Discussion
In this retrospective analysis using multiparametric CMR, we were able to monitor the response to levosimendan infusion in patients with systolic heart failure. In the study cohort, a single dose of levosimendan resulted in significant reduction in LV size, shown by a significant decrease in LVEDV and LVESV (Fig
1). This resulted in unchanged LVEF values, as well as a non- significant increase in CO.. Most importantly, in this cohort, we could not show a significant increase in deformity parameters that might be representative of the positive inotropic effects of the drug on the LV myocardium using CMR strain imaging. Despite the fact that baseline GLS was significantly reduced in patients with ICMP when compared to patients with valvular disease, levosimendan infusion had no effect on LV strain values. Furthermore, the absolute amount of LV scarring assessed by LGE showed no correlation to changes in LV size, function, and strain after levosimendan infusion. Therefore, the amount of LV scaring seems to have no clinical impact on short-term treatment response to levosimendan infusion.
Levosimendan for the treatment of acute heart failure has been shown to be safe and efficient compared to placebo and dobutamine [
1,
18]. However, its usefulness in the pre- or post-operative setting in in patients with low EF and/or conditions of acute heart failure in clinical practice remains controversial, as randomized studies have shown no impact on the length of intensive care unit stay or mortality [
3,
4]. Therefore, appropriate indications and patient selection, as well as timing of the infusion in the setting of potential cardiac surgery, might be of clinical importance [
6]. From available imaging modalities to date, only echocardiography has been used to monitor LV function with the use of the drug [
19,
20]. With the exception of a case report, CMR, including tissue characterization for treatment monitoring, has not been reported in the literature thus [
21].
In our cohort, using CMR, treatment response to levosimendan resulted in a significant decrease in LV size after a single infusion over 24 hours. This resulted in only a non-significant change in LV EF. These findings with CMR in our cohort differ from data reported from echocardiograms or studies using pulmonary catheterization [
1,
22,
23]. This might be based on the difference in the evaluation of the EF primarily when using non-enhanced 2D echocardiography [
24]. For example, Navarri et al. showed significant changes after levosimendan infusion in LVEF with echocardiography, but, at the same time, nonsignificant changes in LV systolic and LV diastolic values [
22]. These finding are in contrast to ours and may be based on the methodology used for assessing LV volumes and LV EF. Imaging reports showed significantly lower EF values in echocardiography when comparing unenhanced echocardiography to CMR, and, despite a good correlation between the modalities, a wide limit of agreement in heart failure patients, especially with regard to LV volumes [
25,
26]. This might favor the use of CMR in heart failure cohorts with severely dilated LVs. Furthermore, we were not able to exactly compare our study cohort to others published in the literature in terms of LV volumes, as most of the studies used invasive pulmonary artery catheterization to report on SV or solely reported on outcomes.
CMR, including LGE for quantification of scar load, has not been applied to response assessment or response prediction after levosimendan administration. In the literature, mostly invasive monitoring using pulmonary artery catheterization or echocardiography has been reported [
1,
20]. In this cohort, LGE was assessed for myocardial viability or for diagnostic proposes in the assessment of LV dilatation [
9,
10]. We evaluated whether there was an influence of scar load on therapy response to levosimendan. The aim of this analyses was to screen for patients who might not profit from levosimendan because of high (ischemic or non-ischemic) scar burden, as this would have an impact on costs and might avoid potential complications of levosimendan infusion, such as hypotension and hypokalemia. However, in our cohort, we were not able to show a correlation between changes in LV parameters and the percentage of scar load. In addition, there was no difference between patients with known ischemic CMP and dilatative CMP based on the presence of valvular disease. These findings are similar to results published by Najjar et al., who showed positive hemodynamic effects of levosimendan in a heart failure cohort, but failed to identify predictors of the response to drug exposure [
7]. Prospective studies are lacking about the influence of LV structure on treatment response or even outcomes.
The assessment of left ventricular strain parameters and changes in these parameters appear to be valuable predictors of treatment response in various cardiac diseases [
12]. The methodology has already shown proven reproducibility with an acceptable inter- and intraobserver variability [
14]. Therefore, strain measurements seem to be a valuable alternative to EF measurements, as these parameters might be more representative of the improvement in myocardial function and deformity than EF measurements based on the ratio between EDV and ESV alone [
27]. Strain measurements, especially GLS, have already been proven as an independent predictor of outcome in CMP and non-ischemic DCM, showing that a reduction in strain of one standard deviation has a severe impact on the hazard ratio for cardiovascular events [
28,
29]. In ischemic heart disease GLS and GCS changes in the early phase after myocardial infarction were predictive of outcome [
15]. More data on changes in strain measurements after medical or vascular interventions are currently lacking in the literature. Despite significant changes in LV volumes after levosimendan infusion, no changes in CMR-derived strain parameters could be observed in our cohort. Therefore, the observed reduction in LV size in our cohort might be based solely on lowering the peripheral resistance and not based on changes in inotropy. Dalla et al. described in his model with septic patients receiving norepinephrine that changes in afterload and preload i.e. an increase in systemic and pulmonary vascular resistance, impact LV EDV, but not ejection fraction or longitudinal strain [
30]. The lack of changes in strain parameters in this cohort could also be explained by a lack of ability in the improvement of deformity based on longstanding LV dilatation, presence of valvular disease, and the presence of fibrosis. The mean percentage of LV fibrosis was 16% in our study group and it has already been shown that the presence of LGE is associated with reduced strain parameters [
14]. Furthermore, the presence of LGE had an impact on GLS values prior to levosimendan infusion in our ICMP subgroup. However, we could not prove a correlation between the absolute amount of scarring by LGE and the changes in GLS in this subgroup, and strain parameters numerically showed even larger changes in the ICMP group when compared to valvular disease. Therefore, LGE does not predict (but also does not prevent) treatment response to levosimendan infusion.
In DCM, imaging studies on short-term changes in LV size are lacking and results of long-term studies are somehow divergent. On the one hand, the presence of LGE has been shown to be a negative predictor of LV reverse remodeling in non-ischemic CMP [
31]. On the other hand, Tayal et al. failed to identify the measurement of cardiac structure or cardiac strain as a marker for LV LV recovery in DCM [
32]. In their work, contractile reserve, as evidenced by dobutamine stress echocardiography, was predictive for a reduction in LV size [
32]. In our cohort, we observed significantly smaller LV volumes after levosimendan infusion regardless of the presence of LV scarring in both groups.
There are several limitations of our study. First, it is a retrospective, single-center cohort study with the focus on imaging parameters. Data on the outcome and/or impact of levosimendan infusion on intermediate and long-term outcome are not available. In addition, the sample size is small and patients were selected non-randomly by a single referring physician, which may have caused a patient selection bias for the analysis. On the other hand by this patient selection was more homogenous than in a multi center setting. Additionally data on hemodynamic changes in case of valvular disease (e.g. changes in regurgitant fraction) are missing as flow measurements have not been implemented in all CMR studies.