Introduction

Heart failure (HF) is a global health problem which affects approximately over 37 million people world wide1. According to data from the Framingham Heart Study, the lifetime risk of developing HF is estimated to be 20% for the ages between 40 and 80 years2. It is caused by structural, but also functional cardiac abnormalities, which result in loss of myocardial function3. However, in the absence of coronary artery disease, which is the leading cause of HF4,5,6, inflammatory cardiomyopathy is common, particularly in HF with reduced ejection fraction (HFrEF). Thereby, according to the report of the 1995 World Health Organization/International Society and Federation of Cardiology Task Force on the Definition and Classification of Cardiomyopathies, inflammatory cardiomyopathy is defined as cardiac dysfunction in co-prevalence with inflammatory disease of the myocardium established by immunological, histological and immunohistochemical criteria7. The exact incidence of inflammatory cardiomyopathy underlying HF is not known. However, according to postmortem analysis, inflammatory cardiomyopathy or myocarditis seems to account for approximately 40% of the sudden cardiac deaths in the young8. Regarding endomyocardial biopsy (EMB) studies myocardial inflammation can be diagnosed in 9 to 23% of the patients with non-ischemic cardiomyopathy9,10. According to the above mentioned definitions, inflammatory cardiomyopathy can only be diagnosed on basis of EMB results11,12. Although the results of studies addressing the immunosuppressive therapy of inflammatory cardiomyopathy remain controversial13,14,15, results of the TIMIC trial suggest that immunosuppressive therapy in patients with virus negative inflammatory cardiomyopathy may be an effective and safe option for recovery of cardiac function in addition to optimal medical therapy16. Inflammatory cardiomyopathy can progress rapidly and may require immediate immunosuppressive therapy to prevent adverse outcomes. Thus, we aimed to identify predictors of myocardial inflammation in HF patients in order to improve diagnostic and therapeutic management.

Methods

Study population

Patients presenting with symptoms of HF at the Heart Failure Outpatient clinic, the emergency department or chest pain unit of the University Medical Centre Mainz between 14/10/2012 and 18/12/2018, who underwent EMB were enrolled in this retrospective monocentric analysis. Decision to obtain EMB of the patients was based on guidelines published of the AHA, the ACC and the ESC in 200717. Ischemic HF, valvular HF and systemic disease with known cardiac involvement were ruled out prior to EMB. The analysis of the patients’ medical records involved the following data: personal medical history, clinical presentation, laboratory values, echocardiography and the finding of the EMB, including viral activity. The exclusion criteria are presented in Fig. 1 and comprise missing CD3+ and/or Mac1 cell counts or presence of relevant virus activity. All data were obtained from individuals enrolled between 2013 and 2018 in the retrospective monocentric Mainz Endomyocardial Biopsy in Heart Failure Study (My Biopsy-HF Study, DRKS #22178), which was approved by the Ethics Committee of Rhineland Palatinate to be in accordance with the legal regulations and the declaration of Helsinki. Informed consent was obtained from all included individuals to use EMB tissue samples for further scientific purpose.

Figure 1
figure 1

Study flow chart.

Full size image

Echocardiography

Transthoracic echocardiography was performed in our echocardiography lab by trained and certified specialists of our department using iE33 Philips, a General Electrics E9 or Siemens Acuson s2000 machines. All examinations were performed with the patient in the standard left lateral position while apnea or quiet breathing. Routinely 2–4 cardiac cycles were obtained at frame rates of 50–100 fps and digitally transferred into a picture archiving and communication system (Xcelera) for offline analysis. Left ventricular (LV) ejection fraction (LVEF) was assessed by biplane Simpson’s method in the apical four and two chamber view. For the assessment of cardiac structure, LV internal diameters were measured. LVEF was categorized into classes LVEF ≥ 50% (HF with preserved LVEF = HFpEF), LVEF 40–49% (HF with mid-range reduced LVEF = HFmrEF) and LVEF < 40% (HFrEF)3. Global longitudinal strain (GLS) was measured offline utilizing QLab 13.0 (Philips Healthcare, Hamburg, Germany) automatic speckle-tracking software. The tracking result was visually controlled and manually readjusted if necessary. In case of persistent poor tracking quality, datasets were excluded from analysis. GLS was automatically calculated by average of GLS in apical four, three and two chamber view.

