3.1 Patient characteristics
A total of 3641 patients diagnosed as CAD admitted to our hospital were enrolled in the present study. The median age was 64 years; 56.1% were male, the median GDF-15 level was 1884 ng/L. The enrolled patients were divided into three groups upon the levels of serum GDF-15 (G1: GDF-15<1200ng/L, G2: GDF-15:1200-1800ng/L, G3: GDF-15>1800ng/L). During 6.4 (median follow-up of 6.4 [interquartile range 5.3–7.6]) years of follow-up, 775 patients had an occurrence of MACEs. In those patients, 158 (15.9%) had values of GDF-15 below 1200 ng/L, 134 (17.8%) between 1200 and 1800 ng/L and 483 (25.2%) above 1800 ng/L. The baseline characteristics of the three groups were shown in Table 1.
Patients with a higher level of GDF-15 were older, had a higher level of total cholesterol (TC), had a history of myocardial infarction and percutaneous coronary intervention or coronary artery bypass graft; and had a higher rate of hypertension, hyperlipidemia, and diabetes. There were no differences between patients included and not included in the analysis regarding other background variables: sex, body mass index (BMI), triglyceride (TG), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), smoking and treatment.
Table 1 Baseline clinical and laboratory characteristics of the study patients according to status of GDF-15
|
Total n=3641
|
Low GDF-15
(<1200ng/L)(n=991)
|
Medium GDF-15
(1200-1800ng/L)
(n=750)
|
High GDF-15
(>1800ng/L)
(n=1900)
|
P value for trend
|
Age, years
|
61.4 (27-95)
|
57.2 (26-95)
|
59.9 (30-92)
|
64.3 (27-95)
|
<0.001
|
Male, n (%)
|
2632 (72.29)
|
735 (74.17)
|
540 (53.33)
|
1357 (71.42)
|
0.484
|
BMI (kg/m2)
|
25.64 (13.3-41)
|
26.02 (13.3-41)
|
25.82 (16.5-37.4)
|
25.38 (17.5-32.1)
|
0.074
|
Current smokers, n (%)
|
1668 (45.82)
|
473 (47.72)
|
345 (46.00)
|
850 (44.70)
|
0.304
|
Hypertension, n (%)
|
2370 (65.09)
|
592 (59.74)
|
474 (63.20)
|
1304 (68.63)
|
<0.001
|
Hyperlipidemia, n (%)
|
1120 (30.76)
|
282 (28.5%)
|
230 (30.70)
|
608 (32.0)
|
0.034
|
Diabetes mellitus, n (%)
|
1163 (31.94)
|
223 (22.50)
|
226 (30.13)
|
714 (37.58)
|
<0.001
|
Previous MI, n (%)
|
254 (6.98)
|
54 (5.45)
|
44 (5.89)
|
156 (8.21)
|
<0.001
|
Previous PCI/CABG, n (%)
|
299 (8.21)
|
66 (6.66)
|
57 (7.58)
|
176 (9.26)
|
<0.001
|
TC (mmol/L)
|
4.03±1.0
|
3.97±1.02
|
4.03±1.09
|
4.10±1.14
|
0.046
|
HDL-C (mmol/L)
|
1.07±0.68
|
1.06±0.43
|
1.09±0.94
|
1.07±0.67
|
0.624
|
LDL-C (mmol/L)
|
2.40±0.91
|
2.36±0.84
|
2.44±0.90
|
2.40±0.96
|
0.201
|
TG (mmol/L)
|
1.62±1.21
|
1.65±1.11
|
1.66±1.05
|
1.60±1.32
|
0.326
|
Medications
|
Aspirin, n (%)
|
3415 (93.79%)
|
942 (95.06%)
|
718 (95.73%)
|
1755 (92.37%)
|
0.067
|
ACEI, n (%)
|
1503 (41.28%)
|
403 (40.67%)
|
305 (40.67%)
|
822 (43.26%)
|
0.289
|
β-blocker, n (%)
|
1629 (44.74%)
|
744 (75.08%)
|
533 (71.07%)
|
1352 (71.16%)
|
0.070
|
Statins, n (%)
|
3442 (94.53%)
|
944 (95.25%)
|
716 (95.47%)
|
1782 (93.79%)
|
0.070
|
3.2 Correlations of serum GDF-15 levels with other clinical biochemical factors
Increasing levels of GDF-15 at presentation were associated with increased age, diabetes, hypertension, hyperlipidemia and a history of previous myocardial infarction and previous PCI /CABG. GDF-15 levels were also related to a lower rate of aspirin use (Table 2). By multiple regression analysis that included all patients’ characteristics shown in Table 2, used the natural logarithm of GDF-15 as the dependent variable, GDF-15 was independently associated with age (P<0.001), diabetes (P<0.001).
