Biomarkers in heart failure: Work in progress

Assessing prognosis in heart failure poses a challenge in clinical practice. Several factors, including age, diabetes mellitus or renal dysfunction, functional class, and left ventricular ejection fraction, among others, have been identified as predictors of adverse cardiovascular events. None of these is a strong predictor, however, and so intense interest has emerged in the predictive value of biomarkers. Two cardiac biomarkers, specifically B-type natriuretic peptides (BNP), a marker for left ventricular overload, and more recently cardiac troponin T (cTnT), a marker for ongoing myocardial damage, have garnered particular attention. In this context,

the study by Miller and Jaffe

provides us with valuable insights about the role of these biomarkers in risk stratification in ambulatory patients with stable chronic heart failure (HF). The key finding was that elevations of cTnT and BNP, detected at any time during clinical follow-up by serial monitoring, are highly associated with increased risk for death or decompensated HF leading to cardiac transplantation, and that the combined elevations of cTnT and BNP substantially add to risk.¹

What are the implications of these observations? The finding that cTnT predicted mortality and HF decompensation above and beyond clinical predictors in patients without clinically overt HF suggests that ongoing cardiomyocyte damage may contribute to progression of HF, thereby providing valuable insights into the pathophysiology of disease and evolution of HF from its earliest stages to more advanced disease.² Another possibility raised by these and previous findings is that these biomarkers may aid clinical decision making by clarifying diagnosis, prognosis, or response to therapy.³ However, 3 key prerequisites are necessary before the utility of these biomarkers in guiding clinical practice might be considered.³ First, the association between biomarkers and outcomes should be strong and independent. In the study by Miller and Jaffe, there was an approximate 3-fold increase in the adjusted risk of death or cardiac transplantation for those with troponin elevations >0.01 ng/mL or elevated BNP and 5-fold increase in risk for combined elevations. The risk increased at a cTnT cut-off of >0.03 ng/mL (hazard ratio [HR] = 8.26) with addition of BNP not adding significantly to risk (HR = 8.58). These findings are consistent with a strong and independent association between these biomarkers and outcomes. Second, the "discrimination" property of the test characterizes how well these biomarkers identify individuals who will experience an adverse outcome. The standard measure of discrimination is the c statistic, computed from the area under the receiver-operating characteristic (ROC) curve for diagnostic or screening tests. Values for the c statistic range from 0.5 (uninformative test) to 1.0 (perfect discrimination). Miller and Jaffe did not report these estimates for risk models incorporating these biomarkers. Thus, whether these biomarkers provide any improvements in risk discrimination over and beyond conventional risk predictors is not discernible from this study. However, previous studies have reported a modest improvement (of questionable clinical importance) in risk discrimination with these biomarkers. For example, in the Val-HeFT study, c statistic increased from 0.685 (for clinical risk factors alone) to 0.697 (for clinical risk factors plus troponins) and 0.702 (for clinical risk factors plus BNP), representing an increment of 0.012 and 0.017, respectively.² Lastly, do cTnT and BNP measurements improve the "calibration" of existing risk models? Calibration refers to the correspondence between predicted probabilities and actual event rates. Again, this information is not forthcoming in this study.

In conclusion, although the data from Miller and Jaffe provide the first evidence in support of a strong association between cTnT and BNP and clinical outcomes in ambulatory patients with stable chronic HF, substantial uncertainty remains regarding 2 key prerequisites—ie, the potential of these biomarkers to add to risk discrimination or improve calibration of conventional risk models. Until the performance of these biomarkers is fully characterized with respect to these attributes, one may reasonably argue that routine assessment of these biomarkers in risk stratification or guiding management of chronic HF patients may not be warranted. Even excellent performance in observational studies may not be sufficient for these recommendations. As with all observational data, caution is needed in attributing causation to these associations. In this regard, trials of BNP-guided therapy in HF that have yielded mixed results^4,5 provide a cautionary note. Ultimately, multiple prospective clinical trials are necessary to establish the clinical utility of these biomarkers. Until then, and the study of Miller and Jaffe notwithstanding, calls for biomarker-guided strategies for management of HF patients remain premature and not justified.