Biomarkers in Advanced Heart Failure: Implications for Managing Patients With Mechanical Circulatory Support and Cardiac Transplantation

While a complete review of the biology of known cardiac biomarkers is out of the scope of this manuscript and is provided elsewhere,^8,9 it is important to highlight changes in these biomarkers after mechanical support. Unfortunately, the landmark randomized trials assessing outcomes following modern LVAD therapy did not report the biomarkers discussed below,^4,10,11 thus most data are derived from smaller, single-center studies. Inclusion of biomarkers of interest in subsequent trials of novel LVADs, at enrollment and during follow-up, would significantly elevate the level of evidence available.

LVAD implantation results in sudden unloading of the left ventricle. Not surprisingly, biomarkers of myocardial stretch—BNP (B-type natriuretic peptide) and NT-proBNP (N-terminal pro-BNP)—decrease rapidly after LVAD implant.^12–14 Significant decreases in concentrations occur by 1 to 4 weeks post-implant with a sustained reduction in concentrations as long as 9 months.^12–19 One study of 9 cf (continuous flow) LVAD (6 HeartMate [HM] II and 3 HeartWare [HVAD]) patients noted a decrease in mean BNP from 1200 pg/mL at baseline to 300 pg/mL by month 3 with subsequent plateau; another of 25 cfLVAD (HMII) patients demonstrated a decrease from 1900 pg/mL to ≈350 pg/mL by month 1 with only slight additional reduction thereafter.^13,20 A similar trajectory has also been noted after initiation of neurohormonal therapy in chronic HF patients,²¹ indicating that natriuretic peptide concentrations can change rapidly regardless of the mechanism of unloading. It is important to note, however, that very few patients with LVAD reach normal natriuretic peptide levels when compared with non-HF patients.^12,20

GDF (Growth differentiation factor)-15 is a member of the transforming growth factor cytokine family. Its expression can be induced in response to pressure overload and inflammatory cytokines.²² GDF-15 levels are elevated in a majority of end-stage HF patients pre-LVAD, and the degree of elevation correlates with degree of myocardial fibrosis on histology at time of LVAD implant.²³ A single-center, retrospective study of 25 primarily HMII LVADs noted a significant decrease in mean serum concentrations of GDF-15 1 month after implant (7.1 ng/mL to ≈2 ng/mL), presumably through a reduction in oxidative stress via restoration of hemodynamics. By 6 months, 75% of patients had achieved laboratory normal levels (<1.2 ng/mL), although these still remained significantly higher than controls.²³

Suppressor of tumorigenicity (ST) 2 is a member of the IL (interleukin)-1 receptor family, which, when bound to its ligand IL33, leads to inhibition of apoptosis, hypertrophy, and fibrosis.²⁴ Soluble ST2 (sST2) serves as a decoy receptor for IL33, resulting in activation of a profibrotic cascade.⁸ A single-center, retrospective study of 38 cfLVAD patients found that mean sST2 concentration was elevated in patients before LVAD implant (74.2 ng/mL with reference range <35 ng/mL), with significant between-group differences when comparing Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) I profile (120.2 ng/mL) versus INTERMACS II/III profile (65.7 ng/mL). Interestingly, while CRP (C-reactive protein) levels correlated with degree of sST2 concentration, numerous other clinical features (right ventricular dysfunction, renal dysfunction, natriuretic peptide levels, filling pressures, cardiac output) did not.²⁵ Similar to the other biomarkers, there was a precipitous decrease of sST2 levels by 1 month post-LVAD and a continued significant decline to month 3 when mean levels were below the reference range; levels stabilized thereafter.²⁵

Galectin (Gal)-3 is a product of the inflammatory response of chronic HF and stimulates pathological remodeling.⁸ One study noted a small but significant decrease in Gal-3 after LVAD implant¹²; others reported a temporary decrease post-LVAD with subsequent increase^13,26; and yet another reported no significant change at 30 days post-LVAD.¹⁸ Additional biomarkers of myocardial fibrosis shown to be elevated in chronic HF (osteopontin and connective tissue growth factor) also do not seem to reliably decrease after LVAD implant.^13,18 These observations suggest that unloading may not end the chronic inflammatory and fibrotic response induced by HF.

