Author + information
- Received July 1, 1996
- Revision received October 31, 1996
- Accepted November 26, 1996
- Published online March 1, 1997.
- John S. Gottdiener, MD, FACCB,* (, )
- Domenic J. Reda, MSA,
- David W. Williams, MSB,
- Barry J. Materson, MDB,1,
- for the Department of Veterans Affairs Cooperative Study Group on Antihypertensive Agents
- ↵*Dr. John S. Gottdiener, Division of Cardiology, 4219 M, Georgetown University Hospital, 3800 Reservoir Road NW, Washington, D.C. 20008.
Objectives. We sought to determine the relations of left atrial (LA) size to blood pressure, obesity, race, age and left ventricular (LV) mass in hypertension.
Background. Although obesity, race and age may influence LV mass, their effects on LA size have not been defined in hypertension.
Methods. Left atrial size was measured in 690 men (58% African-Americans) with mild to moderate hypertension (mean [±SD] blood pressure 152 ± 15/98 ± 6 mm Hg) and a high prevalence of LV hypertrophy. Effects of LV mass, adiposity, race, age, physical activity, height, weight, sodium excretion, plasma renin activity and heart rate were examined.
Results. Left atrial size was greater (p ≤ 0.0001) in obese (44.2 ± 5.7 mm) than in overweight (41.6 ± 5.9 mm) or normal weight (38.9 ± 6.2 mm) patients. Left atrial enlargement (≥43 mm) was present in 56% of obese patients compared with 42% of overweight and 25% of normal weight hypertensive men. As age increased, white patients had a greater LA size than African-American patients. Although there was no relation between LV mass and LA size in normal weight patients, there was a significant positive relation in obese patients. On multiple regression analysis, obesity was the strongest independent predictor of increased LA size.
Conclusions. Obesity is the strongest predictor of LA size in patients with hypertension and amplifies the relation between LA size and LV mass. Race influences effects of age and hypertension on LA size. Because increased LA size and LV mass (also influenced by obesity) are associated with an adverse outcome, these findings underscore the importance of obesity, race and age with regard to the cardiac effects of hypertension.
(J Am Coll Cardiol 1997;29:651–8)
Left ventricular (LV) mass () and left atrial (LA) size ([1–3]) are both increased in patients with hypertension. Although LV mass has been shown to be a predictor of cardiovascular events, including myocardial infarction and death, LA size is associated with the likelihood of developing atrial fibrillation ([4, 5]) and stroke ().
Recently, evaluation of a free-living, population-based cohort () showed modest relations between systolic blood pressure, pulse pressure and echocardiographically determined LA size. The relative contribution of blood pressure to LA size was substantially less than that of age and body mass index. Inclusion of LV mass in statistical models of LA size further weakened the association of blood pressure.
However, the relative contributions of blood pressure, body weight, age and LV mass to LA size have not been determined in patients with established hypertension. Moreover, although it has been shown that race influences the effects of hypertension on LV wall thickness ([8–11]), the effects of race on LA size in patients with hypertension are unknown. Therefore, we used echocardiography to assess LA size in 690 men with mild to moderate hypertension and determined the relations of LA size to blood pressure, obesity, race and LV mass.
The patients included in the present study were part of a cohort of 1,292 men recruited for a randomized, prospective trial of monotherapy in mild to moderate (diastolic blood pressure 95 to 109 mm Hg) hypertension () uncomplicated by clinically evident cardiac or systemic disease. Systolic blood pressure ranged from 120 to 199 mm Hg.
Echocardiograms in 680 men (53%), acquired at 14 centers, were evaluated for this study. This number reflects the product of the overall acquisition rate (72%) and the proportion of technically acceptable studies (75%) for the estimation of LA size and LV mass.
After 6 to 12 weeks of withdrawal of prior antihypertensive medication (460 patients), or 4 to 12 weeks of baseline observation in the 232 patients not receiving therapy at study entry, two-dimensional targeted M-mode echocardiography was performed. To minimize variability, measurements of LA dimension, ventricular septum, LV cavity and posterior wall dimensions from paper stripchart recordings were taken by a single reader, using an off-line image analysis system, according to American Society of Echocardiography (ASE) criteria ().
To determine LA dimension, the maximal dimension was measured between the leading edge of the posterior aortic wall to the leading edge of the posterior wall of the left atrium at end-systole. Interobserver error for LA dimension, determined by comparing 121 echocardiograms read by the core laboratory sonographer for this study with those measurements made by a “reference reader” (J.S.G.), was 3.2%. Intraobserver error was 2.8% for sonographer and 1.7% for “reference” reader measurements.
