Left ventricular mass and incident hypertension in individuals with initial optimal blood pressure

Abstract

Objective

Metabolic abnormalities have been shown to predict 8-year incident arterial hypertension in individuals with optimal blood pressure. As echocardiographic left ventricular mass has also been reported to predict incident hypertension in individuals with baseline blood pressure of less than 140/90 mmHg, we determined whether left ventricular mass predicts 4-year incident hypertension also in individuals with initial optimal blood pressure (<120/80 mmHg), independent of metabolic factors influencing blood pressure.

Methods

We studied 777 of 3257 members of the American Indian population-based Strong Heart Study cohort with optimal blood pressure (34% men, 45% obese, and 35% diabetic), aged 57 ± 7 years, and without prevalent cardiovascular disease.

Results

Over 4 years, 159 individuals (20%, group H) developed hypertension (blood pressure ≥140/90 mmHg). They had a greater baseline BMI, waist girth, and blood pressure (112/69 vs. 109/68 mmHg, all P<0.03) than those remaining normotensive (group N), with similar lipid profile and renal function. At baseline, left ventricular mass was significantly greater in group H than in group N (P<0.004). The difference in left ventricular mass was confirmed after controlling for initial BMI, systolic blood pressure, homeostatic model assessment index, and diabetes. The probability of incident hypertension increased by 36% for each standard deviation of left ventricular mass index (P=0.006), independent of covariates. Participants with left ventricular mass of more than 159 g (75th percentile of distribution) had 2.5-fold (95% confidence interval, 1.4-3.6; P<0.001) higher adjusted risk of incident hypertension than those below this value.

Conclusion

Left ventricular mass predicts incident arterial hypertension in individuals with initially optimal blood pressure. This association is independent of body build, prevalent diabetes, and initial blood pressure.

Arterial hypertension is a complex, multifactorial hemodynamic condition driven by a variety of diseases and abnormalities of cardiovascular homeostasis, making it difficult to prevent its development and reduce its incidence by interventions on single aspects [1-4]. Metabolic risk factors, including obesity, dyslipidemia, and impaired glucose metabolism, can have vascular and hemodynamic effects that ultimately contribute to the development of arterial hypertension [5-7]. There is also evidence that unfavorable changes in these factors over time can potentiate the development of arterial hypertension [7]. Those associations, however, do not clarify why some but not other individuals exposed to these risk factors develop arterial hypertension.

Genetic factors, acting through different pathways, may contribute to the development of hypertension [1,8-10]. Several earlier studies suggested that higher left ventricular (LV) mass in normotensive individuals can predict subsequent hypertension [11-13], an observation that revitalized the concept that a more forceful cardiac pump could contribute to the development of high blood pressure (BP) in the presence of an arterial tree unable to either reduce peripheral resistance or increase conduit artery compliance or both [14]. Unfortunately, previous studies were conducted in population samples with normal or high-normal BP, but not optimal BP. Thus, they could not answer the question of whether the association between LV mass and incident hypertension was driven by initial BP in the range classified today as high-normal or prehypertensive, which may have already caused altered cardiac structure and function [15]. Whether LV mass is associated with the development of arterial hypertension in the presence of initial optimal BP is, therefore, unknown. Accordingly, the present study was designed to determine whether, in the presence of optimal baseline BP, a higher LV mass predicts incident arterial hypertension, independent of established cardiovascular risk factors.

Methods

Population

The Strong Heart Study (SHS) is a population-based longitudinal cohort study of cardiovascular risk factors and disease in American Indians from communities in Arizona, Southwestern Oklahoma and South and North Dakota, as extensively described [16-20].

During the second examination, in the years 1993-1996, participants also underwent a standard transthoracic echocardiographic study. Those with available information on body size and fat distribution, diabetes status, lipid profile, BP, and LV mass were first selected for the present analysis (n=3257). Participants with prevalent cardiovascular disease, including stroke, congestive heart failure, myocardial infarction, and other manifestations of coronary heart disease (assessed by coronary angiography or a combination of typical symptoms with positive treadmill tests or abnormal imaging stress test, or need for revascularization procedures), were excluded (n=394). Prevalent cardiovascular events were confirmed by the Strong Heart Study Mortality and Morbidity Committees, using specified criteria for causes of fatal and non-fatal cardiovascular events [21]. Diabetic and obese participants were included; however, similar to the criteria previously reported [7], only untreated participants with optimal baseline BP (<120/80 mmHg) [22-24] were included in the studied population sample. Thus, after exclusions, 777 participants with an optimal baseline BP at the time of the second SHS examination (261 men or 33.6%) and with a mean age of 57±7 years were considered in the present analysis.

