Introduction
Growth differentiation factor-15 (GDF15) is a member of the transforming growth factor beta superfamily known to be up-regulated in response to oxidative stress, tissue injury, and subsequent inflammation.1 Previous studies suggested robust associations between GDF15 and a variety of diseases, including heart failure (HF), cancer, cognitive decline, pulmonary diseases, diabetes, and renal disease.2–6 Its remarkable cardioprotective properties7 and prognostic significance in patients with HF has promoted extensive investiagtions on clinical application as a biomarker.8
GDF15 has also been characterized as a biomarker of aging.9,10 Its levels increase with aging in healthy individuals.11,12 Because age serves as a non-modifiable confounding factor in all cardiovascular diseases, it is difficult to discern whether the overexpression of GDF15 and its prognostic significance are more reflective of chronological aging or disease state in patients with chronic HF (CHF). Although most previous studies accounted for patient age in their statistical modelling, it is uncertain if the prognostic significance of GDF15 persists to a similar extent across a wide age spectrum. Given that HF encompasses a myriad of biological pathways signified by various risk factors and biomarkers across a wide range of ages, the present study was designed to explore the clinical correlates and prognostic significance of GDF15 across the age spectrum of patients from the SUPPORT (supplemental benefit of an angiotensin receptor blocker in hypertensive patients with stable heart failure using olmesartan) trial.13,14
Methods
Patient population
The rationale, designs, and results of the SUPPORT trial have been previously described.13,14 In brief, the SUPPORT trial was a prospective, randomized, open-label blinded endpoint study performed in 17 participating institutions in the Tohoku District of Japan. The trial examined the incremental benefit of olmesartan in hypertensive patients with stable HF treated with angiotensin-converting enzyme (ACE) inhibitors and/or beta-blockers. From October 2006 to March 2010, the trial enrolled a total of 1147 patients aged between 20 and 80 years. Patients were randomized into either a group of 5–10 mg/day of olmesartan (up-titrated to 40 mg/day, if tolerable) or a control group with standard treatment of HF without the use of any angiotensin receptor antagonists in a 1:1 ratio, and they were followed for incidence of cardiovascular endpoints. Baseline characteristics were obtained at the time of enrolment. The trial adheres to the ethical principles described in the Declaration of Helsinki, and all participants provided written informed consent.
In the present post hoc study of the SUPPORT trial, a total of 1044 patients with baseline GDF15 levels at enrolment were studied. Patients were categorized into four age groups: under 50, 50–59, 60–69, and 70–79 years. Demographic and clinical data collected by an anamnestic interview, physical examination, standardized transthoracic echocardiography, and laboratory examinations performed at the time of enrolment were used for analysis.
Biomarker measurements
N-terminal pro-brain natriuretic peptide (NT-proBNP), high-sensitivity troponin T, and GDF15 were retrospectively measured using the blood samples collected and stored at the time of enrolment. Samples collected at enrolment were stored below −20°C at all times. Biomarkers were measured with an electrochemiluminescence sandwich immunoassay using a Cobas analyser (Roche Diagnostics, Indianapolis, IN, USA). The measuring range of GDF15 was between 400 and 20 000 pg/mL. The internal precision assessed with coefficients of variation for repeatability and intermediate precision provided by the vendor were between 1.1–1.4% and 1.8–2.3%, respectively.
Study endpoints
The primary endpoint of the present study was the composite of HF hospitalization or all-cause death. The secondary endpoint was the composite of all-cause death, HF hospitalization, non-fatal myocardial infarction, or non-fatal stroke, namely, the major adverse cardiac events (MACE). The outcome events were adjudicated by the Endpoint Evaluation Committee, which consisted of two cardiologists and neurologists.
Statistical analysis
Baseline characteristics of patients are presented as means ± standard deviations for normally distributed variables and medians with 25th and 75th percentiles for non-normally distributed variables. For categorical variables, the number of patients with percentages is reported. The elevated GDF15 level was defined as ≥1200 pg/mL based on the previous literature,15,16 and its proportion in each age group is presented. Comparisons of covariates across the age groups were examined using the Kruskal–Wallis test. The χ2 test was used for categorical variables.
Two statistical approaches were employed to identify the clinical correlates of GDF15. As a standard approach, a multivariable linear regression with a bidirectional, AIC (Akaike information criteria)-based stepwise variable selection approach was used to determine the clinical correlates of GDF15 in overall and each age group (the stepAIC function from the MASS package in R). Candidates for clinical correlates included age, sex, body mass index (BMI), New York Heart Association (NYHA) class, heart rate, systolic blood pressure, diastolic blood pressure, history of diabetes, dyslipidaemia, ischaemic heart disease, cardiomyopathy, atrial fibrillation, left ventricular (LV) diastolic and systolic dimensions, LV ejection fraction (LVEF), estimated glomerular filtration rate (eGFR), haemoglobin, haemoglobin A1c, the use of ACE inhibitors, beta-blockers, diuretics, and statins, NT-proBNP, troponin, and high-sensitivity C-reactive protein (hsCRP) (all data at enrolment). Using the same covariates, an advanced variable selection approach using LASSO (least absolute shrinkage and selection operator) regression was used to identify the clinical correlates of GDF15 across the continuous age spectrum as a supplemental analysis. The LASSO regression is a prediction modelling approach suited for identifying a parsimonious set of predictor variables using a regularization technique. Candidates for clinical correlates were first examined for their independent associations with GDF15 and further tested for interactions with continuous age.
