The association of plasma osteoprotegerin levels and functional outcomes post endovascular thrombectomy in acute ischemic stroke patients: a retrospective observational study

Study design

This study was a retrospective observational study. Our institutional review board approved this study, and we received informed consent from all patients or their closest relatives (EUMC 2014-04-023, 2018-10-036). The informed consent during the data collection process was explained directly to the patient by the researcher and was then signed. Withdrawal of consent was possible at any time, and blood samples from the patients who withdrew consent were not included in the final analysis. After collecting the research data, data management and analysis were performed in an anonymized state by entering the patient’s de-identified study number.

Subjects

From April 2014 through December 2020, 413 patients diagnosed with acute ischemic stroke and who underwent EVT were prospectively screened from Ewha Womans University Seoul Hospital and Ewha Womans University Mokdong Hospital in Korea. All screened patients were (1) admitted within 24 h after the onset of neurological symptoms or the last known normal time, (2) diagnosed with acute ischemic stroke confirmed by brain computed tomography (CT) or magnetic resonance image (MRI), and (3) performed EVT after confirming vessel occlusion and perfusion-diffusion mismatch. We excluded patients who had uncommon stroke risk factors, including cancer (n = 8), autoimmune disease (n = 2), or a history of bone fracture (n = 4) within the last 2 months because OPG is closely involved with systemic inflammatory reactions and bone metabolism (Kiechl et al., 2006; Song et al., 2012). Any samples that could not be analyzed for this study due to poor blood sampling quality were excluded (n = 1), and patients who refused (n = 33) or withdrew consent were also excluded (n = 5) (Fig. 1). Finally, we included 360 patients in this study and collected blood samples for plasma OPG measurement after fasting for 12 postoperative hours after EVT.

We evaluated all patients using our institution’s standard stroke evaluation protocol based on the American Heart Association and Korean clinical practice guidelines (Jauch et al., 2013; Ko et al., 2019), which included routine blood tests (blood glucose level at admission, total cholesterol, low-density lipoprotein, triglyceride, white blood cell count, hemoglobin, creatinine, total calcium, phosphate, albumin, C-reactive protein, and vitamin D 25(OH)D), chest radiography, 12-lead electrocardiography, and brain imaging including CT and CT angiography and/or MRI and MR angiography or digital subtraction angiography (DSA) (Chang et al., 2018).

Clinical and radiologic assessment

Detailed information regarding risk factors is shown in the supplementary methods (Chang et al., 2018). The National Institute of Health Stroke Scale (NIHSS) was used to evaluate the severity of neurological symptoms. Successful recanalization was defined as a grade of 2b or 3, assessed by modified treatment in cerebral ischemia based on the final DSA (Higashida et al., 2003). Hemorrhagic transformation was defined according to the European Cooperative Acute Stroke Study (ECASS)-3 trial classification and was confirmed in follow-up brain images by consensus discussions (Hacke et al., 2008; Higashida et al., 2003). Stroke subtypes were classified into three categories according to the Trial of Org 10172 in Acute Stroke Treatment classification system (Adams Jr et al., 1993).

Acute stroke treatment

Administration of intravenous tissue plasminogen activator (IV tPA) and EVT were performed according to current guidelines for acute stroke management (Hacke et al., 2008). IV tPA administration was performed under the supervision of a neurologist according to standard guidelines within 4.5 h of symptom onset (Hacke et al., 2008). EVT also adhered to standard guidelines, and the decision on whether to perform EVT was arrived at through discussions between neurologists and neuroradiologists (Jauch et al., 2013; Kernan et al., 2014). If the appropriate time window for performing EVT was uncertain, CT or MR angiographic, brain MR diffusion-perfusion, diffusion-fluid attenuated inversion recovery (FLAIR), and diffusion-clinical mismatch findings were considered to inform the decision (Jauch et al., 2013; Kernan et al., 2014). EVT primarily involved mechanical thrombectomy. If reocclusion or distal embolization was present, glycoprotein IIb/IIIa antagonists could be administered as adjuvant therapy. Noncontrast brain CT was routinely performed 24 h after EVT, and brain MR (diffusion, FLAIR, gradient recalled echo image) and MR angiography were also routinely performed after 48–72 h.