Laboratory tests and classification of participants

BP measurements were accurately standardized in the SHS [25]. Briefly, BP was measured in the sitting position after a 5-min rest. After the first measurement, another two measurements were taken after raising the arm for 5 s and resting it on the table for another 25 s. The last two measurements were averaged to generate the reported BP value.

Fasting plasma glucose and lipid profile were measured by standard methods [18]. Diabetes was diagnosed by 1997 American Diabetes Association recommendations [26]. Homeostatic model assessment (HOMA) index was used to estimate insulin resistance [27]. Obesity was identified by BMI of at least 30 kg/m² based on 1998 National Institutes of Health (NIH) guidelines [28]. Waist circumference was used as an indicator of central adiposity according to 1998 NIH guidelines [28]. Glomerular filtration rate (GFR) was estimated by the simplified modification of diet in renal disease (MDRD) formula [29].

Echocardiographic measures

Measurements of interventricular and posterior wall thickness, and LV diastolic internal diameter, were taken using M-mode tracings or linear two-dimensional measures taken from the parasternal long axis view according to a well standardized method [30-32]. LV mass was calculated by a necropsy validated formula [33] and was normalized for body surface area and height^2.7 [34]. Left ventricular hypertrophy (LVH) was defined according to population-specific partition values (47.24 g/m^2.7), which have been prognostically validated [34].

Statistical analysis

Data were analyzed using SPSS 12.0 software (SPSS Inc, Chicago, Illinois, USA). Data are expressed as mean±one standard deviation (SD). Indicator variables were included in all multivariate analyses for the three field centers, Arizona, South/North Dakota, and Oklahoma. Incident hypertension was defined either as BP of at least 140 mmHg or 90 mmHg or both, or ongoing antihypertensive treatment at the time of the third SHS examination 46±9 months after the second examination. Descriptive statistics were obtained by one-factor analysis of variance (ANOVA) or χ² distribution. Changes over time of potential metabolic predictors of arterial hypertension were compared between the groups with or without incident hypertension by analysis of covariance (ANCOVA) for repeated measures, adjusting for age and sex. Adjusted means are displayed. The main potential confounders associated with follow-up hypertension were evaluated in binary logistic regression models. To evaluate whether LV mass index significantly added to the prediction of incident hypertension obtained with the main factors, a hierarchical model was generated by entering a primary block of main covariates, including age, sex, systolic blood pressure, BMI, diabetes, and HOMA index. LV mass index was, therefore, added to the primary model (secondary block). Likelihood functions were compared between the models without or with LV mass index [35]. The difference between the two likelihood functions has a χ² distribution, which, for this comparison, has one degree of freedom (i.e. one variable added to the model) [35].

Results

At the time of the second examination, 44.5% of the 777 participants were obese (50% of women and 33% of men) and diabetes was present in 34.7% of participants (38% of women and 28% of men, both P<0.0001).

By the time of the third SHS examination, incident arterial hypertension had developed in 159 participants (20% of men and 21% of women), 98 of whom (62%) were on medications. At baseline, participants developing hypertension during the follow-up were slightly more obese (P<0.02), had 3 mmHg higher average systolic BP with borderline higher diastolic BP; they also had higher plasma glucose and HOMA index, and prevalence of diabetes, than participants maintaining normal BP (Table 1; all P<0.0001). No differences were found between groups in lipid profile, GFR, or heart rate. LVH was found in 62 participants (8%) and was 1.97-fold more prevalent among participants who developed hypertension than among those remaining normotensive [95% confidence interval (CI), 1.12-3.47; P<0.03].

Table 1.

