The past decade has witnessed substantial advances in the treatment of acute ischemic stroke (AIS). Evidence-based reperfusion therapies such as intravenous thrombolysis (IVT) and mechanical thrombectomy have been shown to improve outcomes in AIS and become the standard care for AIS patients. However, despite these major improvement, only less than half of patients achieve functional independence (mRS 0–2) at 90 days as a result of treatment, leaving the others at a high risk of disability and death [1]. Most of the poor 90-day outcomes in AIS after reperfusion therapy (i.e., IVT and/or endovascular treatment) are reported to be largely associated with early neurological deterioration (END). END generally refers to a $\ge$4 point increase in the National Institutes of Health Stroke Scale (NIHSS) score between baseline and 24 hours after treatment. The estimates of END incidence vary widely in AIS patients, ranging from 8% to 28% of patients after IVT [2, 3, 4, 5, 6], and from 35% to 42% of patients after endovascular treatment (EVT) [7, 8, 9]. However, the reasons of END occurrence were not fully understood, some of which are related to symptomatic intracranial haemorrhage (sICH), malignant edema and early recurrent ischemic stroke, while others are remaining unexplained. Since END is closely related to the increase of disability and mortality following reperfusion therapy, it is desirable to identify factors associated with END and to explore the underlying mechanism, so that possible prevention and treatment could be done to improve clinical outcomes.

Thus far, END has only been addressed in either IVT or EVT setting. Few studies considered IVT and EVT as a whole for reperfusion therapy. Unfortunately, the latter is more like the cases in the real world. Furthermore, the exact rate and predictors of END have not been systematically investigated. Therefore, we present here a systematic review and meta-analysis of the predictors of END following reperfusion therapy (IVT and/or EVT) in AIS patients. Specifically, we categorized the predictors of END into three groups, IVT, EVT and overall reperfusion therapy, to study the treatment-specific risk factors.

The search yielded 6833 relevant records from PubMed, Web of Science, and EBSCO. Finally, a total of 29 studies involving 65,960 AIS patients suffering from END within 24 hours following IVT and/or EVT were included [1, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]. The specific study selection process was reported in a PRISMA flowchart (Fig. 1). The basic characteristics of all included articles were shown in Table 6 (Ref. [15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]).

The main findings from this meta-analysis focused on investigating the predictors of END occurrence in AIS patients who receiving reperfusion treatments. Elderly, systolic blood pressure, glucose levels at admission, treatment onset, history of previous diseases (i.e., hypertension, diabetes mellitus, arterial fibrillation), internal carotid artery occlusion, and middle cerebral artery M2 occlusion significant associated with END in AIS patients who experienced overall reperfusion therapy. In addition, elderly, SBP, glucose level at admission, OTT, history of hypertension, DM, AF, large artery atherosclerosis, internal carotid artery occlusion, middle cerebral artery M2 occlusion and bridging therapy associated with END in patients who experienced IVT, and four risk factors including SBP, history of hypertension and internal carotid artery occlusion were found to be related with END in patients witnessed EVT.

The definition of END was not clear in existing studies because of the degree of symptom worsening and the time frame of the deterioration. Some literature adopted an increase of $\ge$2 points of the NIHSS score within the pre-defined time frame ($\mathrm{\Delta }$NIHSS $\ge$2) as the definition of END. View the fact that $\mathrm{\Delta }$NIHSS equals 2 seems to be too small changes in NIHSS out of a total score of 42 points and reflects inadequate reliability of the score itself rather than real symptom worsening, especially for severe stroke with high scores [44, 45]. We used $\mathrm{\Delta }$NIHSS $\ge$4 as the definition of END in this systematic review and meta-analysis, which was consistent with most studies on END. In terms of the time frame for END, although a number of END cases caused by malignant edema tended to occur beyond the first 24 hours, we adopted within 24 hours as the timeframe of END, which was consistent with most of the studies.