Left ventricular mass (ASE) was calculated as described elsewhere (). These measurements were converted to Cornell-Penn convention values using published regression equations (). For determination of the presence of LV hypertrophy, LV mass (Cornell-Penn) was indexed to body surface area, computed by nomogram (); 134 g/m2was selected as the partition value for LV hypertrophy (). In addition, to avoid the flattening contribution of obesity to the estimation of LV hypertrophy prevalence, LV mass (ASE) was also indexed to height according to the Framingham convention (); 164 g/m was selected as the partition value for LV hypertrophy. Echocardiographic characteristics of this group pertinent to LV dimensions, LV mass and prevalence of LV hypertrophy (63% Framingham criteria; 46% Cornell-Penn criteria) have been described previously ().
Left atrial size was analyzed as a linear dimension as well as indexed by height in an attempt to account for allometric, physiologic differences in LA size based on body size. Indexation by body surface area, as has sometimes been done, was considered inappropriate because we wished to determine the effects of obesity on LA size. Indexation by body surface area, principally determined by body weight, would have interfered with analysis of the effects of obesity on LA size.
Body mass (Quetelet) index was selected as a measure of adiposity determined as weight (kg) divided by height squared (m2). “Normal” weight was defined by body mass index ≤27 kg/m2(269 patients); “overweight” as 27 to 30 kg/m2(200 patients); and “obesity” (223 patients) as >30 kg/m2([18, 19]). These categories have been described elsewhere (). In the present study, the “obesity” category corresponds with a moderate risk of overall mortality, and “overweight” corresponds with a 20% increase in “desirable” weight determined from Metropolitan Life tables.
1.4 Clinical measurements
Casual blood pressure was measured with a cuff sphygmomanometer after 15 min of rest sitting upright with the back supported. Plasma renin was determined by iodine-125 radioimmunoassay (Clinical Assays, Travenol Division, Genentech Diagnostics, Inc.). Physical activity index was obtained by an administered questionnaire that queried work and recreational physical activity for the 6-month period preceding study recruitment. A numeric value was derived from the estimated average duration and intensity of activity. Sodium intake was determined from one 24-h measurement of total urinary sodium excretion (mmol).
1.5 Statistical methods
All values for quantitative measures are expressed as mean value ± SD. Left atrial enlargement was defined as LA dimension ≥43 mm. This partition value was defined by the Framingham study () as the 90th percentile value in normotensive, nonobese men. Left atrial enlargement was arbitrarily considered to be mild to moderate if the LA dimension was between 43 and 49 mm, and severe if the left atrial dimension was >50 mm. Comparisons of data () for outcome variables (e.g., LA dimension) between two groups (e.g., black men and white men) were performed using the Student ttest for independent groups. Comparisons across more than two groups (e.g., normal weight, overweight, obese) were done using analysis of variance (ANOVA) for unbalanced data. Pairwise comparisons of groups were made using the Tukey procedure if the ANOVA statistic was significant at the 0.05 level.
Correlations were obtained to assess univariate relations of clinical variables to LA dimension, as well as indexed LA dimension. The independent contribution of clinical descriptors, including LV mass, adiposity (body mass index), race, age, systolic blood pressure, physical activity, sodium excretion, plasma renin activity and heart rate, to LA dimension and indexed LA dimension was determined by stepwise multiple regression. For multiple regression analyses, models were created using predictor variables singly and by examining all possible two-way interactions.
Comparison of the distribution of noncontinuous variables across groups was accomplished using the chi-square test of homogeneity for contingency tables. If significant differences were present, the Grizzle-Starmer-Koch () approach was used to identify which pairs of groups were different.
The statistical computer package SAS, version 6, was used to generate the statistical analyses (). All statistical tests were two-tailed, and p ≤ 0.05 was used to identify statistically significant results.
2.1 Patient characteristics
The average age of the patients was 58 ± 10 years; the average blood pressure was 152 ± 15/98 ± 6 mm Hg; and the average LV mass was 329 ± 93 g. Fifty-eight percent of the subjects were African-American. The prevalence of LV hypertrophy was 63% by Framingham criteria and 46% by Cornell-Penn criteria. Baseline height, weight and echocardiographic characteristics subset by age and race are shown in Table 1. Within each age group there were no significant differences for any of the characteristics.