Characteristics of participants with optimal blood pressure

Characteristics	Remaining normotensive (n=618)	Developing hypertension (n=159)	P value
Age (years)	57.0±6.7	57.3±7.0	NS
Men/women	412/206	104/55	NS
BMI (kg/m2)	29.5±5.9	30.7±6.0	≤0.02
Waist circumference (cm)	102.3±14.9	105.3±13.5	≤0.03
Systolic BP (mmHg)	108.8±8.0	112.0±7.1	≤0.0001
Diastolic BP (mmHg)	67.8±6.9	68.9±6.2	NS
Heart rate (bpm)	71.3±10.6	72.9±10.2	NS
Fasting glucose (mg/dl)	135.4±73.1	177.0±91.6	≤0.0001
Diabetes n (%)	180 (29)	88 (55)	≤0.0001
HOMA index	6.7±9.7	9.7±10.8	≤0.0001
Total cholesterol (mg/dl)	191.0±37.7	189.7±37.2	NS
Triglycerydes (mg/dl)	148.9±108.4	145.7±89.2	NS
HDL cholesetrol (mg/dl)	40.9±12.3	41.1±13.2	NS
LDL cholesterol (mg/dl)	121.6±33.4	119.8±32.0	NS
Non-HDL-cholesterol (mg/dl)	149.9±37.9	148.8±38.8	NS
GFR (ml/min)	86.4±29.5	86.2±28.9	NS

Comparison has been done between those remaining normotensive and those exhibiting arterial hypertension at the time of third Strong Heart Study examination. BMI, body mass index; BP, blood pressure; bpm, beats per minute; GFR, glomerular filtration rate; HDL, high-density lipoprotein; HOMA, homeostatic model assessment; LDL, low-density lipoprotein; NS, not significant.

Table 2 shows changes between the time of the second (baseline) and the third SHS examination in the metabolic profile of participants with or without hypertension detected at the time of the third SHS examination. Body weight increased in the entire population sample during follow-up. Adjusting for age and sex, individuals developing hypertension gained slightly more weight than those remaining normotensive (P<0.03). During follow-up, plasma glucose decreased significantly in the group developing hypertension (P<0.002), due to the effect of treatment of diabetes detected during the second examination, but remained higher than in those with persistent normal BP. No significant differences were detected in changes of lipid profile or GFR.

Table 2.

Changes of anthropometric measures and metabolic risk factors during 4 year follow-up in subjects developing hypertension or remaining normotensive

	Remaining normotensive (n=618)		Developing hypertension (n=159)		P value
	Second examination	Third examination	Second examination	Third examination
Body weight (kg)	80.2±17.0	83.4±17.6	80.4±17.5	83.8±18.0	≤0.03
Waist girth (cm)	102.3±15.0	105.4±15.2	101.8±13.5	105.5±14.8	NS
Plasma glucose (mg/dl)	134.8±73.0	130.1±63.6	174.2±90.8	150.8±72.1	<0.002
HDL cholesterol (mg/dl)	40.9±12.3	41.1±12.8	43.0±12.5	44.4±15.1	NS
LDL cholesterol (mg/dl)	121.9±33.7	120.0±32.3	120.7±32.5	118.5±34.2	NS
Total cholesterol (mg/dl)	191.6±37.8	189.1±38.3	191.3±35.5	190.6±40.9	NS
Plasma triglycerydes (mg/dl)	149.8±109.7	141.0±85.4	141.2±74.4	139.6±89.1	NS
GFR (ml/min)	86.2±29.3	86.6±28.3	89.5±28.3	87.6±30.1	NS

P values refer to the difference between the interaction effect of time and grouping characteristics. GFR, glomerular filtration rate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; NS, not significant.

Baseline left ventricular geometry

At baseline, LV mass was higher in participants developing hypertension than in those remaining normotensive, in absolute terms and after normalization for body size (Table 3; all P<0.0001). Because the LV mass difference could be due to differences in clinical and metabolic factors (Tables 1 and 2), the comparison was repeated with adjustment for baseline systolic BP, BMI, HOMA index, and presence of diabetes. Table 3 also shows the adjusted means of baseline LV mass by different types of normalization for body size in participants with or without follow-up hypertension. Initial LV mass remained higher in participants developing arterial hypertension, irrespective of the method used to normalize for body size and after adjustment for main confounders.

Table 3.

Baseline left ventricular mass in participants remaining normotensive or developing hypertension

	Remaining normotensive (n=618)		Developing hypertension (n=159)		P value
	Unadjusted	Adjusted	Unadjusted*	Adjusted
LV mass (g)	138.0±28.4	138.6±27.6	148.4±33.6	147.8±28.2	<0.009
LV mass index (g/m2 BSA)	74.4±12.7	74.4±13.3	78.5±15.5	79.0±13.6	<0.007
LV mass index (g/m2.7 height)	36.1±7.2	36.3±6.8	39.0±9.0	38.5±6.9	<0.008

Mean values are shown as raw values and after adjusting for confounding risk factors (see also text). P values are adjusted. LV, left ventricular; BSA, body surface area.

