ABSTRACT
Background
In the EXSCEL trial, exenatide did not reduce major adverse cardiovascular events (MACE), but heterogeneity of benefit and the role of cardiac biomarkers remain uncertain. We evaluated the prognostic value of baseline and 1-year changes in N-terminal pro B-type natriuretic peptide (NT-proBNP) and high-sensitivity cardiac troponin I (cTnI), and whether baseline biomarker concentrations modified exenatide effects.
Methods
EXSCEL randomized 14,752 adults with type 2 diabetes to exenatide 2 mg weekly (EQW) or placebo. In a biomarker cohort, 4,292 participants had serial NT-proBNP or cTnI at baseline and 1 year. Biomarkers were log transformed and Cox models related baseline concentrations and 1-year change to MACE, all-cause mortality (ACM), cardiovascular (CV) death, hospitalization for heart failure (hHF), adjusting for clinical covariates and the alternate biomarker. Treatment interaction was tested with biomarker by treatment terms.
Results
Over median 1,480 days follow-up, 529 MACE, 310 all cause deaths, 193 CV deaths, and 157 hHF events occurred. Baseline NT-proBNP was strongly prognostic (adjusted HR per 1 integer unit 1.63 for MACE, 1.85 for ACM, 2.17 for CV death, and 2.17 for hHF; all P <.001). Baseline cTnI was also prognostic with a nonlinear pattern, with risk rising mainly above the median. Per SD rise in NT-proBNP over 1 year predicted later MACE (HR 1.85) and CV death (HR 2.81; both P <.001). Baseline NT-proBNP didn’t modify treatment effects. Baseline cTnI didn’t modify EQW treatment effect on MACE but lower rates of CV deaths and hHF with EQW were observed at higher cTnI concentrations.
Conclusions
NT-proBNP and cTnI were strong prognostic markers of adverse outcomes in patients with type 2 diabetes and their 1-year increases signaled higher subsequent risk. Baseline cTnI may mark heterogeneity of EQW response, but mortality interactions are hypothesis generating and require confirmation.
A growing body of evidence demonstrates the benefits of glucagon-like peptide1 (GLP-1) receptor agonists in reducing major adverse cardiovascular events (MACE) and other adverse outcomes in people living with type 2 diabetes. However, it remains uncertain whether the treatment effects vary based on the presence or absence of established cardiovascular (CV) disease. In the Exenatide Study of Cardiovascular Event Lowering (EXSCEL) trial, once-weekly exenatide (EQW) resulted in nonsignificant reduction in MACE, compared with placebo. However, in a prespecified subgroup analysis in participants with known CV disease at baseline, EQW resulted in a 10% relative risk reduction in MACE. This underscores the importance of identifying subgroups within the type 2 diabetes population where EQW might offer greater benefits.
Cardiac biomarkers, including N-terminal pro B-type natriuretic peptide (NT-proBNP) and high-sensitivity cardiac troponin-I (cTnI), are commonly used in clinical practice to guide diagnosis and management of various conditions as they reflect underlying cardiac dysfunction and injury. NT-proBNP and cTnI are being increasingly recognized for their roles in predicting risk of future CV events, even among those without apparent CV disease. Given their ability to identify patients at elevated CV risk, NT-proBNP and cTnI may be useful for stratifying benefits from various treatments according to their individual CV risk profiles. While serial biomarker assessment has prognostic value for adverse CV events, the lack of consistent biomarker changes with some effective treatments suggests a need for further investigation to clarify the relationship between biomarker trajectories and clinical outcomes.
This posthoc analysis aimed to evaluate the effects of baseline and longitudinal changes in NT-proBNP and cTnI on CV adverse outcomes in people living with type 2 diabetes enrolled in the EXSCEL trial. Additionally, we sought to explore the effect of EQW on NT-proBNP and cTnI over time and the efficacy of EQW on CV outcomes in relation to baseline NT-proBNP and cTnI.
