Degenerative mitral valve disease, usually related to mitral valve prolapse, is responsible for most cases of isolated mitral regurgitation (MR) in the United States (1). There is no effective medical therapy for isolated MR, and therefore, surgery is recommended in patients with severe MR and symptoms or evidence of progressive left ventricular (LV) dysfunction (2). The natural history of MR is progressive LV dysfunction and adverse LV remodeling, eventually leading to heart failure. Initially, LV dilation and augmented stroke volume occur, facilitated by an increase in preload and by ejection into the relatively low-pressure left atrium. These changes are accompanied by increased sympathetic drive early in MR in both animal models (4) and humans (6). However, prolonged excessive adrenergic stimulation has a cytotoxic effect on cardiomyocytes (7), resulting in increased LV end-systolic dimension or volume and wall stress. In the canine model of isolated MR, chronic β₁-adrenergic receptor (AR) blockade has been shown to improve cardiomyocyte and LV function (8). In patients with severe MR and normal LV ejection fraction (EF), β-AR blockade is associated with a survival benefit with or without coronary artery disease (10).

In humans with moderate to severe degenerative MR, 14-day treatment with β₁-AR blockade decreases LV work, but with increases in LV end-diastolic volume (EDV) and end-systolic volume (ESV) and no change in LV EF (11). However, the beneficial effect of β₁-AR blockade on LV function in heart failure is achieved after long-term therapy, and there has not been a human trial of extended β₁-AR blockade on LV remodeling and function in patients with isolated MR. Therefore, the current randomized, controlled study used magnetic resonance imaging (MRI) with tissue tagging and 3-dimensional (3D) analysis as a surrogate outcome to evaluate the effects of long-term β₁-AR blockade on LV remodeling and function in patients with chronic, isolated MR.

Eligible patients had moderate or severe MR documented by color flow Doppler, LV EF of more than 55%, LV end-systolic dimension of <40 mm, and echocardiographic thickening of the mitral valve leaflets and prolapse. Patients were excluded with New York Heart Association functional class III or IV symptoms, previous myocardial infarction, significant coronary artery disease by exercise testing with myocardial perfusion imaging, significant other valvular disease, serum creatinine level of more than 2.5 mg/dl, and hypertension requiring medical treatment. The study was approved by the University of Alabama Institutional Review Board, and all subjects gave written informed consent.

We conducted a randomized, double-blind study with 2 years of treatment of β₁-AR blockade with Toprol XL (AstraZeneca, London, United Kingdom) (range: 25 to 100 mg/day) versus placebo in patients with moderate to severe MR. Toprol XL was administered with a starting dose of 12.5 to 25 mg/day and was titrated as tolerated at 2-week intervals to a maximum of 100 mg/day. After randomization, patients underwent MRI scanning, which was repeated at 6, 12, 18, and 24 months after randomization. MRI also was performed in control volunteers (mean age: 52 ± 11 years, range: 35 to 70 years) who had no prior history of cardiovascular disease and were not taking any cardiovascular medications.

As previously described (12), MRI was performed on a 1.5-T MRI scanner (Signa GE Healthcare, Milwaukee, Wisconsin) optimized for cardiac application. The short axis volumes defined by the contours, excluding papillary muscles, were summed to calculate LV volume (12). LV volume-time curves were constructed and differentiated with respect to time to obtain peak diastolic filling rates (13).

Three-dimensional LV geometric parameters were measured from surfaces fit to endocardial and epicardial contours manually traced near end-diastole and end-systole. 3D wall thickness was computed by measuring the perpendicular distance from a point on the endocardial surface to the closest point on the epicardial surface.

Tagged MRI scans were acquired with repetition times of 8 ms, echo times of 4.4 ms, and tag spacing of 7 mm. 3D LV strain was measured from tagged images at end-systole, which was defined by visual inspection of the image data as the time frame with maximum contraction (14). Two-dimensional strain rates were measured using harmonic phase analysis (15), which measures local myocardial 2-dimensional strain based on the local spatial frequency of tag lines. Strain and strain rates were computed at each segment in the American Heart Association 17-segment model and were averaged over the mid-ventricular segments.

Analyses of MRI scans were performed blinded and side by side in 1 sitting for each patient. MRI outcome variables were categorized according to the following: 1) LV geometry: LV end-diastolic volume (EDV)/body surface area (BSA), LV ED mass/LV EDV, 3D LV ED radius/wall thickness (midwall); 2) LV systolic function: LV EF, LV ESV/BSA, 2-dimensional LV systolic longitudinal strain rate; and 3) LV diastolic function: LV peak early filling rate (EDV/s).