Within the currently acceptable timeframe and degree of worsening of END, multiple risk factors identified in this study were in line with several previous studies. Age has been identified as a risk factor for END onset in Birschel’s study [46]. In our study, the pooled data analysis showed that older AIS patients were at higher risk of END after revascularization treatment, especially more obviously presented underwent IVT treatment in the subgroup analysis, whereas no significant association between age and the occurrence of END was observed in patients receiving EVT treatment. This result might relate to a small number of studies included in the EVT group, limited by the number of patients. In addition, another reason might be that patients receiving EVT treatment have a more strict age restrictions and a smaller age span. Initial NIHSS score is used widely to measure a level of consciousness on admission, specifically, a larger initial NIHSS score indicated a declined consciousness which is related to a greater chance of END achieved [47]. However, in our analysis, the initial NIHSS score was not statistically significant in predicting the occurrence of END. In previous studies, the initial NIHSS score was a significant predictor of END24, both of which used a liberal definition of END24 (worsening of Scandinavian Stroke Scale, which might be more sensitive than NIHSS, and $\mathrm{\Delta }$NIHSS $\ge$1, respectively). Since a severe neurological deficit strongly predicts sICH and malignant oedema. In contrast, another study [41] using a more conservative definition (namely, $\mathrm{\Delta }$NIHSS $\ge$4) found the inverse association, that is, less severe deficits predicted END24, which might be explained by a ‘ceiling effect’ such that higher admission scores are less likely to further increase. Thus, the risk of END24 associated with admission to NIHSS might depend on the definition of END, including whether absolute or relative changes are considered. This is consistent with the definition of END in the included studies in our analysis, so the results are similar. Systolic blood pressure is a significant predictor of increased END risk in our meta-analysis. In both overall and subgroup analyses, in AIS patients treated with IVT or EVT, higher SBP was associated with a greater incidence of END. Augmented systolic blood pressure has been mentioned to predict early neurological deterioration in the preceding study [48]. The exact cause is not yet clear, but it may be related to the fact that high blood pressure can easily aggravate cerebral edema, which occurs after acute cerebral infarction, particularly in patients with extensive oligaemia due to proximal occlusion. This hypothesis has not been directly tested so far. Serum Glucose is a significant predictor of increased END risk.

History of diabetes mellitus is associated with a higher risk of END occurring [49, 50]. Notably, the poor neurological outcome (i.e., END) has resulted from persistently high level of serum glucose following reperfusion treatments [51]. One potential mechanism could be that it increased blood-brain barrier disruption and promoted sICH. However, hyperglycaemia might also facilitate neuronal damage. OTT (Onset to Treatment Time) was another significant predictor of increased END risk in our meta-analysis. Patients with longer OTT were more likely to have END, which is consistent with the concept of “time is brain” that we have known before. In the subgroup analysis, it was more obviously found in the IVT group, but in the EVT group, the association between age and the occurrence of END was not statistically different. We speculated that the reason was the large heterogeneity in OTT among patients receiving EVT. With the improved concept of thrombectomy, many AIS patients beyond the time window with salvage of brain tissue still received EVT. This might be the reason why OTT time did not fully predict the occurrence of END in EVT patients.

In addition, Kwan’s study [52] found atrial fibrillation aggrandized the risk of END attained, which was also a significant predictor of increased END risk in our meta-analysis. It might be due to the stroke etiology in patients with atrial fibrillation was cardiogenic embolism. The occlusion site was usually large vessel, with less collateral compensation and a larger infarct size. However, in another study [53], which focused on non-thrombolysis mild strokes, the incidence of END24 was 9% in lacunar infarcts as compared with 31% and 23% in large atherosclerosis and cardioembolic strokes, respectively, which showed a non-significant difference. Since stroke was caused by many reasons, the mechanism of END in various subtypes of stroke was completely different, yet remains largely unknown. Proximal arterial occlusion was observed more frequently in patients with END in our meta-analysis. One possible explanation for this association would posit that proximal occlusion predicts stroke severity, and therefore also END through its association with sICH and malignant oedema.