Endomyocardial biopsy and myocardial inflammation

EMB was performed using a biopsy forceps (Medwork bioptom, 180 cm, 1.8 mm, Cat.-No. BIO-C4-18-180) by taking up to eight biopsies either from the right ventricular septum or from the lateral wall of the LV. In case of LV biopsy the procedure was carried out via the right or left femoral artery or via trans-radial cardiac catheterization as reported before18. In case of right heart biopsy, the femoral vein was used. Immediately after sampling the biopsy, the specimens were stabilized in a solution to preserve ribonucleic acid (RNA) integrity (RNAlater, Ambion Inc., Austin, Texas) and were sent for further examination to a specialized laboratory approved by Food and Drug Administration [Institut Kardiale Diagnostik und Therapie (IKDT), Berlin, Germany]. For immunohistological evaluation, heart muscle tissue probes fixated in RNAlater were embedded in Tissue Tec (SLEE, Mainz, Germany) and immediately snap‐frozen in methyl butane which had been cooled in liquid nitrogen, and then stored at – 80 °C until processing. Embedded specimens were cut serially into cryosections of 5 mm thickness and placed on 10% poly‐l‐lysine‐pre‐coated slides. As part of a routinely performed work up in case of unclear heart failure histological and immunohistochemical examination as well as virus detection was conducted. Using specific antibodies inflammatory processes were detected by identifying immune cell infiltration and expression of cell adhesion molecules. The specimens of the patients in this study were tested for CD3-positive lymphocytes (Dako; dilution 1:25), CD11a+/LFA-1+ lymphocytes (ImmunoTools; dilution 1:250), macrophages Mac-1 macrophages (ImmunoTools; dilution 1:500), Perforin positive (cytotoxic T Cells) (clone dG9, BD Bioscience; dilution 1:150) and the expression of adhesion molecules HLA-1 (Dako; dilution 1:2000) and ICAM-1 (ImmunoTools; dilution 1:800). Myocardial inflammation was defined as ≥ 7.0 CD3+ lymphocytes/mm2 and/or ≥ 35.0 Mac1 macrophages/mm2 in accordance with previous published studies19. Diagnosis of relevant myocardial virus activity was performed by polymerase chain reaction (PCR) detecting the genomic sequences of viruses that most commonly cause myocarditis. This included: enterovirus, adenovirus, human cytomegalovirus, herpes simplex virus, Epstein–Barr virus, human herpesvirus 6, parvovirus B19 and influenza A and B viruses. Virus load was calculated via quantitative PCR methods. In case of relevant myocardial virus activity patients were excluded from analysis.

Statistical analysis

Descriptive statistics for relevant baseline comparisons of symptomatic non-valvular, non-ischemic heart failure patients, who underwent EMB, stratified according inflammation (≥ 7.0 CD3+ lymphocytes/mm2 and/or ≥ 35.0 Mac1 macrophages/mm2) are provided as median and interquartile range (IQR) or absolute numbers and corresponding percentages. We tested the continuous variables of the groups (inflammation vs. no inflammation) using the Mann–Whitney-U test and categorical variables with the Fisher’s exact or the chi2 test, as appropriate.

Receiver operating characteristics (ROC) curves were calculated for TnI, BNP, CRP and GLS with regard to myocardial inflammation and the area under the curve (AUC) is presented with the corresponding 95% CI. Additionally, patient cohort-optimised cut-off values of these mentioned laboratory markers with regard to myocardial inflammation were calculated based on ROC analyses using Youden index quantification. The software SPSS (version 23.0; SPSS Inc., Chicago, Illinois) was used for computerized analysis. p values of < 0.05 (two-sided) were considered to be statistically significant.

Univariate and multivariate logistic regression models were analyzed to investigate predictors of myocardial inflammation in heart failure patients, thereby using patient-optimised cut-off values derived from the ROC analyses. Results are presented as odds ratio (OR) and 95%CI. The multivariate regression models were adjusted for (1) age and sex, obesity and (2) cardiovascular risk factors (CVRF) including history of smoking, arterial hypertension, diabetes mellitus and hyperlipoproteinaemia.

