Expression of different cytokines and growth factors after myocardial injury has been associated with fibroplasia and dilatation versus reverse remodeling and myocardial repair. Specifically, the proinflammatory/fibrotic mediators: interleukin (IL)-6, epidermal growth factor, and fibroblast growth factor (FGF)-2 cause fibroplasia, whereas reparative cytokines including: IL-1α, IL-1β, IL-4, and IL-13 can limit fibrosis. In appropriate patients, cardiac resynchronization therapy (CRT) reverses cardiomyopathy and improves outcome. However, a significant proportion will not respond to this therapy. We conducted this study to assess the association of proinflammatory/fibrotic and/or reparative immune response mediators at baseline with outcome after CRT. In the multicenter RISK study, plasma samples were collected prospectively before CRT implantation. Plasma IL-6, epidermal growth factor, FGF-2, IL-1α, IL-1β, IL-4, and IL-13 were evaluated by Luminex technology. The primary outcome was predefined as freedom from heart failure hospitalization or death and a decrease in echocardiographic end-systolic volume of >15% at 12 months. To determine associations with the outcome, multivariate logistic regression models including baseline clinical characteristics and the specific cytokines and growth factors were constructed. On multivariate analysis of 257 patients, detectable reparative cytokine IL-13 was significantly associated with the primary outcome (odds ratio 3.79; 95% CI 2.10 to 6.82, p <0.0001). In contrast, detectable proinflammatory/fibrotic growth factor FGF-2 was negatively associated (odds ratio 0.31; 95% CI, 0.14 to 0.68; p = 0.004). In conclusion, in CRT recipients, baseline levels of inflammatory mediators affecting cardiac fibrosis versus repair were associated with subsequent clinical outcome.
Plasma biomarkers may provide insight into future clinical response to cardiac resynchronization therapy (CRT). Myocardial injury associated with inflammation results in fibroplasia and irreversible remodeling of the myocardium. Several cytokine and growth factor pathways act in a coordinated manner to drive a profibrotic cardiac environment. The cardiac inflammatory milieu may be detected in the peripheral blood plasma compartment. We hypothesized that a proinflammatory immune response to cardiac injury by interleukin (IL)-6, epidermal growth factor (EGF), and fibroblast growth factor (FGF)-2 would counter the salutary effect of CRT. We also considered the reparative immune responses to cardiac injuries by IL-1α, IL-1β, IL-4, and IL-13 would allow CRT to effect reverse remodeling. In this study, we evaluated the baseline plasma concentrations of specific previously identified mediators in CRT recipients and their association with subsequent remodeling and clinical outcome.
Methods
We conducted a multicenter study (RISK) to assess the potential role of biomarkers in risk stratifying patients receiving CRT. The method and primary results of the study have been previously reported. Patients with standard clinical indications for a transvenous CRT system at the time were recruited. After giving written informed consent, patients were enrolled and followed for 12 months after implant. Patients with both ischemic and nonischemic cardiomyopathy were included. To qualify for CRT-D, patients met standard indications during the timeframe of the study to include heart failure (HF), ejection fraction ≤35%, and QRS duration ≥120 msecs and stable, optimal medical therapy. For the purpose of the study, stable therapy for both β blockers and angiotensin converting enzyme inhibitor/angiotensin receptor blocker was considered to be: dose increases no more than 100% greater than initial dose at 30 days before study enrollment or decreases in dosing <50% of initial dose at 30 days before study enrollment with no changes in the 30 days preceding enrollment. Patients were excluded if they refused or withdrew consent, did not receive a transvenous CRT system, and had recent (<30 days) acute ischemic syndrome or acute decompensation of HF. At baseline visit, patients underwent a history and physical examination, responded to a standard HF quality of life questionnaire (Minnesota Living with Heart Failure Questionnaire) and a structured questionnaire for determination of New York Heart Association (NHYA) functional class, underwent a standard echocardiogram, had a blood sample drawn for proteomic analysis and underwent a 6-minute hall walk test. At 6 and 12 months, the same assessment including echocardiography was repeated. Echocardiogram acquisition at participating sites was standardized by protocol. All echocardiograms were analyzed in a blinded manner at the central core laboratory (University of Pittsburgh, Pennsylvania). Left ventricular end-diastolic and end-systolic volumes were quantified by biplane Simpson’s rule. HF hospitalization (HFH) satisfied both following criteria: (1) admission to hospital for >24 hours with one of the following HF worsening symptoms: increased HF class, orthopnea, paroxysmal nocturnal dyspnea, edema, dyspnea on exertion, or gastrointestinal symptoms attributable to HF and (2) receipt of any of the following for HF within 24 hours of admission: intravenous diuresis or intravenous inotropic medications. All admissions were blindly adjudicated by 2 members of the steering committee. An a priori determination was made to assess the primary outcome as survival rate with freedom from HFH and end-systolic volume decrease by at least 15% at 12 months compared with baseline. Institutional review boards of participating centers and central core labs approved this study.