Methods
Trial conduct and study population
EXSCEL was a double-blind, randomized controlled trial that studied the cardiovascular effects of subcutaneous injection of either 2 mg of EQW or matching placebo in patients with type 2 DM at 687 sites in 35 countries. The trial enrolled a total of 14,752 patients between June 2010 and September 2015 with a median follow-up of 3.2 years. In brief, the study included patients with a hemoglobin A1c (HbA1c) of 6.5%-10%, 73% of whom had a previous CV event, including previous coronary, cerebrovascular, or peripheral vascular events or stenosis. Key exclusion criteria were history of recurrent severe hypoglycemia, an estimated glomerular filtration rate (eGFR) < 30 ml/min/1.73 m 2, or a previous treatment with a GLP-1 receptor agonist. Given the biomarkers were assessed at varying time points throughout the study, we only included patients who had serial assessments of NT-proBNP or cTnI at baseline or 1-year follow up (8,248 samples from 4,292 participants, with 4,129 samples at baseline and 4,119 samples at 1 year). The study protocol was approved by ethics committees at participating trial sites. All patients provided written informed consent.
Biomarker measurement
NT-proBNP (e411, Roche Diagnostics) and cTnI (i1000sr Architect, Abbott) were measured using the manufacturers assays, calibrators and quality control material. Over three concentrations of control, interassay coefficients of variation were ≤10.9% for NT-proBNP and ≤7.8% for cTnI. The limit of detection (lowest reported value) for NT-proBNP was 5 ng/l. There was no sample with a concentration below the limit of detection value of NT-proBNP. The limit detection for cTnI was 1.2 pg/ml, and samples with a concentration less than this were assigned to have a value of 0.6 pg/ml (n = 188). Samples deemed insufficient were set to missing for the corresponding biomarker (n = 15 for cTnI and n = 5 for NT-proBNP). Due to extreme outliers within cTnI measurements, values of 5 standard deviations (SD) above mean (1,199.98 ng/l) were removed from the analysis (n = 10). All NT-proBNP values were included. Biomarker concentrations were log-transformed before all analyses.
Study outcomes
The primary outcome of interest was the EXSCEL original composite outcome of CV death, nonfatal myocardial infarction, or nonfatal stroke (three-component MACE outcome). Secondary outcomes included all-cause mortality (ACM), CV death, and hospitalization for heart failure (hHF). A blinded, independent clinical events classification committee adjudicated all the components of the primary composite outcome and secondary outcomes.
Statistical analyses
Baseline characteristics were summarized according to baseline NT-proBNP and cTnI (stratified as greater than or equal to, or less than the median). Categorical variables are shown as counts and percentages, and continuous variables are shown as means (SD) or medians (Q 1 -Q 3 ), as appropriate. Differences in the baseline characteristics were evaluated with the Pearson χ2 test for categorical variables and the Student’s t-test for continuous variables. To evaluate the changes in biomarker concentrations, we performed a paired t-test on the log-transformed biomarker values at baseline and 1 year, overall as well as in each treatment group. Separately, we ran a linear mixed model with random intercepts for participants, using biomarker concentration as the outcome and including visit, treatment, and an interaction term between them as independent variables.
Cox proportional hazard models were constructed to assess the relationship between baseline NT-proBNP and cTnI values with time to endpoints of interest. We first fitted a univariate Cox regression with each biomarker modeled as a linear term or a restricted cubic spline (RCS) using R package splines with two knots placed at the 33rd and 66th percentiles and compared the model fits. Across all outcomes, cTnI showed a better fit using a RCS (Anova P -value < 3 × 10 −6), whereas an RCS did not improve the model fit for NT-proBNP over a linear term (Anova P -value >.38). Therefore, from here on we modeled NT-proBNP with a linear term while using an RCS for cTnI. Our multivariate model adjusted for age, sex, region, smoking, prior CV events, prior revascularization, history of myocardial infarction, history of cerebrovascular disease, NYHA class, chronic respiratory disease, atrial fibrillation, body mass index, HbA1c, estimated glomerular filtration rate, diastolic blood pressure, systolic blood pressure, hypertension, hyperlipidemia, diabetes duration, and planned treatment. We also included a model that included all above covariates, plus an indicator for whether the other biomarker is above or below their median baseline level.