Multivariable Cox proportional hazard models adjusted for commonly selected covariates from both the stepwise and LASSO regression approaches were used to determine the risk of study endpoints in overall patients. The interaction term between age and GDF15 was tested in the adjusted model. The associations between GDF15 and study endpoints are demonstrated using cubic spline regression. Statistical indices of C-statistics [area under the curve–receiver operating characteristic curve (AUC–ROC)], likelihood ratio, and R2 are presented across the prediction models of the study endpoints. Demographic variables included age, sex, BMI, NYHA class, systolic blood pressure, diastolic blood pressure, history of diabetes, ischaemic heart disease, end-systolic LV diameter, LVEF, eGFR, haemoglobin, and the use of ACE inhibitors, beta-blockers, diuretics, and statins. The biomarker adjustments were made with log-transformed concentrations of NT-proBNP, troponin, and CRP. The C-statistics (AUC–ROC) comparisons were performed with DeLong's test, and the log likelihood was tested with the likelihood ratio test. The estimated differences in the probabilities of study endpoints between models with and without GDF15 on top of demographics and other biomarkers are demonstrated across the continuous age spectrum. The optimal cut-off values of GDF15 for discriminating between those with and without study endpoints during the follow-up period were estimated based on maximally selected rank statistics in the overall patients. The risks of study endpoints were examined by the determined cut-off level in a Cox regression model adjusted for demographics and other biomarkers. All statistical analyses were performed using R (4.3.2), and the functions and packages used are explained in the supporting information. A two-tailed P-value of <0.05 was considered statistically significant.
Results
Baseline characteristics
Baseline characteristics of the patients are presented in Table 1. The median age was 68 years, and 26% were female. Younger patients aged <50 years had the highest BMI and a high prevalence of dyslipidaemia. In these younger patients, fewer patients had a history of ischaemic heart disease, while the prevalence of cardiomyopathy was the highest across the age groups. In patients aged above 60 years, nearly half of the patients had a history of diabetes. The history of ischaemic heart disease exceeded 50% in age groups beyond 60 years. Although their renal function and haemoglobin levels were lower than those of patients aged below 60 years, they remained relatively preserved. The use of ACE inhibitors and beta-blockers was the lowest in patients aged over 70 years. No statistically significant differences were found for patients with NYHA class III across the age groups. HF phenotypes defined by LVEF were not different across the age groups, with 64% of those having HF with preserved ejection fraction (LVEF ≥ 50%) in overall patients. Serum levels of NT-proBNP, troponin, and GDF15, but not those of hsCRP, increased along with aging (Figure 1). The proportion of patients with higher GDF15 levels [above the established upper limit of normal (1200 pg/mL)] was 51% in overall patients and was the highest in patients in their 70s across the age groups (72%, P < 0.001). The correlations between the biomarkers and age were the strongest for GDF15 (Rho = 0.50) compared with NT-proBNP, troponin, and hsCRP (Supporting Information, Table S1; Rho = 0.25, 0.28, and 0.05, respectively).
Table 1 Baseline characteristics in overall and age groups
| Baseline characteristics | Overall | Age groups | ||||
| Under 50 years | 50–59 years | 60–69 years | 70–79 years | |||
| N | 942 | 73 (7.7%) | 158 (16.8%) | 296 (31.4%) | 415 (44.1%) | - |
| Demographics | ||||||
| Age (years), median [Q1, Q3] | 68.0 [60, 74] | 44 [39, 47] | 56 [54, 58] | 65.5 [63, 80] | 75 [72, 77] | NA |
| Female, n (%) | 243 (25.8) | 13 (17.8) | 31 (19.6) | 78 (26.4) | 121 (29.2) | 0.007 |
| BMI (kg/m2), median [Q1, Q3] | 24.2 [22.1, 26.7] | 26.2 [23.9, 29.4] | 24.6 [21.6, 27.1] | 24.2 [22.3, 26.2] | 23.8 [21.7, 26.1] | <0.001 |
| NYHA class III, n (%) | 66 (7.0) | 4 (5.5) | 8 (5.1) | 16 (5.4) | 38 (9.2) | 0.152 |
| Heart rate (b.p.m.), median [Q1, Q3] | 70 [62, 79] | 71 [62, 80] | 69 [60, 77] | 69 [62, 80] | 70 [61, 79] | 0.631 |
| Systolic BP (mmHg), median [Q1, Q3] | 129 [116, 140] | 129 [117, 140] | 127 [112, 139] | 130 [116, 140] | 128 [116, 142] | 0.511 |
| Diastolic BP (mmHg), median [Q1, Q3] | 73 [67, 81] | 77 [70, 88] | 75 [70, 83] | 74 [68.0, 83.0] | 70 [65, 79] | <0.001 |
| Medical history, n (%) | ||||||
| Diabetes | 449 (47.7) | 28 (38.4) | 73 (46.2) | 147 (49.7) | 201 (48.4) | 0.356 |
| Dyslipidaemia | 490 (52.0) | 33 (45.2) | 82 (51.9) | 172 (58.1) | 203 (48.9) | 0.061 |