Peripheral venous blood samples from participating patients were collected at baseline before CRT implantation. Plasma sample collection and initial processing techniques for all centers were standardized by training of site personnel and providing them written instructions for sample preparation, handling, and shipping to the central laboratory at University of California Los Angeles. Blood samples (10 ml) were drawn by venipuncture into vacutainer tubes containing EDTA and immediately placed on ice. Samples were then centrifuged at 3,400 rpm (1383 g) for 30 minutes at 4°C using a table top centrifuge. Plasma samples were centrifuged as soon as possible after drawing and no later than 30 minutes after blood drawing. Plasma was then drawn off using a disposable pipette without disturbing the buffy coat. Plasma was then immediately frozen to −20°C until shipped on ice to the core laboratory. At the core laboratory, the plasma was kept frozen at −80°C until thawed for cytokine analysis.
Human plasma protein concentrations of IL-6, EGF, FGF-2, IL-1α, IL-1β, IL-4, and IL-13 were quantitated using Luminex bead assays (Millipore, Billerica, MA) according to the manufacturer’s protocol. Human IL-6, EGF, FGF-2, IL-1α, IL-1β, IL-4, and IL-13 had the lowest detectable limit of 0.2, 2.8, 7.6, 9.4, 0.06, 0.42, and 0.18 pg/ml, respectively.
Descriptive statistics of the baseline characteristics are presented as mean ± SD. Comparisons between CRT responders and nonresponders for categorical data were compared with a chi-square test and continuous data were compared using independent samples t -test. Univariate logistic regression models for the predetermined objective clinical outcome of HFH free survival with decrease in end-systolic volume of at least 15% at 12 months were first performed for baseline cytokines and growth factors (IL-6, EGF, FGF-2, IL-1α, IL-1β, IL-4, and IL-13) with a binary input (detectable vs nondetectable) to evaluate the significance of association. Based on the exploratory nature of this predefined study, the final multivariate logistic regression model was conducted using a stepwise selection method including cytokines and growth factors found to approach significant association with the CRT response from the univariate models (p <0.15) and predetermined plausible demographic and clinical variables associated with clinical outcomes from CRT therapy (age, gender, left ventricular ejection fraction, QRS, ischemic origin, and NYHA class). Covariates were selected in a stepwise manner with significance level of 0.15. Covariates were entered into and removed from the model in such a way that each forward selection step can be followed by one or more backward elimination steps. The stepwise selection process terminated if no further covariates could be added to the model or if the current model was identical to a previously visited model. In addition, we constructed a multivariate model for the candidate biomarkers adjusting for potential confounders identified in the study cohort and the previously specified clinical variables in the first model with and without the same stepwise selection method. Subgroup analyses were carried out using the logistic regression models including candidate biomarkers, subgroup variables (gender, ischemic origin, and EF threshold of 20%) and their interactions. Survival curves were assessed for the candidate biomarkers with significant association on multivariate analysis.
Results
Between November 2005 and May 2011, 376 patients from 33 centers ( Appendix 1 ) were enrolled in the study. Of those, 257 had blood samples drawn for the baseline proteomic analysis and completed the required 1 year follow-up. At 12 months, 123 patients (48%) met the primary outcome of survival free from HFH and end-systolic volume decrease by at least 15%.
