A and B, Spline plots of associations between biomarkers and death from all causes (unadjusted). C, Strength of association between biomarkers and all-cause death in an adjusted model including all baseline characteristics and all biomarkers. GDF-15 indicates growth differentiation factor-15; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
Biomarkers (log-transformed and standardized) in relation to specific causes of death. Hazard ratios (HRs) reflect upper vs lower quartile adjusted for baseline characteristics. CRP indicates C-reactive protein; GDF-15, growth differentiation factor-15; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
eTable 1. Definitions of causes of death
eTable 2. Median biomarker levels at baseline, and number of patients in whom each biomarker was assessed
Customize your JAMA Network experience by selecting one or more topics from the list below.
Lindholm D, James SK, Gabrysch K, et al. Association of Multiple Biomarkers With Risk of All-Cause and Cause-Specific Mortality After Acute Coronary Syndromes: A Secondary Analysis of the PLATO Biomarker Study. JAMA Cardiol. 2018;3(12):1160–1166. doi:10.1001/jamacardio.2018.3811
Do biomarkers provide prognostic information on cause-specific mortality in patients with acute coronary syndromes?
In this secondary analysis of a randomized clinical trial of 6 biomarkers in 17 095 patients with acute coronary syndromes, N-terminal pro-B-type natriuretic peptide and growth differentiation factor-15 were markers associated with death due to heart failure, as well as from arrhythmia and sudden cardiac death. Growth differentiation factor-15 had the strongest associations with death due to other vascular or nonvascular causes and tended to be associated with death due to bleeding.
N-terminal pro-B-type natriuretic peptide and growth differentiation factor-15 provide prognostic information in patients with acute coronary syndromes, and their measurement may be warranted to identify high-risk patients with possible benefit from more intense secondary prevention measures.
Mortality remains at about 5% within a year after an acute coronary syndrome event. Prior studies have assessed biomarkers in relation to all-cause or cardiovascular deaths but not across multiple causes.
To assess if different biomarkers provide information about the risk for all-cause and cause-specific mortality.
Design, Setting, and Participants
The Platelet Inhibition and Patient Outcomes (PLATO) trial randomized 18 624 patients with acute coronary syndrome to ticagrelor or clopidogrel from October 2006 through July 2008. In this secondary analysis biomarker substudy, 17 095 patients participated.
Main Outcomes and Measures
Death due to myocardial infarction, heart failure, sudden cardiac death/arrhythmia, bleeding, procedures, other vascular causes, and nonvascular causes, as well as all-cause death.
At baseline, levels of cystatin-C, growth differentiation factor-15 (GDF-15), high-sensitivity C-reactive protein, high-sensitivity troponin I and T, and N-terminal pro-B-type natriuretic peptide (NT-proBNP) were determined.
The median (interquartile range) age of patients was 62.0 (54.0-71.0) years. Of 17 095 patients, 782 (4.6%) died during follow-up. The continuous associations between biomarkers and all-cause and cause-specific mortality were modeled using Cox models and presented as hazard ratio (HR) comparing the upper vs lower quartile. For all-cause mortality, NT-proBNP and GDF-15 were the strongest markers with adjusted HRs of 2.96 (95% CI, 2.33-3.76) and 2.65 (95% CI, 2.17-3.24), respectively. Concerning death due to heart failure, NT-proBNP was associated with an 8-fold and C-reactive protein, GDF-15, and cystatin-C, with a 3-fold increase in risk. Regarding sudden cardiac death/arrhythmia, NT-proBNP was associated with a 4-fold increased risk and GDF-15 with a doubling in risk. Growth differentiation factor-15 had the strongest associations with other vascular and nonvascular deaths and was possibly associated with death due to major bleeding (HR, 4.91; 95% CI, 1.39-17.43).
Conclusions and Relevance
In patients with acute coronary syndrome, baseline levels of NT-proBNP and GDF-15 were strong markers associated with all-cause death based on their associations with death due to heart failure as well as due to arrhythmia and sudden cardiac death. Growth differentiation factor-15 had the strongest associations with death due to other vascular or nonvascular causes and possibly with death due to bleeding.
ClinicalTrials.gov Identifier: NCT00391872.
Although several medical and interventional advances have improved outcomes in patients with acute coronary syndromes (ACS), the residual risk of mortality remains at about 5% within a year after a diagnosis of ACS.1-3 Among patients with ACS, there may be several different pathophysiologic processes eventually leading to death. Knowledge about the underlying mechanisms for the specific causes of death and biomarkers reflecting these mechanisms might allow for preemptive treatment. However, most prior studies have focused on the relationship between biomarker levels and deaths from all causes or from cardiovascular diseases overall, to our knowledge.