| Ischaemic heart disease | 464 (49.3) | 18 (24.7) | 52 (32.9) | 160 (54.1) | 234 (56.4) | <0.001 |
| Cardiomyopathy | 240 (25.5) | 31 (42.5) | 62 (39.2) | 71 (24.0) | 76 (18.3) | <0.001 |
| Atrial fibrillation | 391 (41.5) | 9 (12.3) | 67 (42.4) | 126 (42.6) | 189 (45.5) | <0.001 |
| LV function | ||||||
| HF phenotype | 0.108 | |||||
| HFrEF (EF ≤ 40%) | 161 (17.2) | 16 (21.9) | 25 (15.9) | 54 (18.4) | 66 (16.0) | |
| HFmrEF (EF > 40% and EF < 50%) | 174 (18.6) | 13 (17.8) | 42 (26.8) | 49 (16.7) | 70 (16.9) | |
| HFpEF (EF ≥ 50%) | 602 (64.2) | 44 (60.3) | 90 (57.3) | 191 (65.0) | 277 (67.1) | |
| LVDd (mm), median [Q1, Q3] | 52.0 [47.0, 58.8] | 55.7 [48.0, 61.0] | 54.0 [48.0, 59.0] | 53.0 [47.0, 59.0] | 51.0 [46.0, 58.0] | 0.007 |
| LVDs (mm), median [Q1, Q3] | 36.3 [30.0, 44.0] | 37.9 [31.0, 48.3] | 39.0 [31.6, 45.0] | 36.5 [30.0, 44.0] | 35.3 [29.3, 43.0] | 0.063 |
| LVEF (%), median [Q1, Q3] | 56.0 [45.0, 66.0] | 55.0 [43.0, 66.0] | 51.5 [44.0, 63.1] | 56.0 [45.0, 65.3] | 57.0 [46.6, 66.8] | 0.276 |
| Laboratory values | ||||||
| eGFR (mL/min/1.73 m2), median [Q1, Q3] | 64.1 [52.4, 75.4] | 76.1 [64.7, 90.4] | 76.8 [61.8, 88.9] | 65.7 [53.9, 75.4] | 57.3 [47.0, 71.5] | <0.001 |
| Haemoglobin (g/L), median [Q1, Q3] | 13.8 [12.8, 15.0] | 14.8 [13.7, 15.7] | 14.7 [13.7, 15.6] | 14.1 [13.0, 15.2] | 13.3 [12.3, 14.4] | <0.001 |
| HbA1c (%), median [Q1, Q3] | 5.7 [5.3, 6.2] | 5.5 [5.2, 6.1] | 5.6 [5.3, 6.5] | 5.6 [5.3, 6.2] | 5.7 [5.4, 6.2] | 0.435 |
| NT-proBNP (pg/mL), median [Q1, Q3] | 361 [142, 848] | 178 [60, 327] | 244 [94, 553] | 379 [136, 840] | 480 [202, 1048] | <0.001 |
| Troponin (pg/mL), median [Q1, Q3] | 0.012 [0.007, 0.018] | 0.006 [0.005, 0.012] | 0.009 [0.006, 0.014] | 0.011 [0.008, 0.018] | 0.013 [0.010, 0.020] | <0.001 |
| hsCRP (mg/L), median [Q1, Q3] | 757 [348, 1840] | 673 [378, 1660] | 696 [363, 1403] | 731 [339, 2003] | 794 [345, 2215] | 0.435 |
| GDF15 (pg/mL), median [Q1, Q3] | 1219 [876, 1794] | 691 [531, 1026] | 855 [675, 1158] | 1114 [865, 1671] | 1516 [1163, 2275] | <0.001 |
| Elevated GDF15 (≥1200 pg/mL) (%), n (%) | 479 (51.1) | 10 (13.7) | 36 (22.9) | 138 (46.8) | 295 (71.6) | <0.001 |
| Medications at baseline, n (%) | ||||||
| ARB (olmesartan)* | 466 (50%) | 34 (46.6) | 66 (41.8) | 159 (53.7) | 269 (49.9) | 0.104 |
| ACE-I | 765 (81.2) | 65 (89.1) | 136 (86.1) | 240 (81.1) | 324 (78.1) | 0.045 |
| Beta-blocker | 671 (71.2) | 65 (89.0) | 112 (70.9) | 224 (75.7) | 270 (65.1) | <0.001 |
| Diuretic | 515 (54.7) | 39 (53.4) | 91 (57.6) | 159 (53.7) | 226 (54.5) | 0.871 |
| Statin | 470 (49.9) | 31 (42.5) | 79 (50.0) | 168 (56.8) | 192 (46.3) | 0.025 |
[IMAGE OMITTED. SEE PDF]
Clinical correlates of growth differentiation factor-15 across the age groups
The AIC-based stepwise approach identified age, sex, systolic blood pressure, history of diabetes, dyslipidaemia, ischaemic heart disease, LV end-diastolic and end-systolic diameters, LVEF, eGFR, haemoglobin, NT-proBNP, troponin, CRP, and the use of ACE inhibitors, diuretics, and statins as clinical covariates of GDF15 in overall patients (Table 2). The LASSO regression analysis also selected most of these covariates as clinical correlates of GDF15 that are independent of age (Supporting Information, Table S2, left column). Sex, systolic blood pressure, history of diabetes, ischaemic heart disease, LV end-systolic dimension, LVEF, eGFR, haemoglobin, NT-proBNP, troponin, CRP, and the use of ACE inhibitors, diuretics, and statins were mutually selected in both approaches. In the stepwise approach, eGFR, haemoglobin, and troponin remained significant correlates across all age groups, while other covariates correlated differently with GDF15 across the age groups (Table 2). For example, in patients aged <50 years, increased BMI and a higher prevalence of diabetes were strong additional correlates of GDF15. In middle-aged patients aged 50–59 and 60–69, history of comorbidities such as diabetes, ischaemic heart disease, and dyslipidaemia positively correlated with GDF15. Reduced LV end-systolic dimension and LVEF were associated with increased GDF15 in these groups of patients. In patients aged 70–79 years, the use of diuretics and/or statins appeared as key determinants of GDF15 in addition to robust associations with eGFR and haemoglobin. The LASSO regression analysis, including the interaction terms with continuous age, identified two covariates as correlates of GDF15 with interaction by age: the history of diabetes and NT-proBNP (Supporting Information, Table S2, right column). On the continuous age spectrum, the associations between GDF15 and the history of diabetes and NT-proBNP were particularly robust in patients aged between 60 and 70 years (Figure 2).
