The LIFE echocardiography sub-study evaluated 960 patients with stage II–III hypertension. To qualify for LIFE, subjects had a screening ECG that demonstrated LVH, by either gender-adjusted Cornell voltage duration criteria or Sokolow–Lyon voltage criteria, as previously described [3, 4]. Patients with an aortic valve mean pressure gradient > 20 mmHg, symptomatic heart failure, or known LV EF < 40% were excluded from the study. The design of the LIFE study has been previously reported [3, 4]. Physical examination, laboratory testing, 12-lead ECG and standard echocardiogram, including color Doppler, were performed annually for five years during either losartan- or atenolol-based anti-hypertensive treatment [4].

For the purpose of the present study, prevalent cases of cardiovascular disease, including myocardial infarction, coronary artery disease, and congestive heart failure were excluded as well as reduced EF < 55% at baseline and aortic/mitral regurgitation (> 1+) or stenosis of any degree identified by Doppler echocardiography at baseline or during follow-up. The present analysis was therefore performed in 680 hypertensive patients (66 ± 7 years; 45% women) meeting eligibility criteria (Table 1). An independent endpoint committee adjudicated inter-current fatal and nonfatal myocardial infarction according to definitions provided in the LIFE study predefined endpoint manual [4]. The specific study endpoint for the present analysis was the development of reduced systolic function, defined as EF < 50% at the end of the study, representing a reduction from baseline of at least five EF points. In order to discriminate clearly from baseline EF values, we did not use a more recently applied sex specific EF reduction of < 52% in men and < 54% in women with uncomplicated hypertension.

Echocardiography was performed following a standardized imaging protocol which included 2-dimensionally guided M-mode echocardiograms and selected 2-dimensional and Doppler recordings as previously reported [5]. All echocardiograms were transferred to the Echocardiographic Laboratory of Weill Cornell Medical College for central reading.

Measurements were performed off-line by junior readers, and quality assured by the same, highly-experienced reader, blinded to clinical data, using computerized review station with digitizing tablet and monitor screen overlay for calibration. Reliability of measurements of both LV geometry and LV functional parameters performed by the laboratory has been previously extensively reported [6].

LV mass was calculated using anatomically validated methods [7, 8] and normalized by height to the allometric power of 2.7. Partition values to define LVH were 49.2 g/m^2.7 for men and 46.7 g/m^2.7 for women [9]. Relative wall thickness was calculated as posterior wall thickness in diastole/LV internal radius [10]. LV EF was calculated using the Teichholz method [11].

Left atrial diameter was measured in the parasternal long-axis view at end-systole [12]. Circumferential end-systolic wall stress, a measure of myocardial afterload, was measured assuming a cylindrical LV geometry [13].

As previously described [14], diastolic transmitral blood flow velocity was measured at the mitral leaflet tips. The leading edge of the transmitral Doppler flow pattern was traced to derive the peak of early diastolic and atrial phase LV filling (“E” and “A”, respectively), and the deceleration time of early diastolic LV filling. Isovolumic relaxation time was measured as the time interval between aortic valve closure and mitral valve opening.

In the main double-blinded LIFE study BP was measured as the average of two recordings with a 1-min interval after patients had been quietly seated for 5 min in the morning hours at through at all regular study visits. Investigators had been directed to use a validated, calibrated manual BP manometer maintained in good condition and a cuff of proper size. The arm was positioned so that the location of the stethoscope head was kept at the level of the heart.

In the echocardiographic sub-study, BP was also measured standardized in the beginning of the annual examination as the average of two recordings with a 1-min interval after a 5 min of quiet rest in the supine position using a manual sphygmomanometer or a semiautomatic sphygmomanometer (Dinamap Model 1846 SX, Critikon, Johnson & Johnson Medical Inc., USA).

Data were expressed as mean ± standard deviation (SD) for continuous variables and as percentages for categorical variables. Patients were categorized into two groups according to the presence of incident EF reduction. Groups were compared by Student’s t-test for continuous variables, chi-square for categorical variables and analysis of covariance (ANCOVA) for echocardiographic variables, controlling for relevant confounders. Significant changes in echocardiographic variables over time were expressed as percentage changes from baseline and compared among groups by ANCOVA, adjusting for age, gender and changes over time in heart rate and systolic BP. Time-varying covariate Cox regression analyses were used to identify in-treatment echocardiographic predictors of EF reduction. The last available echocardiogram before reduced EF occurred was used in these models. To assure stability of the time-varying covariate Cox regressions, models were set to have the ratio of number of covariates to number of end points ≈ 1:10. Accordingly, a propensity score variable was obtained through a preliminary logistic regression for baseline EF from anthropometric and clinical data (including age, gender, body mass index, heart rate, systolic and diastolic BP) [15, 16]. A propensity score was included mainly to ensure reduced possibility for selection bias and thus perform optimal statistical analysis in light of the small no. of 37 patients who developed reduced systolic function. The resulting variable was added to the time-varying Cox regression model together with baseline EF, treatment arm, and in-treatment time-varying systolic BP to represent the influence of BP control. Subsequently, non-significant covariates from the previous model were replaced by in-treatment time-varying heart rate, and in-treatment echocardiographic parameters by a stepwise procedure [including in-treatment time-varying LV mass, in-treatment LV diastolic diameter, in-treatment left atrium (LA) diameter, in-treatment mitral E/A ratio and in-treatment mitral E wave deceleration time] to identify echocardiographic predictors of EF reduction.

