Glycoprotein inhibitors as a first line rescue treatment after unsuccessful recanalization of endovascular thrombectomy: A systematic review and meta-analysis

Abstract

Background

Intracranial atherosclerotic disease (ICAD) is a major cause of stroke with a high rate of re-occlusion following mechanical thrombectomy (MT). Among the available rescue options, glycoprotein IIb/IIIa inhibitors (GPI) have shown promise as a potential therapeutic strategy. This systematic review and meta-analysis examine studies exploring the use of glycoprotein inhibitors as a first-line treatment for refractory occlusion or high-grade stenosis following EVT in the setting of ICAD.

Methods

A systematic review and meta-analysis were performed. Studies using GPI as the first-line rescue treatment (GPI-rt) after failed thrombectomy or in the setting with high-grade stenosis (>50%) were included. The primary outcome of interest was good clinical outcomes (defined as a modified Rankin Scale (mRS) score of 0–2 at 90 days). Secondary outcomes of interest were successful recanalization (TICI 2b-3), symptomatic intracranial hemorrhage (sICH), and mortality by 90 days.

Results

Our study processed 2111 articles, which yielded eight relevant studies for review, four single and four double arm. These studies comprised 763 patients, divided into GPI-rt (535 patients) and non-GPI-rt (228 patients) cohorts. The GPI-rt group had higher rates of mRS ≤ 2 at 90 days (58.5% vs 38.9%, p = 0.002) and lower mortality rates (7.8% vs 17.5%, p = 0.04) compared to the non-GPI-rt cohort. mTICI 2b-3 rates and rates of sICH were not significantly different between the cohorts.

Conclusions

First line GPI-rt demonstrates significant clinical benefit and significantly lower mortality without a rise in rates of sICH. GPI are a potential first line rescue treatment of ICAD.

Introduction

Large vessel occlusion (LVO) secondary to intracranial atherosclerotic disease (ICAD) has a prevalence ranging from 10% to 48%. ¹ Mechanical thrombectomy (MT) alone using stent retrieval, aspiration, or both is inadequate in reversing subintimal atheroscleroma. Thrombectomy may also cause further damage to the vascular intima, further exposing the underlying atheroscleroma, resulting in a thrombogenic nidus. Increased thrombus formation coupled with stenosis of the vessel in response to previous occlusion places patients with ICAD at increased risk of vessel re-occlusion. ²

Treatment for vessel re-occlusion has historically involved mechanical intervention such as balloon angioplasty and/or stenting.^3–5 However, there are risks involved with mechanical intervention including rupture of vessel, stent-stenosis, new embolic strokes, and new-perforator strokes.^4,6 Pharmacologic treatment modalities such as long-term antiplatelet therapy have also been a mainstay in treatment of re-occlusion. Glycoprotein IIb/IIIa inhibitors (GPI) were originally recognized for their utility in percutaneous coronary intervention and more recently as a useful option for post-stroke care, especially in ICAD related LVO. ⁷ In the event of re-occlusion, GPI has shown promise in improving clinical outcomes, particularly when utilized in conjunction with MT.^7–10 However, there has been no randomized trial investigating the safety and efficacy of GPIs as a primary rescue treatment nor is there a consolidation of all the retrospective studies of this intervention specifically, with the European stroke organization guidelines meta-analysis including only three studies. ¹¹

We performed a systematic review and meta-analysis of the literature on the efficacy and safety of GPI as an acute standalone post-EVT rescue treatment for failed or threatened thrombectomy in the setting of ICAD-LVO.

Methods

The search strategy was developed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. ¹² Using the common evidence medicine framework PICOS (Patient Population, Intervention, Control/Comparison, Outcome, and Study type), we specified our research question: “In patients with AIS who underwent endovascular thrombectomy but failed to recanalize or who had severe stenosis, are GPIs a effective and safe as a first-line rescue treatment option?” An electronic search of PubMed, EMBASE, Cochrane, and OVID databases was performed from date of database inception to March 2023. No review protocol currently exists, and the following search query was performed without language restriction to identify all relevant titles and abstracts using key-words: glycoprotein, antiplatelet, tirofiban, eptifibatide, large vessel occlusion, stroke, ischemic stroke, thrombectomy, endovascular intervention, aspiration, rescue therapy, rescue, refractory, stenosis. These studies were screened for duplicates across databases and were secondarily screened for duplicates and for concordance with inclusion and exclusion criteria within Covidence (Melbourne, Victoria, Australia) systematic review software.

