Relationship of the electrocardiographic strain pattern to left ventricular structure and function in hypertensive patients: the LIFE study

Subjects

The Losartan Intervention For End point (LIFE) study (12,13)enrolled hypertensive patients with ECG LVH by Cornell voltage-duration product (14,15)and/or Sokolow-Lyon voltage (1)on a screening ECG in a prospective double-blind study to determine whether appreciable reduction in mortality and morbid events is associated with the use of losartan as opposed to atenolol (12). Eligible patients for LIFE were hypertensive men and women aged 55 to 80 years with a seated blood pressure of 160 to 200/95 to 115 mm Hg after one and two weeks on placebo. Study inclusion and exclusion criteria have been previously published (12,13). A previously described (16)representative sample of the whole LIFE study, totaling 964 patients, underwent baseline echocardiograms; 21 patients had unmeasurable echocardiographic LV mass and 57 had incomplete baseline ECG measurements. There were 361 eligible women and 525 eligible men whose mean age was 66 ± 7 years.

Electrocardiography

All ECGs were interpreted at the Core Laboratory at Sahlgrenska University Hospital/Östra in Göteborg, Sweden, by investigators blinded to the clinical information (12,13). The product of QRS duration × Cornell voltage (R_aVL+ S_V3, with 8 mm added in women [14,15]) was used with a threshold value of 2,440 mm · ms to identify LVH. After design of LIFE, studies were published suggesting a smaller gender adjustment (6,17), and feedback from LIFE investigators showed that otherwise eligible patients had ECG LVH by highly specific but insensitive Sokolow-Lyon voltage criteria (1), but not by Cornell product. Accordingly, changes were made in ECG entry criteria: the gender adjustment of Cornell voltage was reduced from 8 to 6 mm and Sokolow-Lyon voltage (S_V1+ RV_5/6) >38 mm was accepted for ECG eligibility (13). Additional ECG measurements were performed at Helsinki University Central Hospital. Repolarization abnormalities in leads V₅and/or V₆indicated typical strain (11)when there was a downsloping convex ST segment with an inverted asymmetrical T-wave opposite to the QRS axis.

Echocardiography

Studies were performed using fundamental imaging with commercially available phased-array echocardiographs as previously described (16,18). Left ventricular internal dimension and wall thicknesses were measured at end-diastole by American Society of Echocardiography (ASE) recommendations (19)on up to three cardiac cycles. When M-mode recordings could not be optimally oriented, correctly oriented linear dimension measurements were made using two-dimensional imaging by leading-edge ASE convention (20). Methods for measurement and evidence of interchangeability of two-dimensional and M-mode LV dimensions have been described in detail (18). Relative wall thickness was calculated as end-diastolic posterior wall thickness/LV radius; LV mass was calculated (21)and was indexed for body surface area and alternatively for height^2.7(22). Under criteria known to predict an adverse prognosis (23–25), LVH was considered present if LV mass index was >104 g/m²in women or >116 g/m²in men. Hypertrophy was considered concentric if LV relative wall thickness was >0.430 and eccentric if relative wall thickness was normal (16); patients with normal LV mass were considered to have normal LV geometry if relative wall thickness was ≤0.43 or to have concentric remodeling if relative wall thickness was increased. Alternative analyses used LV mass/height^2.7partition values of 46.7 and 49.2 g/m^2.7in women and men, respectively.

Myocardial contractile performance was assessed by examining LV systolic shortening in relation to circumferential end-systolic stress (ESS) (26). The relation of LV midwall shortening to midwall circumferential ESS at the LV minor axis was calculated as previously described (27), and was expressed as a percent of the value predicted from circumferential ESS by an equation derived in apparently normal adults (28). This variable, stress-corrected midwall shortening was considered low if <89.2%, the 5th percentile in a separate reference population of 280 normal adults (28). To estimate myocardial oxygen demand, the ESS-LV mass-heart rate product was calculated as previously described (29). Fractional and midwall shortening could be determined in 100% of patients, whereas circumferential ESS, stress-corrected midwall shortening and ESS-LV mass-heart rate product could be determined in 97% (857/886) of patients in this study.

Statistics

Patients were classified as having CHD if they had self-reported angina or myocardial infarction, diagnostic Q-waves by Minnesota code on the ECG or segmental wall motion abnormalities on the two-dimensional echocardiogram. Differences in prevalence were compared using χ²analyses and mean values of continuous variables by unpaired ttest. Echocardiographic variables were further compared using analysis of covariance to adjust for differences in age, gender, race, systolic pressure, cholesterol and high density lipoprotein (HDL) cholesterol, creatinine and prevalent diabetes between groups. Independent relations of echocardiographic LVH, abnormal LV geometry and abnormal stress-corrected midwall shortening to ECG strain were determined using stepwise logistic regression analyses and the same covariates. A two-tailed p < 0.05 was required for statistical significance.