Results

Overall, 517 patients with symptomatic non-valvular, non-ischemic HF who underwent EMB between 2012 and 2018 were screened. Results from EMB with CD3+ and/or Mac1 cell count were available in 447 cases. 50 cases were excluded due to relevant virus activity (Fig. 1). Thus, 397 patients remained in this study and were analysed. Patient characteristics are displayed in Table 1. In brief, median age was 54 years [IQR 43/64] and 71.3% of the patients were males. Median left ventricular ejection fraction (LVEF) was calculated with 30.0% [20.0/40.0] and median left ventricular enddiastolic diameter (LVEDD) with 5.9 mm [5.2/6.6] accompanied by an elevated median BNP value of 447 pg/ml [137.0/1188.5]. GLS was reduced with − 8.8% [− 12.25/− 6.0]. Regarding symptoms, 71.7% patients were admitted with dyspnea (43.7% NYHA functional class III or IV), 32.4% complained about angina pectoris at time of admission and in 19.1% edema were detectable. The duration of symptoms differed: In 36.4% of the patients symptoms started not longer than 2 weeks, 30.3% had symptoms between 2 weeks and 3 months and 32.6% suffered from symptoms more than 3 months duration before admission. In immunohistochemistry, median counts of CD3+ and Mac1 were 6.470/mm2 [IQR 2.0/16.085] and 33.6/mm2 [16.9/58.1), respectively, defining a cut-off of ≥ 7.0 CD3+ lymphocytes/mm2 and/or 35.0 Mac1 macrophages/mm2.

Table 1 Baseline characteristics.
Full size table

Comparison of symptomatic non-valvular, non-ischemic heart failure patients, who underwent EMB, stratified for presence of myocardial inflammation

In 230 patients (57.9%), myocardial inflammation was diagnosed. Patients with and without inflammatory cardiomyopathy were of comparable age and revealed no differences regarding presented symptoms at admission. History of smoking (52.4% vs. 39.8%, p = 0.013) was more prevalent in HF patients with myocardial inflammation. While the groups did not differ regarding mean values of LVEF, GLS (Table 2, Fig. 2) and mean values of TnI and BNP (Table 2, Fig. 3), mean values of CRP (5.4 [2.1/19] vs. 3.7 [1.5/10], p = 0.003) were higher and median hemoglobin levels (14.1 [12.6/15.2] vs. 14.7 (13.5/15.7] g/dl, p = 0.001) were lower in patients with myocardial inflammation (Table 2, Fig. 3).

Table 2 Comparison of patients with inflammation and no inflammation.
Full size table
Figure 2
figure 2

Boxplots showing 1st quartile, median and 3rd quartile values of: (A) LVEF = left ventricular ejection fraction in % (not significant); (B) GLS = global longitudinal strain in % (not significant);

Full size image
Figure 3
figure 3

Boxplots showing 1st quartile, median and 3rd quartile values of: (A) Troponin I in pg/ml (not significant); (B) Brain natriuretic peptide (BNP) in pg/ml (not significant); (C) C-reactive protein (CRP) in mg/l (p = 0.003); (D) Hemoglobin (Hb) in g/dl (p = 0.001).

Full size image

Predictors of myocardial inflammation

ROC curves demonstrated only a moderate prognostic performance of CRP for prediction of myocardial inflammation with acceptable specificity and positive predictive values (Table 3, supplementary Figure 1). The computed best cut-off (according Youden-Index calculation) for CRP was 8.15 mg/l. In contrast, TnI, BNP and GLS were accompanied by a low prognostic performance to predict myocardial inflammation shown in the ROC curves. The computed best cut-off (according Youden-Index calculation) for TnI was 136.5 pg/ml, for BNP 1030 pg/ml and for GLS -13.95% (Table 3). Nevertheless, when calculating best cut-offs (according Youden-Index calculation), the cut-offs were able to differentiate between myocardial inflammation vs. no cardiac inflammation.

Table 3 Prognostic performance of TnI, BNP, CRP and GLS for prediction of myocardial inflammation.
Full size table

In multivariate logistic regression analyses after adjustment for CVRF, elevated TnI values (≥ 136.5 pg/ml) (OR 3.011 [95%CI 1.215–7.464]; p = 0.017) as well as elevated CRP values (≥ 8.15 mg/l) (OR 1.985 [95%CI 1.160–3.397], p = 0.012) were independently associated with myocardial inflammation. No association was detected for elevated BNP (≥ 1030.5 pg/ml) values (OR 1.811 [95%CI 0.873–3.757]; p = 0.111). Regarding echocardiographic parameters, LVEF categories and reduced GLS (≥ − 13.95%) were not associated with myocardial inflammation (Table 4, Fig. 4). However when combining laboratory parameters with echocardiographic parameters, a strong association with myocardial inflammation was found for the combined variable consisting of reduced GLS and elevated TnI (OR 9.633 [95%CI 2.027–45.769]; p = 0.004). Additionally, but to a smaller extent, the combination of TnI and CRP was also independently associated with myocardial inflammation (OR 5.761 [95% 1.240–26.771], p = 0.025) (Supplementary table 1). Furthermore, supplementary table 2 illustrates the prognostic performance of combined parameters to predict myocardial inflammation. While all parameter-combinations had only moderate sensitivity (21–39%), the specificity was better: 0.71 (0.62–0.79) for GLS and elevated CRP, 0.95 (0.91–0.98) for positive TnI and elevated CRP and 0.96 (0.90–0.99) for the combination of GLS and TnI. Remarkably, highest positive predictive value was also detected for the combination of GLS and TnI 0.90 (0.74–0.96). The negative predictive values were only moderate in all of these combinations (41–46%). 