In sum, myocardial unloading by LVAD results in robust and rapid decreases in the concentrations of natriuretic peptides, GDF-15, and sST2 with some inconsistent changes in Gal-3 (Figure 1). Knowledge of this expected response could provide a clinical role for routine measurements of these biomarkers, with deviations suggestive of possible clinical complications, which we review in the following.

A prospective cohort study of 136 patients with cfLVADs (HMII) evaluated whether routine assessment of outpatient natriuretic peptides could be utilized to predict general adverse events after LVAD implant.²⁷ In this study, the authors retrospectively assigned patients to following 3 cohorts based on postimplant clinical course: (A) serious adverse events requiring hospitalization, (B) complications requiring outpatient therapy, and (C) uncomplicated course. Patients in group A had higher absolute values of peak BNP (measured at the time of the clinical event) and higher peak BNP/preimplant BNP ratios than the other groups. When comparing patients having any complication (group A or B) to those who had an event-free course (C), mean BNP concentration even at the first outpatient visit was significantly elevated. They then assessed diagnostic performance of BNP for specific clinical conditions. For infection, a BNP peak of >413 pg/mL demonstrated a sensitivity of 76%, but specificity of only 50% with a receiver operator curve area under the curve of 0.68. Performance was better for HF, where BNP >418 pg/mL had 86% sensitivity, 53% specificity, and area under the curve of 0.75, and pump thrombosis, where BNP >782 pg/mL had 100% sensitivity, 77% specificity, and area under the curve of 0.93.²⁷

A BNP-guided care strategy may be beneficial in the immediate postsurgical period.²⁸ In a single-center study of 85 cfLVAD patients, a clinical protocol was instituted involving measuring BNP concentrations almost daily in the postsurgical ICU. Consecutive increases in BNP levels (of at least 100 pg/mL) prompted the care team to further investigation (echocardiography) or interventions (increasing diuretic/inotrope dose, increasing LVAD speed) guided by individual clinical assessment. The investigators then assessed outcomes in the BNP-guided group as compared with historical controls. The BNP-guided strategy group had a significant, 5 day shorter length of stay post-LVAD but no change in 90 mortality or re-admission rates.²⁸ Given the lack of a contemporaneous control group, the beneficial findings of the strategy may be confounded by improvements in standard care for patients with LVAD but provide hypothesis-generating data for postoperative LVAD management.

Preimplant Gal-3 concentration is a prognostic marker for death post-LVAD implant. A retrospective study of older generation LVADs revealed that Gal-3 plasma concentrations at time of LVAD implantation are significantly higher in those who subsequently die of multiorgan failure post-implant (13.4 versus 9.6 ng/mL),¹⁸ and, in a more contemporary study of 57 cfLVAD patients, Gal-3 levels >30 ng/mL at time of LVAD implant have been associated with lower 2 year survival (76.5% versus 95%).²⁶

Right ventricular (RV) failure remains common after LVAD implant, with incidence ranging from 20% to 30% even in contemporary cohorts.^4,29 Patients who suffer right ventricular failure (RVF) require longer inotropic support, longer intensive care unit stays, longer hospitalizations, and importantly, have higher 1 and 2 year mortality.²⁹ Many risk scores have been derived to try to predict this complication of LVAD implant, most recently from European Registry for Patients with Mechanical Circulatory Support (EUROMACS) database, EUROMACS-RHF. This score incorporates several clinical risk factors as being predictive of postoperative RVF, including need for multiple inotropic agents, INTERMACS class 1-3, RV dysfunction on echocardiography, and a ratio of right atrial to pulmonary wedge pressure of >0.54. The authors assessed significance of a number of routine laboratory assessments (eg sodium, potassium, blood urea nitrogen, platelets, serum bicarbonate, etc) but after multivariate analysis only found a hemoglobin of ≤10 g/dL as predictive of RVF.²⁹ While, they did not assess for the significance of natriuretic peptides in their model, others with more complex predictive models of RVF derived through Bayesian analysis have included elevated preoperative natriuretic peptide concentrations as risk factors for both acute and delayed RVF.³⁰