Patient characteristics

Strain was present in 132 patients (15%) and was more common in patients with CHD than in those without (29%, 51/175 vs. 11%, 81/711, p < 0.001). Clinical and demographic characteristics of patients with or without strain on the ECG, stratified by CHD status, are shown in Table 1. Among patients without CHD, those with typical strain were more likely to be men and African-American with higher systolic pressures, lower HDL cholesterol and higher serum creatinine levels than patients without strain. Among patients with evident CHD, those with strain similarly were more likely to be African-American, have higher serum creatinine, a slightly lower HDL cholesterol, and a greater prevalence of diabetes, but were similar to patients without strain with regard to gender, systolic pressure and serum sodium and potassium levels.

Echocardiographic findings in relation to ECG strain

Relations of LV structure and function to the presence or absence of ECG strain are examined in Table 2. In patients with or without CHD, strain on the ECG was associated with greater LV wall thicknesses, LV mass and LV mass indexed for either body surface area or for height^2.7. Among patients without CHD, LV internal dimensions were also greater in patients with strain. In both patient groups, because strain was associated with a greater increase in LV wall thickness than cavity dimension, strain was also associated with increased relative wall thickness. Strain was associated with similar LV performance at the endocardium, as measured by fractional shortening, and with similar levels of circumferential ESS, but patients with strain had lower LV midwall performance than predicted for the level of ESS, as reflected by lower stress-corrected midwall shortening. Electrocardiographic strain was associated with increased wall mass-stress-heart rate product only among patients without CHD.

Prevalences of echocardiographic LVH, abnormal LV geometry and LV midwall function in patients with or without ECG strain are shown in Table 3. In this population selected on the basis of ECG LVH by Cornell product and/or Sokolow-Lyon voltage, strain was associated with a higher prevalence of LVH defined by LV mass/body surface area in patients with and without CHD. Alternative analyses using gender-specific LV mass/height^2.7partitions confirmed higher prevalences of LVH in patients with ECG strain, both in those with CHD (92% vs. 84%, p = 0.226) and without CHD (82% vs. 70%, p = 0.039). Independent of the presence or absence of evident CHD, patients with strain were less likely than those without strain to have normal LV geometry or concentric remodeling and had a similar prevalence of eccentric LVH, but were much more likely to have concentric LVH than patients without strain on the rest ECG. Depressed LV contractility, as estimated by stress-corrected midwall shortening <89.2%, was more common in CHD patients with strain, but did not differ in prevalence in patients without CHD according to the presence or absence of strain.

Because patients with and without strain differed with respect to demographic and clinical variables that could affect LV structure and function (Table 1), independent relations of echocardiographic findings to the presence of strain were examined in patients with and without CHD by analysis of covariance (Table 4). After adjusting for baseline differences, strain remained strongly associated with increased LV wall thicknesses, relative wall thickness, LV mass and LV mass indices in patients with and without CHD. Left ventricular internal dimensions, fractional shortening and ESS were similar in patients with or without strain, but midwall shortening remained moderately depressed in those with strain. Similarly, after taking into account differences in systolic pressure and normalizing for differences in LV dimension between groups, ESS was similar in patients with and without strain, but stress-corrected midwall shortening remained lower in patients with strain, attaining statistical significance in patients with CHD. Wall stress-mass-heart rate product was elevated in patients without evident CHD after adjusting for baseline differences.

In parallel fashion, relations of echocardiographic LVH, abnormal LV geometry and depressed LV contractility to strain were further examined using separate logistic regression analyses in patients with and without CHD (Table 5). After adjusting for baseline differences, ECG strain was associated with a nearly threefold greater risk of LVH in patients without CHD and a greater than fivefold increased risk of echocardiographic LVH in patients with CHD. In nominal regression analyses where the risk of abnormal LV geometry compared with normal geometry was determined for the presence of strain, ECG strain was most strongly related to increased risk of concentric LVH, was strongly associated with presence of eccentric LVH and was not related to concentric remodeling in patients with or without evident CHD. Strain remained associated with decreased LV contractility, as determined by stress-corrected midwall shortening <89.2%, in patients with CHD but not in patients without evident CHD. Of note, logistic regression analyses performed in the entire study population using an indicator variable for CHD confirmed these relations, demonstrating that ECG strain was strongly associated with echocardiographic LVH, especially of concentric pattern, and was modestly related to depressed stress-corrected midwall shortening.