We repeated these univariate and multivariate Cox models to evaluate the relationship of changes in biomarker values to time to events of interest. Log-transformed biomarkers were scaled using only baseline samples, then these attributes (mean and SD) were used to scale the 1-year samples before the differences were calculated. Observations with events before 1 year (defined by time-to-event < 450 days to be consistent with EXSCEL documentation) were excluded before analyses separately for each outcome of interest. We adjusted for the same set of covariates as in the baseline analysis as well as the baseline biomarker level in all models.
Finally, to determine whether baseline biomarker values affected the time to the outcome of interest differently based on treatment group (EQW vs placebo), we ran Cox models with the baseline biomarker value, treatment group, and an interaction term between them as covariates. The proportional hazards assumption was tested in all Cox models. Significance was considered at nominal P <.05. All analyses were performed using R version 4.5.1.
Results
Participant characteristics
NT-proBNP data at baseline were available in 4,128 participants with a median concentration of 130.9 (66.4-298.6) pg/ml (Supplemental Figure 1). cTnI data at baseline were available in 4,122 participants with a median concentration of 4.3 (2.7-7.6) pg/ml. Participants with NT-proBNP ≥ 130.9 pg/ml were older, had longer duration of diabetes, higher prevalence of CV comorbidities and LV systolic dysfunction at baseline, compared with those with NT-proBNP < 130.9 pg/ml ( Table 1 ). Participants with cTnI ≥ 4.3 pg/ml were older and more likely to be male, had longer duration of diabetes, higher prevalence of CV comorbidities and LV systolic dysfunction at baseline, compared with those with cTnI < 4.3 pg/ml. Characteristics of participants without biomarker data at baseline and 1 year not included in the analysis (n = 10,623) are presented in Supplemental Table 1.
Table 1
baseline characteristics.
| Variables |
NT-proBNP at baseline
(median = 130.9 pg/ml) (n = 4,128) |
cTnI at baseline
(median = 4.3 pg/ml) (n = 4,122) |
||||
|---|---|---|---|---|---|---|
|
≥ median
cohort |
< median
cohort |
P |
≥ median
cohort |
< median
cohort |
P | |
| (n = 2,065) | (n = 2,063) | (n = 2,074) | (n = 2,048) | |||
| Demographics | ||||||
| Age, years | 65.0 ± 8.6 | 58.8 ± 9.1 | <.001 | 64.2 ± 8.9 | 59.6 ± 9.3 | <.001 |
| Male sex, n (%) | 1,257 (60.9) | 1,203 (58.3) | 0.1 | 1,464 (70.6) | 991 (48.4) | <.001 |
| Geographic region, n (%) | <.001 | <.001 | ||||
| Asia Pacific | 143 (6.9) | 213 (10.3) | 174 (8.4) | 181 (8.8) | ||
| Europe | 1,255 (60.8) | 1,334 (64.7) | 1,252 (60.4) | 1,332 (65.0) | ||
| Latin America | 153 (7.4) | 154 (7.5) | 131 (6.3) | 176 (8.6) | ||
| North America | 514 (24.9) | 362 (17.5) | 517 (24.9) | 359 (17.5) | ||
| Race, n (%) | .007 | <.001 | ||||
| Asian | 100 (4.8) | 154 (7.5) | 117 (5.6) | 136 (6.6) | ||
| Black | 30 (1.5) | 25 (1.2) | 34 (1.6) | 21 (1.0) | ||
| Hispanic | 160 (7.7) | 165 (8.0) | 132 (6.4) | 193 (9.4) | ||