Selection criteria

To meet inclusion criteria, studies must present a patient cohort with the following: (1) diagnosed with AIS; (2) Stroke diagnoses were predominantly anterior circulation; (3) EVT performed without successful recanalization (modified TICI ≤ 1) or with residual stenosis  ≥ 50% concerning for re-occlusion; (4) majority of cases identified as being predominantly related to intracranial atherosclerosis; (5) received GPI either as intra-arterial (IA) or intravenous (IV) as primary rescue treatment after unsuccessful or threatened recanalization; (6) clinical outcome data following rescue GPI treatment available; (7) patient cohort older than 18 years of age. Reports were excluded if the language indicated that patients received GPI and stenting or angioplasty as a combined first line rescue therapy. Reports were excluded if angioplasty/stenting made up a majority of GPI-rt cases. Further, we excluded articles that were editorials, letters, expert opinions, reviews without original data, animal studies, pediatric cohorts, and articles not available in the English language.

Data extraction and statistical analysis

Patient demographics were collected when provided. Outcome data were extracted from the texts, tables, and figures from all articles. Descriptive statistics of the studies were summarized. A random intercept logistic regression model (generalized linear mixed model) was used for analyses. We used the “metaprop” function within the “meta” package ¹³ in R 4.2.2 (R Core Team, Vienna, Austria) and Rstudio (Posit Software, Boston, MA, USA). Forest plots were used to display results. These analyses facilitated assessments for group differences, and tests for heterogeneity across studies were done using the I² statistic. Funnel plots were also generated and visually assessed to evaluate potential for publication bias.

Results

A PRISMA flow diagram of the study selection process is shown in Figure 1. The literature search returned 2643 articles after a preliminary duplicate screen. After removing 531 additional duplicates, 2111 articles were retained. After screening titles and abstracts based on the selection criteria, 43 articles remained. After a full-text review, 35 articles were excluded, leaving a total of 8 articles. Four of these studies were single arm. There was a total of 763 patients, 535 in the GPI-rt arm and 228 non-GPI-rt. There was evidence of heterogeneity in all four of our outcome variables. Significant heterogeneity (I²≥ 50%) was seen in Modified Thrombolysis in Cerebral Infarction (mTICI), modified Rankin Scale (mRS), and mortality.

Figure 1.

PRISMA flow diagram.

Study characteristics

The study characteristics can be seen in Table 1. All the studies included in our analysis were retrospective in nature. GPI-rt data were collected from all eight of the studies and control arm data was collected from four of the studies. The largest study consisted of 140 patients and the smallest contained 39 patients. Patient demographics were collected from seven of the eight studies. Demographic information was not available from the ICAD sub-group in Zhang et al. The weighted average age among the GPI-rt studies was 67.1 years (range 61.9–71.0 years) and 68.3 (range 63–72.2 years) among non-GPI-rt. The weighted average NIHSS was 14.1 among GPI-rt cohorts and 15.3 among non-GPI-rt cohorts. Infarct territory was anterior circulation in 85.1% of the pooled GPI-rt cohort and 87.5% of the pooled non-GPI-rt group. The rate of intravenous thrombolysis was 41.3% in the pooled GPI-rt cohort and 46.9% in the pooled non-GPI-rt cohort. Intra-arterial tirofiban was utilized in all studies and was the single intervention of investigation in seven of the GPI-rt arms, with one study also reporting intra-arterial abciximab (see table). Intravenous tirofiban was reported in only two of the studies (Yan et al. and Baek et al.). Among the 228 control rescue treatment arm patients, 118 (51.8%) received angioplasty and/or stenting. Other rescue treatments described included multiple passes, switching thrombectomy modality, and medical management.

Table 1.

Characteristics of included studies.

