Can Circulating Biomarkers Identify Heart Failure Patients at Low Risk?

This article refers to ‘Biomarkers and low risk in heart failure? Data from COACH and TRIUMPH’, by W.C. Meijers et al., published in this issue on pages XXX–XXX. Biomarkers have multiple potential applications in disease management. They may aid in diagnosis, prognosis, selection of treatment, monitoring of response to therapy, provision of a selection criterion, and/or a surrogate endpoint in therapeutic trials, and they may even direct attention towards new therapeutic targets. Biomarkers, specifically the brain type cardiac natriuretic peptides (BNP and NT-proBNP) have assumed relevance in the clinical management of heart failure (HF). Data and experience accumulated over more than 20 years have culminated in acceptance of measurements of BNP or NT-proBNP in the diagnosis, prognosis, and guidance of treatment in HF.1, 2 Many other markers provide some independent risk stratification in HF.1 In this issue of the journal, Meijers et al. ask the important question ‘can a marker or markers identify a very low risk group of HF patients among those recently admitted with acutely decompensated HF (ADHF) who are safe for discharge’?3 This is a truly urgent need given the global recognition that readmission and mortality rates in the days following discharge after admissions for ADHF remain unacceptably high.4, 5 The authors make the important point that biomarkers are generally used to define high risk, and little is known about which markers at what thresholds may usefully indicate low risk. Notably, this approach has borne fruit in the management of patients with chest pain presenting to the Emergency Department and requiring ruling out of acute coronary syndromes. In that context, accelerated diagnostic protocols incorporating biomarker (cardiac troponin) measurements have improved triaging to the extent of identifying between 20% and 40% of such presentations as at such low risk that they may be safely discharged within a few hours for continued assessment as necessary on an outpatient basis.6-8 Can this strategy be applied to identify low risk patients admitted with ADHF who are suitable for discharge? Meijers et al. have undertaken the difficult logistic and technical exercise of measuring 29 circulating markers in a subset (592 out of 1023 patients) of participants in the COACH trial which recruited patients admitted with ADHF.3, 9, 10 Such studies enabling head to head comparison of multiple markers are all too rare. The authors have used the novel screening strategy of ranking markers according to the positive predictive value of levels below the 10th percentile within the 592 COACH patients for the absence of death or readmission with ADHF over the 180 days from discharge. This analysis resulted in a striking finding with respect to plasma galectin-3. Remarkably, the 60 or so COACH patients with the lowest decile of galectin-3 measurements in samples obtained pre-discharge (13 ± 10 days after admission with ADHF) incurred no deaths and only 1 (1.7%) episode of admission for recurrent ADHF at 180 days from discharge. This contrasts with 11 (19%) events in those with pre-discharge plasma NT-proBNP below the 10th percentile (626.8 pg/mL) and 8 (14%) events for cardiac troponin I (TnI; Supplementary table S2). In a Cox regression multivariate model incorporating the COACH risk model and NT-proBNP, galectin <11.8 ng/mL remained independently indicative of low risk, with an odds ratio for low risk at 180 days of 7.68 (95% confidence interval 1.04–50; P = 0.045). If 20th and 30th percentiles of markers are inspected (as in Supplementary table S3) rather than the 10th, galectin values remain associated with low event rates with this marker, ranked among the top two or three, with the best performance in this regard along with low circulating erythropoietin and tumour necrosis factor-alphaR1a levels. The exciting implication raised by the report is that a pre-discharge plasma galectin-3 concentration of <11.8 ng/mL may be incorporated into an algorithm identifying ADHF patients at very low risk of adverse events within the subsequent 180 days and therefore fit for discharge. A role for low plasma galectin-3 in reflecting low risk has a good biological plausibility. Early pre-clinical studies highlighted galectin-3 as the most robustly overexpressed gene in failing vs. functionally compensated hearts among transgenic TGRmRen2-27 rats.11 Its candidacy as a novel mediator of adverse cardiac remodelling has been well reviewed, and it is well recognized that galectin-3 is up-regulated in cardiac hypertrophy and that its effects upon macrophage migration and fibroblast proliferation underpin a role in cardiac fibrosis.12 The relationship of plasma and/or tissue galectin-3 to circulating markers of cardiac extracellular matrix turnover has been the subject of mixed results. Although reports generally agree that galectin-3 is increased in cardiac hypertrophy, there is less consistent data with respect to correlations between this marker and circulating procollagens, matrix metalloproteases (MMPs), and tissue inhibitors of metalloprotease (TIMPs).13, 14 However, the relationship of increasing plasma galectin-3 to poorer function and poor prognosis in HF with and without preserved EF is clear.15-18 How firm is the suggestion that low galectin-3 might be used as a particularly strong indicator of low risk? It is important to ascertain whether the marker offers information above that routinely already available, and multivariate analyses are intended to address this issue. The multivariate analysis presented by Meijers et al. incorporates the COACH risk engine as its clinical base model. The risk engine is not comprehensive with respect to inclusion of accepted predictors of risk following an episode of ADHF. For example, the presence or absence of discharge heart rate, anaemia, antecedent hypertension, and discharge prescription of evidence-based anti-HF medications are not considered in this model. Therefore, further testing of the marker in a more comprehensive predictive model is warranted. In addition, NT-proBNP is part of the COACH risk engine but, in the current analyses, it is unclear whether NT-proBNP was included in this model as a continuous variable (as in the originally described COACH risk engine) or dichotomized at the 10th percentile, which might well reduce its predictive performance. Then, are the findings from the COACH analysis generalizable? Proof of generalizability is always a challenge. The authors have appropriately sought to address this by consideration of marker performance in an independent group of 285 HF patients from the TRIUMPH study, a cohort established specifically to explore the utility of candidate biomarkers in HF (NTR 1893; http://www.trtialregister.nl/trialreg/admin/rctview.asp? TC1893). The comparison of COACH and TRIUMPH is partly limited by intercohort differences. The validation group is half the size of the derivation group which compromises the power of subsequent comparisons. Prescription rates of key anti-HF drugs are significantly lower in TRIUMPH (Supplementary table S11), and NYHA I/II patients comprised 47% of COACH patients compared with 73% of TRIUMPH patients (P < 0.001), suggesting the likelihood of a higher proportion of low risk patients in TRIUMPH. Low galectin-3 was still associated with low event rates, but the clear distinction from other markers, such as NT-proBNP and hsTnI, seen in COACH, was as not as striking in TRIUMPH. Galectin-3 did not add significantly to the clinical risk model plus NT-proBNP for identifying low risk in the TRIUMPH cohort. Therefore, findings in the validation cohort do not fully support the discovery highlighted from COACH data and it remains necessary to pursue further validation of the key finding reported by Meijers et al.3 This report should stimulate further rigorous evaluation of galectin-3 in adequately powered validation cohorts. Before clinical application can be mandated, candidate markers must repeatedly pass tests of robust validity including receiver operator analysis, multivariate modelling, and measurement of net reallocation improvement (NRI) in independent discovery and validation cohorts followed by pragmatic controlled trials in real-life clinical settings. A search for markers of low risk to aid optimal timing of discharge after admission for ADHF is important. This report, albeit in need of further corroboration, constitutes an early effort to address a crucial unmet need.3 Conflict of interest: none declared.