Hemorrhagic complications

Increase in hemorrhagic complications were mainly associated with overt hyperglycemia (p = 0.045) (Table 2). When glycemic subgroup was incorporated into a logistic regression model for occurrence of parenchymal hematoma type 2 (Table 4), only overt hyperglycemia showed an independent association (OR 9.28, CI 1.66–51.88, p = 0.011) referenced by normoglycemia. In a subgroup analysis according to reperfusion status, overt hyperglycemia was associated with parenchymal hematoma type 2 only in the reperfusion group (OR 12.34, CI 1.60–

95.07, p = 0.016) as referenced to normoglycemia, but the association was insignificant in the non-reperfusion group. Overt hyperglycemia was also confirmed to be as an independent predictor of parenchymal hematoma type 1–2 (OR 3.75, CI 1.25–11.23, p = 0.018), and any type of hemorrhagic transformation (OR 3.10, CI 1.39–6.92, p = 0.006) with adjustment of the same covariables.

Table 4. Logistic regression model of hyperglycemia as a risk factor for parenchymal hematoma type 2 according to reperfusion status after endovascular revascularization therapy.

Overall (n = 341) Reperfusion subgroup (n = 272) Non-reperfusion subgroup (n = 69)

OR (95% CI) p-value OR (95% CI) p-value OR (95% CI) p-value

ICA T Reference Reference Reference

ICA I 0.00 (0.00) 0.998 0.00 (0.00) 0.998 - 1.000

Normoglycemia Reference Reference Reference

Overt hyperglycemia 9.28 (1.66–51.88) 0.011 12.34 (1.60–95.07) 0.016 - 1.000 OR, odds ratio; CI, confidence interval; NIHSS, National Institutes of Health Stroke Scale; IV tPA, intravenous tissue plasminogen activator; ICA, internal carotid artery; MCA, middle cerebral artery; DWI, diffusion-weighted imaging; mTICI, modified Treatment In Cerebral Ischemia.

IV. DISCUSSION

The present study demonstrates that hyperglycemia at hospitalization are independently associated with poor outcomes, infarct growth, and hemorrhagic complications; furthermore, the glucose level cut-off points to predict these results varied. Overall, hyperglycemia was associated with a poor functional outcome and infarct growth, while overt hyperglycemia was associated with hemorrhagic complications. In the non-reperfusion subgroup, the association between hyperglycemia and infarct growth was dominant, while the association between overt hyperglycemia and hemorrhagic complications were dominant in the reperfusion subgroup. The comorbidity of DM was 10% in the normoglycemia group, 18% in the moderate hyperglycemia group, and 80% in the overt hyperglycemia group. A similar trend was seen for mean HbA1c levels, suggesting that overt hyperglycemia on presentation may be representative of sustained hyperglycemia, while moderate levels of hyperglycemia might be more representative of hyperglycemia under non-diabetic or pre-non-diabetic conditions.

Post-procedural infarct volumes were significantly larger in the moderate and overt hyperglycemia groups than in the normoglycemia group, and both groups were independently associated with poor outcome. This corresponds to a relatively low cut-off glucose level of >110mg/dL. While infarct growth seems to be a mediator between hyperglycemia and a poor outcome, there was no correlation between admission glucose and infarct growth nor an interaction between glucose level and infarct growth in terms of poor outcomes (data not shown). This suggests that such associations may not be linear, and an approach using cut-off values may better represent the phenomenon. The significance of glucose levels as an independent predictor for outcomes and infarct growth disappeared in multivariable analysis when patients with successful reperfusion were analyzed, whereas both significant associations were still present in patients with unsuccessful reperfusion. This finding

implies that successful reperfusion may overcome the negative effects of hyperglycemia. This observation is partly supported by analysis of MR CLEAN data, which showed no interaction of hyperglycemia and EVT effect when compared with non-endovascular treatment, indicating that admission hyperglycemia is not a contraindication for EVT (Osei et al., 2017).

Several mechanisms via which moderate hyperglycemia leads to poor outcomes and infarct growth can be postulated. First, possibility is the concept of stress hyperglycemia. Meta-analyses demonstrate a strong correlation between glucose levels > 110 to 126 mg/dL and poor outcomes only in non-diabetic patients with acute ischemic stroke (Capes et al., 2001). Glucose levels are known to increase with increasing stroke severity via activation of the hypothalamic-pituitary-adrenal axis (Feibel et al., 1977; O'Neill et al., 1991). Second, direct contribution of hyperglycemia itself in aggravation of ischemic stroke needs to be considered.