Different biomarkers reflect different pathophysiological processes, eg, N-terminal pro-B-type natriuretic peptide (NT-proBNP) levels reflect myocardial dysfunction4; troponin levels, myocardial damage; C-reactive protein, inflammation; and cystatin-C, kidney dysfunction. Growth differentiation factor-15 (GDF-15), a marker of oxidative stress, has further emerged as a strong bleeding marker in coronary heart disease and atrial fibrillation.5-8
We hypothesized that the levels of these different biomarkers, indicative of different pathophysiological processes, would provide information about the risk associated not only with all-cause but also with the risk of cause-specific mortality in patients with ACS.
The design and results of the Platelet Inhibition and Patient Outcomes (PLATO) study (ClinicalTrials.gov identifier: NCT00391872) have been published previously.1,9 Briefly, the PLATO trial randomized 18 624 patients presenting with ACS between October 2006 and July 2008 and demonstrated that ticagrelor compared with clopidogrel reduces ischemic events, including mortality, at the expense of an increase in non–coronary artery bypass surgery–related major bleeding. All patients gave their written informed consent to participate, and the trial was approved by ethical review boards and adhered to the Declaration of Helsinki.10 In this study, we included patients who further consented to participation in the preplanned biomarker substudy (N = 17 095).
Blood samples were obtained via direct venipuncture at the time of randomization. This occurred at a median of about 10 hours after admission. Plasma was frozen in aliquots and stored at −70°C until analysis. Levels of high-sensitivity cardiac troponin T (troponin T), GDF-15, and NT-proBNP were determined with sandwich immunoassays on Cobas Analytics Immunoanalyzers (Roche Diagnostics). Cystatin-C, high-sensitivity C-reactive protein, and cardiac troponin I (troponin I) were analyzed on the Architect platform (Abbott Diagnostics).
In the PLATO trial, all deaths were classified as either vascular or nonvascular by the event adjudication committee. As previously described,11 these deaths were subcategorized by 2 independent reviewers blinded to treatment assignment. Cases of major disagreement were resolved in meetings with at least 3 reviewers present (of which 2 were not the original reviewers). In the present study, we assessed death caused by myocardial infarction, heart failure, sudden cardiac death/arrhythmia, bleeding, procedures (mainly coronary artery bypass grafting surgery), other vascular causes, and nonvascular causes—the definitions of which are found in eTable 1 in the Supplement.
Patient characteristics, medical history, invasive management, and final diagnosis are presented with continuous variables shown as medians and interquartile range and categorical variables as number and percentage.
To allow for comparison, all biomarkers were first log-transformed, then centered and standardized to a common scale, such that for a log-transformed biomarker x, each biomarker value xi is transformed:
where x̅ is the mean and σ, the SD of the biomarker.
The associations between biomarkers and all-cause mortality were assessed in unadjusted and adjusted Cox models in which the biomarkers were entered as restricted cubic splines with 4 knots to allow for nonlinear associations. In the adjusted analyses, covariates used were age, sex, ST-elevation myocardial infarction, planned invasive management, diabetes mellitus, hypertension, previous myocardial infarction, previous heart failure, history of peripheral arterial disease, and chronic kidney disease. The results are presented as hazard ratios (HR) and 95% CIs for the upper vs lower quartile. The unadjusted models are also presented as spline plots.
In an additional analysis of all-cause mortality, all biomarkers were added in an adjusted Cox model (adjusting for the covariates above) to assess the effect of each biomarker when all other biomarkers were taken into account. Each biomarker’s strength of association with outcome (ie, X2−df) was calculated in an analysis of variance.
Associations between biomarkers and death due to specific causes were modeled by Cox models, adjusted for the same covariates as for all-cause death, except for procedure-related and bleeding deaths, for which history of peripheral artery disease was excluded from the list of covariates, to avoid overfitting due to the low number of events. The results are presented as a forest plot. All analyses were performed using R, version 3.3.2 (The R Foundation for Statistical Computing). As this is an exploratory and hypothesis-generating study, no adjustments were made for multiple comparisons.
