We performed a systematic search of PUBMED and SCOPUS for all studies with prognostic evaluation of microRNAs in HF patients, including all stages of HF and clinical settings (Figure 1).

Figure 1.

Flow diagram showing study identification to inclusion.

(0.53MB).

Publications were limited between 1 January 2010 and 31 July 2020. The search strategy on PUBMED included a MESH search: ((“Biomarkers”[Mesh]) OR (“MicroRNAs”[Mesh] OR “Circulating MicroRNA”[Mesh])) AND (“Heart Failure”[Mesh] OR “Heart Failure, Diastolic”[Mesh] OR “Heart Failure, Systolic”[Mesh]), and the filter English was not used. The search strategy for SCOPUS encompassed key-words search: ALL (“microRNAs”) OR (“circulating microRNAs”) AND (“heart failure”) AND (Limit-To (DOCTYPE, “ar”)) AND (Limit-To (SUBJAREA, “medi”)). Authors were contacted when relevant for missing prognostic performance data. All publications of interest were collected in EndNote. After removing duplicates, publication titles and abstracts were reviewed by one assessor (R.F.) to check they met the inclusion criteria, before full-text screening. Secondly, studies were assessed for risk of bias by R.F., R.A., C.S. PRISMA was used to include all items that are part of a systematic review19 (Figure 1). Tools recommended for producing a systematic review were adopted (QUADAS-220 and Cochrane Handbook for Systematic Reviews for Diagnostic Test Accuracy21).

Time period. Studies published from 1 January 2010 to 31 July 2020 in English to focus on recent evidence written in English concerning human participants.

Biomarker types and sample types. Studies focusing on microRNA as an individual biomarkers or multiple biomarkers were included. Studies that did not report statistical significance or quantitative prognostic measures (hazard ratio (HR), odds ratio (OR), or relative risk (RR)) of the biomarker's prognostic performance were excluded. Studies measuring miR as a biomarker in medium other than serum or plasma samples were excluded. Studies using methods other than RT_PCR for miR quantification or studies that used RNU6 as a normalizer were excluded.

Study population. Studies in adult subjects with HF diagnosis were included. To minimize the risk of excluding promising early-stage research studies, the inclusion cut-off was studies with >50 human participants in total. There were no criteria for the number of participants per group.

Study types. Studies with any cohort design, including cross-sectional studies were included. As recommended by the Cochrane Handbook for Systematic Reviews for Diagnostic Test Accuracy, we excluded all case-control design studies.21 Reviews, systematic reviews, meta-analysis, books and documents, case reports, comments, news, and editorials were excluded.

To assess the true potential of miRs in HF, we analyzed miRs expression as: (i) The potential to discern between HF with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF); (ii) a marker for HF severity (including associations between echocardiographic measurements, New York Heart Association (NYHA) class and NP).

Lastly, to assess the capacity of miRs as prognostic biomarkers, three outcomes were established: (i) Cardiovascular hospitalization (CV hospitalization) (including the risk of CV patients, focusing on HF patients being admitted or re-admitted to hospital), in order to comprise one endpoint without death; (ii) Cardiovascular hospitalization and/or death (including the risk of patients with HF diagnosis being hospitalized/re-hospitalized or death after HF diagnosis) and (iii) all-cause death (including HF patients’ risk of death, regardless of the cause).

The data collection from included articles comprised: the study design; study population; number of subjects; female and male percentage; mean age±SD; initial diagnosis; variables included in multivariate models (Table 1) and the associated measures (HR, OR and RR) for different outcomes and statistical analysis performed (univariate or multivariate).

To assess the quality of the studies and identify potential bias, we applied the QUADAS-2 tool. Considering the risk of bias analysis, study design, patient selection, miR index test and standard reference, studies were scrutinized for each outcome of interest.