A large single-center study of 189 patients with cfLVAD measured BNP concentration before and 2 weeks after LVAD implant and characterized patients into those who had improvement, or those who did not. Patients who had reductions in BNP post-operatively also had significant decreases in central venous pressure and improvement in creatine and bilirubin, whereas those without BNP reductions did not. This nonresponder group also had higher rates of right ventricular failure (53% versus 39%), longer length of stay, and worse overall survival.³¹

Data from echocardiographic ramp studies provide a potential link between RVF, inadequate LV unloading, and elevated BNP. One group performed routine hemodynamic ramp studies on 17 stable outpatients a median of 16 months post HeartWare (HVAD) implant. In those patients who underwent clinically guided speed adjustments, follow-up echocardiography revealed improvements in noninvasive parameters of RV function, including RV fractional area change and RV longitudinal peak systolic strain along with significant reductions in NT-proBNP (3162–2294 ng/L).³²

While newer generation centrifugal devices have extremely low rates of pump thrombosis, these events continue to have important sequelae, including need for pump exchange and ischemic stroke.⁴ At present, LDH (lactate dehydrogenase) remains a ubiquitous and useful biomarker for assessment of acute pump thrombosis, with better performance than serum-free hemoglobin, but other assays continue to be developed.^33,34

Pump thrombosis is mediated by interactions between the blood-prosthesis interface. A novel assay, the platelet activity state (PAS) assay, evaluates the effect shear-stress exposure on platelet activity and provides information on clotting independent from routine measures of coagulation, with elevated activity indicative of propensity to clot. One study assessed utility of PAS in risk stratification and diagnosis of thromboembolic complications of LVAD therapy in 68 patients with cfLVADs (HMII, HVAD, HMIII) over a median follow-up of 602 days. PAS is reported as a normalized value, with mean levels of 0.48% in healthy volunteers and a median level of 0.45% in patients with LVADs.³⁵ Notably, patients with HMII had higher mean PAS activity than those with centrifugal LVADs, potentially providing a mechanistic link for the decreased thromboembolic complications seen in HMIII.⁴

Overall, 6 patients suffered a thrombotic complication; PAS measured at the time of event was significantly (15-fold) higher than baseline values (6.67% versus 0.45%).³⁵ Of note, only one of these patients (patient with HMII with pump thrombosis) had a significantly elevated LDH, suggesting that PAS is a more sensitive marker of thromboembolic disease. Interestingly, these same patients had elevated preoperative PAS activity (1.9% versus 0.42%) implying that patient-specific predisposition may play a role in thrombosis. Thus, PAS assay could be measured (1) preimplant to help decide antiplatelet strategies and/or (2) after a major bleed to assess if aspirin can be safely discontinued.

Myocardial recovery after LVAD implant is a rare phenomenon³⁶ but remains an important goal in selected patients. The INTERMACS Cardiac Recovery Score (I-CARS) was developed to identify patients with higher likelihood of myocardial recovery and relies primarily on clinical and echocardiographic information. Utilization of biomarkers to augment this score would be of clinical interest.

As mentioned, natriuretic peptide concentrations decrease after LVAD implant, but rarely reach normal. However, in an older study of 17 pfLVADs, patients who had rapid recovery of BNP to normal levels within 1 week of LVAD implant had higher rates of myocardial recovery.³⁷ Similarly, other groups have found that patients who experienced myocardial recovery had significantly lower levels of BNP at 1 month¹⁹ and 3 months after LVAD implant.³⁸ In contrast to these findings, however, a separate study found that neither the concentration of BNP or change of BNP from pre-VAD levels at postoperative day 90 were correlated to 180 day outcomes after multivariable adjustment.³⁹ Finally, the trial used to derive the I-CARS score revealed nonsignificant decreases in BNP levels in those successfully explanted.³⁶