Table 1 describes the baseline characteristics of our cohort. In patients who met the primary outcome (group I), there was a significantly higher proportion of women, a significantly longer QRS duration, and they were less likely to have history of paroxysmal atrial fibrillation, myocardial infarction, or diabetes, compared with others (group II). Table 2 describes clinical and echocardiographic outcomes in both groups. By definition, there were no deaths or HFH in group I. There was significant improvement in echocardiographic parameters shown at 6- and 12- month follow-up indicating early and sustained favorable reverse remodeling.
| Baseline Variable | All Enrolled Patients (N=257) | Group I (N=123) | Group II (N=134) | p-value |
|---|---|---|---|---|
| Age (years) | 66 ± 11 | 66 ± 11 | 66 ± 11 | 0.7113 |
| Female | 54 (21.0%) | 36 (29.3%) | 18 (13.4%) | |
| Male | 203 (79.0%) | 87 (70.7%) | 116 (86.6%) | |
| Ejection Fraction | 25 ± 8 | 26 ± 8 | 25 ± 8 | 0.2940 |
| LV End Systolic Volume | 132 ± 63 | 130 ± 59 | 134 ± 67 | 0.6701 |
| QRS (msec) | 154 ± 27 | 160 ± 27 | 149 ± 27 | 0.0022 |
| NYHA Class | 0.5374 | |||
| I | 6 (2.3%) | 3 (2.5%) | 3 (2.2%) | |
| II | 36 (14.0%) | 19 (15.6%) | 17 (12.7%) | |
| III | 193 (75.1%) | 90 (73.8%) | 103 (76.9%) | |
| IV | 8 (3.1%) | 2 (1.6%) | 6 (4.5%) | |
| Ischemic Cardiomyopathy | 137 (53.3%) | 59 (48.0%) | 78 (58.2%) | 0.1002 |
| Myocardial Infarction | 108 (42.0%) | 41 (33.3%) | 67 (50.0%) | 0.0068 |
| Coronary Artery Bypass Surgery | 96 (37.4%) | 42 (34.1%) | 54 (40.3%) | 0.3085 |
| Paroxysmal Atrial Fibrillation | 55 (21.4%) | 20 (16.3%) | 35 (26.1%) | 0.0542 |
| Diabetes | 41 (16.0%) | 13 (10.6%) | 28 (20.9%) | 0.0239 |
| Hypertension | 180 (70.0%) | 87 (70.7%) | 93 (69.4%) | 0.8163 |
| Minnesota Living With Heart Failure Questionnaire | 51 ± 26 | 51 ± 26 | 50 ± 26 | 0.7657 |
| 6-Minute Hall Walk Test (Feet) | 938 ± 348 | 931 ± 352 | 944 ± 344 | 0.7591 |
| Pharmacologic Therapy | ||||
| Beta Blockers | 224 (87.2%) | 106 (86.2%) | 118 (88.1%) | 0.6525 |
| ACE-I/ARB | 209 (81.3%) | 104 (84.6%) | 105 (78.4%) | 0.2031 |
| Aldosterone Inhibitors | 22 (8.6%) | 13 (10.6%) | 9 (6.7%) | 0.2701 |
| Diuretics | 104 (40.5%) | 45 (36.6%) | 59 (44.0%) | 0.2245 |
| Detectable Reparative Cytokines | ||||
| Inter-Leukin_1α (≥9.4 pg/ml) | 83 (32.3%) | 47 (38.2%) | 36 (26.9%) | 0.0520 |
| Inter-Leukin_1β (≥ 0.06 pg/ml) | 229 (89.1%) | 117 (95.1%) | 112 (83.6%) | 0.0030 |
| Inter-Leukin_4 (≥ 0.42 pg/ml) | 164 (63.8%) | 88 (71.5%) | 76 (56.7%) | 0.0135 |
| Inter-Leukin_13 (≥ 0.18 pg/ml) | 119 (46.3%) | 75 (61.0%) | 44 (32.8%) | <.0001 |
| Detectable Pro-inflammatory Cytokines | ||||
| Inter-Leukin_6 (≥ 0.2 pg/ml) | 253 (98.4%) | 121 (98.4%) | 132 (98.5%) | 0.9312 |
| Epidermal Growth Factor (≥ 2.8 pg/ml) | 167 (65.0%) | 81 (65.9%) | 86 (64.2%) | 0.7786 |