Few studies analyzed miRs as potential biomarkers in a differential diagnosis between HF with preserved or reduced ejection fraction. Watson et al. demonstrated that serum expression of miR-30c, miR-221, miR-328, and miR-375 were able to distinguish HFrEF from HFpEF (all area under curve (AUC) >0.7).22 Wong et al. performed a similar analysis and identified four miRs that are able to differentiate between the two HF entities (miR-125a-5p, -190a, -550a-5p, and -638) (AUC miR panel 0.8).23

Previous studies suggest that specific miR could help in HF severity evaluation. From the articles included in our review, only Klenke et al. did not find any association between BNP or NYHA status.24 Vogel et al. in order to test if miR expression might correlate with disease severity, compared systolic function (mild-moderate and severe) and concluded that miRNAs expression patterns in control groups are different from HF group and significant different between the severity groups. To summarize miRs expression might correlate with different HF stages or severities. Also, significant correlations between left ventricular (LV) EF and miR-622, miR-520d, miR-519, miR-200b, miR-122, and miR-588 were found (miR expression levels increased with lower values of (LV)EF), but not with NYHA functional class.25 MiR-150-5p expression was significantly correlated with symptoms severity and adverse remodeling degree. Also, miR-150-5p was inversely correlated with NYHA classes (p=0.004) and log NT-proBNP).26 MiR-21 and miR-132 expression levels increase alongside NYHA grade and consequently so does HF severity.27,28 On the other hand, negative correlations between miR-145 and cardiac function were found and lower levels of miR-22-3p and Ln_miR-145 (p<0.0001) were associated with HF worsening.27,29 Ovchinnikova et al. found that acute HF is associated with the analyzed miR (let-7i-5p, miR-18a-5p, miR-18b-5p, miR-223-3p, miR-301a-3p, miR-423-5p miR-652-3p) shown by a consistent pattern of decreased miR expression levels with increased HF severity, thus reaching the conclusion that miR expression levels were higher at discharge than at admission.30 In 2017, Vegter et al. discovered that miRNA levels decreased in parallel with clinical manifestations of atherosclerotic disease, therefore contributing to HF severity.31

In this systematic review, several miRs were identified as potential biomarkers for HF diagnosis, prognosis and disease severity, supporting miR's pivotal role in cardiac physiology.

Several miR were related with HF. MiR-21, miR-22 and miR-132 were described as important biomarkers in HF pathophysiology and progression. They are also clinically correlated to NYHA classes, volume status and fluid overload. Although none of the studied miR were revealed to have a better potential as a biomarker, after combining miR expressions with BNP and NT-proBNP, their potential to distinguish HFrEF from HFpEF was prominent, and even better than NP. MiR-423, let-7i-5p, miR-223-5p, miR-1306-5p and miR-22-3p were analyzed in more than one article and described as potential biomarkers for HF prognosis (Table 5).

Heart failure with preserved ejection fraction and HFrEF cannot be differentiated on clinical grounds and imaging tests are essential for a correct diagnosis. HFpEF is not a straightforward matter because it is diagnosed by a constellation of clinical signs and symptoms, most of which relate to the fact that the ventricle operates at a higher filling pressure.39,40

The distinction between HFpEF vs. HFrEF is based on echo derived left ventricular ejection fraction (LVEF). However, LVEF has limitations in detecting subtle abnormalities of systolic function. Previous studies using strain showed a reduction of radial and longitudinal despite normal ejection fraction. Furthermore, echography tests involve well trained sonographers and equipment. A serum-based biomarker that could help distinguis between HFpEF vs. HFrEF would facilitate easy and fast differential diagnosis in clinical practice. Studies have tried to use NP levels to distinguish HF entities, but the use of NP lacks specificity and does not allow a correct differentiation of HF subtypes.10,22,41,42

Natriuretic peptides are directly related to ventricular function; they are produced in response to ventricular diastolic stretch. The main trigger for their release in the bloodstream is left ventricular (LV) end-diastolic wall stress, present in a dilated LV in HFrEF but not necessarily in HFpEF.43,44