Thus far, few biomarkers in the LVAD population have undergone rigorous investigation or have robust enough evidence to recommend broad application (Table 1). While these observational studies allow for interesting hypotheses, much work needs to be done to solidify a role for routine measurement of biomarkers in management of patients with LVADs. Most critically, the degree of evidence must progress from single-center, observational studies to multicenter, prospective, randomized trials. This can be done by incorporation of biomarkers of critical interest into existing registries, such as INTERMACS, to allow more robust assessment of their utility, both at initial LVAD implant and with subsequent complications. Studies should feature homogenous populations (eg, HMIII patients who are destination-therapy, HVAD patients who are bridged to transplant, etc) to allow appropriate application of results into clinical practice. Ideal trial design should include randomization to a biomarker-guided strategy, as has been done in chronic heart failure,⁴¹ to determine whether such a strategy can (1) improve diagnosis of disease, (2) alter downstream testing, and, ultimately, (3) improve outcomes.

Use of biomarkers, in particular, seems well suited for assessment of LVAD complications. For assessment of RV dysfunction, for example, patients could be randomized to periodic natriuretic assessments versus routine care. Elevations in natriuretic peptides could prompt further workup (echocardiography) and possible changes to clinical management (initiation of pulmonary vasodilators or inotropes, speed adjustments, etc). Natriuretic peptides seem particularly attractive as they are relatively inexpensive and widely available. For thromboembolism, patients could be randomized to periodic screening of PAS, with elevations prompting assessment of possible complications and higher anticoagulation goals. Similarly, after a bleeding episode, patients could be randomized to PAS-guided anticoagulation targets versus routine care and then assessed for a reduction in subsequent bleeding and thrombosis. Until higher-quality observational (prospective, multi-center) or randomized studies can be completed, the role of biomarkers in the management of patients with LVAD remains uncertain.

Acute rejection results in cell death of donor myocardium, releasing donor DNA into recipient circulation. Donor and recipient DNA can be differentiated by unique single nucleotide polymorphisms. Therefore, quantification of donor DNA in recipient circulation can serve as a surrogate of myocardial injury.

A prospective study assessing the ratio of circulating cell free donor DNA (cfdDNA) in 65 patients post-transplant found that ratio of cfdDNA was significantly higher during episodes of rejection and could discriminate between mild and moderate-severe acute cellular rejection (ACR). In this study, a cfdDNA ratio of ≥0.25% yielded an area under the curve of 0.83 with sensitivity of 58% and specificity of 93% for moderate-severe rejection (>2R/3A by ISHLT criteria) or significant antibody-mediated rejection. Unfortunately, the test cannot distinguish antibody-mediated rejection and ACR, mandating follow-up biopsies if positive. cfdDNA was significantly elevated for samples collected up to 5 months before the rejection episode, providing possible lead time to adjust immunosuppression.⁴³ This technology shows promise but is still resource-intensive. Novel technologies using limited single nucleotide polymorphisms with next-generation sequencing are under development and have already been validated in kidney transplantation (AlloSure test).⁴⁴

Gene expression profiling (GEP), on the other hand, has proven efficacy in heart transplant recipients and is commercially available as the AlloMap test. The test relies on the observation that during acute rejection, one can expect a gene expression signature of immune activation and leukocyte trafficking in recipient immune cells.⁴⁵ In the CARGO (Cardiac Allograft Rejection Gene Expression Observational) study, a set of 11 discriminatory genes were identified and assigned a score reflective of degree of activation. A score of <30 effectively ruled-out high grade ACR (≥3A/2R) with negative predictive value of 99.6%.⁴⁵ A rejection screening strategy utilizing GEP was compared with routine endomyocardial biopsies in the IMAGE (invasive monitoring attenuation through gene expression) trial.⁴⁶ In this trial, transplant recipients >6 months from surgery without complications were randomized to undergo rejection screening via GEP or biopsy. Patients who had a GEP score above threshold underwent subsequent biopsy to confirm presence and type of rejection. This threshold was changed during the trial from 30 to 34, as further analysis revealed a higher threshold would maintain negative predictive value while decreasing triggered biopsies. At 2 years, the groups showed no difference in a composite of adverse outcomes (death, re-transplantation, rejection with hemodynamic compromise, or nonspecific graft dysfunction) but did have less biopsies.⁴⁶ Of the 34 episodes of rejection identified by GEP, all but 6 patients had symptoms or imaging evidence of graft dysfunction. In cases with clinical signs of rejection, endomyocardial biopsy remains the test of choice for rapid identification (GEP results can take several days) and initiation of treatment. However, for the asymptomatic patient who has been stable post-transplant, GEP may have a role in reducing the need for endomyocardial biopsy. A GEP-based screening strategy has been validated in a real-world registry with 35 centers and >1500 patients and shows excellent short-term survival with similar negative predictive values (98%) for high grade ACR.⁴⁷