The t test (for continuous variables) and Fisher exact test (for categorical variables) were used to compare demographic characteristics, clinical characteristics, and outcome variables at baseline between the 2 groups. Treatment differences in the rate of progression, assessed on an intention-to-treat basis, was the focus of the comparisons over time between the treatment groups. Rate of progression for each outcome was modeled with SAS PROC MIXED (SAS Institute, Cary, North Carolina) assuming the best working correlation structure based on the Bayesian information criterion from the choices of autoregressive lag 1, heterogeneous autoregressive lag 1, compound symmetry, and heterogeneous compound symmetry. Compound symmetry was found to be the best structure for all outcomes except for LV EDV/BSA, where autoregressive lag 1 was found to be the best structure. A significant interaction effect between the linear component of time and the treatment group was the measure of treatment effect, assessed at p < 0.05.

MRI examinations were performed at 6-month intervals over the 2-year study period. To achieve a more accurate representation of a patient's rate of progression over time, the actual time of the MRI relative to randomization that is closest to a 3-month interval (i.e., 0, 3, 6, 9, 12, 15, and so on) was used in the analysis, instead of the target 6-month visit. For example, if a patient was scheduled to have a 12-month MRI but actually underwent an MRI examination 2.5 months late from the target visit, the time associated with the MRI data for this visit was considered to be at 15 months, instead of 12 months.

Demographic data are summarized in (Table 1). Thirty-eight patients were enrolled: 19 in the placebo group and 19 in the Toprol group. One patient in the placebo group dropped out soon after the randomization, and 1 patient in the Toprol group died of pulmonary embolus after cosmetic surgery shortly after the 12-month visit. Thus, 36 patients had 2 years of follow-up data. However, these patients were included in the random effects models on an intention-to-treat principle.

Baseline age, sex, race, heart rate, or blood pressure did not differ between groups. Of the 38 MR patients, 10 had holosystolic murmurs, with 5 in treated patients and 5 in untreated patients. No patients had a flail leaflet. No patients had atrial fibrillation, and all patients were New York Heart Association functional class I or II, with 90% and 95% of placebo and treatment groups being in New York Heart Association class I, respectively.

At baseline, MR patients had 35% higher LV EDV/BSA, 50% higher LV ESV/BSA, 33% higher LV stroke volume/BSA, and 18% lower LV ED mass/EDV ratio, whereas LV EF was slightly lower in MR patients versus control subjects ((6), 6).

First-order and second-order polynomials for time effects were fitted for all outcome variables, and coefficients associated with the second-order terms were not found to be significant. Therefore, a first-order (linear) time effect model was used. (Figure 29_gr1) display the raw data by treatment group for each LV outcome with the fitted lines as well as the fit with 95% confidence bands based on individual confidence intervals for the means at each time point. (Table 2) displays the estimates and standard errors for the annual rates of progression (slope × 12 months) for each group including the p values, which compare the estimates between the 2 groups.

There were no significant treatment effects found on LV EDV/BSA (p = 0.457), LV ED mass/EDV (p = 0.1967), or LV ED 3D radius/wall thickness (p = 0.55). Significant treatment effects were identified on LV EF (p = 0.006), whereas no significant treatment effects were found on LV ESV/BSA (p = 0.214) or LV systolic longitudinal strain rate (p = 0.16). The slope of LV EF per month for the Toprol group was estimated to be 0.04 (95% confidence interval [CI]: −0.08 to 0.16, not significantly different from 0), whereas the slope of the placebo group was estimated to be −0.21 (95% CI: −0.33 to −0.08, significantly different from 0, p = 0.001). This implies that the LV EF for the placebo group after 2 years is expected to decrease by as little as 1.92% and as much as 7.92%, on average, with 95% confidence. Significant treatment effects were detected for LV diastolic function measured by the peak early filling rate from the MRI volume time curves (p = 0.001). The slope for the Toprol group was estimated to be 0.008 (95% CI: −0.002 to 0.017), whereas the slope for the placebo group was estimated to be −0.015 (95% CI: −0.02 to −0.005). In addition, there was a treatment effect for heart rate (p = 0.006) and systolic blood pressure (p = 0.03) in the placebo group versus the Toprol-treated MR patients (6).

Sensitivity analyses were performed to account for a possible effect in the statistical analyses of participants who had undergone mitral valve surgery. These patients are identified (black lines) in (Figure 29_gr1). Surgery is performed because it is expected that the patient's condition will improve. Therefore, it is logical to assume that the data observed after surgery of the patients who underwent mitral valve repair or replacement is better than what it would have been if they did not undergo surgery; hence, the data after surgery were replaced with the worst value for the visit within the treatment group. The conclusions were the same.

