Serum hepatocyte growth factor is probably associated with 3-month prognosis of acute ischemic stroke

Abstract

Background and Purpose

Serum hepatocyte growth factor (HGF) is positively associated with poor prognosis of heart failure and myocardial infarction, and it can also predict the risk of ischemic stroke in population. The goal of this study was to investigate the association between serum HGF and prognosis of ischemic stroke.

Methods

A total of 3027 acute ischemic stroke patients were included in this post hoc analysis of the China Antihypertensive Trial in Acute Ischemic Stroke. The primary outcome was composite outcome of death or major disability (modified Rankin Scale score ≥3) within 3 months.

Results

After multivariate adjustment, elevated HGF levels were associated with an increased risk of primary outcome [odds ratio, 1.50; 95% confidence interval (CI), 1.10–2.03; P_trend=0.015] when two extreme quartiles were compared. Each standard deviation increase of log-transformed HGF was associated with 14% (95% CI: 2–27%) increased risk of primary outcome. Adding HGF quartiles to a model containing conventional risk factors improved the predictive power for primary outcome (net reclassification improvement: 17.50%, P<0.001; integrated discrimination index: 0.23%, P=0.022). The association between serum HGF and primary outcome could be modified by heparin pre-treatment (P_interaction=0.001) and a positive linear dose-response relationship between HGF and primary outcome was observed in patients without heparin pre-treatment (P_linearity<0.001) but not in those with heparin pre-treatment.

Conclusions

Serum HGF levels were higher in the more severe stroke at baseline, and elevated HGF levels was probably associated with 3-month poor prognosis independently of stroke severity among ischemic stroke patients, especially in those without heparin pre-treatment. Further studies from other samples of ischemic stroke patients are needed to validate our findings.

Introduction

Hepatocyte growth factor (HGF) is a pleiotropic cytokine which regulates growth, motility, and morphogenesis of multiple types of cells, and play important roles in tissue protection and repair of various organ¹. It is reported that HGF is an endothelium-specific growth factor with proangiogenic effect and its mitogenic activity on endothelial cell growth is the most potent among various growth factors². Dysfunction of endothelial cells is well known in cardiovascular diseases³, thus serum HGF secretion might be elevated in response to cardiovascular diseases as a counter-system against endothelial cell damage. Indeed, several epidemiological and clinical studies had indicated that elevated serum HGF was associated with increased risk of hypertension⁴, heart failure⁵, and myocardial infarction⁶, and predicted poor clinical outcomes in patients with heart failure⁷ or myocardial infarction⁸. In terms of stroke, serum HGF was reported to be positively associated with the incidence of ischemic stroke^{9, 10}. However, the relationship between serum HGF and prognosis of ischemic stroke has not yet been studied. The aim of this study is to investigate the association between serum HGF and prognosis of ischemic stroke among patients from the China Antihypertensive Trial in Acute Ischemic Stroke (CATIS).

Methods

Study participants

Data are available to researchers on request for purposes of reproducing the results or replicating the procedure by directly contacting the corresponding author (yhzhang@suda.edu.cn). This study was conducted on the basis of the CATIS, a randomized clinical trial conducted in 26 hospitals across China. Details of the trial design and methods have been reported previously¹¹. Briefly, from August 2009 to May 2013, a total of 4071 patients aged over 22 years who had ischemic stroke confirmed by computed tomography or magnetic resonance imaging of the brain within 48 hours of symptom onset, and with an elevated systolic blood pressure (SBP) between 140 mm Hg and 220 mm Hg were recruited. Patients with a SBP ≥220 or diastolic blood pressure (DBP) ≥120 mmHg, severe heart failure, acute myocardial infarction or unstable angina, atrial fibrillation, aortic dissection, cerebrovascular stenosis, or resistant hypertension; those in a deep coma; and those treated with intravenous thrombolytic therapy were excluded. A further 989 patients were excluded because they refused to offer blood samples, or some collected samples were hemolyzed in storage or transport. After 3-month follow-up, 55 patients were lost and the follow-up rate was 98.2%. Finally, a total of 3027 patients were included in this analysis. Most baseline characteristics were balanced between patients who were assayed for HGF and those not assayed (Supplemental Table I), indicating that those assayed basically represented the total participants of CATIS.