A total of 17 095 patients were included in this biomarker substudy, of whom 6428 (37.6%) had a final diagnosis of ST-elevation myocardial infarction, and 4905 (28.7%) were women (Table 1). The median levels of the biomarkers are found in eTable 2 in the Supplement. There were 782 deaths (4.6%), of which 322 (1.9%) were due to myocardial infarction, 61 (0.4%) of heart failure, 151 (0.9%) of sudden cardiac death/arrhythmia deaths, 23 (0.1%) due to bleeding, 9 (0.05%) procedure related, 120 (0.7%) due to other vascular causes, and 93 (0.6%) due to nonvascular causes.
In the unadjusted analyses, all biomarkers except troponin I were associated with death due to all causes (Table 2 and Figure 1A and B). Most biomarkers remained significantly associated with outcome in the adjusted analyses. The strongest markers of all-cause mortality were NT-proBNP and GDF-15, with adjusted HRs well above 2 when comparing upper vs lower quartile. In a model including baseline characteristics and all biomarkers, NT-proBNP and GDF-15 remained the most important biomarkers (Figure 1C).
The associations between biomarkers and cause-specific mortality are shown in Figure 2. All markers were associated with death due to myocardial infarction (which occurred in 322 patients [1.9%]) with the strongest associations seen for NT-proBNP and GDF-15. For deaths due to heart failure, which were less common (61 [0.4%]), NT-proBNP was the biomarker with the highest hazard ratio (HR, 8.20; 95% CI, 2.60-25.88) for upper vs lower quartile. C-reactive protein, cystatin-C, and GDF-15 were also associated with death due to heart failure, with about a 3-fold increase for upper vs lower quartile for each marker. For sudden cardiac death/death due to arrhythmia, NT-proBNP had the strongest association, with GDF-15 associated with outcome to a lesser extent.
For death caused by bleeding, which occurred in 23 patients (0.1%), there was a possible signal of association between GDF-15 and outcome (HR, 4.91; 95% CI, 1.39-17.43). No such trend could be seen for the other biomarkers. Procedure-related deaths were so rare (9 cases [0.05%]) that no conclusions could be drawn with regard to biomarker associations.
There were associations between all biomarkers except troponins with death due to other vascular causes, with GDF-15 and NT-proBNP having the highest point estimates. For death caused by nonvascular causes, troponin T, NT-proBNP, and GDF-15 showed associations with outcome, with about a 2-fold increased risk at the upper vs lower quartile.
In this study, we assessed 6 different biomarkers in patients with ACS and their association with mortality. We confirmed the strong association between NT-proBNP and all-cause death after an ACS but also for deaths due to heart failure, arrhythmia, and sudden cardiac death. Our study is the first to demonstrate a possible association between GDF-15 and death due to bleeding, to our knowledge.
N-terminal pro-B-type natriuretic peptide is a well-established prognostic biomarker. Increased levels are associated with adverse outcomes in ACS,4,12-14 stable coronary disease,15-18 and heart failure.19 Furthermore, in the pivotal trial of angiotensin-neprilysin inhibition in heart failure (Prospective Comparison of ARNI With an ACE-Inhibitor to Determine Impact on Global Mortality and Morbidity in Heart Failure [PARADIGM-HF]),20 patients were selected based on levels of natriuretic peptides. Elevated NT-proBNP levels post-ACS, when assessed in their clinical context, could trigger further diagnostic workup to identify heart failure, to allow for medical optimization, perhaps even using neprilysin inhibition. In our study, NT-proBNP was not only associated with death due to heart failure, but also due to arrhythmia and sudden cardiac death. This is likely due to the increased risk of ventricular arrhythmia associated with heart failure. Elevated NT-proBNP levels could possibly strengthen the decision for primary prevention with an implantable cardioverter defibrillator to prevent sudden cardiac death, as prior studies have shown a strong association between natriuretic peptide levels and risk of ventricular tachyarrhythmia.21,22
Growth differentiation factor-15, a member of the transforming growth factor β family, is a circulating protein that under physiological conditions seems involved in the regulation of food intake and body weight, acting through the recently identified receptor glial cell line-derived neurotrophic factor family receptor α-like.23-25 While weakly expressed during physiological conditions, GDF-15 is markedly induced in response to oxidative stress,26 aging,27 and inflammation.28 Growth differentiation factor-15 also seems involved in hemostasis by inhibiting platelet integrin activation and thrombus formation in vitro.29 Levels of GDF-15 are associated with fatal and nonfatal cardiovascular events in patients with ACS,12,30 stable coronary disease,31 and in the general population.32 Growth differentiation factor-15 is also a risk indicator of death due to nonvascular causes.33 The present findings corroborate GDF-15 as an important risk indicator for fatal events and that the combined finding of elevated NT-proBNP and GDF-15 levels might constitute an important indicator of need for further protection against fatalities after ACS. Furthermore, GDF-15 has emerged as a strong marker of bleeding in patients with ACS receiving dual antiplatelet therapy5,6 and in patients with atrial fibrillation receiving anticoagulant therapy.7 In addition, if a change in GDF-15 occurs during follow-up after an ACS event, this seems to reflect a change in bleeding risk.6 The present study is in line with this, being the first to our knowledge to find a possible signal between GDF-15 levels and death due to bleeding in ACS although with considerable uncertainty given the small number of deaths due to bleeding in the PLATO trial. Also, the association with death due to other causes and the fact that GDF-15 also is associated with nonfatal cardiovascular events, highlights the difficulty of finding a marker that strictly reflects bleeding risk only. Thereby, the risk information obtained from GDF-15 should be further appraised together with other (possibly multiple) markers and clinical factors, preferably in a clinical risk prediction model, to better separate bleeding from ischemic risk.