| Fibroblast Growth Factor_2 (≥7.6 pg/ml) | 214 (83.3%) | 98 (79.7%) | 116 (86.6%) | 0.1392 |
| Outcome Variable | Group I (N=123) | Group II (N=134) | p-value |
|---|---|---|---|
| Clinical Variables at 12 months: | |||
| Death | 0 (0%) | 17 (12.7%) | <0.0001 |
| Heart Failure Hospitalization | 0 (0%) | 24(17.9%) | <0.0001 |
| Minnesota Living With Heart Failure Questionnaire (Improvement) | 24 ± 24 | 19 ± 24 | 0.1123 |
| 6 Minute Hall Walk Test (Improvement, feet) | 62 ± 326 | 117 ± 284 | 0.2038 |
| Incident AF | 15 (12.2%) | 23(17.2%) | 0.29 |
| Echocardiogram: Baseline | |||
| Ejection Fraction (%) | 29 ± 8 | 29 ± 8 | 0.6601 |
| Left Ventricular End Diastolic Volume (mm 3 ) | 180 ± 67 | 185 ± 77 | 0.5816 |
| Left Ventricular End Systolic Volume (mm 3 ) | 130 ± 59 | 134 ± 67 | 0.6701 |
| Mitral Regurgitation : moderate or greater | 19 (37.3%) | 25 (38.5%) | 0.8942 |
| Left Atrium Dimension (mm) | 43 ± 10 | 44± 8 | 0.3660 |
| Echocardiogram: 6 Months | |||
| Ejection Fraction (%) | 40 ± 11 | 31 ± 9 | <0.0001 |
| Left Ventricular End Diastolic Volume (mm 3 ) | 146 ± 61 | 189 ± 71 | <0.0001 |
| Left Ventricular End Systolic Volume | 90± 50 | 132 ± 61 | <0.0001 |
| Mitral Regurgitation: moderate or greater | 10(19.6%) | 15 (27.3%) | 0.3530 |
| Left Atrial Dimension (mm) | 41 ± 7 | 44 ± 7 | 0.0022 |
| Echocardiogram: 12 Month | |||
| Ejection Fraction (%) | 44 ± 11 | 30 ± 9 | <0.0001 |
| Left Ventricular End Diastolic Volume (mm 3 ) | 134 ± 57 | 192 ± 69 | <0.0001 |
| Left Ventricular End Systolic Volume (mm3) | 78 ± 47 | 138 ± 60 | <0.0001 |
| Mitral Regurgitation: moderate or greater | 8 (19.5%) | 16 (33.3%) | 0.1431 |
| Left Atrial Dimension (mm) | 41 ± 7 | 44 ± 9 | 0.0011 |
Preimplantation baseline plasma concentrations of the anti-inflammatory cytokine IL-13 were detectable in 119 (46.3%) patients ( Table 1 ). Importantly, there were baseline detectable concentrations of IL-13 in 75 of patients in group I (61.0%), but only 44 of 134 in group II (32.8%, Table 1 ). In a univariate logistic regression model, detectable IL-13 was associated with clinical outcome (odds ratio [OR] 3.20, 95% CI 1.92 to 5.33, p <0.0001, Figure 1 ). Plasma concentrations of IL-1α, IL-1β, and IL-4 were detectable in 83 (32.3%), 229 (89.1%), and 164 (63.8%) patients ( Table 1 ). Assessed by outcome group, there were baseline detectable concentrations of IL-1α, IL-1β, and IL-4 in 47 (38.2%), 117 (95.1%), and 88 (71.5%), respectively, in group I. Baseline detectable concentrations of IL-1α, IL-1β, and IL-4 were also found in 36 (26.9%), 112 (83.6%), and 76 (56.7%), respectively in group II ( Table 1 ). Using univariate logistic regression, detectable concentrations of IL-1α, IL-1β, and IL-4 were associated with favorable outcome (OR 1.68, 95% CI 0.99 to 2.85, p = 0.05; OR 3.83, 95% CI 1.50 to 9.80, p = 0.005; and OR 1.92, 95% CI 1.14 to 3.23, p = 0.014, respectively, Figure 1 ).
