Serum Growth Differentiation Factor 15 Levels Are Associated With Depression After Ischemic Stroke.

Background The effect of serum growth differentiation factor 15 (GDF-15) on poststroke depression (PSD) remains unknown. This study aimed to investigate the association between serum GDF-15 and PSD among patients with ischemic stroke. Methods and Results This study was based on a random sample from CATIS (China Antihypertensive Trial in Acute Ischemic Stroke). A total of 572 patients from 7 participating hospitals with GDF-15 levels were included in this analysis. The study outcome was depression (Hamilton Depression Rating Scale score ≥8) at 3 months after ischemic stroke. A total of 231 (40.4%) patients with stroke experienced PSD within 3 months. The multivariate-adjusted odds ratio of PSD associated with the highest tertile of serum GDF-15 was 2.92 (95% CI, 1.36-6.27) compared with the lowest tertile. Each SD increase in log-transformed GDF-15 was associated with a 42% (95% CI, 2%-97%) increased risk of PSD, and a linear association between serum GDF-15 and the risk of PSD was observed (P for linearity=0.006). Conclusions Elevated serum GDF-15 levels in the acute phase of ischemic stroke were independently associated with PSD, suggesting that GDF-15 may be a valuable prognostic biomarker for PSD.

Stroke is the main cause of mortality and long‐term disability and has become an important public health concern worldwide. Among all complications of stroke survivors, poststroke depression (PSD) is the most frequent psychiatric problem. In 2005, a systematic review conducted by Hackett et al estimated that the incidence rate of PSD was up to 33% (95% CI, 29%–36%). The consequences of PSD can extend beyond just mental health; it can also adversely affect rehabilitation and quality of life and increase the risk of death after stroke. The predictive capacities of the traditional risk factors for PSD have been well established, but these established risk factors could not fully explain the development of PSD. Therefore, the identification of novel risk factors to predict PSD early is urgently required to provide clinical evidence for better prevention and intervention of PSD and thus contribute to better stroke outcomes.

Growth differentiation factor 15 (GDF‐15), also known as macrophage inhibitory cytokine‐1, is a distant member of the transforming growth factor β cytokine superfamily. The expression levels of GDF‐15 could significantly increase in response to inflammation and oxidative stress. GDF‐15 is independently associated with coronary atherosclerosis presence, which affects the incidence of ischemic stroke. In patients with ischemic stroke, circulating levels of GDF‐15 appear to be elevated and closely related to poor prognosis, including neurological outcomes. However, studies on the relationship between GDF‐15 levels and depressive symptoms are disputed, and the effect of serum GDF‐15 levels on the risk of PSD remains unclear. Herein, we aimed to investigate the association between serum GDF‐15 levels and PSD among patients from CATIS (China Antihypertensive Trial in Acute Ischemic Stroke).

Methods

Requests for data access may be sent to the corresponding author. Detailed methods are described in Data S1.

Study Design, Participants, and Data Collection

This study was conducted among patients with ischemic stroke from CATIS. Details of the trial design, methods, and data collection are presented in Data S1. The present observational study was based on a preplanned ancillary study, which aimed to investigate whether early antihypertensive treatment would reduce poststroke cognitive impairment and PSD in patients with acute ischemic stroke at 3 months after randomization among a random sample of CATIS. In this ancillary study, 660 CATIS participants were systemically selected before randomization from 7 participating hospitals for cognitive function and psychological state evaluation at their 3‐month follow‐up visit. In the present study, we further excluded 88 participants without GDF‐15 data or available follow‐up. Finally, 572 participants were included in this analysis (Figure S1).

This study was approved by the ethical committees at Soochow University and the institutional review boards at Tulane University. Written consent was obtained from all study participants or their immediate family members. CATIS is registered at clinicaltrials.gov (identifier: NCT01840072).

Outcome Assessment

The study outcome was depression at 3 months after stroke onset, which was assessed by trained neurologists using the validated version of the Hamilton Rating Scale for Depression (HRSD‐24). The HRSD‐24 has been translated into Chinese and confirmed as a screening tool for depression in the Chinese population. It is widely accepted that a total HRSD score of 8 is the cutoff point for diagnosing depressive symptoms. In addition, the severity of depression was categorized as follows: a score of ≤7 indicated the absence of depression, a score between 8 and 19 indicated mild depression, and a score of >20 indicated severe depression.