^*All P<0.0001.

Prediction of incident hypertension

After adjusting for sex, baseline BMI, systolic BP, HOMA index, and presence of diabetes or impaired fasting glucose, the probability of developing arterial hypertension over a 4-year follow-up increased by 39% for each 7.9 g/m^2.7 (corresponding to one SD in the whole population sample) of initial LV mass index [relative risk (RR), 1.39/7.9g/m^-2.7 (1.11-1.73, P=0.004)]. In addition to the LV mass index, higher initial systolic BP [RR, 1.07/mmHg (1.03-1.10; P<0.0001)] and presence of diabetes [RR, 2.59 (1.43-4.70; P=0.002)] were also independent predictors of incident hypertension, whereas no effect was found for HOMA index, BMI, and impaired fasting glucose.

Because of the significant effect of diabetes in the prediction of subsequent hypertension, diabetic and non-diabetic participants were also separately analyzed. LV mass index was significantly higher in the presence of diabetes (38.08±7.59 vs. 36.18±7.70 g/m^2.7; P<0.02, adjusted for systolic BP) and was confirmed as a predictor of incident hypertension in nondiabetic participants [RR, 1.84/7.9g/m^-2.7 (1.35-2.51; P<0.0001)] independent of the significant effect of systolic BP [RR, 1.06/mmHg (1.02-1.12; P=0.009)], without other significant contributions from sex, BMI, and HOMA index. In contrast, among diabetic participants, the only significant independent predictor of incident hypertension remained initial BP [RR, 1.07/mmHg (1.02-1.12; P=0.004)].

To examine whether the addition of LV mass index improved the prediction of incident hypertension in the whole population, we also used the χ² distribution to compare the likelihood functions of the model including all covariates but excluding LV mass index with the same model into which LV mass index was forced on the top of the primary variables. The -2log likelihood function was 592 in the model without LV mass and 583 in the model including LV mass, a difference that was statistically significant (P<0.005).

Figure 1 shows the progressively higher likelihood of developing hypertension in progressively higher deciles of LV mass/height^2.7 obtained from the resulting functions from the logistic regression model. Participants in the highest decile of the distribution had a more than 30% likelihood of developing hypertension in the next 4 years. Participants with LV mass higher than the 75th percentile of the distribution in the SHS cohort (159 g) had a 2.5-foldhigher adjusted risk of developing arterial hypertension (1.42-3.60, P<0.0005) in the next 4 years than participants with LV mass 159 g or less, independent of the covariates used in the logistic regression model.

Fig. 1.

Adjusted probability of 4-year incident hypertension in relation to initial deciles of left ventricular mass index. LVMI, left ventricular mass index.

Discussion

The possibility that LV mass predicts future hypertension has been already reported in studies in relatively unselected populations [11-13] and in children with a family history of hypertension [36]. However, the entry criteria of those studies reflected old classifications of BP (i.e. <140/90 mmHg) [11,13] that included participants with high-normal and prehypertensive BP in the normotensive group [24,37]. According to recent guidelines, normal (optimal) BP values should be below 120/80 mmHg. Clinical evidence of optimal BP might be reasonably considered to confer a low risk of future high BP. In addition, in all the previous studies, LV mass was measured using only M-mode echocardiograms, a method that results in the exclusion of a large number of individuals due to body size, age, and other factors [38], whereas the combined M-mode and 2D approach used in the SHS allows measurement of LV mass in almost all participants, even if most are obese [38].

Thus, the present study is the first prospective analysis examining whether initial values of LV mass are associated with incident hypertension in a population-based cohort of individuals with initial optimal BP values and a broad range of body size and related metabolic abnormalities. Our analysis demonstrated that even when BP is optimal, the magnitude of LV mass measured by transthoracic echocardiography strongly predicts the probability of incident arterial hypertension.

