences in the determinants of left ventricular mass in childhood: the Medical College of Virginia Twin Study. Circulation. 1992;85:1661–1665.
Post WS, Larson MG, Myers RH, Galderisi M, Levy D. Heritability of left ventricular mass: the Framingham Heart Study. Hypertension. 1997;30:1025–1028.