Outcomes

In the present study, a modified Rankin Scale (mRS) at 3 months (90 days ± 14 days) after index stroke occurrence was used to assess functional outcomes. The mRS was evaluated by a neurologist with more than 5 years of experience and assessed based on face-to-face visits when the patients visit the outpatient clinic; if they could not visit in person, we provided a teleconsultation. Poor outcomes were defined by mRS ratings > 2.

Measurement of plasma OPG levels

At admission, we collected venous samples in EDTA from all enrolled patients. Samples were immediately centrifuged at 1,900 g for 15 min at the correct temperature (4 °C), followed by storing at −80 °C until analysis. OPG DuoSet (R&D Systems, Abingdon, UK), an enzyme-linked immunosorbent assay kit, was used to measure plasma OPG levels. The detection range was 31.25–4000 pg/mL. One of our research team members measured and averaged OPG levels in duplicate (Song et al., 2012). Coefficients of variability were 3.1% for inter-assay comparisons and 2.5% for intra-assay comparisons.

Sample size estimation

According to previously published studies, the rates of poor functional outcome defined as below mRS 2 are reported to be between 20 and 40% (Appelros, Nydevik & Viitanen, 2003; Song et al., 2012). We applied a formula for sample size calculation for descriptive studies with proportions. Using a 30% prevalence rate, a precision value of 0.05 and a Z-value of 1.96, the formula yielded a minimum sample size of 323. Taking into account a 10% loss to follow up, this study planned to include a total of 355 participants.

Statistical analysis

Continuous variables and categorical variables were analyzed using independent t-tests, the Mann–Whitney test, Kruskal–Wallis test, chi-square test, or Fisher’s exact test, as appropriate. The Shapiro–Wilk test was performed for evaluating normality, and continuous variables in our dataset were normally distributed except for the NIHSS. Univariate and multivariate binary logistic regression and ordinal logistic regression analyses were performed to investigate the association of plasma OPG levels with poor functional outcomes. For multivariate analyses, sex, body mass index, and variables with p values < 0.1 in the univariate analysis (age, NIHSS, thrombolysis methods, number of trials for thrombectomy, successful recanalization, any hemorrhagic transformation, blood glucose level at admission, hemoglobin, total cholesterol, C-reactive protein, and vitamin D 25(OH)D) were adjusted. Because of over-fitting in the logistic regression, hemorrhagic transformation was entered after dichotomization. For the sensitivity analysis, plasma OPG levels were entered in the multivariate analysis model as continuous variables, per standard deviation and categorical variable (tertiles), respectively. For investigating the predictability of the plasma OPG levels for the prognosis, receiver operator characteristic (ROC) curves, area under the curve (AUC), integrated discrimination index (IDI) and net reclassification improvement (NRI) were performed to evaluate the performance of OPG levels for predicting poor outcomes (Pencina, D’Agostino Sr & Demler, 2012). We performed a subanalysis to check whether the association between OPG levels and functional outcome was consistent according to the door-to-puncture time. Using 8 h of door-to-puncture time as a cut-off, we divided patients into two groups (Jahan et al., 2019). All variables with p values < 0.05 were considered statistically significant. Statistical analyses were performed using SPSS Statistics for Windows, version 21.0 (IBM Corp., Armonk, NY, USA).