Study	Total number of patients			Age mean (SD) or median [IQR]		Pre-NIHSS median n = (%)		IV tPA n = (%)		GPI usage		Other rescue treatments
	Total	GPI-rt	non-GPI-rt	GPI-rt	non-GPI-rt	GPI-rt	non-GPI-rt	GPI-rt	non-GPI-rt	IA GPI dosing	IV GPI usage	GPI-rt	non-GPI-rt
Sun 2019	39	39	0	61.9 (10.9)	-	14 [9–20]	-	13 (33.3)	-	0.25–0.5mg	None described	Angioplasty 11 cases; Permanent stenting 6 cases;	-
Yan 2020	98	50	48	68.1 (13.9)	70.5 (13.8)	19 [13–25]	21 [13–30]	43 (86)	34 (70.1)	0.4–0.5mg	0.4–0.5mg/hr ×24 h	Balloon angioplasty 4 cases;	Balloon angioplasty 5 cases
Kang 2014	40	40	0	67.2 (11.1)	-	14.6 (4.96)	-	14 (35)	-	0.5–1mg	None described	NA	-
Kang 2019	140	68	72	68 [60.25–75]	66 [57–73.5]	15 [11–20]	11 [9–15]	27 (39.7)	28 (38.9)	0.5–1mg	None described	intracranial Angioplasty/stenting 6 cases; carotid stenting 4 cases	Angioplasty/stenting 72 cases; IA fibrinolytics; carotid stenting 1 case
Kim 2020	119	59	59	71 [61–78]	63 [55–75]	14 [10–20]	15 [12–21]	30 (33.9%)	29 (49.2)	0.5–2mg	None described	Balloon angioplasty 2 cases; stenting 6 cases	Balloon angioplasty 9 cases; stenting 12 cases
Baek 2021	133	84	49	65.8 (13.2)	72.2	12.0 [7.0; 15.0]	16.25 (0.7)	25 (29.8)	16 (32.7)	0.3–1.5mg Tirofiban or 5–10mg abciximab	73 (87% of GPI group); no specified dosing regimen	0 stenting/angioplasty	Stenting 25 cases; Medical management 24 cases
Choi 2022	115	115	0	67.0 (59.0–75.0)	-	13.0 [9.0–19.0]	-	36 (31)	-	0.5–2mg	None described	0 stenting/angioplasty	-
Zhang 2019	80	80	0	-	-	-	-	-	-	0.5–1mg	NA	Stenting/angioplasty unspecified number of cases	-

Risk of bias

There was considerable risk for bias among published studies. Publication bias was present, based on non-symmetrical forest plots, in all four outcome variables, including mRS, mTICI, mortality, and symptomatic intracranial hemorrhage (sICH).

Modified Rankin score of 0–2 at 90 days

All eight studies including all 535 patients in the GPI-rescue group reported 90-day mRS, while the non-GPI group had four studies, including 228 patients, reporting mRS. The overall rate of mRS ≤ 2 in our random effects analysis was 52.1%. The GPI-rescue group demonstrated a significantly higher rates of mRS ≤ 2 at 90 days when compared to the non-GPI group (58.5% (95% CI [54.3; 62.6%]) vs 38.9% (95% CI [28.4%; 50.5%]), p = 0.002) (Figure 2). There was significant heterogeneity in this outcome measure (I² = 69.8%) and presence of publication bias as seen by the asymmetry of the funnel plot (Figure 3).

Figure 2.

Forest-plot of meta-analysis for proportions of 90-day mRS 0–2.

Figure 3.

Funnel-plot demonstrating asymmetry of studies’ results of 90-day mRS 0–2.

Mortality

Seven studies, representing 495 GPI-rt patients, reported mortality. In the non-GPI-rt, there were four studies, including 228 patients, reporting mortality. The overall mortality rate was 10.6%. The GPI-rt group demonstrated a significantly lower mortality rate when compared to the non-GPI group (7.8% (95% CI [5.1%; 11.7%]) vs 17.5% (95% CI [12.0%; 24.8%]), p = 0.04) (Figure 4). There was significant heterogeneity in this outcome measure (I² = 66.3%) and presence of publication bias as seen by the asymmetry of the forest plot (Figure 5).

Figure 4.

Forest-plot of meta-analysis for proportions of mortality rates.

Figure 5.

Funnel-plot demonstrating asymmetry of studies’ results of mortality.

Modified TICI score

There were five studies in the GPI-rt group, including 366 patients, and three studies in the non-GPI-rt group, representing 180 patients, that reported mTICI. The overall rate of mTICI2b-3 was 90.9%. The GPI rescue group had non-statistically different rates of mTICI 2b-3 when compared to the non-GPI group (94.5% (95% CI [83.8%; 98.3%]) vs 79.1% (95% CI [42.8%; 95.1%]), p = 0.142) (Figure 6). There was significant heterogeneity in this outcome measure (I² = 89.7%) and presence of publication bias as seen by the asymmetry of the forest plot (Figure 7).

Figure 6.

Forest-plot of meta-analysis for proportions of mTICI 2b-3.

Figure 7.

Funnel-plot demonstrating asymmetry of studies’ results of mTICI 2b-3.