The CATIS trial is registered at clinicaltrials.gov (Identifier: NCT01840072). The study was approved by the Institutional Review Boards at Soochow University in China and Tulane University in the United States. Written consent was obtained from all study participants or their immediate family members.

Data collection

Baseline data on demographic characteristics, lifestyle risk factors, medical history, clinical features, and imaging data were collected at the time of enrollment. The National Institutes of Health Stroke Scale (NIHSS) was used to evaluate stroke severity by trained neurologists at baseline¹². According to the symptoms and imaging data of the patients, ischemic stroke was classified as large artery atherosclerosis (thrombotic), cardiac embolism (embolic) and small artery occlusion lacunae (lacunar)¹³. Three blood pressure measurements were obtained at baseline while the patient was in the supine position using a standard mercury sphygmomanometer. Routine laboratory determinations (blood glucose, blood lipids, etc.) were performed for all enrolled patients in each participating hospital at admission. Dyslipidemia was defined as total cholesterol ≥6.22 mmol/L or triglyceride ≥2.26mmol/L or low density lipoprotein cholesterol ≥4.14 mmol/L or high density lipoprotein cholesterol <1.04mmol/L or self-reported history of physician-diagnosed dyslipidemia according to Chinese guidelines on the prevention and treatment of dyslipidemia¹⁴.

Serum HGF measurement

Blood samples were collected after at last 8 h of fasting within 24 hours of hospital admission. All blood samples were stored at −80°C until tested. We used a quantitative sandwich enzyme-linked immunosorbent assay with the Quantikine Human HGF Immunoassay kit (Catalog: SHG00; R&D Systems, Inc, Minneapolis) to test HGF levels in serum. Intra- and interassay coefficients of variation were less than 5.6% and 7.0%, respectively. Laboratory technicians who performed these measurements were blind to the clinical characteristics and outcomes of the study participants.

Outcome assessment

The primary outcome was composite outcome of death or major disability [modified Rankin Scale (mRS) score, 3–6] within 3 months. Secondary outcomes were separately those of death and major disability (mRS score of 6 and 3–5, respectively), and stroke recurrence. We also included an ordered seven-level categorical score of the mRS as a secondary outcome for neurologic functional status based on a recommendation for acute stroke trials¹⁵. Death certificates were obtained for deceased patients, and hospital data were abstracted for stroke recurrence. A trial-wide outcomes assessment committee, blinded to treatment assignment, reviewed and adjudicated subsequent outcomes based on the criteria established in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial.

Statistical Analysis

All participants were categorized into four groups according to quartiles of serum HGF concentrations. Baseline variables were compared among the four groups by analysis of variance or Kruskal-Wallis test for continuous variables and chi-squared test for categorical variables. Multivariate logistic regression analyses were performed to assess the association between HGF and clinical outcomes. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated for upper quartiles of HGF with the lowest quartile as a reference. The potential covariates such as age, sex, time from onset to hospitalization, current smoking, alcohol consumption, dyslipidemia, blood glucose and SBP at baseline, history of hypertension, history of coronary heart disease, history of diabetes mellitus, family history of stroke, ischemic stroke subtypes, heparin pre-treatment, receiving immediate blood pressure reduction, and baseline NIHSS score were included in the multivariate model. The effects of serum HGF on mRS shift were analyzed using multivariable ordinal logistic regression model adjusted for the aforementioned variables. Furthermore, we calculated the continuous net reclassification index (NRI) and integrated discrimination improvement (IDI) to evaluate the predictive value of adding HGF to conventional risk factors¹⁶.

In the secondary analyses, we conducted subgroup analyses to assess the association between HGF and primary outcome. Interactions between serum HGF and subgroup variables on the primary outcome 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. It has been suggested that HGF concentrations are dramatically increased in response to heparin administration¹⁷. Therefore, we further evaluated the pattern and magnitude of associations between serum HGF and primary outcome using a logistic regression model with restricted cubic splines among the patients without or with heparin pre-treatment during the 24 hours before blood sampling, with three knots placed at the 10th, 50th, and 90th percentiles of HGF, and the 5% highest and lowest HGF observations were trimmed¹⁸. Two-tailed P <0.05 was considered to be statistically significant. All analyses were conducted using SAS statistical software (version 9.4, Cary, NC, USA).