Death of donor myocardial cells results in release of cellular contents and myocardial stress, making the ubiquitous biomarkers of cTn (cardiac troponin) and NT-proBNP of interest. One study analyzed stored serum samples in 98 transplant patients and retrospectively matched them to endomyocardial biopsies. High-sensitivity (hs) cTnI concentrations were significantly higher in patients with rejection; a threshold of 15 ng/L yielded a sensitivity of 94% and specificity of 60% for diagnosing ≥2R/3A ACR, antibody-mediated rejection, or graft dysfunction requiring treatment.⁴⁸ Quantitatively, levels of cTn are also observed to rise in parallel with worsening rejection severity.⁴⁹ Although it cannot be used post-operatively because of perioperative injury and ischemia, cTn may have similar negative predictive value to GEP for long-term surveillance of stable transplant recipients.

Intraindividual increases in NT-pro-BNP have also been found to be predictive of high grade (≥2R/3A) ACR. In a retrospective single-center study, 10-fold increase in NT-pro-BNP conferred an odds ratio of 27.7 for high-grade rejection.⁵⁰ A combined biomarker approach could potentially improve test characteristics of these individual nonspecific biomarkers. The obvious benefits of troponin and natriuretic peptide testing over cfdDNA or GEP are their widespread availability, rapid turnaround, and lower cost.

Cardiac allograft vasculopathy (CAV) is a common cause of late morbidity and mortality⁴² in heart transplant recipients, but screening can be invasive and resource intensive.⁵¹ Noninvasive biomarkers would be beneficial if they could help guide frequency of CAV assessment or subsequent therapy.

Several biomarkers of inflammation have been investigated for their utility as diagnostic markers of CAV. A single-center prospective study of 150 patients assessed hsCRP (high-sensitivity C-reactive protein) at time of angiography and found that elevated levels of hsCRP and a long-term rise in hsCRP over time were risk factors for development of CAV.⁵² Another single-center prospective study of 101 patients evaluated several markers of the inflammatory milieu in patients with CAV assessed by intravascular ultrasound and virtual histology. On multivariate analysis, CRP>1.5 mg/L, VCAM-1 (vascular cell adhesion molecule) >391 ng/mL, and neopterin (marker of monocyte/macrophage activation) >7.7 nmol/L were independently associated with mean maximal intimal thickness >0.5 mm.⁵³ Finally, elevated levels of CRP, NT-proBNP, and cTnI in the first year post-transplant are risk factors for the development of CAV and subsequent graft failure and mortality.^54,55

The development of GEP and cfdDNA as screening methods for transplant rejection can reduce patient burden associated with repeat procedures and can effectively rule out high degrees of ACR. GEP, in particular, has been tested in prospective trials and validated in large registries and serves as a model for future biomarker strategy trials in the transplant population. Use of GEP passes the bar for clinical recommendation. While no specific marker of CAV has been developed, CRP shows promise in being a useful predictor (Table 2). Utilization of CRP could risk stratify patients to more aggressive angiographic screening, more frequent use of intravascular ultrasound, or perhaps early incorporation of alternative immunosuppression (sirolimus). A trial assessing outcomes based on a biomarker-guided screening approach, however, is needed before universal adoption of such a strategy.