Although BP is a pathophysiological continuum, from the practical standpoint, the evidence that LV mass is still a predictor of hypertension when the baseline variability of BP is constrained to the optimal range and, therefore, substantially limited (the coefficient of variability of systolic BP was as low as 7%) is relevant. In addition, unlike previous studies [11], our analysis also considered metabolic risk factors that others and we [5-7] have shown to be associated with the rise of BP over time, probably due to the progressive alteration of the arterial tree due to atherosclerosis. The conservative hierarchical approach we have used, giving LV mass a secondary value in the regression model compared with BP, body weight and abnormal glucose metabolism, confirmed that LV mass significantly adds to the prediction of 4-year incident hypertension obtained with the factors used as primary predictors. This result was also consistent in analyses with absolute values of LV mass or LV mass normalized for different measures of body size. Interestingly, the effect was substantially driven by the nondiabetic population with lower levels of LV mass, and therefore, with a more identifiable association than the subgroup with diabetes, which already presents with a more severe impairment of the cardiovascular system (33% of diabetic participants developed hypertension as compared with 14% of nondiabetic participants, independent of LV mass).

There are possible explanations of our findings. The SHS population has a high prevalence of central obesity [39], and masked hypertension is found more frequently in patients with central obesity [40]. In our study, arterial hypertension has been detected using office and not ambulatory BP. Thus, the possibility that at least some participants with incident office arterial hypertension could have had masked hypertension [40] at the time of initial examination could not be ruled out. The prevalence of masked hypertension is reported to be about 10% in unselected populations [41,42]. This would imply that 70-80 participants in our study could have had masked hypertension. In addition, because of the variability of BP in day to day measures, there is also the possibility that the higher LV mass reflects a prolonged sustained exposure to higher BP, which cannot be captured by single office measurements, even if taken with a very standardized protocol. This has also been recently suggested by Kokkinos et al. [43] who reported that a change in BP from rest to five metabolic equivalents (i.e. a workload corresponding to most daily activities) is the strongest predictor of LV hypertrophy, whereas fitness protects from an excessive rise of BP during activity [44].

Although the above explanation is simple and plausible, this study also supports the possibility of reverse causation, not necessarily alternative to the possibility that some or many of our participants had masked hypertension. The concept that the power developed by the heart represents one of the mechanisms responsible for the increase in BP is supported by the evidence that hypertension can be predicted by the increase in BP during exercise. This exercise-induced increase in BP is in turn associated with a high LV mass [45], a finding that parallels the detection of masked hypertension by ambulatory monitoring during daily activities, a time when the blood supply requirements are higher than during the resting conditions at the office visit [46,47]. An equilibrium among pump force, circulating flow, conduit artery compliance and elasticity and peripheral resistance to blood flow is a requisite for the circulatory system to maintain fluid pressure [48-50]. Thus, the development of arterial hypertension requires a convergence of alterations of different factors, including cardiovascular musculature, heart strength and magnitude of circulating volume, which result in BP augmentation when the elasticity of conduit arteries and peripheral resistance are impaired by the adverse effect of atherogenic risk factors, all elements that can be found in our analyses [7].

In our previous study [7], abdominal obesity and its progression were strongly associated with 8-year incident arterial hypertension, despite the relatively small range of variability of these parameters in this population, consistent with the evidence that obesity and the associated neurohormonal patterns are well documented predictors of future hypertension in epidemiological and clinical studies [51,52]. In the present shorter follow-up, obesity and a higher BMI were also associated with incident hypertension in univariate analysis, but their effect in the model was obscured by other cofactors, including LV mass index, suggesting that an increased LV mass might integrate at least a part of the effect of alteration of body size on incident hypertension. This finding is also in line with previous evidence that genes common to LV mass and body weight significantly influence their covariance and that as much as 90% of the LV mass-body weight relation might be due to common genes [53].

In conclusion, we provide evidence that LV mass is a predictor of 4-year incident arterial hypertension in a population-based sample with high prevalence of obesity but initial optimal BP. This prediction is independent of metabolic and anthropometric factors already demonstrated to be associated with incident hypertension.

The possibility to refine the identification of a phenotype at high risk of arterial hypertension, by pooling metabolic information with the measure of LV mass, increases the chance to target individuals that might benefit from preventive treatment, as already proposed for high-normal BP values. Further research is needed to quantify the possible impact of masked hypertension on the association between LV mass and incident office hypertension and clarify their cause-effect relationship.

Abbreviations

BMI

Body mass index

GFR

Glomerular filtration rate

HOMA

Homeostatic model assessment

LVH

Left ventricular hypertrophy

MDRD

Modification of diet in renal disease

SHS

Strong Heart Study