A Systematic Review and Meta-Analysis of Animal Studies Testing Intra-Arterial Chilled Infusates After Ischemic Stroke

Abstract

Background: As not all ischemic stroke patients benefit from currently available treatments, there is considerable need for neuroprotective co-therapies. Therapeutic hypothermia is one such co-therapy, but numerous issues have hampered its clinical use (e.g., pneumonia risk with whole-body cooling). Some problems may be avoided with brain-specific methods, such as intra-arterial selective cooling infusion (IA-SCI) into the arteries supplying the ischemic tissue. Objective: Our research question was about the efficacy of IA-SCI in animal middle cerebral artery occlusion models. We hypothesized that IA-SCI would be beneficial, but translationally-relevant study elements may be missing (e.g., aged animals). Methods: We completed a systematic review of the PubMed database following the PRISMA guidelines on May 21, 2020 for animal studies that administered IA-SCI in the peri-reperfusion period and assessed infarct volume, behavior (primary meta-analytic endpoints), edema, or blood-brain barrier injury (secondary endpoints). Our search terms included: "focal ischemia" and related terms, "IA-SCI" and related terms, and "animal" and related terms. Nineteen studies met inclusion criteria. We adapted a methodological quality scale from 0 to 12 for experimental design assessment (e.g., use of blinding/randomization, a priori sample size calculations). Results: Studies were relatively homogenous (e.g., all studies used young, healthy animals). Some experimental design elements, such as blinding, were common whereas others, such as sample size calculations, were infrequent (median methodological quality score: 5; range: 2-7). Our analyses revealed that IA-SCI provides benefit on all endpoints (mean normalized infarct volume reduction = 23.67%; 95% CI: 19.21-28.12; mean normalized behavioral improvement = 35.56%; 95% CI: 25.91-45.20; mean standardized edema reduction = 0.95; 95% CI: 0.56-1.34). Unfortunately, blood-brain barrier assessments were uncommon and could not be analyzed. However, there was substantial statistical heterogeneity and relatively few studies. Therefore, exploration of heterogeneity via meta-regression using saline infusion parameters, study quality, and ischemic duration was inconclusive. Conclusion: Despite convincing evidence of benefit in ischemic stroke models, additional studies are required to determine the scope of benefit, especially when considering additional elements (e.g., dosing characteristics). As there is interest in using this treatment alongside current ischemic stroke therapies, more relevant animal studies will be critical to inform patient studies.

Keywords: MCAO (middle cerebral artery occlusion); focal ischemia; intraarterial cooling; meta-analysis; neuroprotection; therapeutic hypothermia (TH).

Figure 1

PRISMA diagram of studies returned by our PubMed search conducted on May 21, 2020. Inclusion criteria limited studies to those in animal models of intracarotid cooling after ischemic stroke that assessed infarct volume and/or behavior and/or edema (English only). Our search included the following terms: (focal ischemia OR stroke OR ischemic stroke OR MCAO OR middle cerebral artery occlusion) AND (intra-arterial saline OR saline infusion OR IACS OR SI-AC OR ICSI OR carotid infusion OR LCI OR local cooling infusion OR LEVI OR local endovascular infusion) AND (hypothermia OR local cooling OR focal hypothermia OR focal cooling OR cooling OR therapy OR therapeutic hypothermia OR stroke therapy OR efficacy OR translation) AND (rat OR mouse OR gerbil OR rodent OR primate OR monkey OR dog OR cat OR lamb OR patient OR human) NOT [coronary (Text Word) OR myocardial (Text Word)].

Figure 2

Analysis of experimental characteristics used in intra-arterial selective cooling infusion (IA-SCI) literature. (A) Distribution of saline infusion volumes used in IA-SCI studies. Upper and lower dotted lines represent our calculated limits for saline infusion volumes. Calculations are based on 1 L of saline being infused into a patient and scaled to brain mass (1.4 mL; lower line) and body mass (4.4 mL; upper line). (B) Breakdown of anesthetics used in IA-SCI neuroprotection studies. (C) Animal sexes used in IA-SCI neuroprotection literature. No studies used exclusively female animals. (D) Analysis of infarct assessment timing. A majority of studies investigated infarct volume between 0 and 14 days post-MCAO (median = 2 days post-MCAO). (E) Analysis of behavioral assessment timing. Similar to infarct volume assessment, a majority of studies assessed behavior between 0 and 14 days post-MCAO (median = 2 days post-MCAO). (F) Evaluation of methodological quality score for IA-SCI studies (max score = 12). MQS, Methodological Quality Score.

Figure 3

Quantitative analysis of studies that assessed infarct volume using random-effects meta-analysis. (A) Forest plot of studies investigating infarct volume (normalized mean difference ± 95% CI). Effect size estimates were heterogeneous, likely owing to study design differences. *Control group in (58) serviced 2 treatment groups, and thus to avoid outcome dependence, we divided the control group in half [procedure described in (61)]. (B) Funnel plot with trim-and-fill analysis to assess publication bias. Closed circles represent actual studies whereas open circles represent results of the trim-and-fill analysis. These results suggest that null and negative studies may be missing from the distribution (i.e., the left side of the funnel is more filled out, favoring beneficial effects of the treatment on infarct volume). Note that the effect size in Panel B is smaller than in Panel A, as this has been corrected for in the trim-and-fill analysis.

Figure 4

Quantitative analysis of studies that assessed behavioral outcomes using random-effects meta-analysis. (A) Forest plot of studies investigating behavioral outcomes (normalized mean difference ± 95% CI). Effect size estimates were very heterogeneous, likely owing to differences in MCAO duration, treatment parameters, and timing of assessment. Note: as behavioral results were combined across a number of tests, the test date could not be presented in the figure because behavioral tasks often had differing timing. (B) Funnel plot with trim-and-fill analysis to assess publication bias. Closed circles represent actual studies whereas open circles represent results of the trim-and-fill analysis. These results suggest that null and negative studies may be missing from the distribution (i.e., the right side of the funnel is more filled out, favoring beneficial effects of the treatment on behavioral outcomes). As in Figure 3, the effect size in Panel B is smaller than in Panel A, because of trim-and-fill correction.

Figure 5

Quantitative analysis of studies that assessed edema using random-effects meta analysis. Random-effects meta-analysis was conducted because of the variability in MCAO durations and treatment infusion parameters. (A) Forest plot of studies investigating behavioral outcomes (Hedge's G standardized mean difference ± 95% CI). One study was not included in the final analysis owing to the exceptionally large effect size observed therein, which caused a high amount of heterogeneity in the meta-analytic model and statistical leverage in the meta-regression model. With the exceptional study removed, study heterogeneity was reduced substantially and the meta-regression model no longer showed signs of statistical leverage. *Control group in (58) serviced 2 treatment groups, and thus to avoid outcome dependence, we divided the control group in half [procedure described in (61)]. Cont n = control group sample size; IA-SCI n = IA-SCI group sample size. (B) Funnel plot with trim-and-fill analysis. There did not appear to be publication bias for studies using this endpoint in our analysis.