Table 4 Predictors of inflammation.
Full size table
Figure 4
figure 4

Predictors for inflammation (Forest plot). OR (Odds Ratio) and 95% CI (Confidence Intervall) displayed in a Forest plot after adjustment for age/sex/CVRF. Absolute values are shown in Table 4.

Full size image

Concerning the interrelation of cardiovascular risk factors with myocardial inflammation, only smoking was associated with myocardial inflammation in crude logistic regression analysis (OR 1.6687 [95%CI 1.113–2.500]; p = 0.013), but this association showed only a trend towards significance in the fully adjusted model (OR 1.659 [95%CI 1.000–2.753]; p = 0.050) (Table 4, Fig. 4).

Discussion

The key findings of the present study are that (I) elevated CRP and TnI levels are associated with inflammatory cardiomyopathy in patients with non-ischemic, non valvular heart failure; (II) TnI levels ≥ 136.5 pg/ml and CRP levels ≥ 8.15 mg/l are potential predictors of myocardial inflammation; (III) common heart failure biomarkers such as BNP and reduced LVEF were not predictive for myocardial inflammation and; (IV) GLS might add a benefit to predict myocardial inflammation when combined with TnI.

A growing body of evidence indicates that myocardial inflammation may lead to severe cardiac damage, deteriorate into dilated cardiomyopathy and be accompanied by poor prognosis. This implies that early diagnosis using predictive markers may avert poor outcome20, since immunosuppressive therapy can significantly improve the prognosis of HF patients with presence of myocardial inflammation16. Thus, readily available markers like the ones identified in our study might help to select patients for EMB and for shorter control follow-up examinations. Although cardiac MRI diagnostics has become increasingly important in the diagnosis of myocardial inflammation21,22, sensitivity and specificity of MRI findings are not unequivocal. Since EMB can be performed easily as part of coronary catheter examination18 and should be carried out as early as possible if there is clinical or laboratory suspicion of inflammatory myocardial disease, our study results might help to accelerate the diagnostic work-up of HF.

The association of elevated CRP and TnI with myocardial inflammation was independent of age and sex, supporting earlier findings by Lauer et al., who reported that patients with elevated troponin T values were more likely to have EMB-proven myocarditis23. Because cardiac troponins are detectable in the blood for approximately one week24 and have a half-life of 2h25, elevated troponins are suitable to indicate permanent myocardial cell injury26,27, which can be explained by inflammatory infiltration with consecutive cell damage documented in the EMB. Contrary to our results, Sramko et al. demonstrated that troponin was not able to distinguish between patients with idiopathic dilated cardiomyopathy or inflammatory dilated cardiomyopathy arguing that troponin indicates myocardial injury without indicating weather its due to myocardial inflammation or simply caused by myocardial wallstress related to heart failure28. However, the study sample size was small, and EMB was only performed on the right ventricle making a sampling error more likely.

Similar to positive troponin values, positive CRP values were associated with detection of inflammation in EMB. These results go in line with observation by Liu et al., who demonstrated that highly sensitive CRP levels were higher in patients with EMB-confirmed myocarditis than in patients without myocarditis29. However, it has to be pointed out that the sensitivity for highly sensitive CRP in detecting myocarditis was only 50.1% whereas the specificity was 80.7%. Since CRP is also increased in acute myocardial infarction30 or in rheumatic valve disease31. CRP should not be interpreted alone, but in conjunction with cardiac troponin. Moreover, as the present study indicates, cut-off values for TnI and CRP seem to be higher with regard to myocardial inflammation compared to cut-off values used for diagnosis of acute coronary syndrom. Further studies are needed to prove those cut-off values in patients with myocardial inflammation.