Results

This study included a total of 3027 patients, including 1924 males and 1103 females, and the average age was 62 years old with a range from 25 to 95 years. A total of 1037 patients were pre-treated with heparin. Among them, 380 used unfractionated heparin, 620 used low molecular weight heparin and 37 used both unfractionated heparin and low molecular weight heparin. The participants with higher HGF tended to be older, and had higher blood glucose, baseline NIHSS score, and prevalence of heparin pre-treatment, thrombotic, and embolic than those with lower HGF, whereas the participants with higher HGF had shorter time from onset to randomization and lower prevalence of lacunar compared with those with lower HGF (Table 1).

Table 1.

Baseline Characteristics of the study participants according to HGF quartiles

Characteristics	HGF, pg/mL
	<1242.14	1242.14–1577.65	1577.66–2115.95	≥2115.96	P value
Number of subjects	756	757	758	756
Demographics
Age, years	60.93 ± 10.05	61.91 ± 10.03	62.48 ± 11.23	63.53 ± 11.74	<0.001
Male	479 (63.36)	498 (65.79)	486 (64.12)	461 (60.98)	0.272
Current cigarette smoking	261 (34.52)	275 (36.33)	283 (37.34)	281 (37.17)	0.653
Current alcohol drinking	229 (30.29)	231 (30.52)	263 (34.70)	218 (28.84)	0.081
Medical history
History of hypertension	589 (77.91)	593 (78.34)	592 (78.10)	596 (78.84)	0.975
History of CHD	71 (9.39)	72 (9.51)	76 (10.03)	94 (12.43)	0.176
History of diabetes mellitus	117 (15.48)	122 (16.12)	138 (18.21)	152 (20.11)	0.073
Family history of stroke	132 (17.46)	143 (18.89)	140 (18.47)	149 (19.71)	0.727
Clinical features
Time from onset to hospitalization, h*	12 (5–24)	12 (5–24)	10 (4–24)	8 (4–24)	<0.001
Systolic BP, mmHg	165.84 ± 16.88	166.14 ± 17.28	166.93 ± 17.27	166.60 ± 16.18	0.609
Diastolic BP, mmHg	97.15 ± 11.00	96.30 ± 10.77	97.23 ± 10.68	96.20 ± 11.72	0.137
Dyslipidemia	400 (52.91)	427 (56.41)	430 (56.73)	440 (58.20)	0.201
Blood glucose, mmol/L*	5.70 (5.00–6.80)	5.60 (5.00–6.90)	5.80 (5.10–7.30)	5.96 (5.20–7.80)	<0.001
Baseline NIHSS score*	4 (2–6)	4 (2–7)	5 (3–8)	6 (3–10)	<0.001
Ischemic stroke subtype†
Thrombotic	546 (72.22)	564 (74.50)	588 (77.57)	625 (82.67)	<0.001
Embolic	18 (2.38)	33 (4.36)	37 (4.88)	58 (7.67)	<0.001
Lacunar	206 (27.25)	179 (23.65)	160 (21.11)	87 (11.51)	<0.001
Heparin pre-treatment	219 (29.01)	190 (25.13)	227 (29.95)	401 (53.04)	<0.001
Receiving immediate blood pressure reduction	393 (51.98)	357 (47.16)	387 (51.06)	375 (49.60)	0.260

Data were presented as mean ± standard deviation or n (%) unless otherwise noted.

^*Data were presented as median (interquartile range).

CHD = coronary heart disease; BP = blood pressure; HGF = hepatocyte growth factor; NIHSS = National Institute of Health Stroke Scale;

^†Ten patients with both thrombotic and embolic subtypes; 68 patients with thrombotic and lacunar subtypes; 4 patients with embolic and lacunar subtypes; 1 patient with all 3 subtypes.