Table 2 Clinical correlates of GDF15 across age groups (based on stepwise approach)
| Covariates | Overall | Age groups | ||||||||
| Under 50 years | 50–59 years | 60–69 years | 70–79 years | |||||||
| Age | 7.8 [3.4] | 0.025 | - | - | 20.4 [15.0] | 0.175 | - | - | - | - |
| Sex (female) | −148.2 [77.1] | 0.055 | −263.1 [149.1] | 0.083 | −173.7 [118.0] | 0.143 | - | - | - | - |
| BMI | - | - | 28.5 [11.4] | 0.015 | - | - | −48.8 [15.2] | 0.001 | −20.1 [14.4] | 0.164 |
| NYHA class III | - | - | - | - | 538.6 [178.5] | 0.003 | - | - | - | - |
| Heart rate | - | - | - | - | 3.6 [2.7] | 0.182 | - | - | - | - |
| Systolic BP | −4.4 [1.8] | 0.011 | −7.9 [4.1] | 0.057 | - | - | −11.4 [4.2] | 0.008 | - | - |
| Diastolic BP | - | - | −7.6 [5.8] | 0.197 | −9.8 [3.3] | 0.004 | 15.7 [6.4] | 0.014 | - | - |
| Diabetes | 218.3 [61.2] | <0.001 | 342.6 [120.3] | 0.006 | 151.0 [88.7] | 0.091 | 399.2 [109.0] | <0.001 | 204.7 [98.7] | 0.039 |
| Dyslipidaemia | 530.8 [208.2] | 0.011 | - | - | - | - | 1379.3 [473.0] | 0.004 | - | - |
| Ischaemic heart disease | 186.3 [70.4] | 0.008 | - | - | 186.6 [93.4] | 0.048 | - | - | - | - |
| Cardiomyopathy | - | - | - | - | - | - | - | - | −267.0 [125.9] | 0.035 |
| Atrial fibrillation | - | - | - | - | 189.5 [89.8] | 0.037 | - | - | −262.4 [108.8] | 0.016 |
| LVDd (mm) | 66.0 [12.7] | <0.001 | −50.6 [24.7] | 0.045 | 35.1 [19.4] | 0.073 | 155.1 [20.3] | <0.001 | −15.4 [6.1] | 0.012 |
| LVDs (mm) | −112.7 [16.4] | <0.001 | 70.2 [37.0] | 0.063 | −55.2 [25.6] | 0.033 | −220.1 [25.4] | <0.001 | - | - |
| LVEF (%) | −41.3 [6.4] | <0.001 | 25.8 [15.9] | 0.111 | −20.9 [10.0] | 0.040 | −78.0 [10.2] | <0.001 | - | - |
| eGFR (mL/min/1.73 m2) | −13.2 [1.8] | <0.001 | −14.1 [2.8] | <0.001 | −4.7 [2.2] | 0.030 | −9.1 [3.1] | 0.003 | −22.7 [3.3] | <0.001 |
| Hb (g/L) | −126.4 [19.6] | <0.001 | −198.6 [40.7] | <0.001 | −61.8 [29.7] | 0.004 | −97.2 [32.3] | 0.002 | −99.5 [31.4] | 0.002 |
| HbA1c (%) | - | - | - | - | 47.3 [34.0] | 0.166 | - | - | - | - |
| lnNT-proBNP (pg/mL) | 74.6 [30.9] | 0.016 | - | - | - | - | 80.9 [54.2] | 0.137 | 89.8 [54.0] | 0.097 |
| lnTroponin (ng/mL) | 329.3 [53.5] | <0.001 | 197.1 [73.7] | 0.009 | 215.6 [55.8] | <0.001 | 504.5 [94.5] | <0.001 | 272.4 [98.1] | 0.006 |
| lnhsCRP (mg/L) | 79.2 [22.8] | <0.001 | - | -- | 60.0 [34.5] | 0.084 | - | - | 111.7 [35.1] | 0.002 |
| ACE-I | −128.5 [76.5] | 0.094 | - | - | - | - | - | - | - | - |
| Beta-blocker | - | - | - | - | 193.9 [89.0] | 0.031 | - | - | - | - |
| Diuretic | 258.7 [68.5] | <0.001 | −214.1 [124.9] | 0.092 | - | - | 271.4 [124.4] | 0.030 | 366.3 [105.7] | <0.001 |
| Statin | −669.2 [207.9] | 0.0001 | - | - | - | - | −1403.0 [469.2] | 0.003 | −256.7 [99.4] | 0.010 |
[IMAGE OMITTED. SEE PDF]
Associations with study endpoints across the age groups
During the mean follow-up of 8.6 years (25th–75th percentiles 5.9–11.5 years), a total of 474 primary endpoints (the composite of HF hospitalization and all-cause death) and 517 secondary endpoints (MACE; all-cause death, HF hospitalization, non-fatal myocardial infarction, or non-fatal stroke) occurred. In overall patients, the highest GDF15 tertile (median 2236 pg/mL, 25th–27th percentiles 1797–2800 pg/mL) had a significantly increased risk of the primary and secondary endpoints compared with the lowest tertile of GDF15 {hazard ratio [HR] 3.79 [95% confidence interval (CI) 2.98–4.80], P < 0.001, and HR 3.32 [95% CI 2.65–4.16], P < 0.001, respectively}. GDF15 on the continuous spectrum was also associated with both study endpoints in the fully adjusted model with demographic and clinical correlates identified earlier [Table 3; HR 1.84 (95% CI 1.45–2.33) and HR 1.59 (95% CI 1.26–2.01), respectively]. There was a significant interaction by age on associations between GDF15 and study endpoints (Pinteraction = 0.0008 and 0.0003, respectively). On the continuous age spectrum, the association between GDF15 and outcomes was significant in patients aged 55–60 years or older. The associations between GDF15 and the composite of HF hospitalization or all-cause death appeared the most prominent in patients aged around 70 years (Figure 3). The extent of associations persisted for patients aged after 70 years for MACE.
