Author + information
- Richard W. Troughton, MBChB, PhD∗ ( and )
- Christopher M.A. Frampton, PhD
- Department of Medicine, Christchurch Heart Institute, University of Otago, Christchurch, New Zealand
- ↵∗Address for correspondence:
Dr. Richard W. Troughton, Department of Medicine, University of Otago, Christchurch, PO Box 4345, Christchurch 8140, New Zealand.
Personalized treatment, namely the ability to target treatments and doses to an individual based on their specific phenotype and risk, has become a holy grail for modern medicine (1). The aim is for patients who are most likely to benefit to receive target doses of a treatment, whereas for those not expected to benefit, unnecessary treatments or side effects from higher dosages may be avoided. For conditions where the disease phenotype is closely linked to a single-gene or -protein defect, a high degree of targeted treatment can be achieved. In contrast, the expression of heart disease in most individuals is complex and influenced by an extensive range of clinical, biochemical, and genomic factors. This is especially true for the heart failure phenotype, which can be a final common pathway for a large array of cardiovascular conditions and is frequently accompanied by significant comorbidity (2).
With modern tools, the phenotypic expression of heart failure in an individual can now be robustly characterized. Detailed assessment and quantitation are now possible for cardiac structure and function, circulating peptides, and genetic markers. There is, additionally, a large body of research supporting our understanding of circulating markers that reflect heart failure pathophysiology and can act as heart failure biomarkers (3). Some biomarkers, such as the B-type natriuretic peptides (BNP) and N-terminal pro–B-type natriuretic peptide (NT-proBNP), have been studied extensively, and we have a thorough understanding of their value in heart failure to guide diagnosis, assess prognosis, and monitor clinical status or guide treatment (4). It is a testament to the value of NT-proBNP and BNP as validated heart failure markers that they are now a criterion regularly included in heart failure studies to ensure a more specific heart failure phenotype.
A more comprehensive phenotyping of individuals, incorporating an array of markers relevant to heart failure pathophysiology, could potentially allow more accurate assessment of an individual’s risk and potential response to therapy (2,5).
In a study in this issue of the Journal, Ouwerkerk et al. (6) address this topic by testing whether biomarkers measured at baseline could be used to identify benefit from treatment for heart failure with reduced ejection fraction (HFrEF). These investigators studied patients with new onset or acute worsening of heart failure, confirmed by elevated BNP or NT-proBNP levels, who were enrolled in the BIOSTAT-CHF (BIOlogy Study to TAilored Treatment in Chronic Heart Failure), a European multicenter observational study. At the time of recruitment, patients were either naïve to evidence-based therapies or were receiving ≤50% of target doses of angiotensin-converting enzyme (ACE) inhibitor/angiotensin receptor blocker (ARB) and beta-blocker (BB). Medications were expected to be titrated to optimal doses during the initial 3 months of follow-up. From baseline plasma samples, investigators measured 161 biomarkers by using conventional assays as well as using more modern techniques that allow a large number of markers to be assayed from a small volume of plasma.
From the original BIOSTAT-CHF cohort of 2,516 patients, the investigators identified 1,802 patients who survived beyond the 3-month medication titration period and who had sufficient baseline biomarker data. Nearly 30% of the study cohort had ACE inhibitor/ARB doses titrated to >50% of target, whereas 18% achieved >50% of BB dose. The statistical analyses used for this data set then extended beyond those familiar to most cardiovascular researchers. The imputation and multivariate regression models with training and validation samples are necessary in this context to show the potential benefits of biomarker-based treatment strategies from an observational data set with limited events and considerable missing data. The multivariate models considered presenting clinical factors, the probability of an individual achieving optimal medication doses and baseline biomarker levels in modeling clinical events. The authors identified differential patterns of biomarker prediction of clinical events depending on whether medications had been successfully up-titrated or not. Notably, many of the biomarkers identified in the current study as predictive of response to therapy or of clinical outcomes, including markers such as NT-proBNP, ST-2, and troponin, have been identified in other studies as independent markers of heart failure outcome (7,8).