Although generally there was a pattern of a higher adverse event or serious adverse event frequency in the placebo group, by design the study was not powered to detect modest differences between groups. Twelve of the 19 patients randomized to placebo and 8 of the 19 patients randomized to Toprol experienced at least 1 adverse event. Seven of the 12 in the placebo group and 3 of the 8 in the Toprol group experienced serious adverse events. A formal test of the hypothesis using the Fisher exact test did not show this difference to be statistically significant (p = 0.33 for adverse event and p = 0.27 for serious adverse event). Six of the 19 patients in the placebo group and only 2 of the 19 in the Toprol group had undergone mitral valve repair or replacement. The Fisher exact test also did not show this difference to be statistically significant (p = 0.23).

In this randomized placebo-controlled study, chronic β₁-AR blockade prevented the progressive decline of LV EF, whereas LV EF slope decreased in the placebo group. At randomization, all patients were within standard echocardiographic guidelines, with an LV end-systolic dimension of less than 40 mm and an LV EF of more than 55% in the absence of symptoms. The beneficial effects of β₁-AR blockade persist on an intention-to-treat basis and with a sensitivity analysis. There is no difference in the rate of progression for systolic wall stress calculated using blood pressure and 3D radius/wall thickness at the base, mid, or distal LV (data not shown). Thus, the tendency for an increase in LV ESV in the placebo group resulting in a decrease in the slope of LV EF may represent a decrease in contractility. In parallel with the effects on LV systolic function, early LV diastolic filling rate demonstrates a decreasing slope in placebo and an increasing slope in treated patients.

No treatment effects were detected for LV remodeling. The discordance between LV remodeling and improved LV systolic function with β₁-AR also has been reported in the canine model of isolated MR (8). The dog model of isolated MR is marked by loss of extracellular matrix components essential to cardiac geometry (16), decrease in protein synthesis (18), and decrease in profibrotic growth factors, including transforming growth factor-β (16). Thus, extracellular matrix loss combined with a less robust hypertrophy response produces LV wall thinning and a decrease in LV EDV mass/volume ratio (16), as in our MR study patients (6). These myocardial responses are a poor match for antifibrotic and antihypertrophic effects of renin-angiotensin system blockade, which explains why this therapy or vasodilators do not attenuate LV remodeling in isolated MR (2). Although β₁-AR blockade improves LV and cardiomyocyte function in the MR dog, interstitial collagen loss is unchanged, which may explain the failure to attenuate LV dilatation (9).

β₁-AR blockade exhibits a trend toward preventing the need for operative intervention; however, the current study is not powered adequately to evaluate this outcome. Of interest, (Figure 29_gr1) demonstrates baseline MRI-derived LV EF of less than 55% in 3 patients who received β₁-AR blockade. The discrepancy between echo-derived LV end-systolic dimension and fractional shortening and MRI volume-based LV EF likely resides in the mid and apical spherical remodeling distal to standard echo-derived LV end-systolic dimension at the tips of the papillary muscles (19). (6) and (6) demonstrate the greater amount of LV apical spherical remodeling, which contributes to increased LV ESV. Nevertheless, LV EF slope was positive in the treated group, further supporting the beneficial effect of β₁-AR blockade in isolated MR.

We do encourage caution that spurious findings may have arisen from the multiple statistical tests conducted for the 7 outcomes considered in this report. Because this was a pilot study, it was not clear at the inception what adjustments were appropriate to protect from the possibility of these spurious findings. However, even the most conservative approach, a Bonferroni adjustment with alpha of 0.0071 (0.05/7), subsequently was shown not to affect the interpretation of the findings, because all factors significant at 0.05 also were significant at this strict level. In addition, as noted in the statistical methods, multiple correlation structures may be used to analyze the data. The choice of an appropriate structure was based on a priori and objective criteria of goodness-of-fit indices for the model using that particular working correlation and were not based on treatment differences in the outcome measures. Thus, it is worth noting that using autoregressive structures considered or using actual times when MRI was undertaken resulted in no significant differences in any of the outcomes.

The current study uses cardiac remodeling as a surrogate outcome to assess the potential beneficial impact of β₁-AR blockade in chronic isolated MR. Although we are convinced that the surrogate outcome of LV EF will be related strongly with important clinical outcomes, including prevention of heart failure and prolonging the need for surgical intervention, the current study does not definitively establish that clinical outcomes will improve, an association that can be assessed only in a Phase III randomized clinical. Nevertheless, this study using LV functional outcome, in addition to other reports of a survival benefit in patients with isolated MR (10), provides empiric support for the use of β₁-AR blockade in patients with chronic degenerative MR. These findings call for a large multicenter clinical trial to confirm these effects.