MiRs have been described as being able to distinguish HFrEF and HFpEF, and several correlated with echocardiographic measurements.25,26,29 Several articles have stated that miR expression has a similar but not superior performance to NP. However, when the same analysis combined miR expressions with BNP and NT-proBNP, the potential for distinguishing HFrEF from HFpEF was prominent, and even better than NP used alone.22,23,37

The articles under study compared diagnostic potential for both NP and miRs. To assess miR diagnostic potential in HFrEF, Vogel et al. compared NT-proBNP accuracy and sensitivity with miR and found a better performance of miR -519, -520d and -622 (miRs AUC 0.81; NT- pro-BNP sensitivity 78% and specificity 44%).25 Wong et al. compared NT-proBNP and miR as diagnostic biomarker between HFrEF and HFpEF and although NT-proBNP showed potential, the combination of miR with NT-proBNP improved efficacy to distinguish between them (AUC NT-proBNP alone 0.83 to AUC NT-proBNP with miR 0.91).23 Watson et al. assessed miR capacity to improve the diagnostic utility of existent BNP and concluded that miR could improve diagnostic performance of BNP in HF (AUC of BNP alone 0.87 vs. AUC of BNP with miR 0.90), demonstrating superiority in distinguishing HF entities (AUC BNP alone 0.66 vs. AUC BNP with miR 0.85).22

MiR-622, -519 and -122 were significantly related with HFrEF. High expression levels of miR-622 and-519 were found in granulocyte cells by Vogel et al., reinforcing the inflammatory processes involvement in HF development and progression. MiR-122 is originated in hepatic cells and has been related to HF due to diminished cardiac output, resulting in liver congestion.25,45,46 Nevertheless, only miR-328, miR-375 and miR-499 expression levels were significantly different between HFrEF and HFpEF.22,33

The articles under study made reference to a gradual increase in specific miR levels in more stabilized HF, chronic HF (CHF) and healthy controls, when compared to acute HF patients.30,31,47 Indeed, low levels of miRs were associated with increased levels of biochemical markers of inflammation, angiogenesis and endothelial dysfunction, thus supporting previous data.31

Negative correlations between miR-22 and the outcome have also been described.27 Scientific research has stated that miR-22-3p has a protective role in the myocardium due to its negative regulation of angiotensin II.27,48 It is mostly expressed by striated muscle tissues and is essential in normal cardiac remodeling after environmental stress.27,47 In fact, among the included studies, several severity-related miRs were described as upregulated28,34 or downregulated26,30 regardless of HF etiology. This suggests that different factors may influence miR serum expression. Volume status variations and fluid overload due to renal impairment and proteinuria (once they are bound to plasma proteins such as albumin, and low filtration rate may lead to the loss of carrier plasma proteins).34,47

Recent literature reports the ability of miRs to predict independently the outcome in HF patients. Several miR were highlighted as prognostic markers and whose targets are signaling pathways already known to be involved in vascular and cellular processes related with HF (Figure 3).

Figure 3.

MicroRNAs targets associated with more than one outcome. MiR-223-3p has been associated with several carcinomas, and is involved in cell proliferation, growth, migration and invasion, furthermore it negatively regulates PRDM1 involved in immune and inflammatory processes. KLF15 promotes cellular growth, regulating negatively TGFβ1 and RASA1, promoting cell migration.67–70 MiR-21 is negatively related to TGFβ1/SMAD7 and TNFα1, enhancing fibrotic processes, thus promoting ventricular remodeling.71–73 Let-7i-5p is known for negatively regulating inflammation and fibrosis processes, partly due to its effect on Ang-II and IL-6 and on E2F2 and CCND2 expression (regulators of cell cycle) and cardiac recovery.49,74 Few studies analysed miR-1254 targets, which is thought to be involved in cardiac remodeling, playing a protective role by inhibiting the PI3K/AKT pathway and SMURF1 expression (important in cell motility, signaling and polarity processes).75 MiR-106-5p might have a protective role, since it is suggested that it affects stress response due to MAPK signaling pathway inhibition and consequently oxidative stress.76,77 MiR-1306-3p inhibits the TGFβ/SMAD pathway, and is involved in pro-apoptotic processes. MiR-423-3p expression is triggered by hypoxia/reoxygenation processes, positively regulating Caspase 3/7, Bax, Caspase C promoting mitochondrial dysfunction.