Demographic data

Demographic data are shown in Table 1 and thrombectomy related variables and stroke subtypes are shown in Table 2. Of the 360 patients enrolled, 185 (51.4%) were men. Their mean age was 74.9 ± 13.9 years. Their mean ± standard deviation plasma OPG level was 200.2 ± 74.4 pg/mL. Of all included patients, 215 (59.8%) and 145 (40.2%) patients had good (mRS ≤2) and poor (mRS > 2) functional outcomes, respectively. Plasma OPG levels were higher in patients with poor outcomes (212.9 ± 63.7 vs. 191.7 ± 79.8 pg/mL, p = 0.008) than in patients with good outcomes. Patients with poor outcome were older (78.1 ± 13.2 vs. 72.7 ± 13.9 years, p < 0.001) and were more likely to have diabetes mellitus (68.3% vs. 41.9%, p < 0.001), undergo mechanical thrombectomy only (67.6% vs. 54.4%, p = 0.017), and undergo hemorrhagic transformation (46.9% vs. 28.8%, p = 0.001), and they less frequently underwent successful recanalization (74.5% vs. 94.4%, p < 0.001) than patients with good outcomes. Compared with patients with good outcomes, patients with poor outcomes also had higher NIHSS scores (median 18, interquartile range [13–22] vs. median 12 interquartile range [7–16], p < 0.001) and a higher mean blood glucose level at admission (148.2 ± 48.1 vs. 134.8 ± 50.9 years, p = 0.013). Patients with poor outcomes had lower mean vitamin D 25(OH)D (19.4 ± 6.7 vs. 21.5 ± 6.8 years, p = 0.004) and hemoglobin (12.9 ± 2.2 vs. 13.5 ± 1.8 years, p = 0.004) levels than patients with good outcomes (Tables 1 and 2). Table S1 shows the correlations between the plasma OPG concentration tertiles and other variables. Higher plasma OPG concentrations were associated with higher NIHSS scores (p = 0.011).

Plasma OPG levels and functional outcomes

After adjusting for sex, body mass index, and variables with p < 0.1 in the univariate analysis (age, diabetes mellitus, NIHSS, thrombolysis methods, number of trials for thrombectomy, recanalization, any hemorrhagic transformation, glucose at admission, hemoglobin, total cholesterol, and C-reactive protein, and vitamin D 25(OH)D), higher plasma OPG levels were independently associated with poor functional outcomes (continuous variable: odds ratios (OR) 1.004, 95% confidence interval (CI) (1.000–1.008), p = 0.031; per standard deviation: OR 1.373, 95% CI (1.039–1.824), p = 0.031; highest tertile vs. lowest tertile: OR 2.121, 95% CI [1.089–4.191], p = 0.037) in the multivariate binary logistic regression analysis (Table 3). Table S2 shows the associations between the poor functional outcome and other variables after adjusting plasma OPG concentration in the univariate and multivariate binary logistic regression analysis.

In the ordinal logistic regression analysis with mRS as a dependent variable, higher plasma OPG concentrations were associated with poor functional outcomes (continuous variable: OR 1.004, 95% CI [1.002–1.007], p = 0.029; per standard deviation: OR 1.388, 95% CI [1.134–1.700], p = 0.029; highest tertile vs. lowest tertile: OR 2.102, 95% CI [1.301–3.412], p = 0.002) (Table 4). Table S3 shows the associations between the poor functional outcome and other variables after adjusting plasma OPG concentration in the univariate and multivariate binary logistic regression analysis. In the ROC comparison, the AUC of the multivariate analysis model with OPG was higher, but not significantly so, than the model without OPG (AUC 0.859 vs 0.855, p = 0.379). In contrast, NRI was significantly increased in the multivariate analysis model with OPG relative to without OPG (0.012, p = 0.027), although IDI did not increase it (0.174, p = 0.104) (Table S4).

In the subgroup analysis, there were no associations between plasma OPG levels and functional outcomes at 3 months after index stroke according to sex, age, hypertension, diabetes mellitus status, body mass index, hypercholesterolemia status, coronary artery disease status, congestive heart failure status, atrial fibrillation status, smoking history, alcohol intake, previous stroke history, NIHSS score, or thrombolysis method (mechanical thrombectomy only vs. tPA and mechanical thrombectomy). Onset-to-puncture time (<480 min vs. ≥480 min, p for interaction = 0.028) was the only variable with a significant association in the subgroup analysis. In the binary logistic regression analysis, the association of plasma OPG levels with poor functional outcomes was significant among patients with door-to-puncture times < 480 min (OR 1.551, 95% CI [1.199–2.030], p = 0.001), but it was not significant in patients with door-to-puncture times ≥480 min (OR 0.863, 95% CI [0.541–1.355], p = 0.526) (Table S5).