Baseline plasma concentrations of FGF-2 were detectable in 214 (83.3%) patients. Concentrations of FGF-2 were detectable in 98 (79.7%) of group I and 116 (86.6%) of group II patients ( Table 1 ). In univariate logistic regression model, the association between undetectable FGF-2 and outcome was not statistically significant (OR 0.61, 95% CI 0.31 to 1.18, p = 0.142, Figure 1 ). Baseline plasma concentrations of IL-6 and EGF were detectable in 253 (98.4%) and 167 (65.0%) patients, respectively ( Table 1 ). There were baseline detectable concentrations of IL-6 and EGF in 121 (98.4%) and 81 (65.9%) of group I ( Table 1 ). However, baseline detectable concentrations of IL-6 and EGF were also found in 132 (98.5%) and 86 (64.2%), respectively of Group II ( Table 1 ). In univariate logistic regression models, detectable IL-6 and EGF concentrations were not associated with a response to CRT.
Based on the previously mentioned results, a multivariate logistic model with stepwise selection method was carried out for the binary clinical outcome with plausible predictive covariates including age, gender, ischemic origin, NYHA class, left ventricular ejection fraction, and QRS duration in addition to the cytokines and growth factors approaching significant association with clinical outcome on univariate analysis (IL-1α, IL-1β, IL-4, IL-13, and FGF-2). We found that patients with detectable concentrations IL-13 had a significantly higher likelihood of meeting the predetermined clinical outcome (OR 3.79, 95% CI 2.10 to 6.82, p <0.0001, Figure 1 ). Conversely, patients with detectable concentration of FGF-2 were less likely to meet the primary outcome measure (OR 0.31, 95% CI 0.14 to 0.68, p = 0.004, Figure 1 ). We also found associations with outcome for QRS duration, per 1 msec increase, (OR 1.02, 95% CI, 1.01 to 1.03, p = 0.002) and female gender (OR 2.21, 95% CI 1.11 to 4.41, p = 0.024).
We constructed a separate multivariate analysis model taking into account potential confounders specific to this cohort. As listed in Table 1 , there were significant differences between groups with regard to QRS duration, female gender, previous myocardial infarction, paroxysmal atrial fibrillation, or diabetes. With respect to these potential confounders, patients with detectable IL-13 were less likely to have had previous myocardial infarction or diabetes (myocardial infarction: 66 [47.8%] vs 42 [35.3%], p = 0.0424; diabetes: 28 (20.3%) vs 13 [10.9%] p = 0.041), whereas none were different between those with or without detectable FGF-2. Using a stepwise selection multivariate analysis model with p <0.15 as entry and exit significance level and including the potential confounders identified in the cohort, the previously mentioned prespecified clinical variables and the candidate biomarkers identified on univariate analysis, IL-13 and FGF-2 remained significant covariates (IL-13: OR 3.640, CI 1.995 to 6.644, p <0.0001 and FGF-2: OR 0.330, CI 0.149 to 0.731, p = 0.0063). The other significant covariates in the model were QRS duration and gender (QRS per 1 msec increment: OR 1.018, CI 1.007 to 1.030, p = 0.0019 and gender: OR 2.526, CI 1.232 to 5.177, p = 0.0114).
In addition, we studied the association of IL-13 and FGF-2 with the primary outcome in prespecified subgroups divided according to gender, ischemic origin, and EF threshold of 20%. Detectable IL-13 remained significantly associated with outcome for all subgroups except in females where there was only a trend toward significant association ( Figure 2 ). In contrast, the association between undetectable FGF-2 and outcome was significant only among the ischemic subgroup ( Figure 2 ). There was no statistically significant interaction between detectable concentration of IL-13 and gender (p = 0.79), ischemic origin (p = 0.16), or EF > or ≤20% (p = 0.24) for the primary outcome measure. For undetectable FGF-2, there was statistically significant interaction with ischemic cardiomyopathy (p = 0.04) but not with gender (p = 0.93) or EF threshold of 20% (p = 0.50).