After 3-month follow-up, a total of 750 patients experienced primary outcome, among which 665 were major disability and 85 were death, and 54 patients developed stroke recurrence (Table 2). The cumulative incidence rates of primary outcome within 3 months among ischemic stroke patients in four quartiles of serum HGF (from low to high) were 16.80, 21.80, 25.20 and 35.32%, respectively. Patients in the fourth quartile of serum HGF had the highest risks of primary outcome and death. After adjustment for potential confounders, the ORs associated with highest quartile of serum HGF were 1.50 (95% CI, 1.10–2.03; P_trend = 0.015), 1.23 (95% CI, 0.91–1.66; P_trend = 0.215), 5.93 (95% CI, 1.76–19.97; P_trend = 0.001), and 0.95 (95% CI, 0.41–2.17; P_trend = 0.979) for primary outcome, major disability, death, and stroke recurrence, respectively. Similarly, on continuous analyses, each standard deviation increase of log-transformed HGF was associated with 14% (95% CI: 2–27%) and 50% (95% CI: 18–90%) increased risk of primary outcome and death, respectively, but not significantly associated with major disability and stroke recurrence. Multivariable ordinal logistic regression analyses showed a significantly worse shift in the distribution of mRS score with higher HGF quartiles in acute ischemic stroke patients (P_trend = 0.004, Figure 1). Moreover, adding HGF quartiles to the conventional model improved the discriminatory ability for primary outcome (NRI: 17.50%; IDI: 0.23%) and death (NRI: 36.41%; IDI: 0.95%) but not for major disability and stroke recurrence (Supplemental Table II).

Table 2.

Odds ratios and 95% confidence intervals of clinical outcomes according to serum HGF quartiles in ischemic stroke patients

	HGF, pg/mL					Each standard deviation increase of log-HGF
	<1242.14	1242.14–1577.65	1577.66–2115.95	≥2115.96	Ptrend
	Primary outcome: death or major disability (mRS 3–6)
No. of cases (%)	127 (16.80)	165 (21.80)	191 (25.20)	267 (35.32)		750 (24.78)
Model 1	1.00	1.36 (1.03–1.80)	1.51 (1.14–1.98)	2.43 (1.85–3.19)	<0.001	1.33 (1.21–1.46)
Model 2	1.00	1.19 (0.88–1.63)	1.19 (0.88–1.62)	1.50 (1.10–2.03)	0.015	1.14 (1.02–1.27)
Secondary outcomes
Major disability (mRS 3–5)
No. of cases (%)	123 (16.27)	149 (19.68)	171 (22.56)	222 (29.37)		665 (21.97)
Model 1	1.00	1.26 (0.95–1.67)	1.36 (1.03–1.80)	1.91 (1.45–2.52)	<0.001	1.21 (1.10–1.34)
Model 2	1.00	1.12 (0.83–1.51)	1.12 (0.83–1.50)	1.23 (0.91–1.66)	0.215	1.05 (0.94–1.17)
Death
No. of cases (%)	4 (0.53)	16 (2.11)	20 (2.64)	45 (5.95)		85 (2.81)
Model 1	1.00	4.06 (1.15–14.39)	4.97 (1.45–17.09)	11.00 (3.34–36.26)	<0.001	1.74 (1.41–2.13)
Model 2	1.00	3.23 (0.90–11.65)	3.43 (0.98–12.05)	5.93 (1.76–19.97)	0.001	1.50 (1.18–1.90)
Stroke recurrence
No. of cases (%)	15 (1.98)	11 (1.45)	12 (1.58)	16 (2.12)		54 (1.78)
Model 1	1.00	0.80 (0.34–1.89)	0.92 (0.41–2.08)	0.93 (0.41–2.10)	0.943	1.23 (0.93–1.62)
Model 2	1.00	0.81 (0.34–1.90)	0.93 (0.41–2.10)	0.95 (0.41–2.17)	0.979	1.24 (0.94–1.64)

Model 1, adjusted for age, sex, time from onset to hospitalization, current smoking, alcohol consumption, dyslipidemia, blood glucose and SBP at baseline, history of hypertension, history of coronary heart disease, history of diabetes mellitus, family history of stroke, ischemic stroke subtypes, heparin pre-treatment, and receiving immediate blood pressure reduction.

Model 2, further adjusted for baseline NIHSS score.

Figure 1.

Distribution of mRS score at 3 months according to serum HGF quartiles in ischemic stroke patients. Multivariable adjusted ORs of ordinal logistic regression analysis, 1.36 (95% CI, 1.11–1.67) for patients in the fourth quartile of HGF compared to the patients in the lowest quartile (P for trend = 0.004). Multivariable model adjusted for the same variables as model 2 in Table 2.