Relationship of strain to LVH

Previous studies have documented a clear relation between strain and anatomic LVH (1–3,5,6)and an increased prevalence of strain with more severe hypertrophy (3). However, most studies did not account for potential effects of concomitant CHD, which is highly covariate with LVH (4,24,25)and could contribute to the genesis of ST/T-wave changes (3,8,10). Schillaci et al. (6)demonstrated low sensitivity (16%) and high specificity (98%) of strain for LVH in 923 white hypertensive patients without clinically evident CHD, but did not examine differences in LV structure and mass in relation to strain. In a subset of 40 patients with severe isolated aortic regurgitation without significant obstructive CHD at angiography, Roman et al. (11)demonstrated that strain was associated with greater LV mass. However, neither study adjusted for other variables that could affect the relation between strain and increased LV mass. The current study significantly extends these observations to a large cohort of patients with hypertension, demonstrating that strain is associated with greater LV mass, indexed LV mass and higher prevalences of echocardiographic LVH in patients with and without CHD, even after taking into account baseline differences that could contribute to differences in the prevalence of ECG strain.

Development of typical ST segment and T-wave changes of the strain pattern in response to hypertrophy is predicted by a distributed dipole model of the ECG response to hypertrophy (30), which demonstrates that T-wave amplitudes are proportional to the square of the cell radius and that the flattening or inversion of the T-wave observed with LVH can be attributed to the contiguity effect of adjacent layers having different transmembrane action potential durations. These findings suggest that greater degrees of ST depression and T-wave inversion will be associated with larger increases in LV mass, consistent with the linear relations of these variables to LV mass (31).

Relationship of strain to LV geometry

The relationship of strain to LV geometry has been less well studied. In 107 patients not on digitalis (3), strain was more strongly associated with an LV internal dimension >55 mm (50% vs. 13%, p < 0.01) than with posterior wall thickness >12 mm (31% vs. 14%, p = 0.06), suggesting a stronger relationship between strain and eccentric as opposed to concentric hypertrophy. Among patients with pure aortic regurgitation (11), typical strain was associated with both increased LV chamber dimension and wall thicknesses, and with an increased relative wall thickness that nonetheless remained within the normal range, such that the overall LV geometric pattern reflected an eccentric hypertrophic response consistent with increased LV volume load. The present study extends these observations to a more heterogeneous population of hypertensive patients, demonstrating that ECG strain was associated with a greater prevalence of concentric as opposed to eccentric LVH due to increased wall thickness out of proportion to slightly increased LV chamber size after controlling for other factors (Table 4).

Relationship of strain to LV performance

The relation of ECG strain to measures of LV function has not been previously examined in hypertensive patients. Devereux and Reichek (3)and Roman et al. (11)found the prevalence of reduced LV systolic function to be higher in patients with the strain pattern. Moreover, Roman et al. (11)reported lower fractional shortening by echocardiography and lower ejection fraction on radionuclide angiography in patients with strain. However, neither study examined measures of LV contractility, nor did they adjust for possible confounders by multivariate analyses. The present study extends these observations to hypertensive patients with CHD demonstrating lower myocardial contractility, as estimated by stress-corrected midwall shortening (Table 4)and a more than 2.5-fold greater prevalence of abnormally low stress-corrected midwall shortening (Table 5), among CHD patients with strain. However, strain was not associated with depressed midwall shortening in patients without evident CHD.

Mechanisms of abnormal repolarization

The present study provides additional support for several possible mechanisms for the repolarization abnormalities of the strain pattern in the absence of CHD. First, the association between strain and a primarily concentric geometric pattern of LVH, due to greater increases in LV wall thickness than in chamber dimension, suggests a possible primary effect of myocardial cell hypertrophy, consistent with the model proposed by Thiry et al. (30). Second, ST depression and T-wave inversion may reflect true subendocardial ischemia in the absence of CHD attributable to increased LV mass and wall thickness (32). This hypothesis is supported by the association of strain with increased wall stress-mass-heart rate product among patients without CHD, providing evidence of a major demand-side predisposition to myocardial ischemia. Indeed, although coronary artery size increases with LV mass and is related to regional myocardial mass (33), compensatory increases in coronary size, as measured by lumen area to LV mass ratio, are frequently inadequate to match the increased mass (33,34). The finding that the ratio of coronary lumen area to regional LV mass can partially normalize with regression of LVH after aortic valve replacement for aortic stenosis (35)suggests that repolarization abnormalities of strain may also diminish or disappear with regression of hypertrophy after antihypertensive therapy.

Implications

These findings have important clinical implications. First, the increased morbidity and mortality associated with anatomic LVH (24,25), and particularly with concentric hypertrophy (25)and abnormal midwall LV mechanics (29), provide new insights into the demonstrated prognostic value of strain (8–10). The association of strain with male gender and African-American race may in part explain the higher event rates in these groups. These findings and the subjective and qualitative nature of strain criteria currently employed provide impetus to develop quantitative, computerized ST segment and T-wave amplitude measurements to identify LVH and stratify risk. Further study is needed to determine whether hypertrophy regression is associated with reversal or reduction of repolarization abnormalities of strain, and whether serial evaluation of quantitative ST segment and T-wave measurements will provide additional insight into this process.