Symptomatic intracranial hemorrhage

sICH was reported in 7 GPI-rt arms and three non-GPI rt arms, representing 476 and 220 patients, respectively. The overall rate of sICH in the cohort was 3.4%. The rates of sICH between the groups were similar (2.6% (95% CI [0.84%; 7.7%]) vs 5.2% (95% CI [2.4%; 11.1%]), p = 0.310) (Figure 8). There was no significant heterogeneity in this outcome measure (I² = 0.6%). There was presence of publication bias as seen by the asymmetry of the forest plot (Figure 9).

Figure 8.

Forest-plot of meta-analysis for proportions of sICH.

Figure 9.

Funnel-plot demonstrating asymmetry of studies’ results of sICH.

Discussion

Our meta-analysis included 683 patients and showed a significantly greater proportion of patients achieving functional independence at 90 days associated with GPIrt. In addition, we demonstrated a significant mortality benefit of GPIrt and a similar hemorrhage risk to non-GPIrt but no difference in rates of successful recanalization. Our analysis expands on the recently published European Stroke Organization guidelines. ¹¹ In their meta-analysis of three studies (Baek et al., 2021; Kim et al., 2020; and Yan et al., 2020),^14–16 they report a statistically significantly greater unadjusted odds of functional independence associated with GPI rescue (OR, 2.97; 95%CI [1.82–4.84]) but a non-significant adjusted OR. Additionally, they report that the unadjusted OR of mortality was significantly lower in the GPI group (OR, 0.24 95% CI [0.11–0.52]) and that unadjusted OR of achieving recanalization were not different between studies. Our analysis includes these three studies and five additional studies; these include two arms from Kang et al., 2019 and four intervention arms from Choi et al., 2022; Sun et al., 2019; Zhang et al., 2019; and Kang et al., 2014.^17–20

There was significant variability and vague details surrounding treatment indications and decision making. Most studies stated that the use of GPI was at the discretion of the treating physician. Many studies listed multiple indications for GPI-rt including re-occlusion following MT, multiple passes (≥3), failure to recanalize with MT, and significant stenosis. Some studies merely included severely stenotic patients (Yan et al., 2020; Kang et al., 2014; Kim et al., 2020)^15,16,20 but the degree of stenosis cited ranged from ≥50% to ≥70%.^15,16,21 This heterogeneity in interventional indications likely indicates heterogeneity in the initial patient cohorts.

There was also variability between studies in regard to treatment groups. The overall NIHSS were relatively similar between GPI-rt and control arm, however ranges of median NIHSS did vary considerably between studies, from as low as 11 in Kang et al., 2019 to as high as 21 in Yan et al.^16,21 Similarly, the rates of thrombolysis, while not significantly different within each study and overall similar in pooled rates, ranged significantly between studies. Baek et al., 2021 reported thrombolysis rate of only 29.8% among GPI-rt patients whereas Yan et al. reported thrombolysis rates of 86% among their GPI-rt group. It should also be noted that our analysis did not study the use of GPI in general but rather the specific treatment of GPI as a first-line treatment. In terms of primary treatment group, we aimed to minimize any extraneous treatment effects by selecting studies in which GPI-rt was initiated as a first line treatment without initial angioplasty or stenting. For example, Baek et al. included a treatment arm of GPI plus stenting as one of the treatment arms, which we excluded from our analysis.

While none of the studies include angioplasty/stenting as a combined first line treatment with GPI, five of the studies (Kim et al., Yan et al., Zhang et al., Kang et al. 2019, and Sun et al.) did have a moderate to significant proportion of angioplasty and stenting in the GPI-rt group.^{15,16,18,19,21} Kim et al., Kang et al. 2019, and Yan et al. included a small proportion of angioplasty/stenting but Sun et al. had a fairly significant proportion at 43.6%, however permanent stenting was used in only 15.4%. Zhang et al. did not report the exact proportion of their atherosclerotic patients who received angioplasty or stenting, but it likely represented a significant proportion although less than 50%. The details on these cases were not clearly defined within the studies and we cannot tell whether the patients receiving angioplasty/stenting did so secondary to failed GPI rescue or simply as a backup measure based on interventionalists preferences.