In subgroup analyses stratified by age, sex, baseline SBP, baseline NIHSS score, time from onset to hospitalization, cigarette smoking, alcohol consumption, receiving immediate blood pressure reduction medication and ischemic stroke subtypes, the modest positive associations between serum HGF and primary outcome were observed in almost all subgroups, and reached significant levels in several subgroups (Supplemental Table III). Statistical tests for interactions between serum HGF and these interested factors on primary outcome were not significant (all P > 0.05). As expected, the association between serum HGF and primary outcome could be modified by heparin pre-treatment (P_interaction = 0.001), and serum HGF was associated with primary outcome only in patients without heparin pre-treatment but not in those with heparin pre-treatment (Supplemental Table III). As shown in Figure 2, with an increase of HGF levels, the risk of primary outcome for patients without heparin pre-treatment increased (P_nonlinearity = 0.317, P_linearity < 0.001). In contrast, we did not observe any association between serum HGF and the primary outcome in patients with heparin pre-treatment (P_nonlinearity = 0.298, P_linearity = 0.134).

Figure 2.

Adjusted ORs of primary outcome according to HGF. OR and 95 % CI derived from restricted cubic spline regression, with knots placed at the 10th, 50th, and 90th percentiles of HGF. Panels adjusted for the same variables as model 2 in Table 2. (A) Patients without heparin pre-treatment; (B) Patients with heparin pre-treatment.

Discussion

In this post hoc analysis of the CATIS trial, we found that serum HGF levels were higher in more severe stroke at baseline, and elevated HGF levels were probably associated with 3-month poor prognosis independently of stroke severity among ischemic stroke patients. We also found that adding serum HGF to a model containing conventional risk factors could improve risk prediction for the poor prognosis of ischemic stroke. In the secondary analysis, we further found that heparin pre-treatment before blood sampling could modify the association between serum HGF and primary outcome, and a positive linear dose-response relationship between HGF and poor prognosis was observed in patients without heparin pre-treatment but not in those with heparin pre-treatment.

Endothelial cells are known to secrete various antiproliferative and vasodilating factors, such as nitric oxide and natriuretic peptides¹⁹. It has been suggested that the presence of endothelial cell dysfunction with loss of substances from endothelial cells might result in the development and progression of abnormal vascular growth, such as atherosclerosis²⁰. As we all known, endothelial dysfunction has been confirmed to be an early pathophysiological event of cardiovascular diseases³ and might be an independent predictor and prognostic factor for cardiovascular diseases²¹. HGF is a member of endothelium-specific growth factors and its mitogenic activity is the most potent among various growth factors², thus serum HGF concentrations might be elevated in response to cardiovascular diseases as a counter-regulatory system against endothelial dysfunction. In fact, serum HGF has been reported to be positively associated with the risk of many cardiovascular diseases^4–6, including ischemic stroke^{9, 10}. Although several observational studies also find that elevated serum HGF is an independent predictor of poor outcomes in patients with heart failure⁷ or myocardial infarction⁸ during the follow-up, no such evidence is available for ischemic stroke.

Herein, we report that elevated serum HGF was associated with poor outcomes at 3 months after an ischemic stroke, which is concordant with the observation that higher HGF levels correlate with worse prognosis of heart failure or myocardial infarction^{7, 8}. Moreover, serum HGF was associated with mortality but not disability at 3 months after ischemic stroke onset, suggesting that the difference of primary outcome across HGF groups is mainly driven by mortality. Previous studies had reported that serum HGF were positively related to some stroke prognostic factors, including increased age, smoking and hypertension^{4, 10}. Interestingly, in our study, elevated HGF concentrations remained independently associated with poor prognosis when these powerful prognostic parameters were included into the multivariate models. From the findings of our study, we hypothesized that HGF might be useful in risk stratification of ischemic stroke prognosis, and could assist the selection of high-risk patients for aggressive monitoring and therapeutic interventions. However, further prospective studies from other samples of ischemic stroke patients are needed to validate our hypothesis. At first glance, these results may appear to contradict the biological function of HGF, including antiapoptotic, mitogenic, antifibrotic and proangiogenic activities¹. However, this phenomenon is very similar to those previously reported for natriuretic peptides. Natriuretic peptides are known to play beneficial effects in ischemic stroke patients, but elevated levels are strong predictors of poor outcomes²². The precise mechanisms explaining the detrimental effect of elevated HGF on the stroke prognosis are not well defined, but may include inflammation²³, accelerated progression of atherosclerotic lesions⁹, plaque neovascularization²⁴, intraplaque hemorrhage and plaque rupture²⁵. In addition, higher serum HGF levels may also reflect a greater degree of baseline endothelial dysfunction⁹ which is associated with subsequent infarct expansion²⁶ and hemorrhagic transformation²⁷, thus increase the risk of poor prognosis.

