Novel Predictors of Left Ventricular Reverse Remodeling in Individuals With Recent-Onset Dilated Cardiomyopathy

Study population

Consecutive DCM patients with a history of symptoms shorter than 6 months were enrolled in this prospective study. They were admitted to a tertiary hospital between November 2008 and November 2010. DCM was defined by the presence of left ventricular dilation (LVEDD >33 mm/m in men and >32 mm/m in women) and LVEF <45% in the absence of coronary artery disease, severe systemic arterial hypertension and primary valve disease (1,7). Individuals with a history of drug abuse or excessive alcohol consumption and/or presenting with persistent supraventricular tachyarrhythmias were not included. The investigation was approved by the local ethics committee. All subjects gave their written informed consent prior to their participation in the study.

Study protocol

At baseline, all subjects underwent a clinical assessment, electrocardiography, echocardiography, cardiopulmonary exercise testing, CMR, and EMB. Peripheral venous blood was obtained in the morning before EMB for the measurement of hs-cTNT, BNP, and galectin-3 and for routine biochemical analysis. Pharmacotherapy of heart failure was optimized according to the latest guidelines (8). Clinical evaluation, electrocardiography, echocardiography, cardiopulmonary exercise testing, routine biochemical analysis, and measurement of BNP were repeated at 3, 6, and 12 months of follow-up. The study was completed by CMR at 12 months of follow-up. Individuals admitted with decompensated heart failure were treated to be euvolemic before any study visit.

LVRR was defined as an absolute increase in LVEF from ≥10% to a final value of >35%, accompanied by a decrease in LVEDD ≥10% as assessed by echocardiography at 12 months. Doses of angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and beta-blockers were expressed as a percentage of the maximum recommended daily dose according to the latest European Society of Cardiology guidelines (8).

Echocardiography and cardiopulmonary exercise testing

Echocardiography was performed by experienced operators in accordance with guidelines of the American Society of Echocardiography (7,9). M-mode, 2-dimensional images, and Doppler recordings were obtained using a Vivid 7 (GE Healthcare, Chalfont St Giles, United Kingdom). LVEF was assessed using Simpson's biplane method. Mitral regurgitation was graded semiquantitatively on a scale of none, trivial, mild, moderate, and severe. Mitral inflow pattern was classified as restrictive in the presence of an E-wave deceleration time <120 ms or a ratio of early transmitral flow velocity to atrial flow velocity ≥2 associated with an E-wave deceleration time ≤150 ms (10). Cardiopulmonary exercise testing was performed using a symptom-limited bicycle ergometry and 25-W increases in workload every 3 min. Minute ventilation, oxygen consumption, and carbon dioxide output were measured by a heated pneumotachograph and mass spectrometry (Sensormedics system, Viasys Healthcare, Conshohocken, Pennsylvania), as reported previously (11).

Cardiovascular magnetic resonance

The CMR studies were performed on a clinical 1.5-T MR scanner (Avanto, Siemens Medical Solutions, Erlangen, Germany). Electrocardiogram-gated steady-state free precision imaging of the left and right ventricles was undertaken (repetition time [TR]/echo time [TE] 65/1.2 ms; flip angle 70°) with a slice thickness of 8 mm in contiguous short-axis and orthogonal long-axis planes. A breath-hold, T2-weighted dark blood sequence (TR/TE = 2 RR/58 ms, TI 140 ms, slice thickness of 8 mm, interslice gap of 2 mm) was acquired in short-axis slices. Images for assessment of early enhancement were acquired in basal, mid-papillary, and apical short-axis slices using a T1-weighted turboFLASH sequence (TR/TE 170/1.05 ms, flip angle 12°, TI 100 ms, slice thickness of 10 mm) before and at 40 to 70 s after intravenous bolus injection of 0.2 mmol/kg of gadobutrol (Gadovist, Bayer Schering, Berlin, Germany). Late gadolinium enhancement (LGE) images were acquired 10 to 15 min after the administration of the contrast in short-axis and orthogonal long-axis planes using phase-sensitive inversion-recovery sequences (TR/TE 690–850/3.2 ms, adjusted TI, slice thickness of 8 mm, interslice gap of 0.8 mm, in-plane resolution of 1.7 × 1.7 mm).