There are a few limitations in this meta-analysis that ought to be considered when interpreting the findings. Although we performed the meta-analysis for AIS cases with EVT, a sufficient subgroup analysis providing reliable results in the EVT group equivalent to IVT and total group may not be possible as scarce EVT studies are included. Due to insufficient data of re-occlusion and collateral circulation reported, it was difficult to analyze these two factors in this meta-analysis. Nevertheless, re-occlusion has been recognized as one of the important prognoses of END by clinical physicians [15], as well as playing a predominant role to improve the neurological functions for stroke patients. Another mechanism of END focuses on the insufficient collaterals with the adverse metabolic consequences among AIS patients in clinical practice [54]. Accordingly, collateral status described a linear relationship of the NIHSS score, implicating such the mechanism moderating symptoms in the hyperacute and post-therapeutic acute stroke phase [22]. Despite the ASPCT score contributing a lot to END occurring [1], we were not able to find a significant association of END as assigning the ASPECT score may vary by different physician.

The strength of this meta-analysis is that several large sample studies were involved to conduct the meta-analysis, providing convincing evidence of the associations, as well as depicting a complete picture to predict END occurring in stroke patients with medical or reperfusion therapies in clinical practice. We not only focused on END occurrence after EVT, but we also determined that the site of occlusion was linked with END after EVT.

This study contains some implications in clinical practice. END at 24 hours was associated with poor outcomes, hinting toward needs of the concentration on neurological deficit management. Understanding and monitoring older age, elevated systolic blood pressure, high levels of serum glucose, hypertension, diabetes mellitus and atrial fibrillation may elevate the early neurological worsening and achieve a satisfactory prognosis. Apart from this, other unmentioned factors, including collateral circulation, ASPECT score, CE stroke, are supposed to be thought about in decision-making protocols before reperfusion therapy conducting. Further studies ought to look into END occurring after EVT deeply because the exact proportion of END onset following EVT, and short-term/long-term prognosis still remain unclear at this stage.

All data generated or analyzed during this study are included in this published article. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. PRISMA checklist is shown in supplementary materials.

JL and WJJ take responsibility for the integrity of the data and the accuracy of the data analysis. Conceptualizing and designing the study—JL, HXS, CL and YQZ. Extracting, analyzing and interpreting the data—All authors. Drafting the study—HXS, CL and JL. Conducting meta-analysis—JL, HXS, CL, YQZ, XL, AFL,YEL. Critical revision of the study for important intellectual content—All authors. Obtaining funding—WJJ. Supervision—JL and WJJ.

Not applicable.

Not applicable.

This study was supported by the grants from National Key Basic Research Program of China (973 program) (grant No. 2013CB733805), and the National Natural Science Foundation of China (grant No. 81871464).

The authors declare no conflict of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/j.jin2202052.

Appendix I: Detailed Search strategy

Pubmed

Free Text

Initial search performed on 3 August 2021

Search updated on 7 December 2022

English language limit applied

Web of Science

Free Text

Initial search performed on 2 August 2021

Search updated on 6 December 2022

English language limit applied

EBSCO

Free Text

Initial search performed on 16 August 2021

Search updated on 10 December 2022

English language limit applied

Appendix II. Quality score

The quality score is composed of 5 items, and each item was allocated 0, 1 or 2 points. This allowed a total score between 0 and 10 points, 10 representing the highest quality. The following items are included in the score:

Objective

0 for no study objectives mentioned

1 for study objectives reported but non-specific

2 for specific study objectives reported

Design

0 for cross sectional studies

1 for case-control studies

2 for longitudinal studies (retrospective or prospective) or interventional studies

Population

Observational studies

0 if n 100

1 if n 100 to 500

2 if n $>$500

Outcome (see table below)

Adjustments

0 if findings are not controlled** for at least age and gender

1 if findings are controlled for age and gender

2 If findings are additionally controlled for covariates:

** ‘Controlled for’ here refers to: adjusted for in the statistical analyses (e.g. with multiple regression); stratified for in the analyses (e.g. males and females separately)