| Indian American or Alaska Native | 4 (0.2) | 3 (0.1) | 6 (0.3) | 1 (0.0) | ||
| Native Hawaiian or Other Pacific Islander | 2 (0.1) | 7 (0.3) | 6 (0.3) | 3 (0.1) | ||
| White | 1,769 (85.7) | 1,709 (82.8) | 1,779 (85.8) | 1,694 (82.7) | ||
| Hispanic ethnicity, n (%) | 296 (10.1) | 206 (10.0) | .97 | 189 (9.1) | 225 (11.0) | .051 |
| Randomized to the treatment arm, n (%) | 1,021 (49.4) | 1,039 (50.4) | .58 | 1,048 (50.4) | 1,012 (49.4) | .55 |
| Medical history | ||||||
| Duration of diabetes, years | 13.9 ± 8.6 | 11.4 ± 7.2 | <.001 | 13.8 ± 8.5 | 11.6 ± 7.5 | <.001 |
| Baseline HbA1c, % | 8.1 ± 0.94 | 8.2 ± 0.9 | .006 | 8.1 ± 0.9 | 8.2 ± 0.9 | .33 |
| Prior CV event, n (%) | 1,648 (79.8) | 1,215 (58.9) | <.001 | 1,637 (78.9) | 1,221 (59.6) | <.001 |
| History of myocardial infarction, n (%) | 834 (40.4) | 456 (22.1) | <.001 | 819 (39.5) | 467 (22.8) | <.001 |
| History of cerebrovascular disease, n (%) | 386 (18.7) | 303 (14.7) | .001 | 386 (18.6) | 302 (14.7) | .001 |
| History of heart failure, n (%) | 577 (27.9) | 296 (14.3) | <.001 | 565 (27.2) | 307 (15.0) | <.001 |
| LVEF category, n (%) | <.001 | <.001 | ||||
| Normal (>55%) | 518 (52.9) | 451 (69.1) | 534 (54.0) | 435 (67.7) | ||
| Mild dysfunction (40%-55%) | 334 (34.1) | 185 (28.3) | 331 (33.5) | 186 (28.9) | ||
| Moderate dysfunction (25%-39%) | 101 (10.3) | 14 (2.1) | 95 (9.6) | 20 (3.1) | ||
| Severe dysfunction (<25%) | 27 (2.8) | 3 (0.5) | 28 (2.8) | 2 (0.3) | ||
| Atrial fibrillation/atrial flutter, n (%) | 296 (14.3) | 39 (1.9) | <.001 | 250 (12.1) | 85 (4.2) | <.001 |
| Smoking status, n (%) | <.001 | <.001 | ||||
| Current | 190 (9.2) | 253 (12.3) | 206 (9.9) | 235 (11.5) | ||
| Former | 828 (40.1) | 729 (35.3) | 882 (42.5) | 674 (32.9) | ||
| Body mass index, kg/m 2 | 33.2 ± 6.5 | 33.7 ± 6.3 | .017 | 33.6 ± 6.5 | 33.3 ± 6.3 | .14 |
| Systolic blood pressure, mmHg | 136.8 ± 16.7 | 134.3 ± 14.7 | <.001 | 137.1 ± 16.5 | 134.0 ± 15.0 | <.001 |
| Diastolic blood pressure, mmHg | 77.7 ± 10.6 | 79.4 ± 9.6 | <.001 | 78.0 ± 10.5 | 79.1 ± 9.7 | <.001 |
| NYHA functional class at baseline, n (%) | .073 | .17 | ||||
| 1 | 167 (28.9) | 94 (31.8) | 160 (28.3) | 101 (32.9) | ||
| 2 | 331 (57.4) | 178 (60.1) | 329 (58.2) | 179 (58.3) | ||
| 3 | 75 (13.0) | 24 (8.1) | 73 (12.9) | 26 (8.5) | ||
| 4 | 4 (0.7) | 0 (0.0) | 3 (0.5) | 1 (0.3) | ||
| Baseline diabetes medication use | ||||||
| Insulin therapy, n (%) | 1,024 (49.6) | 867 (42.0) | <.001 | 1,069 (51.5) | 820 (40.0) | <.001 |
| Sulfonylurea, n (%) | 727 (35.2) | 790 (38.3) | .043 | 698 (33.7) | 816 (39.8) | <.001 |
| Metformin, n (%) | 1,412 (68.4) | 1,659 (80.5) | <.001 | 1,427 (68.8) | 1,637 (80.0) | <.001 |
| Thiazolidinedione, n (%) | 93 (4.5) | 76 (3.7) | .21 | 79 (3.8) | 90 (4.4) | .39 |
| DPP-4 inhibitors, n (%) | 262 (12.7) | 339 (16.4) | .001 | 276 (13.3) | 324 (15.8) | .025 |
| Other GLP-1 receptor agonists, n (%) | 1 (0.0) | 0 (0.0) | 1 | 0 (0.0) | 1 (0.0) | .99 |
| SGLT-2 Inhibitors, n (%) | 2 (0.3) | 4 (0.6) | .54 | 5 (0.7) | 1 (0.2) | .28 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