Decreased reperfusion and penumbral salvage are associated with hyperglycemia, resulting in infarct growth (Venables et al., 1985; Kruyt et al., 2010). Further, hyperglycemia results in reduced penumbral salvage in patients with perfusion-diffusion mismatch (Kagansky et al., 2001; Parsons et al., 2002). A third mechanism worth discussing is that presenting hyperglycemia may represent pre-existing abnormalities in glucose metabolism. Significant number of ischemic stroke patients without history of DM have impaired glucose metabolism, insulin resistance, or DM at follow-up (Kernan et al., 2005; Vancheri et al., 2005). Thus presenting hyperglycemia may reflect insulin resistance and comprise the metabolic syndrome, known to be associated with poor leptomeningeal collateral status in acute ischemic stroke (Menon et al., 2013). However, pre-diabetic conditions were not addressed in this study, and further studies are needed to address this issue.

In the present study, overt hyperglycemia, or diabetic levels of hyperglycemia, was associated with severe intracerebral hemorrhage, particularly in

the reperfusion subgroup. The results suggest that reperfusion injury can be exacerbated by chronic sustained hyperglycemia. The association between DM and ICH in patients with ischemic stroke after IV thrombolysis is well recognized (Demchuk et al., 1999). However, IV tPA was not associated with the clinical outcomes in our endovascular population. In endovascular populations, associations between sustained hyperglycemia or DM and hemorrhagic transformation have been recently reported; however, in one study, the endovascular devices and methods used were somewhat outdated (Nogueira et al., 2015), and in another study using stent retrievers, the significance for hyperglycemia was not shown, but DM was confirmed to be significant (Jiang et al., 2015). Oxidative stress and activation of inflammation are reported to be aggravated in overt hyperglycemia, resulting in dysfunction of the blood-brain barrier (Bouchard et al., 2002; Kamada et al., 2007; Martini and Kent, 2007). Moreover, severe hyperglycemia significantly worsens cortical intracellular acidosis and results in mitochondrial dysfunction in the ischemic penumbra (Anderson et al., 1999). Such mechanisms can lead to hemorrhagic transformation and extensive hemorrhage (de Courten-Myers et al., 1992). Elevated erythrocyte sedimentation rates were seen in patients with overt hyperglycemia in our study, which may indicate such inflammatory reactions.

In terms of glucose cut-off values, a previous large-scale study of patients who received IV tPA yielded results similar to those of the present study. The Safe Implementation of Treatment in Stroke International Stroke Thrombolysis Register (SITS-ISTR), involving over 16,000 acute ischemic stroke patients treated with thrombolysis, show that glucose levels > 120 mg/dL were associated with increased mortality, while levels > 180 mg/dL were associated with symptomatic ICH (Ahmed et al., 2010). The ORs for mortality and functional dependence were significantly higher for glucose levels > 120 mg/dL in non-diabetic patients, while the values for mortality and functional independence were 181–200 mg/dL and 160 mg/dL,

higher in both diabetic and non-diabetic patients with glucose levels of 181–200 mg/dL compared with that for patients who had lower glucose levels (Ahmed et al., 2010). These results and our present findings both provide similar ranges of dual glucose cut-off values, and the lower value is associated with poor outcome and the higher value is associated with hemorrhagic complications. Our present research further reveals that higher glucose cut-off values may be associated with sustained hyperglycemia and comorbid DM.

There are limitations to the present study. First, although our analysis includes multicenter data, it is limited by the observational study design. Nonetheless, the population was medium-sized and presented with acceptable revascularization profiles. Second, the retrospective study design precluded the use of symptomatic ICH as an endpoint, more widely used in EVT trials. However, parenchymal hematoma type 2, used in the current study as a hemorrhagic outcome is strongly associated with clinical deterioration (Berger et al., 2001). Third, patients were only included when both pre-procedural and post-procedural MRI data were available.