Table 3 The risk of study endpoints by growth differentiation factor-15 and interaction with age
| Study endpoints | Number of events (per 100 person-years) | Unadjusted model | Adjusted model | |
| HR (95% CI) | HR (95% CI) | |||
| Composite of HF hospitalization or all-cause death | 474 (6.60) | 2.80 (2.40–3.29) | 1.84 (1.45–2.33) | 0.0008 |
| MACE | 517 (7.53) | 2.48 (2.14–2.88) | 1.59 (1.26–2.01) | 0.0003 |
[IMAGE OMITTED. SEE PDF]
The C-statistics (AUC–ROC), likelihood ratio, and R2 were studied across the prediction models for identifying patients with study endpoints (Table 4). The P-values of the pair-wise comparison of AUC–ROC across all models are available in Supporting Information, Table S3. The model with GDF15 added to the fully adjusted model by demographics and other biomarkers (NT-proBNP, troponin, and CRP) showed a marginal improvement of the model performance for HF hospitalization or all-cause death, but not for MACE (AUC–ROC 0.803 vs. 0.796, P = 0.045, and 0.785 vs. 0.781, P = 0.182, respectively). Improvement of the model performance was marginal between the models with and without interaction terms (between GDF15 and age) for both study endpoints (AUC–ROC 0.808 vs. 0.803, P = 0.200, and 0.789 vs. 0.785, P = 0.250, respectively). Estimated differences in the probability of the study endpoints between models with and without GDF15 on top of demographic and other biomarker adjustments appeared similar across a wide range of age spectrums (Supporting Information, Figure S1).
Table 4 Model performance indices across models with different predictors
| AUC–ROC | Log likelihood | ||
| HFH or all-cause death | |||
| (1) Model with demographics | 0.747 | −535.72 | 0.246 |
| (2) Model with demographics + biomarkers | 0.796* | −495.49† | 0.338 |
| (3) Model with demographics + GDF15 | 0.771* | −518.14† | 0.287 |
| (4) Model with demographics + biomarker + GDF15 | 0.803* | −489.97† | 0.350 |
| (5) Model with demographics + biomarker + GDF15 + GDF15 * age | 0.808* | −484.13† | 0.363 |
| MACE | |||
| (1) Model with demographics | 0.728 | −546.06 | 0.211 |
| (2) Model with demographics + biomarkers | 0.781* | −509.71† | 0.298 |
| (3) Model with demographics + GDF15 | 0.749* | −534.52† | 0.239 |
| (4) Model with demographics + biomarker + GDF15 | 0.785* | −507.06† | 0.304 |
| (5) Model with demographics + biomarker + GDF15 + GDF15 * age | 0.789* | −500.65† | 0.319 |
Optimal cut-off value of growth differentiation factor-15
In overall patients, the optimal cut-off value for discriminating patients with and without both study endpoints was 1397 pg/mL. The cut-off value of 1400 pg/mL precisely discriminated between patients with and without study endpoints (Figure 4). In the fully adjusted model with demographics and biomarkers, patients with GDF15 ≥ 1400 pg/mL had significantly higher risks of both endpoints compared with those <1400 pg/mL [adjusted HR 2.01 (95% CI 1.59–2.54), P < 0.001, and HR 1.79 (95% CI 1.43–2.24), P < 0.001, respectively]. There were no interactions between GDF15 ≥ 1400 pg/mL and age for both study endpoints (Pinteraction = 0.384 and 0.383, respectively).
[IMAGE OMITTED. SEE PDF]
Discussion
The participants of the SUPPORT trial reflected the typical HF population in Japan in terms of both demographics and clinical characteristics.14 In this trial, the majority of participants (93%) were in NYHA class II, and their extensive follow-up data have enabled us to better understand their prognostic factors. In this explorative post hoc study in patients with stable CHF, GDF15 was robustly and positively associated with age. Both the classic stepwise regression and the more advanced approach using LASSO regression demonstrated variations in clinical correlates of GDF15, in which the extent of the associations between GDF15 and some clinical correlates is dependent on age. The extent of associations between GDF15 and cardiovascular endpoints may also differ across age, with the most robust associations in patients aged around 70 years. To the best of our knowledge, this is the first study demonstrating the long-term prognostic significance of GDF15 with a special reference to age in CHF.
GDF15 has been well recognized as a biomarker of aging that is associated with various diseases, including HF.17 Therefore, understanding the clinical correlates of elevated GDF15 across a wide age spectrum with different backgrounds and comorbidities is critical for understanding the biological mechanisms involved in elevated GDF15 in CHF. In the present study, consistent with the previous literature, in overall patients with CHF, many clinical covariates such as blood pressure, comorbidity history, LV dimensions and systolic function, renal function, anaemia, and biomarkers including NT-proBNP and troponin were associated with GDF15 independent of age.18–20 The associations with LV dimensions, systolic function, and other biomarkers are consistent with previous studies demonstrating the essential roles of GDF15 in cardiomyopathy and metabolic diseases via cardiac fibrosis and inflammation.21–24 Seminal studies also suggest the vital role of elevated GDF15 in the progression of cardiac remodelling.25,26 The present study also revealed possible variations in the strength of associations across patient ages. In particular, the history of diabetes and NT-proBNP presented robust associations in patients aged between 60 and 70 years. The biological rationale for the difference in the extent of associations between GDF15 and the history of metabolic disease and/or biomarkers of cardiac-specific features across the age spectrum remains inconclusive; however, the present study corroborates the multifaceted involvement of GDF15 in CHF.