Ventricular volumes, mass, and ejection fractions were measured from cine images using Segment version 1.8 (12). The presence and distribution of LGE were independently determined by 2 expert radiologists (D.K. and J.W.) in a blinded fashion. The agreement between the 2 radiologists was as follows: kappa = 0.75, 95% confidence interval: 0.53 to 0.97, p = 0.001. The extent of LGE, defined by a signal intensity >2 SDs above the mean of the remote reference myocardium (13), was quantified semiautomatically on the short-axis contrast images using Matlab (Mathworks, Natick, Massachusetts) (Fig. 1). The LGE extent was expressed as a weight of tissue and as a percentage of LV mass (indexed LGE extent). Quantitative assessment of the myocardial edema ratio (T2 index) and the early enhancement were performed as described previously (4,14). The myocardial edema ratio of ≥1.9 and the early enhancement of ≥45% were regarded as abnormal (4,14). The extent of LGE, the myocardial edema ratio, and the early enhancement were analyzed by a single observer (M.S.) blinded to the results of EMB. The corresponding intraobserver relative variability was 5.2 ± 4.5%, 6.9 ± 6.8%, and 10.1 ± 7.3%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Quantification of LGE in an Individual With LVRR at Baseline and at 12 Months Follow-Up

The figure illustrates quantification of late gadolinium enhancement (LGE) in 1 individual with left ventricular reverse remodeling (LVRR) at baseline (A and B) and 12 months of follow-up (C and D). A and C show corresponding native images of LGE on a short-axis slice. B and D demonstrate the semiautomatic quantification of LGE. The yellow areas indicate a signal intensity >2 standard deviations above the mean of the remote reference myocardium.

Laboratory methods

Plasma levels of BNP were measured immediately using a chemiluminescent immunoanalysis (Architect BNP, Abbott Diagnostics, Abbott Park, Illinois). The lower limit of detection was 10 ng/l, intra-assay coefficient of variation (CV) <3.8%, and interassay CV <5.3%. Serum was frozen at −30°C and preserved until the hs-cTNT and galectin-3 assays. Hs-cTnT was detected using an electrochemiluminescent immunoassay (T-hs-STAT, Cobas e411, Roche Diagnostics, Mannheim, Germany). The lower limit of detection was 5 ng/l, CV at 13 ng/l was 10%, intra-assay CV <3.2%, and interassay CV <6.2%. The upper reference limit was set at 13.5 ng/l. Galectin-3 was measured using an enzyme-linked immunosorbent assay (BioVendor, Brno, Czech Republic). The lower limit of detection was 0.12 μg/l, intra-assay CV <6.4%, interassay CV <11.4%, and expected serum levels 0.62 to 6.25 μg/l. Glomerular filtration rate was estimated using the Modification of Diet in Renal Disease formula, as reported previously (15).

Endomyocardial biopsy and processing of samples

Before EMB, each patient underwent coronary angiography to exclude coronary artery disease. EMB was performed via the internal jugular vein using a flexible bioptome (diameter 7-F, Cordis Corporation, Bridgewater, New Jersey) under fluoroscopic guidance. Eight to 10 samples were obtained from the right ventricular side of the interventricular septum. There were no complications related to the EMB procedure. The processing and analysis of EMB specimens were described previously (16). In brief, histopathological diagnosis of myocarditis followed the Dallas criteria (17). Immunohistochemical criteria of myocardial inflammation were based on detection of mononuclear infiltrates consisting of >7 CD3-positive T lymphocytes/mm² or >14 CD3-positive T lymphocytes or CD68-positive macrophages per 1 mm² (18). The extent of fibrosis was assessed visually using an ocular reticule. Quantitative polymerase chain reactions were performed to detect human cytomegalovirus, Epstein-Barr virus, human herpesvirus 6, parvovirus B19, adenoviruses, and enteroviruses (comprising coxsackieviruses and echoviruses) in EMB samples as described previously (16).

Statistical analysis

Categorical data were expressed as percentages and compared using a chi-square test or Fisher exact test. Normally distributed continuous variables were expressed as the mean and standard deviation. Abnormally distributed continuous variables were given as median and interquartile range. Continuous variables were compared using the Student t test or by the nonparametric Mann-Whitney U test, where appropriate. Variables that significantly differed between groups with and without LVRR were entered in a stepwise logistic regression analysis to identify independent predictors of LVRR at baseline, 3 months, and 6 months of follow-up. To address the risk of overfitting in multivariate logistic regression models, univariate logistic regression was performed for all variables included in these models (Table 1), and new multivariate models were created using 2 predictors of LVRR with the best F statistic at baseline and at 3 and 6 months of follow-up. This approach identified the same variables and resulted in the same models as stepwise logistic regression, which confirmed the reliability of the aforementioned models. Receiver-operating characteristics (ROC) analysis was performed to assess the performance of selected variables to predict LVRR. Areas under the ROC curve were calculated also for combinations of independent predictors of LVRR using logistic regression. For all tests, a probability value of p < 0.05 was considered significant. All analyzes were performed using the statistical software SPSS (Chicago, Illinois) version 17.0.