Statistical Analysis

All participants were categorized into 3 groups according to tertiles of serum GDF‐15 levels. Baseline characteristics among different tertiles of serum GDF‐15 were compared by ANOVA or the Kruskal‐Wallis test for continuous variables and χ2 test for categorical variables. Multivariate logistic regression models were used to assess the association between baseline serum GDF‐15 levels and PSD. ORs and 95% CIs for the upper tertiles of GDF‐15 compared with the lowest tertile, 1‐SD increment of log‐transformed, and raw GDF‐15 level were calculated. The covariates included in the multivariable model were age, sex, education, time from onset to randomization, clinical center, current smoking, alcohol consumption, systolic blood pressure (BP), blood glucose, galectin‐3, body mass index, baseline National Institutes of Health Stroke Scale (NIHSS) scores, medical history (hypertension, hyperlipidemia, and coronary heart disease), family history of stroke, ischemic stroke subtype, and immediate BP reduction. The effect of serum GDF‐15 levels on PSD severity was analyzed using an ordinal logistic regression model adjusted for the aforementioned variables. We further conducted subgroup analyses to assess the robustness of the association between serum GDF‐15 levels and PSD according to other risk factors (Figure S2). We categorized participants into low and high GDF‐15 groups, and the cutoff point (885.68 ng/L) for serum GDF‐15 level was obtained from the receiver operating characteristic curve using Youden index. Interactions between serum GDF‐15 levels and subgroup variables on PSD were tested in the models with interaction terms by the likelihood ratio test, adjusting for the aforementioned covariates unless the variable was used as a subgroup variable. To test the effect of BP variability on PSD, we conducted sensitivity analyses by further including the ratio of systolic BP at 12 hours, 24 hours, 3 days, 7 days, and 3 months after stroke to baseline systolic BP in multivariable‐adjusted models.

Afterward, we evaluated the pattern of the association between serum GDF‐15 and PSD using a logistic regression model with restricted cubic splines. The median of the lowest tertile of serum GDF‐15 (616.03 ng/L) was used as the reference point and 4 knots were placed at the 5th, 35th, 65th, and 95th percentiles of serum GDF‐15. The number of knots used in the cubic spline model was chosen based on maximizing goodness of fit. Furthermore, to assess improvement in the risk prediction of poststroke cognitive impairment, C statistics, net reclassification index (NRI), and integrated discrimination improvement (IDI) were utilized to assess the incremental predictive value of serum GDF‐15 levels in the risk of PSD beyond conventional risk factors. We calculated the relative IDI with reference to Pencina et al. All P values were 2‐tailed, and a significance level of 0.05 was used. Statistical analysis was conducted using SAS statistical software (version 9.4, SAS Institute Inc).

Results

Baseline Characteristics

Detailed results are described in Data S1. Most baseline characteristics were balanced between the participants who did and did not undergo assay (Table S1). Baseline characteristics of patients according to quartiles of GDF‐15 are shown in Table S2.

Association Between Serum GDF‐15 and PSD

At the 3‐month follow‐up, a total of 231 patients (40.4%) had PSD. The serum GDF‐15 level in patients with PSD was higher than that in patients without PSD (median [interquartile range], 1035.58 ng/L [753.89–1481.27 ng/L] versus 896.06 ng/L [679.86–1310.61 ng/L]; P=0.003). Patients in the third tertile of serum GDF‐15 had the highest incidence of PSD (48.2%; P=0.002). After adjustment for confounders, the OR of PSD associated with the highest tertile of serum GDF‐15 was 2.92 (95% CI, 1.36–6.27; P trend=0.006; Table) compared with the lowest tertile. In continuous analysis, a per‐SD increase in raw GDF‐15 and log‐transformed GDF‐15 was associated with 43% (95% CI, 3%–98%) and 42% (95% CI, 2%–97%) increased risk of PSD, respectively. Moreover, multivariate ordinal logistic regression analysis showed a significant association between serum GDF‐15 and PSD severity (OR, 2.09; 95% CI, 1.07–4.10 [P trend=0.030]) (Table and Figure 1). In addition, multivariable‐adjusted spline regression models showed a linear dose‐response association between serum GDF‐15 and PSD (P for nonlinearity=0.140 and P for linearity=0.006, respectively) (Figure 2). In addition, we also found that baseline NIHSS score and clinical center were significantly associated with risk of PSD in the multivariable analysis (Table S3). For more detailed results, please see Data S1 and Table S4.