In intravascular thrombus-associated disorders, it was reported that release of endogenous heparin during mast cell degranulation stimulated the production and release of serum HGF²⁸, and HGF was also called heparin-binding growth factor. It has been demonstrated that HGF levels increase markedly and rapidly after heparin administration in humans¹⁷. Therefore, we consider heparin pre-treatment before blood sampling as a confounder of the associations between HGF and poor outcomes, and the positive associations remain significant after including heparin pre-treatment into the multivariate models. Recently, one study showed a conflicting result regarding the predictive value of HGF in acute coronary syndrome patients pre-treated with intravenous heparin, namely, they found that serum HGF elevation was associated with a favorable prognosis²⁹. Administration of heparin has already been used to evaluate the functional reserve of HGF in the organism and the capability of secreting HGF into circulation³⁰, and exogenous supplementation of HGF by a gene transfer technique has been shown to reduce intimal hyperplasia after balloon injury³¹. This may explain the apparently conflicting observation. In patients without heparin pre-treatment, serum HGF is naturally HGF which produced in response to endothelial dysfunction, so higher HGF reflects a greater degree of baseline endothelial dysfunction and is associated with worse prognosis. In patients with heparin pre-treatment, serum HGF is stimulated HGF whose secretion is induced by heparin administration, so elevated HGF is associated with good prognosis through its protective effect in tissue repair²⁹. Similarly, we found that heparin pre-treatment could modify the association between serum HGF and prognosis of ischemic stroke. In the present study, serum HGF was positively associated with poor prognosis in patients without heparin pre-treatment, whereas there was an inversely trend of the association between HGF and poor outcomes in patients with heparin pre-treatment although it did not reach statistical significance. Patients with heparin pre-treatment are only half the number of those without heparin pre-treatment, which might be underpowered to detect the significant inverse association in those pre-treated with heparin.

Our study has some strengths. First, this is the first study to investigate the association between serum HGF and clinical outcomes in acute ischemic stroke patients. Second, this is an observational study based on the patients from CATIS, a randomized clinical trial with strict quality controls in data collection and outcome assessment, and a relatively large sample size enables us perform analysis with high statistical power. Some limitations should be discussed here. First, this study is a post hoc analysis of the CATIS trial and a selection bias may exist. However, baseline characteristics of participants in this study were similar to those from the China National Stroke Registry³², suggesting that the selection bias may be minimal. Second, serum HGF concentrations were tested only once at baseline. Therefore, we don't know how the levels of HGF change with differences in baseline characteristics and further studies are needed to investigate this. We also have no data to examine the association between HGF changes and prognosis of acute ischemic stroke although HGF levels are reported to be relatively stable over several years¹⁰. Third, the number of patients with heparin pre-treatment which account for 1/3 of total participants is relatively few, which might be underpowered to detect the significant inverse association in those pre-treated with heparin. Fourth, we did not consider the lesion volume when we examined the association of HGF with prognosis. However, lesion volume is highly correlated with NIHSS score³³ and we have adjusted baseline NIHSS score in the analysis. Finally, relatively few number of stroke recurrence (maybe because some fatal stroke recurrences were classified into death events) were observed during 3 month follow-up of relatively short time, which might limited our power to detect significant association between serum HGF and stroke recurrence. Further long-term follow-up studies are required to verify these relationships.

Conclusions

In summary, we found that serum HGF levels were higher in more severe stroke at baseline, and elevated HGF levels were probably associated with 3-month poor prognosis independently of stroke severity among ischemic stroke patients, especially in those without heparin pre-treatment. Further prospective studies from other samples of ischemic stroke patients are needed to validate our findings.