2.2 Left atrial dimension and enlargement
For all patients the average LA dimension was 41.4 ± 6.4 mm (range 20 to 63), whereas indexed LA dimension averaged 23.5 ± 3.6 mm/m (range 11.6 to 36.5). Left atrial enlargement was noted in 271 men (40%). Of these, 210 men (77%) had mild to moderate enlargement (average 45.9 ± 1.9, range 43 to 49) and 61 (23%) had severe enlargement (average 52.8 ± 3.0 mm, range 50 to 63).
The strongest correlations of LA size (Fig. 1) were, in decreasing order, with body weight, body mass index, LV mass, LV cavity dimension, height, septal thickness and posterior wall thickness. Correlations were not significant with systolic blood pressure, diastolic blood pressure, age, plasma renin activity, heart rate, sodium excretion or physical activity score.
For indexed LA size (Fig. 1), correlations were strongest for body mass index, body weight, LV mass, LV cavity dimension, height (negative), septal thickness, systolic blood pressure and age. Relations were not significant with plasma renin activity, heart rate, posterior wall thickness, diastolic blood pressure, physical activity or sodium excretion.
2.3 Relation of obesity to left atrial size
Left atrial size (Fig. 2, top) differed significantly (p ≤ 0.0001 by ANOVA) across obesity categories. Based on the Tukey procedure, obese patients had a greater LA size (44.2 ± 5.7 mm) than overweight (41.6 ± 5.9 mm) or normal weight (38.9 ± 6.2 mm) patients. Indexed LA size (Fig. 2, bottom) also differed significantly (p ≤ 0.0001 by ANOVA) across obesity categories, averaging 25.3 ± 3.4 mm/m for obese, 23.6 ± 3.3 mm/m for overweight and 22.1 ± 3.5 mm/m for normal weight patients (all three groups significantly different by the Tukey procedure). Left atrial enlargement was present in 55.8% of obese patients, 42.2% of overweight patients and 25.0% of normal weight patients (p < 0.005, all pairwise comparisons). The average body mass index for patients with a normal LA size was 27.6 ± 4.0 kg/m2in contrast to 30.4 ± 5.0 kg/m2for patients with LA enlargement (p < 0.0001). Of patients with LA enlargement, 45% were obese compared with 24% of patients with a normal LA size.
2.4 Effect of obesity on relation between left ventricular mass and left atrial size
Using two-way ANOVA, a significant interaction (p = 0.02) was found between body mass index and LV mass in their relation to LA size (Fig. 3, top). For normal weight individuals, there was no relation between LV mass and LA size. Left atrial size averaged 38.9 ± 5.5 mm for patients with LV mass <225 g, 38.7 ± 6.9 mm for patients with LV mass between 225 and 275 g and 39.1 ± 6.8 mm for patients with LV mass >275 g. For obese individuals, a significant relation was found between LV mass and LA size. Left atrial size for patients with LV mass <225 g averaged 41.1 ± 5.2 mm. By pairwise comparison, this was significantly less than the average of 44.4 ± 5.7 mm for patients with LV mass between 225 and 275 g, and 45.6 ± 5.7 mm for patients with LV mass >275 g.
Similarly, a significant interaction (p = 0.003) was found between body mass index and LV mass in their relation to indexed LA size (Fig. 3, bottom). For normal weight individuals, no relation between LV mass and indexed LA size was found, whereas obese patients with LV mass <225 g had significantly (p < 0.05 pairwise comparison within ANOVA) smaller indexed LA size compared with patients with LV mass between 225 and 275 g and those with LV mass >275 g.
2.5 Relation between race, age and left atrial size
For men <55 years, LA size (Fig. 4, top) averaged 41.1 ± 5.5 mm for whites compared with 41.0 ± 6.4 mm for African-Americans. In whites 55 to 64 years old, LA size averaged 42.7 ± 6.6 mm compared with 40.4 ± 6.0 mm for African-Americans of comparable age. Among men 65 years or older, LA size averaged 43.9 ± 6.9 mm in whites compared with 40.0 ± 6.2 mm in African-Americans. Two-way ANOVA of age and race and their interaction found the interaction to be significant (p = 0.01), indicating that the age relation to LA size differed between races. Pairwise comparison found no race difference among younger men but a significant difference among the oldest group.
For indexed LA size (Fig. 4, bottom), two-way ANOVA of age and race and their interaction also found the interaction to be significant (p = 0.04). Pairwise comparison indicated no race difference among the younger patients but a significant race effect in the oldest group. For men <55 years, indexed LA size averaged 23.4 ± 3.0 mm/m for whites compared with 23.0 ± 3.5 mm/m for African-Americans. Among those 55 to 64 years old, indexed LA size averaged 24.3 ± 3.7 mm/m for whites compared with 23.2 ± 3.6 mm/m for African-Americans. In the oldest age group, the averages for indexed LA size were 25.1 ± 3.8 mm/m for whites and 22.8 ± 3.7 mm/m for African-Americans.