Follow-up data

A total of 39 patients (89%) completed the 12-month follow-up, whereas 4 individuals (9%) required implantation of a ventricular assist device during the first 3 months, and another patient (2%) underwent urgent heart transplantation at 4 months. In addition, 6 patients (14%) received a single-chamber implantable cardioverter-defibrillator within the primary preventive indication. At 12 months, LVRR was observed in 20 individuals (45%). However, only 3 patients (7%) reached a LVEF higher than 50%.

Prediction of LVRR from baseline data

Nine of 70 baseline variables showed some predictive value for the subsequent LVRR (Table 2) in the univariate analysis. Compared with the remaining individuals, LVRR was heralded by lower serum levels of hs-cTNT, higher plasma levels of sodium, and surprisingly, by a worse renal function. Other predictive parameters included a higher prevalence of myocardial inflammation in biopsy and 2 CMR variables, namely a smaller extent of LGE and a higher myocardial edema ratio derived from T2 imaging. Neither the presence of viral genomes nor the extent of fibrosis in EMB samples was related to LVRR. A multivariate analysis identified 2 independent predictors of LVRR at baseline: a lower value of the indexed LGE extent and a higher value of the myocardial edema ratio (Table 1). A simultaneous positivity of both variables predicted LVRR with a sensitivity of 70% and a specificity of 78% (Table 3). The higher prevalence of myocarditis (50% vs. 21%, p < 0.05) and elevated values of the myocardial edema ratio on CMR in individuals with LVRR suggest that resolving myocarditis may contribute to the process of LVRR.

Prediction of LVRR from the follow-up data

Fifteen of 60 follow-up variables predicted LVRR in the univariate analysis. These predictors included a better exercise capacity, a lower LVEDD, a higher LVEF, less severe mitral regurgitation, a smaller left atrial volume, and lower values of noninvasive indicators of the LV filling pressure (E/E′ ratio, presence of restrictive mitral pattern, BNP plasma level) (Table 4,Fig. 2). Importantly, there were no differences in the use of heart failure medication in both groups. At 3 months, the latest plasma level of BNP was the most powerful predictor of LVRR (Table 1). Specifically, BNP level <344 ng/l predicted LVRR with a sensitivity of 95% and a specificity of 50% (Table 3). The conventional methods (LVEDD index and E/E′ ratio) became the strongest predictors of LVRR as late as after 6 months of follow-up. Table 5 illustrates improvement in areas under the ROC curve for combinations of follow-up variables and even for combinations of baseline CMR variables with the follow-up data. Surprisingly, both the baseline indexed LGE extent and the myocardial edema ratio improved the predictive performance of BNP plasma levels at 3 months and the echocardiographic variables at 6 months (Table 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Longitudinal Changes in Median BNP levels and LVEDD Index in Individuals With and Without LVRR

The light blue columns illustrate significantly lower plasma levels of B-type natriuretic peptide (BNP) in the group with left ventricular reverse remodeling (LVRR) compared with the remaining individuals (dark blue columns) at 3, 6, and 12 months of follow-up. LVEDD = left ventricular end-diastolic dimension.

CMR findings at 12 months of follow-up

A follow-up CMR was available in 30 individuals. At baseline, LGE was present in 20 (67%) of 30 subjects with the follow-up CMR. At 12 months, LGE persisted in 13 individuals (43%) and disappeared in 7 individuals (23%). Of 10 subjects who were LGE negative at baseline, 8 individuals remained LGE negative (27%), and 2 individuals (7%) developed a new LGE lesion. There were no statistically significant differences in the baseline and follow-up distribution of LGE in the entire group or within subgroups defined according to LVRR (data not shown). Importantly, both the LGE extent (5.5 [0 to 11.6] g vs. 1.3 [0 to 6.0] g, p = 0.001) and the myocardial edema ratio (1.53 ± 0.37 vs. 1.29 ± 0.25, p = 0.002) decreased during follow-up. Although the baseline values of the LGE extent and the myocardial edema ratio were significantly different between the aforementioned groups (Table 2), the absolute or relative changes in these 2 variables were similar in both groups (data not shown).