Table 1

ORs and 95% CIs for the Risk of PSD According to GDF‐15 Tertiles

	GDF‐15, ng/L			P value for trend	Each SD (0.23 ng/L) increase in log‐GDF‐15	Each SD (792.79 ng/L) increase in GDF‐15
	<792.94	792.94–1234.66	≥1234.66
PSD*
Events, n (%)	62 (32.6)	77 (40.3)	92 (48.17)
Unadjusted	1.00	1.39 (0.92–2.12)	1.92 (1.27–2.91)	0.002	1.29 (1.09–1.54)	1.28 (1.06–1.53)
Model 1	1.00	1.35 (0.88–2.07)	1.75 (1.11–2.77)	0.017	1.24 (1.03–1.50)	1.22 (1.01–1.48)
Model 2	1.00	1.63 (0.81–3.28)	2.92 (1.36–6.27)	0.006	1.42 (1.02–1.97)	1.43 (1.03–1.98)
PSD severity †
Unadjusted	1.00	1.32 (0.87–2.00)	1.82 (1.20–2.75)	0.004	1.29 (1.09–1.54)	1.32 (1.12–1.56)
Model 1	1.00	1.27 (0.83–1.95)	1.68 (1.07–2.66)	0.025	1.25 (1.04–1.51)	1.28 (1.08–1.53)
Model 2	1.00	1.21 (0.65–2.26)	2.09 (1.07–4.10)	0.030	1.26 (1.02–1.55)	1.28 (1.01–1.81)

Odds ratios (ORs) are derived from ordinal regression. GDF‐15 indicates growth differentiation factor 15.

Model 1 was adjusted for age, sex, and education level.

Model 2 was adjusted for model 1 and further adjusted for time from onset to randomization, clinical center, current smoking, alcohol consumption, systolic blood pressure (BP), blood glucose, galectin‐3, body mass index, baseline National Institutes of Health Stroke Scale scores, medical history (hypertension, hyperlipidemia, and coronary heart disease), family history of stroke, ischemic stroke subtype, and immediate BP reduction.

Poststroke depression (PSD): Hamilton Rating Scale for Depression (HRSD) score ≥8.

Severity of depression categorized as normal (HRSD score: 0–7), mild depression (HRSD score: 8–19), and severe depression (HRSD score ≥20).

Figure 1

Serum growth differentiation factor 15 (GDF‐15) and poststroke depression severity.

Adjusted odds ratio of ordinal logistic regression analysis for highest vs lowest tertile of serum GDF‐15: 2.09 (95% CI, 1.07–4.10; P value for trend=0.030) for Hamilton Rating Scale for Depression score.

Figure 2

Association of serum growth differentiation factor 15 (GDF‐15) with the risk of poststroke depression.

Odds ratios (ORs) and 95% CIs derived from restricted cubic spline regression, with knots placed at the 5th, 35th, 65th, and 95th percentiles of the distribution of serum GDF‐15. The reference point for serum GDF‐15 is 616.03 ng/L. ORs were adjusted for the same variables as model 2 in Table.

Subgroup and Sensitivity Analyses

High serum GDF‐15 levels were associated with PSD (OR, 1.91; 95% CI, 1.04–3.51 [P=0.038]) (Figure S2) after adjustment for potential confounders. In subgroup analyses stratified by age, sex, education, time from onset to hospitalization, systolic BP, history of hypertension, baseline NIHSS score, body mass index, and receiving immediate BP reduction, ORs of PSD were significant in older participants, patients with higher education, baseline NIHSS score, body mass index, history of hypertension, and receiving immediate BP reduction. No significant interaction between serum GDF‐15 levels and these interesting factors on PSD was observed (all P for interaction >0.05). Sensitivity analyses showed that after considering BP variability at different time points, the significance of the association between GDF‐15 and PSD still exists (Table S5).

Discussion

To the best of our knowledge, this is the first multicenter study to examine the association between baseline serum GDF‐15 concentration and PSD. In this study, we demonstrated a significant association between serum GDF‐15 level and increased risk of subsequent depression at 3 months after acute ischemic stroke, even after adjustment for established risk factors. A multiple‐adjusted spline regression model suggested a dose‐response relationship between serum GDF‐15 and PSD, and subgroup analyses and sensitivity analyses further confirmed these associations. Furthermore, adding serum GDF‐15 levels to conventional risk factors significantly improved the predictive power for PSD. These findings suggest that serum GDF‐15 might be a valuable biomarker in the prediction of PSD, but it still needs to be further replicated by other studies from various populations.