The cases of angioplasty, stenting, and medical management were all lumped into our control arm. There has been growing interest in the use of stenting for refractory thrombus. While the SAMPPRIS trial showed worse outcomes after stenting in the setting of mild to moderately symptomatic stenosis this is not necessarily generalizable to ICAD LVOs. ⁴ A recent meta-analysis on the utility of rescue stenting reported an significant clinical benefit and improved safety outcomes associated with rescue stenting. ²² In our study, one of the clearest examples of the potential benefit was the retrospective study by Baek et al., 2021 which analyzed prospectively collected data in Chinese patients with acute ICAD-related large-vessel occlusion. Their study demonstrated that GPI-rt was associated with better functional outcomes compared to stenting alone. But they also demonstrated that the combination of rescue stenting and concomitant IA-GPI infusion resulted in significantly better outcomes in terms of functional independence at 90 days compared to either GPI-rt or stenting alone. ¹⁴ While we excluded this combination treatment group from our analysis, this combined treatment modality may be a future direction of research that should be investigated further.

GPI for EVT

Multiple studies and meta-analyses have reported the potential benefits of concomitant GPI injection either intravenously or intra-arterially for improving revascularization rates and functional outcomes in patients undergoing endovascular treatment for stroke, especially in the presence of ICAD. One such study was a matched-control study by Ma et al., 2022, which evaluated the safety and efficacy of combined MT and eptifibatide for acute ischemic stroke. Their analysis revealed that patients who received eptifibatide concomitantly with EVT had higher recanalization rates as well as greater odds of achieving favorable functional outcomes with no increased risk of sICH or 3-month mortality. ²³ Multiple meta-analysis on the use of GPI in conjunction with endovascular thrombectomy reported benefits from the usage of GPI including improved odds of favorable functional outcomes, increased recanalization rates, and decreased mortality rates.^8,24,25 The observed benefits reported in these studies are attributed to the specific mechanism of GPIs by disrupting platelet-to-platelet crosslinking, thus disrupting the activation of the platelet cascade and eventual coagulation cascade. ²⁶ This effect may be particularly beneficial in patients with intracranial atherosclerosis, as the underlying pathology involves platelet dependent clot formation along a highly thrombotic plaque resulting in higher rates of re-occlusion following MT. A randomized clinical trial of IV tirofiban before thrombectomy failed to show functional benefit across all patients although the rates of rescue treatment required were significantly lower among patients pre-treated with IV tirofiban as compared to the placebo (8.4% vs 12.0%, p = 0.04). ²⁷ Furthermore, post-hoc analysis on the included ICAD patients demonstrated that tirofiban was associated with significantly greater odds of functional independence (adjusted odds ratio 1.68; 95% CI 1.11–2.56, p = 0.02). The authors also showed that 20% of this improvement in functional outcomes was explained by the significant reduction in number of thrombectomy passes required per patient. ²⁸

While the target of action is similar across the three GPI, there are also differences in medication pharmacokinetics that should be noted. Tirofiban, the most frequently reported drug in our study, has high specificity for GP IIb IIIa receptor and binds reversibly in a competitive inhibitory manner. It has a short time of action, and will dissociate rapidly from platelets, resulting in reversed antiplatelet effects after 3 h of discontinuation. ²⁶ Abcixmab, on the other hand, is a monoclonal antibody with a high binding affinity and low reversibility.^26,29 These differences lead to real-world implications. A meta-analysis of the safety of these three medications in IV in the setting of stroke reported that both tirofiban and abciximab were associated with lower risk of 90-day mortality. They did report that abciximab was associated with higher risk of sICH and eptifibatide with lower risk. Tirofiban was not associated with higher risk of sICH except for old patients and those with high NIHSS. Exposure to EVT did not alter tirofiban's risk of sICH. ³⁰ Similar to the results of the previous meta-analysis, we found no association between GPI-rt and elevated odds of sICH. We also found a robust mortality and functional outcomes benefit associated with GPI-rt despite a non-significant difference in mTICI 2b-3 rate.

The delivery of GPI, IA vs IV should also be considered. A retrospective propensity-score matched cohort analysis of patients receiving either IV or IA tirofiban as an adjunct with EVT by Yang et al. reported an association of IV tirofiban with greater recanalization rates, and 3-month rates of mRS 0–2 whereas IA delivery of tirofiban was associated with increased hemorrhage, mortality, and poor functional outcomes. However, the treatment groups were unequal as the IA tirofiban cohort represented a generally more complicated patients often requiring multimodality rescue treatment. ³¹ All the articles included in our analysis report the use of local intra-arterial tirofiban. The most common dose described was 0.5–1 mg, however the range spanned from 0.25 to 2 mg. There was one study that utilized intra-arterial abciximab (5–10 mg) for some patients as well as tirofiban in other patients, but the patients were all pooled together under GPI-rt. ¹⁴ There was heterogeneity reflected in the utility of IV GPI with only two studies describing an IV tirofiban drip.^14,16 The preference of tirofiban seems to reflect the trends in the greater stroke literature as well, with tirofiban being the most frequently cited GPI in stroke research. ³⁰