Table 2 Spearman’s correlation coefficients between GDF-15 and clinical and biochemical parameters
|
spearman correlation
|
|
|
coefficient
|
P
|
|
Age, years
|
0.267
|
<0.001
|
|
Male, n (%)
|
-0.025
|
0.133
|
|
BMI (kg/m2)
|
-0.024
|
0.100
|
|
Current smokers, n (%)
|
-0.025
|
0.127
|
|
Hypertension, n (%)
|
0.079
|
<0.001
|
|
Hyperlipidemia, n (%)
|
0.053
|
0.031
|
|
Diabetes mellitus, n (%)
|
0.137
|
<0.001
|
|
Previous MI, n (%)
|
0.064
|
<0.001
|
|
Previous PCI/CABG, n (%)
|
0.096
|
<0.001
|
|
TC (mmol/L)
|
0.017
|
0.315
|
|
HDL-C (mmol/L)
|
0.007
|
0.656
|
|
LDL-C (mmol/L)
|
0.014
|
0.403
|
|
TG (mmol/L)
|
﹣0.022
|
0.192
|
|
Medications
|
|
|
Aspirin, n (%)
|
-0.057
|
0.001
|
|
ACEI, n (%)
|
0.025
|
0.134
|
|
β-blocker, n (%)
|
-0.031
|
0.050
|
|
Statins, n (%)
|
-0.032
|
0.055
|
|
3.3 Clinical Outcomes
Primary endpoint
A composite of major adverse cardiovascular events was analyzed during follow-up (Figure 1).
In this way, 775 patients had an occurrence of MACEs. Of those patients, 158 (15.9%) had values of GDF-15 below 1200 ng/L, 134 (17.8%) between 1200 and 1800 ng/L and 483 (25.2%) above 1800 ng/L. The MACEs rate was significantly higher in the group of patients with GDF-15 values > 1800 ng/L compared with those with GDF-15 levels between 1200 and 1800 ng/L and patients with GDF-15 values<1200 ng/L (25.2% vs 17.8% vs 15.9%, P<0.001). Kaplan–Meier curve of the incidence of the primary endpoint is presented in Fig. 2a. The incidence of the primary endpoint in the G3 group was significantly higher than that in the G1, G2 group (log-rank P < 0.001). Univariate Cox proportional analyses revealed that GDF-15 values > 1800 ng/L were significantly associated with the incidence of MACEs (unadjusted HR=1.92; 95% CI 1.37-2.52; P<0.001). After adjusted for basic clinical risk factors (age, sex, smoking hypertension, body mass index, diabetes mellitus and hyperlipidemia), in multivariate analysis, GDF-15 values > 1800 ng/L was associated with the incidence of MACEs with an HR of 1.74 (95% CI: 1.44–2.02; P < 0.001) (Table 3).
Table 3 Relation of the GDF-15 level and MACEs in univariate and multivariate survival analysis
Independent Predictors of Major Adverse Cardiac Events
|
|
Univariate Models
|
Multivariate Models
|
|
HR
|
95%CI
|
P
|
HR
|
95%CI
|
P
|
Age
|
1.02
|
1.013-1.026
|
0.00
|
1.02
|
1.009-1.022
|
0.00
|
Sex
|
0.00
|
0.001-0.028
|
0.00
|
0.01
|
0.001-0.057
|
0.00
|
Smoking
|
1.10
|
0.933-1.296
|
0.23
|
﹣
|
﹣
|
﹣
|
BMI
|
0.98
|
0.959-1.000
|
0.05
|
1.00
|
0.977-1.020
|
0.89
|
Hypertension
|
1.03
|
0.646-0.880
|
0.00
|
0.82
|
0.703-0.967
|
0.02
|
Hyperlipidemia
|
1.15
|
0.98-1.34
|
0.09
|
﹣
|
﹣
|
﹣
|
DM
|
0.79
|
0.680-0.911
|
0.00
|
0.89
|
0.767-1.032
|
0.14
|
GDF-15 ≤ 1200ng/L
|
1.05
|
0.87-1.44
|
0.16
|
﹣
|
﹣
|
﹣
|
1200<GDF-15≤1800 ng/L
|
1.09
|
0.90-1.65
|
0.08
|
﹣
|
﹣
|
﹣
|
GDF-15>1800ng/L
|
1.92
|
1.37–2.52
|
<0.001
|
1.74
|
1.44-2.02
|
<0.001
|
Secondary endpoint
During 6.4 years of follow-up (median follow-up of 6.4 [interquartile range 5.3–7.6] years), 275 patients died. Patients with GDF-15 levels<1200 ng/L had a low mortality rate of 3.2% (32). Patients with GDF-15 levels between 1200 and 1800 ng/L had the same mortality rate of 3.2% (24), whereas those with GDF-15 levels>1800 ng/L had a very high mortality rate of 11.3% (219) (P<0.001). Kaplan–Meier curve of the incidence of the primary endpoint is presented in Fig. 2b. The incidence of all-cause death the G3 group was significantly higher than that in the G1, G2 group (P log-rank < 0.001). Univariate Cox proportional analyses revealed that GDF-15 >1800 ng/L were significantly associated with the incidence of all-cause death (table 3). After adjustment for potential confounders, in multivariate analysis, higher GDF-15 values >1800 ng/L were still independently associated with all-cause death (adjusted HR 2.04; 95% CI 1.57–2.61; P<0.001) (Table 4).