Comparison with previous studies of LVRR

LVRR in individuals with recent-onset DCM was described by Dec et al. in 1985 (19). A larger study by Steimle et al. (20), conducted before the era of beta-blockers, revealed an incidence of LVRR (defined as an increase of LVEF ≥15% at 12 months) of 27%. LVRR was predicted by a shorter symptom duration and less severe hemodynamic decompensation. Interestingly, EMB results, fulfilling the Dallas criteria of myocarditis in 9%, played no role in prediction of LVRR (20). Over time, advances in heart failure pharmacotherapy have increased the incidence of LVRR both in the general heart failure population (21,22) and in patients with recent-onset DCM (23,24). Along these lines, McNamara et al. (23,24) reported data from Intervention in Myocarditis and Acute Cardiomyopathy trials IMAC-1 and IMAC-2, showing the incidence of LVRR (defined as an increase of LVEF ≥10%) of 56% at 12 months and 70% at 6 months, respectively. The prevalence of LVRR in our study (45%) was similar to that in the IMAC-1 study. The epidemiological IMAC-2 trial reported the following predictors of LVRR: a lower New York Heart Association functional class and LVEDD together with a higher systolic blood pressure and non-black race. However, none of these variables had sufficient power for clinical decision making. Importantly, our study revealed novel predictors of LVRR with an acceptable predictive capacity, namely, variables derived from CMR and serial BNP testing. In addition, we evaluated the predictors of LVRR, not only at baseline, but also longitudinally. As a result, we could identify specific cutoff points for each follow-up period and showed that the novel predictors provided an earlier prediction of LVRR than the conventional methods.

The role of CMR in DCM patients

Most studies assessed the prognostic value of CMR, specifically of LGE, in individuals with a long-term established diagnosis of DCM (25,26). There is only 1 CMR study that evaluated individuals with recent-onset DCM and attempted to predict improvement in LV systolic function defined as an increase in LVEF >5% at 5 months of follow-up (27). Leong et al. (27) demonstrated in that study an inverse correlation between the LGE extent and improvement in LVEF. However, their study population was different from our cohort because individuals with suspected myocarditis having either abnormal troponin I levels or evidence of myocardial edema on T2-weighted CMR were excluded. Tachycardia-induced cardiomyopathy, however, was not considered to be an exclusion criterion.

In our study, we excluded individuals with suspected tachycardia-induced cardiomyopathy and alcoholic cardiomyopathy, which are known as highly reversible disease entities. We found no good reason not to recruit troponin I–positive individuals with recent-onset DCM who represented almost one-third of our consecutive patients. We believe that our inclusion criteria, the definition of LVRR, and the structured 1-year follow-up support the applicability of our results in clinical practice.

Therapeutic implications

The role of CMR, specifically the extent of LGE, to predict a response to pharmacotherapy in chronic heart failure patients was reported by Bello et al. (28) already in 2003. This was the first study that opened the door to the personalized medicine in this field. We extended the available evidence by inclusion of T2-weighted CMR and serial BNP testing into prediction of LVRR in the specific setting of recent-onset DCM. Improved identification of DCM patients with subsequent LVRR might be helpful for the cost-effective use of implantable cardioverter-defibrillators in the primary prevention of sudden cardiac death, and it could also be used for optimal timing of transplantation referral in DCM patients. Therefore, we propose to include CMR into the baseline assessment of patients with recent-onset DCM and to use the serial BNP measurement during the follow-up.

Study limitations

First, the small study group size that reflects rather complex study design with the use of EMB and repeated controls may limit general applicability of the study results. Second, although the serial assessment of hs-cTNT could have improved the prediction of LVRR (29), this was not known at conception of the study design. Third, because the accuracy of regression coefficients is driven by the number of events, multivariate models including more than 2 independent variables for prediction of 20 events may be inaccurate (30). To address the risk of overfitting in multivariate logistic regression models, univariate logistic regression was performed for all variables included in these models, and new multivariate models were created using 2 predictors of LVRR with the best F statistic at baseline and at 3 and 6 months of follow-up. This approach identified the same variables and resulted in the same models as stepwise logistic regression, which confirmed the reliability of the aforementioned models. Last, the study results have to be considered specific for the present subject sample and may vary in different populations.