The authors then calculated survival probabilities, event rates, and the number of events potentially avoided for hypothetical scenarios depending on whether either ACE inhibitor/ARB or BB was successfully up-titrated. They report that a scenario of successful ACE inhibitor/ARB titration and BB up-titration-to-target in all subjects would prevent 9.8 and 1.3 events per 100 treated patients at 24 months, respectively. Up-titration of ACE inhibitor/ARB titration and BB based on a scenario of a biomarker-based treatment selection model would prevent 9.9 and 4.7 events per 100 treated patients, respectively.
There are significant limitations to the dataset and the analyses that have been performed in this study, and the authors have been frank in acknowledging them. The cohort is modest in size and subject to selection bias, both through the recruitment process and the exclusion of patients who died early or had insufficient data. Additionally, there were very few patients who did not benefit from ACE inhibitor/ARB up-titration, which provides a sample scenario in which biomarkers can offer few insights. The investigators tested a large number of biomarkers, many of which are undoubtedly interrelated and some of which may have been quantified with limited precision. Even the appropriate use of Bonferroni correction of p values to reduce the likelihood chance findings and the use of sparse regression models to avoid model overfitting may not adequately account for spurious or idiosyncratic findings in this particular cohort. Most importantly, and as stated by the authors, treatment in this study was not randomized, and despite appropriate statistical methods to adjust for potential confounding, it is unlikely that the analyses accounted for inherent biases related to lack of randomization to treatment strategies. As a consequence, the assumptions in regard to the estimated outcomes if medications had been titrated to target might not have held true either in real-world situations or in more rigorous randomized controlled trial settings. The authors are correct to consider these scenarios hypothetical. They also appropriately endorse up-titration of guideline-based medications in all patients with HFrEF.
Some of the specifics of the study findings may well be debated, but the findings should be considered as hypothesis generating, and the authors should be commended for stimulating us to consider how the use of biomarker panels may assist in identifying patients who may benefit from heart failure treatments.
How can we advance our understanding of how biomarkers could be used to personalize treatment of heart failure? Performing analyses similar to the current one but in much larger heart failure cohorts would provide greater statistical power and confidence. Ideally such cohorts should include some prospective evaluation of changes in biomarker levels during therapy titration. If possible, a similar analysis of biomarker panels in plasma samples collected from landmark studies of proven therapies now endorsed in guidelines (such as ACE inhibitor/ARB or BB) would also provide important insights in a setting where treatment allocation was randomized. We should encourage the inclusion of pre-specified biomarker substudies, including the potential use of biomarker panels, in pivotal phase 3 studies of new heart failure therapies. Although it would be difficult to perform a prospective randomized controlled trial to test the hypotheses generated by the current study in relation to ACE inhibitor/ARB and BB up-titration, there may be opportunities to examine past studies that produced a neutral result for a new heart failure therapy. Biomarker analysis could potentially identify a cohort for whom benefit might be expected.
There are salutary lessons from studies testing BNP/NT-proBNP-guided treatment of heart failure that should encourage us to pursue the concept of biomarker guided monitoring and treatment, but also highlight to us the high bar that biomarker guided studies need to cross to prove utility beyond intensive guideline based therapy (9,10). Although BNP/NT-proBNP-guided studies have consistently shown that changes in BNP/NT-proBNP predict future clinical outcome with treatment that lowers NP levels being consistently associated with improved clinical outcomes, based on current evidence, heart failure guidelines are not likely to endorse routine use of BNP/NT-proBNP-guided treatment.
Ultimately, a better understanding of specific heart failure phenotypes based at least in part on biomarker profiles such as those described in the current study may lead to more specific treatments that target specific mechanisms of cardiac dysfunction and to more effective targeting of treatment to individuals most likely to benefit.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Troughton has received grants from Roche Diagnostics; and consulting fees from Merck. Dr. Frampton has reported he has no relationships relevant to the contents of this paper to disclose.
- 2018 American College of Cardiology Foundation
- Blaus A.,
- Madabushi R.,
- Pacanowski M.,
- et al.
- Braunwald E.
- Chow S.L.,
- Maisel A.S.,
- Anand I.,
- et al.
- Richards A.M.
- Shah S.J.,
- Kitzman D.W.,
- Borlaug B.A.,
- et al.
- Ouwerkerk W.,
- Zwinderman A.H.,
- Ng L.L.,
- et al.
- Tang W.H.,
- Wu Y.,
- Grodin J.L.,
- et al.
- Felker G.M.,
- Anstrom K.J.,
- Adams K.F.,
- et al.