2.6 Combined effects of demographic, clinical and echocardiographic descriptors on left atrial size and indexed size
Multiple regression analysis considering demographic, clinical and echocardiographic characteristics was done to model LA size. The following potential predictors were evaluated: body mass index, age, race, systolic blood pressure, heart rate, physical activity, plasma renin, sodium excretion, LV mass and relative wall thickness. Two models were developed. The first evaluated which of these variables were independently significant to assess their relative importance in predicting LA size. In the second model, all possible two-way interactions of the variables selected in the first model were also considered. By convention, if an independent variable became nonsignificant with the addition of the interaction terms, it was retained in the model if it was also represented in a significant interaction.
The second model is presented in Table 2. Body mass index was the most important predictor, explaining 13% of the variation in LA size. This was followed, in descending order, by race, LV mass–body mass interaction, age–race interaction, age and LV mass. The final model explained 20% of the variation in LA size. Body mass index and its interaction with LV mass accounted for 76% of the variation explained by the model.
A similar modeling process was done for indexed LA size. Body mass index independently accounted for 15% of the variation in indexed LA size. This was followed by race, age, LV cavity dimension, age–race interaction, LV cavity dimension–body mass index interaction, systolic blood pressure, sodium excretion–body mass index interaction, race–LV diastolic dimension interaction, sodium excretion and LV diastolic dimension–sodium excretion interaction. The final model explained 24% of the variation in indexed LA size. Body mass index and its interactions with LV cavity dimension and sodium excretion accounted for 68% of the variation explained by the model.
Although echocardiographic studies of hypertension have evaluated the structure and function of the left ventricle, effects of hypertension on the left atrium are also important. Left atrial size is of clinical relevance for the occurrence of atrial fibrillation (), stroke ([6, 22]) and congestive heart failure. In an early echocardiographic study, Savage et al. () found LA enlargement in 5% (in comparison with increased LV mass in 51%) of hypertensive patients, even though mean LA size was not increased. Left atrial dimension was weakly correlated (r = 0.20) with systolic blood pressure, as was LV mass (r = 0.23). Another echocardiographic study of hypertension (Systolic Hypertension in the Elderly Program [SHEP]) found LA enlargement on the echocardiogram in 51% of elderly patients with systolic hypertension, but also in 30% of age-matched, normotensive control subjects (). It is uncertain, however, whether the LA enlargement noted in these normotensive control subjects represented normal, age-related increases in LA size in comparison with a younger reference group, or pathologic increases in LA size consequent to impaired LV filling or other degenerative cardiac changes. There were weak but significant relations in the SHEP study between LA size and LV mass indexed by body surface area (r = 0.20), as well as with relative wall thickness (r = 0.31). In the Framingham Study of a free-living cohort (), blood pressure and LV mass were both correlated with LA enlargement on the echocardiogram, although on multivariable analysis most of the variance in LA dimension was explained by LV mass and body weight.
Our study extends the findings of the Framingham group in noting that even in patients with established hypertension and with a high prevalence of LV hypertrophy, obesity is nonetheless the most important predictor of LA size, LA size indexed by height and LA enlargement. Moreover, obesity magnifies the relation between LV mass and LA size. However, the relation of systolic blood pressure to LA size is weak when controlling for obesity, LV mass and race and their interactions.
3.1 Race and left atrial size
Not noted in the Framingham Study, possibly because of the relatively small proportion of African-American subjects available for study, was our finding of a relation between race and LA size in older hypertensive men, even after adjustment for covariates. The finding of a greater LA size and a higher prevalence of LA enlargement in older white than in older African-American men is somewhat surprising. In addition, the expected direct relation of age to LA size was maintained only in white hypertensive men. Although LV mass did not differ by race (), relative wall thickness, itself a predictor of outcome (), was greater in the African-American patients (). Conceivably, in African-American men, a parallel increase in LA wall thickness may limit distensibility and hence cavity size of the left atrium. Because LA wall thickness is not measurable by echocardiography, this hypothesis is not testable in the present data set.
A recent preliminary report from the Cardiovascular Health Study (), which evaluated predictors of LA size in a community-based cohort of 5,201 elderly subjects, also found a greater LA size in whites than in African-Americans. This relation retained significance even after adjustment for obesity, blood pressure and LV mass. Notably, body mass index was also the most important predictor of LA size, even in an elderly cohort.