The clinical correlates of GDF15 in younger patients appear particularly distinct from other age groups. In addition to the common clinical correlates of impaired renal function, lower haemoglobin, and elevated troponin levels, the classic stepwise variable selection approach identified an increased BMI and prevalent diabetes as key determinants of elevated GDF15 in patients aged under 50 years. This strongly reminds us that GDF15 is not only an aging biomarker but also serves as a metabolic mediator via suppression of food intake and inflammation.27 Furthermore, the association with BMI is extremely intriguing, as it appears to have a non-linear relationship across the age distribution (the directionality of the association is not consistent across the age groups). The variations in the extent of associations between GDF15 and clinical correlates indicate the need for further studies from a multitude of aspects (i.e. sex, comorbidities, underlying aetiology of HF, and treatment strategies) to determine the multifaceted yet essential roles of GDF15 in HF.
The present results further support the prognostic utility of GDF15 in addition to other established risk factors in CHF, with the most prominent impact in patients aged 60–70 years. The absence of a profound difference in the additive value of GDF15 across the age spectrum suggests the use of a universal cut-off value, regardless of age, which would be particularly useful in the clinical setting. Nevertheless, careful consideration should be given to its potential use as a diagnostic marker, as we demonstrate substantial age-dependent differences in GDF15 concentration in patients with CHF. Notably, the median level of GDF15 in patients over 70 years was more than twice as high as that of those under 50 years in the present study. Importantly, these differences in GDF15 levels were observed regardless of HF severity (i.e. NYHA class) or LVEF across the age groups. These findings exhibit a high burden in utilizing a universal GDF15 level as a diagnostic tool for HF across a wide age spectrum. Thus, more attention should be paid to the age-specific cut-off values of GDF15 before it is utilized for diagnostic purposes.
The present study embraces several strengths and future perspectives. Establishing the prognostic significance of GDF15 in non-elderly patients with modest frailty is important while a clinical trial of ponsegromab, a monoclonal antibody of GDF15, is underway targeting patients with HF and evidence of cachexia (ClinicalTrials.gov Identifier: NCT05492500). In the present study, we were able to demonstrate the prognostic significance of GDF15 in CHF, in addition to that in ischaemic heart disease.28 The prognostic cut-off level of GDF15 (1400 pg/mL) found in the present study was lower than the level (1800 pg/mL) reported in a recent meta-analysis of eight trials with patients with ischaemic heart disease.29 Furthermore, the present study also provides novel evidence on the long-term prognostic significance of GDF15 in CHF, which should be useful in future practice.
Several limitations should be mentioned for the present study. First, the general interpretation of the findings is largely limited by the ethnicity and age of the studied patients, wherein the SUPPORT trial patients were Japanese patients aged between 20 and 80 years. Although the mean age of overall patients (66 ± 10 years) aligns with previous studies performed and used in the individual meta-analysis,29 the majority of the patients were aged between 60 and 79 years. Second, it is important to note that patients in the SUPPORT trial were not on any angiotensin receptor blockers at baseline. Potential statistical power loss due to substantial differences in the number of patients across the age groups may have resulted in the discrepancy of the suggested clinical correlates between the stepwise and LASSO approaches. Further studies are needed to validate our findings and corroborate the interaction by age on associations between GDF15 and clinical correlates such as diabetes and NT-proBNP. Third, in comparison with previous research,30,31 GDF15 levels were relatively lower for patients with HF across all age groups in the present study. This may reflect the unique patient characteristics of this cohort, consisting of patients with stable conditions with intensive treatment with either or both ACE inhibitors and beta-blockers. Fourth, although the suggested cut-off value (i.e. 1400 pg/mL) is comparable to the values suggested in the previous literature, the lack of external validation in the present study limits the generalizability of the finding. Above all, the use of data-driven analytical approaches (either the LASSO regression or the cubic spline models) is at risk of overfitting and therefore requires careful interpretation of the findings and validation with external sources. Finally, because the GDF15 measurement was retrospectively performed with stored blood samples from the time of enrolment to the trial, the measurement levels may have been influenced by the sample stability over time, despite appropriate measures taken to minimize such an impact (i.e. proper storage).
Conclusions
GDF15 was associated with a variety of clinical correlates with possible effect modification by age. GDF15 is a powerful determinant of cardiovascular endpoints in addition to other established risk factors, particularly in patients aged around 70 years. The additive prognostic value of GDF15 appeared consistent across the age spectrum, suggesting the use of a universal cut-off level.
Acknowledgements
We would like to acknowledge the contributions of all site investigators and clinical co-ordinators.
Conflict of interest
H.S. has received lecture fees from Bayer Yakuhin, Ltd. (Osaka, Japan) and Daiichi Sankyo Co., Ltd. (Tokyo, Japan). The remaining authors have nothing to disclose.
Funding
This study was supported in part by a grant from Roche Diagnostics K.K. Funding to pay the Open Access publication charges for this article was provided by the author.