Strengths of our study

While the role of GPI in stroke has been studied in many retrospective studies, there is significant variability in the drugs’ use, with many meta-analyses investigating the utility and safety of GPI with EVT as a combination treatment, or even as a bridging-treatment before EVT.^8,9,25,32 Among the literature of rescue treatment there is also considerable heterogeneity and imbalance in the reported studies of GPI-rt, with many studies including vague language in regard to GPI use and indications. Several studies reported stenting and angioplasty, or distal embolization as indications for GPI which we thereby excluded from our analysis as it was not consistent with our pre-specified PICO question. There was also considerable heterogeneity among studies in regard to non-rescue comparison arms. Many studies were vague in their description of the control arm, and it was unclear whether the control arms were true controls in the sense that they had severe stenosis or re-occlusion. We thoroughly analyzed each paper for adequate comparison arms but found only four which indicated sufficient qualifications of a non-GPI-rt control arm. We decided to include single interventional arms from studies that did not meet satisfactory pre-defined definitions for control to capture a greater representative sample of the published literature. Several of these studies included a second arm but did not meet sufficient requirements for inclusion as a comparison group in our analysis. For example, Sun et al. grouped patients by stroke etiology as well as the use of GPI. While they included an ICAD group with non-GPI-rt, these patients were never specified to have required rescue treatment at all and were therefore not a legitimate comparison group. ¹⁸ Kang et al., 2014 created a complex patient analysis by first subcategorizing by the presence of re-occlusion requiring GPI and then further partitioning all ICAD and non-ICAD into groups for comparison. ²⁰ Because the majority of GPI-rt were categorized as ICAD, and likewise the majority of ICAD patients received GPI-rt, we considered this to be a single arm study.

The decision to use a random effects model enhances the validity of our analysis by accounting for heterogeneity across studies, providing more conservative estimates and wider confidence intervals, better adjusting for confounding variables, and improving the generalizability of our findings. These factors set our analysis apart from the prior meta-analysis which reported unadjusted odds ratios, which may be more prone to bias and limited in its generalizability.

Limitations

Given the retrospective nature of the studies included, there is high publication bias among these studies as observed in our funnel plots. The findings must be taken with this in mind. Reporting bias also likely underrepresents poor outcomes in the literature leading to a potential overestimation of true effect size and inability to generalize results across more heterogeneous populations. There is also a lack of consistent reporting across studies as noted in our results section. This severely limits our comparisons of mortality, sICH, and recanalization rates. In addition, while all studies were retrospective, the variability in study designs limits the comparability of these studies and introduces significant bias and possible confounders. For instance, in Kang et al., 2019 the rescue treatment groups were based on the hospital of treatment, with one hospital treating with GPI-rt and the other with angioplasty and stenting. ²¹ This inherently introduces potential for both selection bias and intervention bias. In addition to bias, there is also inherent noise in terms of the radiographic diagnosis and grading of stenosis. ³³ Most of the studies investigating GPI in stroke, including all of the studies in our meta-analysis, have been conducted in Asian populations which have reportedly higher rates of ICAD as compared to American or European patient populations. ³⁴

Regarding our study design, the inclusion criteria of studies was also very specific and limits the generalizability of these outcomes. The inclusion of single-arm studies is also a limitation of design. Our meta-analysis did not adjust for potential confounding factors such as age, comorbidities, and stroke severity, which could have influenced the outcomes of interest. It is possible that our search strategy missed articles that would be appropriate for inclusion and analysis. We attempted to mitigate this by including a wide search across multiple databases using a broad range of search terms, as well as searching for relevant references within the articles that escaped our search queries.

Conclusion

The observed clinical improvement and significant decrease in mortality linked to GPI-rt, without an associated increased risk of sICH, are promising. However, these results are limited by low quality evidence and significant bias in publications. Considering the high percentage of GPI-rt patients also undergoing angioplasty/stenting, our understanding of GPI as a standalone treatment is still incomplete. GPI administered intra-arterially may serve as a practical initial therapy, possibly reducing the immediate need for angioplasty or stenting, thus preventing related complications. Further rigorous investigation, including randomized controlled trials and combined rescue studies is necessary to confirm these preliminary findings and improve treatment approaches for patients with intracranial atherosclerosis and acute ischemic stroke.