The clinical implications of our study, which in contrast to previous studies evaluated associations between biomarkers and specific causes of death, include the use of NT-proBNP as a marker associated with death due to multiple different cardiovascular causes. This could be potentially actionable information, whereby an estimated increased risk of death due to heart failure could instigate additional work-up, medical optimization and closer follow-up, as well as prevention of sudden cardiac death. For GDF-15, the possible signal of association with death due to bleeding is in line with prior results indicating a association between GDF-15 levels and major bleeding in both ACS and atrial fibrillation,5-7 which could trigger preventive measures. However, like NT-proBNP, GDF-15 was also associated with all-cause mortality as well as with death due to myocardial infarction, heart failure, arrhythmia, other vascular causes, and nonvascular causes. While the possible bleeding signal was the main feature distinguishing GDF-15 from NT-proBNP, these other associations should be taken into account, further highlighting the importance of interpreting biomarker values in clinical context. Further study is needed incorporating biomarker determination into clinical decision making to evaluate the potential impact on clinical outcomes.
There are limitations of this study. Although this study involved multiple biomarkers in a large cohort, when subdividing mortality into specific causes, there is a loss of power to detect associations, as was especially evident in the low number of deaths from procedures and from bleeding. Subsequently, a caveat regarding the association between GDF-15 and bleeding is that there is substantial uncertainty about this result (very wide 95% CI). However, this is in line with prior findings of GDF-15 as a bleeding marker. Second, the subclassification of causes of death was performed after the study was completed and after the overall results were known. However, this subclassification was performed by independent assessors blinded to treatment assignment. Third, this study assessed only biomarkers obtained at the time of randomization in the PLATO study. In the acute phase of an ACS, levels of troponins and C-reactive proteins are certainly influenced by the acute illness. It is acknowledged that measurements at other times and that other biomarkers could provide additional prognostic information or attenuate the associations.
In patients with ACS, the baseline levels of NT-proBNP and GDF-15 were markers of all-cause death based on their associations with death due to heart failure as well as from arrhythmia and sudden cardiac death. Growth differentiation factor-15 had the most significant associations with death due to other vascular or nonvascular causes and was possibly associated with death due to bleeding.
Corresponding Author: Daniel Lindholm, MD, PhD, Uppsala Clinical Research Center, Dag Hammarskjölds väg 38, SE-751 85 Uppsala, Sweden (firstname.lastname@example.org).
Accepted for Publication: September 14, 2018.
Published Online: November 14, 2018. doi:10.1001/jamacardio.2018.3811
Author Contributions: Drs Lindholm and Wallentin had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Lindholm, James, Storey, Himmelmann, Cannon, Mahaffey, Steg, Siegbahn, Wallentin.
Acquisition, analysis, or interpretation of data: Lindholm, James, Gabrysch, Storey, Cannon, Steg, Held, Siegbahn, Wallentin.
Drafting of the manuscript: Lindholm.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Lindholm, Gabrysch, Wallentin.
Obtained funding: James, Himmelmann, Wallentin.
Administrative, technical, or material support: James, Himmelmann, Cannon.
Supervision: Himmelmann, Cannon, Mahaffey, Siegbahn, Wallentin.