The present study evaluated a relatively large number of hypertensive patients with a high proportion of African-American participants. Hence, the study was sufficiently powered to determine the relative influences of clinical and LV structural characteristics on LA size.
3.2 Study limitations
However, it is important to note several limitations. Because men with a relatively high prevalence of LV hypertrophy were evaluated, the findings of the present study may not be generalizable to women or to patients with less severe hypertensive disease.
Clinically important mitral regurgitation usually results in LA dilation. Systematic Doppler examination was not performed in the present study to determine the presence and severity of coincidental mitral regurgitation. However, the presence of clinically evident valvular heart disease was a criterion for exclusion from the study. Hence, it is unlikely that a sufficient number of patients with mitral regurgitation severe enough to bias the results would have been recruited.
Much of the variation in LA size in this study was not accounted for by the factors examined in the multivariate models. However, in large-scaled studies it is not unusual to see regression models that account for only a small part of the variation in the response measured. Unmeasured factors that may account for variation in LA size include biologic variability of the subjects and methodologic variability in echocardiographic acquisition, as well as other undiscovered factors that might have a direct relation to LA size. Also, despite a positive correlation of LA size to height, height-indexed LA size displayed a negative correlation even after adjustment for other covariates. This suggests that algebraic indexation of LA size by height does not remove the influence of height from the measurement.
Lastly, a single LA dimension obtained from M-mode echocardiography, representing the anteroposterior diameter, may not accurately reflect LA volume in all patients, particularly if LA enlargement is restricted to superoinferior or mediolateral planes. Also, in some patients, dilation of the left atrium may occur asymmetrically or may be greatest distal to the relatively proximal placement of the cursor in relation to the mitral valve plane. However, the placement of the M-mode cursor was guided by two-dimensional visualization of the LA area in two orthogonal anatomic planes so as to traverse the maximal diameter of the chamber. Hence, it is unlikely that there was misrepresentation of the LA volume by inappropriate placement of the M-mode cursor in a substantial number of cases. Furthermore, preliminary data have suggested that even in different disease states, LA enlargement occurs in all three anatomic planes (). In fact, numerous studies establishing relations between LA size and biologic covariates have all employed two-dimensional guided M-mode echocardiography ([2–7, 22, 27]).
3.3 Clinical implications
Previously, the importance of obesity in hypertension was underscored by its relation to the prevalence and severity of hypertension as well as its independent effect on LV mass, itself an important predictor of clinical outcome. The present study extends these observations by showing that LA size, also an important predictor of clinical events, is strongly related to obesity. Moreover, obesity was the strongest predictor of LA size, followed by LV mass, age in whites and LV cavity size. In contrast, the effect of systolic blood pressure was marginal, even in this hypertensive cohort with a wide range of systolic blood pressures. In addition, greater LA size in older white hypertensive men suggests the possibility of enhanced morbidity associated with LA size (i.e., atrial fibrillation and embolic stroke).
- analysis of variance
- American Society of Echocardiography
- left atrial
- left ventricular
- Systolic Hypertension in the Elderly Program
- Received July 1, 1996.
- Revision received October 31, 1996.
- Accepted November 26, 1996.
- The American College of Cardiology
- Pearson AC,
- Gudipati C,
- Nagelhout D,
- Sear J,
- Cohen JD,
- Labovitz AJ
- Henry W,
- Morganroth J,
- Pearlman A,
- et al.
- Gottdiener JS,
- Reda DJ,
- Materson BJ,
- et al.
- Mayet J,
- Shahi M,
- Foale RA,
- Poulter NR,
- Sever PS,
- Thom SA
- Chaturvedi N,
- Athanassopoulos G,
- McKeigue PM,
- Marmot MG,
- Nihoyannopoulos P
- Moser M,
- Setaro JF
- the Committee on M-Mode Standardization of the American Society of Echocardiography,
- Sahn DJ,
- DeMaria A,
- Kisslo J,
- Weyman A
- Troy BL,
- Pombo J,
- Rackley CE
- Devereux RB
- Gray D
- ↵(1989) SAS/STAT Users Guide, Version 6 (SAS Institute, Cary (NC)), 4th ed.
- Woolson RF
- Savage DD,
- Drayer JIM,
- Henry WL,
- et al.
- Tsang SM,
- Sholinsky PD,
- Gardin JM,
- Smith VE,
- Gottdiener JS,
- Manolio TA
- Abascal VM,
- Patel AR,
- Weyman AE,
- Davidoff R