Post-procedural MRI data were not available for patients with severe pathology or for those who expired, resulting in selection bias. Nevertheless, the analysis of infarct growth through both pre-procedural and short-term post-procedural DWI is not easy to perform in clinical studies, and we believe that this data has a novel value. Finally, the management of hyperglycemia might have differed between hospitals and attending physicians. Management of hyperglycemia is another unresolved issue, and therapeutic trials such as the Stroke Hyperglycemia Insulin Network Effort (SHINE) currently focus in resolving such issues (Bruno et al., 2014).

In conclusion, in Korean patients with acute intracranial large artery occlusion who underwent EVT, moderate to overt hyperglycemia on admission was associated with a poor outcome and infarct growth in the total population and those who did not achieve successful reperfusion. Additionally, overt hyperglycemia was

associated with significant hemorrhagic complications in the total population, and patients with successful reperfusion. These findings should be confirmed in future large-scale prospective cohorts.

PART III. Potential Neuroprotection

Co-administration of Metformin and the DPP-4 Inhibitor Evogliptin Reduces Cerebral Infarct Volume in Diabetic

Rat Brain

I. INTRODUCTION

Diabetes mellitus (DM) is a major risk factor for cardiovascular events including stroke. The patients not only experience more severe stroke (Els et al., 2002) but also have worse outcome. Morbidity and mortality is increased in diabetics after stroke (Candelise et al., 1985). Hyperglycemia rather than type 2 DM is consistently associated with more severe stroke and poor outcomes (Capes et al., 2001). In animal models of focal ischemia in hyperglycemic rats, transient ischemia consistently causes bigger infarct volume, and apoptotic cell death is markedly increased in core and penumbra (Muranyi et al., 2003). Oxidative stress seems to play an essential role in the overall pathogenesis of cerebral ischemic reperfusion injury (Niizuma et al., 2009).

We have previously shown that early neurological deterioration is increased with prediabetic levels of glycated hemoglobin and DM, and fibrinogen levels are associated with neurological deterioration. Furthermore, a moderate level of presenting hyperglycemia is associated with poor outcomes and infarct growth patients that where treated with endovascular reperfusion therapy for acute occlusion of the intracranial large artery in the anterior circulation. These clinical results suggest that insulin resistance and beta cell dysfunction, which is the two cornerstone of type 2 DM, may adversely affect the outcomes of acute ischemic stroke even at the pre-diabetic levels. A recent cohort study has indeed shown that beta cell dysfunction assessed by homeostatic model assessment (HOMA) 2-ß% (Pan et al., 2017) and insulin resistance as assessed by homeostatic model assessment-insulin resistance (HOMA-IR) (Jing et al., 2017) was both associated with 12-month poor prognosis in nondiabetic patients with ischemic stroke. Thus, therapeutic strategies to modify insulin resistance and beta cell dysfunction may be needed, and neuroprotective roles of anti-diabetic medications could be pursued through such approach.

Metformin is an antidiabetic drug that acts through transient inhibition of the mitochondrial respiratory chain complex I, resulting in indirect activation of AMP-activated protein kinase (AMPK) (Sanders et al., 2007). By inhibiting gluconeogenesis, induces muscle cell glucose uptake, and lowering blood glucose and insulin levels (Dowling et al., 2011), metformin counteracts insulin resistance.

Neuroprotective roles of metformin pretreatment have been reported, with decrease of apoptosis through AMPK activation (Ashabi et al., 2014) or pre-activation of AMPK dependent autophagy (Jiang et al., 2014).

Evogliptin is a potent long acting DPP-4 inhibitor for treatment of type 2 diabetes mellitus (Kim et al., 2012). Increased levels of glucagon-like peptide-1 (GLP-1) stimulates insulin synthesis at the translational level, maintaining beta cell secretory capacity and insulin stores (Perfetti and Merkel, 2000). Furthermore, GLP-1 signaling may promote beta cell proliferation while enhancing resistance to apoptosis (Li et al., 2003). Neuroprotective role of GLP-1 receptor analogues (Briyal et al., 2012) and DPP-4 inhibitors (Darsalia et al., 2013) have been reported.

Herein, we sought to reproduce the clinical situation of glucotoxicity in acute stroke situations of ischemia-reperfusion by using streptozotocin (STZ) induced rat models of transient middle cerebral artery occlusion (tMCAO) and evaluated the potential benefits of two potential neuroprotective antidiabetic medication, metformin and evogliptin, with main mode of action on insulin resistance and beta cell dysfunction, respectively.

II. METHODS