Wollert KC, Kempf T, Wallentin L. Growth differentiation factor 15 as a biomarker in cardiovascular disease. Clin Chem 2017;63:140‐151. doi:
Siddiqui JA, Pothuraju R, Khan P, Sharma G, Muniyan S, Seshacharyulu P, et al. Pathophysiological role of growth differentiation factor 15 (GDF15) in obesity, cancer, and cachexia. Cytokine Growth Factor Rev 2022;64:71‐83. doi:
Chai YL, Hilal S, Chong JPC, Ng YX, Liew OW, Xu X, et al. Growth differentiation factor‐15 and white matter hyperintensities in cognitive impairment and dementia. Medicine (Baltimore) 2016;95: [eLocator: e4566]. doi:
Lankeit M, Kempf T, Dellas C, Cuny M, Tapken H, Peter T, et al. Growth differentiation factor‐15 for prognostic assessment of patients with acute pulmonary embolism. Am J Respir Crit Care Med 2008;177:1018‐1025. doi:
Adela R, Banerjee SK. GDF‐15 as a target and biomarker for diabetes and cardiovascular diseases: A translational prospective. J Diabetes Res 2015;2015: [eLocator: 490842]. doi:
Nair V, Robinson‐Cohen C, Smith MR, Bellovich KA, Bhat ZY, Bobadilla M, et al. Growth differentiation factor‐15 and risk of CKD progression. J Am Soc Nephrol 2017;28:2233‐2240. doi:
Ago T, Sadoshima J. GDF15, a cardioprotective TGF‐beta superfamily protein. Circ Res 2006;98:294‐297. doi:
Arkoumani M, Papadopoulou‐Marketou N, Nicolaides NC, Kanaka‐Gantenbein C, Tentolouris N, Papassotiriou I. The clinical impact of growth differentiation factor‐15 in heart disease: A 2019 update. Crit Rev Clin Lab Sci 2020;57:114‐125. doi:
Tanaka T, Biancotto A, Moaddel R, Moore AZ, Gonzalez‐Freire M, Aon MA, et al. Plasma proteomic signature of age in healthy humans. Aging Cell 2018;17: [eLocator: e12799]. doi:
Liu H, Huang Y, Lyu Y, Dai W, Tong Y, Li Y. GDF15 as a biomarker of ageing. Exp Gerontol 2021;146: [eLocator: 111228]. doi:
Kempf T, Horn‐Wichmann R, Brabant G, Peter T, Allhoff T, Klein G, et al. Circulating concentrations of growth‐differentiation factor 15 in apparently healthy elderly individuals and patients with chronic heart failure as assessed by a new immunoradiometric sandwich assay. Clin Chem 2007;53:284‐291. doi:
Welsh P, Kimenai DM, Marioni RE, Hayward C, Campbell A, Porteous D, et al. Reference ranges for GDF15, and risk factors associated with GDF15, in a large general population cohort. Clin Chem Lab Med 2022;60:1820‐1829. doi:
Sakata Y, Nochioka K, Miura M, Takada T, Tadaki S, Miyata S, et al. Supplemental benefit of an angiotensin receptor blocker in hypertensive patients with stable heart failure using olmesartan (SUPPORT) trial—Rationale and design. J Cardiol 2013;62:31‐36. doi:
Sakata Y, Shiba N, Takahashi J, Miyata S, Nochioka K, Miura M, et al. Clinical impacts of additive use of olmesartan in hypertensive patients with chronic heart failure: The supplemental benefit of an angiotensin receptor blocker in hypertensive patients with stable heart failure using olmesartan (SUPPORT) trial. Eur Heart J 2015;36:915‐923. doi:
Wollert KC, Kempf T, Peter T, Olofsson S, James S, Johnston N, et al. Prognostic value of growth‐differentiation factor‐15 in patients with non‐ST‐elevation acute coronary syndrome. Circulation 2007;115:962‐971. doi:
Bonaca MP, Morrow DA, Braunwald E, Cannon CP, Jiang S, Breher S, et al. Growth differentiation factor‐15 and risk of recurrent events in patients stabilized after acute coronary syndrome: Observations from PROVE IT‐TIMI 22. Arterioscler Thromb Vasc Biol 2011;31:203‐210. doi:
Conte M, Giuliani C, Chiariello A, Iannuzzi V, Franceschi C, Salvioli S. GDF15, an emerging key player in human aging. Ageing Res Rev 2022;75: [eLocator: 101569]. doi:
Anand IS, Kempf T, Rector TS, Tapken H, Allhoff T, Jantzen F, et al. Serial measurement of growth‐differentiation factor‐15 in heart failure: Relation to disease severity and prognosis in the Valsartan Heart Failure Trial. Circulation 2010;122:1387‐1395. doi:
Chan MM, Santhanakrishnan R, Chong JP, Chen Z, Tai BC, Liew OW, et al. Growth differentiation factor 15 in heart failure with preserved vs. reduced ejection fraction. Eur J Heart Fail 2016;18:81‐88. doi:
Sharma A, Stevens SR, Lucas J, Fiuzat M, Adams KF, Whellan DJ, et al. Utility of growth differentiation factor‐15, a marker of oxidative stress and inflammation, in chronic heart failure: Insights from the HF‐ACTION study. JACC Heart Fail 2017;5:724‐734. doi:
Xu X, Li Z, Gao W. Growth differentiation factor 15 in cardiovascular diseases: From bench to bedside. Biomarkers 2011;16:466‐475. doi:
Lok SI, Winkens B, Goldschmeding R, van Geffen AJ, Nous FM, van Kuik J, et al. Circulating growth differentiation factor‐15 correlates with myocardial fibrosis in patients with non‐ischaemic dilated cardiomyopathy and decreases rapidly after left ventricular assist device support. Eur J Heart Fail 2012;14:1249‐1256. doi:
Ho JE, Mahajan A, Chen MH, Larson MG, McCabe EL, Ghorbani A, et al. Clinical and genetic correlates of growth differentiation factor 15 in the community. Clin Chem 2012;58:1582‐1591. doi:
Wong CM, Hawkins NM, Petrie MC, Jhund PS, Gardner RS, Ariti CA, et al. Heart failure in younger patients: The Meta‐analysis Global Group in Chronic Heart Failure (MAGGIC). Eur Heart J 2014;35:2714‐2721. doi:
Santhanakrishnan R, Chong JP, Ng TP, Ling LH, Sim D, Leong KT, et al. Growth differentiation factor 15, ST2, high‐sensitivity troponin T, and N‐terminal pro brain natriuretic peptide in heart failure with preserved vs. reduced ejection fraction. Eur J Heart Fail 2012;14:1338‐1347. doi:
Wang F, Guo Y, Yu H, Zheng L, Mi L, Gao W. Growth differentiation factor 15 in different stages of heart failure: Potential screening implications. Biomarkers 2010;15:671‐676. doi:
Wang D, Day EA, Townsend LK, Djordjevic D, Jørgensen SB, Steinberg GR. GDF15: Emerging biology and therapeutic applications for obesity and cardiometabolic disease. Nat Rev Endocrinol 2021;17:592‐607. doi:
Schopfer DW, Ku IA, Regan M, Whooley MA. Growth differentiation factor 15 and cardiovascular events in patients with stable ischemic heart disease (The Heart and Soul Study). Am Heart J 2014;167:186‐192.e1. doi:
Kato ET, Morrow DA, Guo J, Berg DD, Blazing MA, Bohula EA, et al. Growth differentiation factor 15 and cardiovascular risk: Individual patient meta‐analysis. Eur Heart J 2023;44:293‐300. doi:
Gaggin HK, Szymonifka J, Bhardwaj A, Belcher A, De Berardinis B, Motiwala S, et al. Head‐to‐head comparison of serial soluble ST2, growth differentiation factor‐15, and highly‐sensitive troponin T measurements in patients with chronic heart failure. JACC Heart Fail 2014;2:65‐72. doi:
Gürgöze MT, van Vark LC, Baart SJ, Kardys I, Akkerhuis KM, Manintveld OC, et al. Multimarker analysis of serially measured GDF‐15, NT‐proBNP, ST2, GAL‐3, cTnI, creatinine, and prognosis in acute heart failure. Circ Heart Fail 2023;16: [eLocator: e009526]. doi:
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2024. This work is published under http://creativecommons.org/licenses/by-nc/4.0/ (the "License"). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Aims
Growth differentiation factor‐15 (GDF15), a cytokine in the transforming growth factor family, is up‐regulated in stress and inflammatory conditions and is elevated in patients with heart failure (HF). However, the age‐specific attributes and prognostic significance of GDF15 across age remain unknown in chronic HF (CHF).
Methods and results
Serum levels of GDF15 were examined in 942 hypertensive patients (median 68 years) with CHF from the SUPPORT trial across the four age groups [under 50 (n = 73), 51–59 (n = 158), 60–69 (n = 296), and 70–79 years (n = 415)] and in the continuous spectrum. Clinical correlates of GDF15 were explored using the classic stepwise and LASSO (least absolute shrinkage and selection operator) regression approaches. Interaction terms with age were tested in the LASSO regression approach. The associations with the composite outcome of HF hospitalization or all‐cause death were investigated across ages. Median GDF15 levels (pg/mL) increased along with aging, from 691 in under 50 years to 855 in 51–59 years, 1114 in 60–69 years, and 1516 in 70–79 years (trend P < 0.001). Age, sex, systolic blood pressure, history of diabetes, ischaemic heart disease, left ventricular (LV) end‐systolic dimension, LV ejection fraction, estimated glomerular filtration rate, haemoglobin, N‐terminal pro‐brain natriuretic peptide (NT‐proBNP), troponin, C‐reactive protein, and the use of angiotensin‐converting enzyme inhibitors, diuretics, and statins were mutually selected as clinical covariates of GDF15. The LASSO regression analysis identified significant interactions between age and the history of diabetes and NT‐proBNP, with particularly robust associations in patients aged between 60 and 70 years. During the mean follow‐up of 8.6 years, 474 composite endpoints of HF hospitalization or death occurred. GDF15 was associated with a higher risk of HF hospitalization or all‐cause death [adjusted hazard ratio 1.84 (95% confidence interval 1.45–2.33)], with a particularly heightened risk in patients aged around 70 years (Pinteraction = 0.0008). The model with GDF15 on top of other established risk factors yielded marginally higher C‐statistics compared with the model without GDF15 (0.803 and 0.796, P = 0.045). The additive value of GDF15 on top of other established risk factors appeared similar across ages. A universal cut‐off value of 1400 pg/mL performed well in discriminating between those with and without HF hospitalization or death.
Conclusions
Some clinical correlates of GDF15 have an interaction with age. GDF15 is an important determinant of cardiovascular endpoints, particularly in patients aged around 70 years. The additive value of GDF15 appeared consistent across ages, suggesting the use of a universal cut‐off value.
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
Details
; Nochioka, Kotaro 2
; Sakata, Yasuhiko 3 ; Nishimura, Kunihiro 4 ; Shimokawa, Hiroaki 5 ; Yasuda, Satoshi 2 1 Department of Biostatistics, National Cerebral and Cardiovascular Center, Osaka, Japan
2 Department of Cardiovascular Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan
3 Department of Clinical Medicine and Development, National Cerebral and Cardiovascular Center, Osaka, Japan
4 Department of Preventive Medicine and Epidemiology, National Cerebral and Cardiovascular Center, Osaka, Japan
5 Department of Cardiovascular Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan, International University of Health and Welfare Graduate School, Narita, Japan