(0.32MB).

Let-7i-5p and miR-223-5p are well known inflammation and fibrosis related miRs.31,49 Authors have suggested that the Let-7 family might have an active role in HF pathogenesis, and it has been related to poor outcomes in dilated cardiomyopathy.50,51 Interestingly and similar to miR-22, recent literature states that let-7i-5p negatively regulates angiotensin II, attenuating cardiac inflammation and fibrosis,49 thus contradicting previous literature and our findings.

Recently identified for the first time in HF patients, miR-1306-5p had been considered a possible biomarker in two large cohorts.33,37 Its function might be related to numerous processes in cells, including proliferation, differentiation and cell cycles.52

Previous literature states that miR-208 and miR-499 overexpression promotes LV hypertrophy, and ultimately HF.53 Positive correlations with HF have been found in miR-208 but not in miR-499, respectively.33

MiR-122, as previously stated, is a liver-specific miR associated with the risk of developing metabolic syndrome and has been linked to HF prognosis.45 Stojkovic et al. pointed to miR-122 as an independent predictor of all-cause mortality after adjustment for NT-proBNP in HFrEF patients. These data might suggest liver involvement in HF pathophysiology.36

Several of the included studies analyzed miR-423-5p. Positive, negative and no association with disease progression were described30–32,36,37,54 MiR-423-5p has been associated with chemotherapy resistance in cancer55–58 and, more recently, with cardiovascular diseases. The literature has pointed to a link between miR-423, vascular endothelial growth factor and nitric oxide (NO), concluding that miR-423 may serve as a biomarker of vascular growth/proliferation or inhibition, depending on the pathological process and target organ.59 These contradicting conclusions could be related to different miR expression in HF patients during an acute phase. In fact, most of the literature describes an association between miR-423-5p and a poor outcome. Cells and blood are the major source of miR suggesting that they might have a paracrine function that could justify miRs serum dysregulation. This indicates that HF may not be the only source of prognostic information, but it may derive from other damaged organs or cells.60

Curiously, negative associations were found between miR-423-5p and CV hospitalization and all-cause death,32 implying a protective role of miR-423. This unexpected result can be explained for different reasons: miR can be diluted and, also, different expression values in collected blood samples can be detected.38 Indeed, some miR can be undetectable in a large portion of samples.37,61

MiR-21 has also been linked to cell proliferation, migration and apoptosis processes and it plays an important role in the pathophysiology of hypertension.62 Despite its apparent role in heart disease, only one study has clinically referred to an association between miR-2163 and HF prognosis.32,38

The studies subject to analysis revealed the need for a common approach to miR quantification and uniformization of methods. The standardization of methods and an adequate normalization of miR levels could clarify disparities in the conclusions.

MiR quantification method. The small noncoding RNA RNU6 genes are the reference genes most used as normalizers. Nonetheless, RNU is not a miR and, thus, it is not able to reflect the biological and chemical features of miRs, among other limitations.64 Several miR have been described and suggested as normalizers (i.e., miR-16) because of their increased blood expression and uniformity across several samples.64 Currently, there is no standardized normalizer miR in the scientific community, however, several studies have demonstrated that the use of more than one reference normalizer increases quantification accuracy. Unfortunately, all the studies included in this review only used one miR as a normalizer. Proportionally, different miRs were identified as altered in serum along with the study groups, making it impossible to measure and compare. This implies that using different normalizers may be responsible for distinct miR detection.

Study population definition criteria. An additional identified limitation was the lack of uniformity regarding a definition of HF and functional class criteria among the included studies. Also, study group selection in analyzed articles was distinct. Indeed, several groups included numerous definitions such as: hospitalized acute HF patients, hospitalized and outpatient chronic HF patients, unspecified HF, among others.