Conflict of Interest Disclosures: Dr Lindholm reports institutional grants from AstraZeneca during the conduct of the study, institutional grants from GlaxoSmithKline outside the submitted work, and consultancy fee to the institution from AstraZeneca outside the submitted work. Dr James reports institutional research grants, honoraria, and consultancy/advisory board fees from AstraZeneca during the conduct of the study; grants and personal fees from Bayer; personal fees from Sanofi; grants and personal fees from The Medicines Company; and consultancy/advisory board fees from Janssen Pharmaceuticals outside the submitted work. Dr Gabrysch reports institutional grants from AstraZeneca during the conduct of the study. Dr Storey reports grants and personal fees from AstraZeneca during the conduct of the study; personal fees from Actelion; grants and personal fees from PlaqueTec; personal fees from Bayer; personal fees from Bristol-Myers Squibb/Pfizer; grants, personal fees, and other support from AstraZeneca; personal fees from Avacta; personal fees from Novartis; personal fees from Idorsia; personal fees from Thromboserin; and consultancy fees from The Medicines Company outside the submitted work. Dr Himmelmann reports grants and personal fees from AstraZeneca and was an employee AstraZeneca during the conduct of the study. Dr Cannon reports grants and personal fees from Amgen; personal fees from Amarin; grants and personal fees from Boehringer Ingelheim; grants and personal fees from Bristol-Myers Squibb; grants from Daiichi Sankyo; grants from Janssen Pharmaceuticals; grants and personal fees from Merck; personal fees from Alnylam; personal fees from Amarin; personal fees from Kowa; personal fees from Pfizer; personal fees from Eisai Co Ltd; personal fees from Sanofi; personal fees from Regeneron outside the submitted work; grants and consultancy/advisory board fees from Arisaph; grants and consultancy/advisory board fees from Takeda; consultancy/advisory board fees from AstraZeneca, consultancy/advisory board fees from GlaxoSmithKline; and consultancy/advisory board fees from LipoMedix outside the submitted work. Dr Mahaffey reports grants from Afferent Pharmaceuticals, Amgen, Apple, AstraZeneca, Cardiva Medical Inc, Daiichi Sankyo, Ferring Pharmaceuticals, Google (Verily), Johnson & Johnson, Luitpold Pharmaceuticals, Medtronic, Merck, Novartis, Sanofi, St Jude, and Tenax during the conduct of the study; personal fees from Ablynx, AstraZeneca, Baim Institute, Boehringer Ingelheim, Bristol-Myers Squibb, Cardiometabolic Health Congress, Elsevier, GlaxoSmithKline, Johnson & Johnson, MedErgy, Medscape, Merck, Mitsubishi, MyoKardia, Novartis, Oculeve, Portola, Radiometer, Springer Publishing, Theravance, University of California, San Francisco, and WebMD; and other support from BioPrint Fitness outside the submitted work. Dr Steg reports personal fees and nonfinancial support from AstraZeneca during the conduct of the study; grants and personal fees from Bayer/Janssen Pharmaceuticals, Merck, Sanofi, and Amarin; personal fees from Amgen, Bristol-Myers Squibb, Boehringer-Ingelheim, Pfizer, Novartis, Regeneron, Eli Lilly and Company, and AstraZeneca; grants and personal fees from Servier outside the submitted work; and speaker/consultancy fees from GlaxoSmithKline outside the submitted work. Dr Held reports grants from GlaxoSmithKline during the conduct of the study and served on the advisory board for AstraZeneca, Bayer, and Boehringer Ingelheim outside the submitted work. Dr Siegbahn reports institutional grants from AstraZeneca and Roche Diagnostics during the conduct of the study; institutional grants from Boehringer Ingelheim, Bristol-Meyers Squibb/Pfizer, and GlaxoSmithKline; and personal fees from Olink Proteomics outside the submitted work. Dr Wallentin reports grants from AstraZeneca and Roche Diagnostics during the conduct of the study; grants from Merck & Co, GlaxoSmithKline, Boehringer Ingelheim, and Bristol-Myers Squibb/Pfizer; personal fees from Abbott outside the submitted work; and has patents EP2047275B1 and US8951742B2 licensed to Roche Diagnostics.
Funding/Support: This research was supported by AstraZeneca, who founded the PLATO trial. Roche Diagnostics supported this research by providing the growth differentiation factor-15 assay for free.
Role of the Funder/Sponsor: AstraZeneca had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation or approval of the manuscript; and decision to submit the manuscript for publication. AstraZeneca did have the opportunity to review, but not prepare or approve, the manuscript. Roche Diagnosits had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Ebba Bergman, PhD, at Uppsala Clinical Research Center, for editorial support. Dr Bergman did not receive compensation outside of her standard salary.