Interestingly GLS was not able to distinguish between EMB proven myocardial inflammation and no inflammation. Escher et al. demonstrated before, that patients with acute myocarditis had reduced GLS values. After a follow-up period of 6.2 months patients with persistent inflammation had worse GLS values than patients without inflammation32. However in contrast to our study GLS values were compared when LVEF had already been improved. Eventhough it has been shown that GLS is a superior predictor of adverse cardiac events compared with LVEF33,34, studies demonstrate that in general and irrespectively of outcome prediction, GLS correlates well with LVEF in heart failure patients35. Therefore the results of our study suggest that GLS alone is not able to identify patients with myocardial inflammation in the presence of reduced LVEF. Nevertheless, the combined evaluation of GLS and Troponin values were the second strongest predictor of myocardial inflammation in multivariate regression analysis. Additionally, the best prognostic performance was found for the combination of GLS and Troponin. In particular, the high positive prognostic value of 90% to predict myocardial inflammation is promising. Thus, the combination of GLS and Troponin may help to identify patients with myocardial inflammation in the future and could add a benefit in the diagnosis and follow up treatment of those patients.

Interestingly, smoking was associated with myocardial inflammation in our study. In a very small study sample, it was shown, that in patients hospitalized for acute myocarditis smoking was the most prevalent cardiovascular risk factor36. Recently, Detorakis et al. reported a significant correlation of smoking habits with late gadolinium enhancement extent in cardiac MRI in patients with clinically suspected myocarditis37. Prolonged exposure to tobacco smoke may cause cardiovascular cell damage, suggesting that increased myocardial cell necrosis and cell death would make myocardial inflammatory infiltration more likely38.

Another finding of the present study was that patients with myocardial inflammation had lower hemoglobin levels. It is well known that anemia of inflammation or anemia of chronic disease is primarily a disorder of iron distribution39 and even low grade inflammation is associated with lower hemoglobin levels40. As experimental models have demonstrated that low cardiac iron levels promote heart failure41,42 the link between the observed lower hemoglobin levels could be an effect of possibly local iron distribution disorders due to myocardial inflammation. Since hemoglobin levels in the present study were generally not significantly decreased, more data are needed to support this observation.

The primary stimulus for natriuretic peptide synthesis by myocardial cells is related to increased LV wall stress induced for example by acute myocardial infarction43, arterial hypertension44, muscle hypertrophy45 or increased pulmonary arterial pressure46. We expected that inflammation would also cause elevated BNP levels as reported earlier47, particularly in patients with myocarditis and dilated cardiomyopathy48,49,50. Increase of BNP has also been detected in systemic inflammation regardless of systolic function51. In the present study, no differences in BNP levels and LVEF were found in patients with and without myocardial inflammation, reflecting a rather compensated disease state of our study population with low rates of hydropic decompensation. Myocarditis and inflammatory cardiomyopathy are a dynamic processes; after inflammation has subsided, a reduced LVEF is due to secondary post-inflammatory fibrosis and myocardial cell death and scars52. As conditions with and without inflammation lead to changes of LVEF and BNP, the present study demonstrates that increase of BNP and the decrease of LVEF do not specifically reflect the presence of inflammation, but rather reflect a general loss of myocardial tissue with consecutive changes in hemodynamic conditions regardless weather inflammation can be detected or not.

Limitation

There are certain limitations of our study which need to be mentioned: Firstly, our study was conducted as a monocentric retrospective data analysis. Therefore, it has to be emphasized that the results of this study can only serve in terms of a hypothesis generating research and the results of the present study should be verified by prospective trials. Secondly, there is no generally applicable definition of myocardial inflammation. Thus, it might be difficult to adopt the results to other endomyocardial biopsy studies, but generating more information about inflammation process was one of the main aims of our study. Finally, a potential limitation regarding EMB interpretation might be the sampling error. Regarding this point, Hauck et al. demonstrated that sampling error was prevalent 45% in Ieft and up to 37% in right EMB53. The authors conclude that only positive results of EMB can be considered diagnostic. However, the cited study included a very small sample of only 36 patients.

Conclusion

Elevations of readily available markers like cardiac TnI and CRP as well as the combination of GLS and TnI may predict inflammatory cardiomyopathy and might be useful to select patients for EMB and for shorter control follow-up examinations. In the diagnostic approach to detect suspected myocardial inflammation, higher cut-off values concerning cardiac and inflammatory biomarkers may have to be applied. Smoking is associated with inflammatory cardiomyopathy in non-ischemic, non valvular HF patients and may indicate individuals at risk to develop inflammatory cardiomyopathy.

Ethics approval

All data were obtained from individuals enrolled between 2013 and 2018 in the retrospective monocentric Mainz Endomyocardial Biopsy in Heart Failure Study (My Biopsy-HF Study, DRKS #22178), which was approved by the Ethics Committee of Rhineland Palatinate to be in accordance with the legal regulations and the declaration of Helsinki.