Table 4 Relation of the GDF-15 level and all-cause death in univariate and multivariate survival analysis
Independent Predictors of All-cause death
|
|
Univariate Models
|
Multivariate Models
|
|
HR
|
95%CI
|
P
|
HR
|
95%CI
|
P
|
Age
|
1.97
|
1.04-1.12
|
<0.001
|
1.07
|
1.01-1.09
|
0.01
|
Sex
|
1.23
|
0.98-1.37
|
0.09
|
﹣
|
﹣
|
﹣
|
Smoking
|
0.87
|
0.80-1.17
|
0.39
|
﹣
|
﹣
|
﹣
|
BMI
|
0.99
|
0.95-1.02
|
0.48
|
﹣
|
﹣
|
﹣
|
Hypertension
|
0.98
|
0.74-1.28
|
0.87
|
﹣
|
﹣
|
﹣
|
Hyperlipidemia
|
1.54
|
1.38-1.76
|
0.02
|
1.48
|
1.38-1.86
|
0.03
|
DM
|
1.11
|
0.87-1.42
|
0.42
|
﹣
|
﹣
|
﹣
|
GDF-15≤1200ng/L
|
1.11
|
0.87-1.42
|
0.42
|
﹣
|
﹣
|
﹣
|
GDF-15≤1800 ng/L
|
1.33
|
0.94-1.56
|
0.13
|
﹣
|
﹣
|
﹣
|
GDF-15>1800ng/L
|
2.54
|
1.99-3.09
|
<0.001
|
2.04
|
1.57–2.61
|
<0.001
|
3.4 Incremental value of GDF-15 over conventional risk factors for MACEs
We analyzed the predictive value of GDF-15 by ROC curve. For MACEs: ROC curve analyses indicated that C index (area under the curve) was 0.583 (95% CI 0.559–0.606) for clinical model (model1), 0.595 (95% CI 0.594–0.641) for GDF-15 alone, 0.628 (95% CI 0.605–0.651) for clinical model including GDF-15 (model2). ROC curve analysis showed non-significant differences in the clinical model alone compared with GDF-15 alone (P=0.093), however, there was a significant difference compared to the clinical model with GDF-15 (P<0.001, Figure 3a). Furthermore, Model2 showed a significantly improved net reclassification improvement (0.578) and IDI (0.021), compared with model 1 (Figure 3). For all-cause mortality: ROC curve analyses indicated that C index were 0.728 (95% CI 0.694-0.761) for clinical model (model1) 0.766 (95% CI 0.735-0.798) for GDF-15 alone, 0.817 (95% CI 0.787-0.846) for clinical model including GDF-15. ROC curve analysis showed significant differences in the clinical model alone compared to the clinical model with GDF-15 alone (P<0.001), and there was a significant difference compared to the clinical model with GDF-15 (P<0.001,Figure 3b). Moreover, Model2 showed a significantly improved net reclassification improvement (0.629) and IDI (0.035), compared with model 1 (Figure 3).
3.5 Prognostic value of GDF-15 in long-term and short-term
During the follow up of 0-6 months, 80 patients had the occurrence of MACEs and 31 patients died. During the follow up of 6-12 months, 98 patients had the occurrence of MACEs and 33 patients died. During the follow up of 12-72 months, 501 patients had the occurrence of MACEs and 174 patients died (Figure 4a, 4c). And the MACEs rates in 0-6 months, 6-12 months, 12-72 months were 2.20%, 2.68%, 14.47% respectively (Figure 4b). All-cause death rates in 0-6 months, 6-12 months, 12-72 months were 0.85%, 0.91%, 4.86 % (Figure 4d). To evaluate the prognostic value of GDF-15, we also performed the Kaplan–Meier analysis upon different GDF-15 levels. The results indicated that the patients with GDF-15 > 1800 ng/L were significantly associated with an increased risk of all-cause death (log rank P=0.021) in 0-6 months. But for MACEs, no significant difference was seen in 0-6 months (log rank P=0.067). However, by the time of 12 months, the patients with GDF-15 > 1800 ng/L were significantly associated with an increased risk of all-cause death (log rank P<0.001) and MACEs (log rank P<0.001), the results were the same when they were followed up for 6 years (Figure 5).