Previous studies have documented that elevated levels of GDF‐15 are significantly implicated in cardiovascular dysfunction and diseases, especially ischemic stroke. , However, little is known about the relationship between baseline serum GDF‐15 levels and PSD. Herein, we conducted this first multicenter study to examine the association between baseline serum GDF‐15 levels and 3‐month PSD. We found that the highest serum GDF‐15 tertile was associated with an ≈2.9‐fold increased odds of subsequent PSD, indicating that serum GDF‐15 might be a valuable marker in predicting PSD onset. Scuteri et al found that depression was associated with greater BP variability, which can increase the risk of arterial damage and accelerated arterial aging, and was an established risk factor for depression. Of note, in our sensitivity analysis, the association between GDF‐15 and PSD was still significant after adjusting for BP variability. Given the high incidence and adverse consequences of PSD, our research has important public health and clinical significance for the early identification and intervention of PSD. According to our results, patients with ischemic stroke who have high serum GDF‐15 levels should receive active monitoring and therapeutic intervention to prevent the occurrence of PSD.

The precise mechanisms underlying the observed association between serum GDF‐15 and PSD are unclear, but several potential pathophysiological processes may explain the relationship. As a pleiotropic cytokine that affects a variety of immune and inflammatory cells, GDF‐15 is overexpressed in macrophages and other cell types under the induction of oxidative stress and inflammation. Elevated levels of inflammatory markers are related to psychological depression. In addition, GDF‐15 may also cause endothelial dysfunction by generating oxidative stress in the blood vessel wall. Endothelial dysfunction plays an important role in the pathogenesis of depression, suggesting that GDF‐15 may be involved in the pathways affecting PSD. GDF‐15 has also been reported to be associated with chronic vascular brain injuries and central systolic BP, both of which are associated with depression. Further functional studies and epidemiology studies are needed to clarify the detailed mechanisms of the association between GDF‐15 and PSD. For example, it is of clinical interest to investigate the specific effects of targeting GDF‐15 on the risk of PSD.

This study has several important strengths. First, this is the first multicenter study to examine the association between baseline serum GDF‐15 concentration and PSD. Second, this study was based on standardized protocols and rigid quality control procedures in data collection and outcome assessment. Third, comprehensive information about relevant covariates was controlled in the multivariable models, so the present study was appropriate and provided a more valid appraisal of the association between serum GDF‐15 and PSD. However, some limitations should also be discussed. First, this study is not a specially designed study for the association between GDF‐15 and PSD, but an observational study based on CATIS. We included only patients with systolic BP 140 to 220 mm Hg, thus a selection bias might exist. Second, serum GDF‐15 concentrations were tested only once at baseline. Therefore, we were unable to explore the association between GDF‐15 changes and PSD, although serum GDF‐15 levels were reported to remain stable within several days after ischemic stroke onset. Third, hypertension and depression often occurred at the same time in the elderly. Data on prestroke depression were not collected in this study, which may have a potential confounding effect on our findings. However, depression is highly correlated with age, sex, stroke severity, and other baseline characteristics. We adjusted for these covariates in the analysis, therefore the potential bias caused by prestroke depression on the results might be minimal. Finally, the current study did not collect data on the use of antidepressants, and the selection of antidepressants is associated with accelerated arterial aging, which may bias our research results.

Conclusions

Elevated serum GDF‐15 levels in the acute phase of ischemic stroke were associated with an increased risk of 3‐month PSD, independent of established conventional risk factors. Further prospective studies conducted among different populations are needed to confirm our findings and clarify the potential biological mechanisms underlying the association between GDF‐15 and PSD.

Sources of Funding

This work was supported by the National Natural Science Foundation of China (grants 82103917, 81803309, and 82020108028), the high‐level personnel project of Jiangsu Province (grant JSSCBS20210712), the Natural Science Research Project of Jiangsu Provincial Higher Education (grant 21KJB330006), the Startup Fund from Soochow University (grant Q413900420), Scientific and Technological Research Program of Chongqing Municipal Education Commission (grant: KJQN201800441), and a Project of the Priority Academic Program Development of Jiangsu Higher Education Institutions.

Disclosures

None.

Data S1

Tables S1–S5

Figures S1–S2

Click here for additional data file.