During a mean follow-up of 4.8 ± 1 years, 41 patients (6%) developed reduced EF. In four of these 41 patients (10 %) an inter-current myocardial infarction was recorded; accordingly, these patients were excluded from further analysis. Analyses from adverse event database could neither detect any concomitant diseases that could explain reduction in EF.

As shown in Table 1, the remaining 37 patients who developed reduced EF without clinically detected myocardial infarction were mostly men (72 %). Compared to patients who did not develop reduced EF (n = 643), those with incident reduced EF had similar baseline age, body mass index, heart rate, and mean systolic and diastolic BP. No difference was found between groups in randomized treatment, prevalence of diabetes, atrial fibrillation, chronic obstructive pulmonary disease, or in laboratory data, including fasting glucose, serum creatinine, total or HDL-cholesterol, or serum electrolytes [all P = not significant (NS)].

ANCOVA for echocardiographic baseline characteristics, adjusting for differences in age, gender, systolic BP and heart rate is shown in Table 2. Patients with subsequent reduced EF had higher LV diameter, LV mass and prevalence of LVH, and lower EF at baseline (all P < 0.05), without significant differences in relative wall thickness, LV midwall shortening, end-systolic stress (ESS), LA diameter or indices of diastolic function compared to patients who did not develop reduced EF.

The timeframe for the development of reduced EF in the study sample is shown in Figure 1. As shown, approximately 70% of patients who developed reduced EF had a normal EF through the 3rd year of follow-up, while nine (24%) patients showed signs of systolic dysfunction within the first 2 years of the study. Patients who developed reduced EF had an average EF reduction of 18% (ranging from –6% to –34%).

Data from the last echocardiographic examination available before the EF reduction or at study end in patients with maintained EF are included in Table 3. Despite similar reduction in mean systolic and diastolic BP, patients with reduced EF showed significantly greater mean increase in LA size (+7.9% vs. –1.0%), LV diameter (+7.2% vs. +3.8%), ESS (+20.8% vs. +10.6%) and mitral E/A ratio (+48.4% vs. +15.5%, all P < 0.05) compared to patients who did not develop reduced EF. These changes were independent of gender, or changes in BP and heart rate. Parallel trends in LV mass and isovolumic relaxation time in the two groups did not attain statistical significance.

Time-varying Cox regression models for the development of reduced EF are included in Table 4. As is shown in Table 4A, a preliminary model included the propensity score for baseline EF derived from clinical characteristics, treatment arm, baseline EF and in-treatment systolic BP. The propensity score, randomized treatment and in-treatment systolic BP were not significantly associated with increased risk of reduced EF. Accordingly, these covariates were replaced in the following models by in-treatment heart rate and by in-treatment echocardiographic parameters. The final model is shown in Table 4B. Independently of baseline EF and of the significant impact of in-treatment heart rate, the strongest predictors for the development of reduced EF were higher in-treatment LV diameter [hazard ratio (HR) = 5.19 per cm; 95% confidence interval (CI): 2.58–10.41] and higher in-treatment mitral E/A ratio (HR = 2.37 per unit; 95% CI: 1.58–3.56; both P < 0.0001). In separate models, excluding in-treatment LV diameter, borderline significant impacts could also be detected for in-treatment LA diameter (HR = 1.67; 95% CI: 1.05–2.72; P = 0.032) without significant association with in-treatment LV mass or in-treatment mitral E wave deceleration time.

Our data suggest a possible pathophysiological progression leading from hypertension to overt LV systolic dysfunction, suggested by the changes in echocardiographic parameters occurring in the sample patient shown in Table 5. After one year of anti-hypertensive treatment a reduction in LV diameter and LV mass occurred. However, after three years, despite normal EF (57%) the LV significantly dilates, ESS increases, and signs of increased LV filling pressures can be identified (increased mitral E/A ratio and LA diameter), preceding the reduction in EF to 47% occurring at year 4. These data suggest a mechanism, shown schematically in Figure 2, whereby chronic pressure overload hypertrophy eventually gives rise to increased heart size and ESS, and begets LV dysfunction.

The LIFE study was a prospective and randomized clinical trial set up to compare cardiovascular outcomes on the angiotensin-receptor blocker losartan vs. the beta-1 selective receptor-blocker atenolol. Though the echocardiographic sub-study in LIFE was pre-specified with separate protocol from the outset, the present investigation of incident EF reduction unrelated to myocardial infarction was an ad hoc analysis. Besides, few parameters were used to characterize LV systolic function, namely EF and the midwall shortening because data were collected in a time-period when global LV strain and mitral ring velocities using tissue Doppler were not available study tools. Further, EF is not the optimal functional parameter especially in hypertensive patients with LVH.