In January 2015, we searched two electronic databases (MEDLINE via PubMed Central, and EMBASE via OvidSP) using the key words “alpha calcitonin gene-related peptide,” “αCGRP,” and “subarachnoid hemorrhage” in combination using the Boolean operator [AND]. The search was restricted to “other animals.” Two investigators (Liam M. C. Flynn and Caroline J. Begg) independently screened the abstracts and titles to identify those that met our inclusion criteria. Any differences were resolved by discussion with a third reviewer (Peter J. D. Andrews). We included in vivo animal studies describing the effect of αCGRP in animal models of SAH where outcome was reported as a change in arterial diameter and the articles were published in English. We also included studies which examined in vivo SAH and αCGRP administration with postmortem in vitro measurements of artery diameter.

The above search was repeated in July 2017 using the same methods. No further studies meeting eligibility criteria were identified.

Two investigators independently extracted data relating to species of animal and weight; method of inducing SAH (single injection, double injection, or clot placement); whether the basilar artery, middle cerebral artery, anterior cerebral artery, or internal carotid artery were measured; the method of measurement (angiography, in vitro measurement or direct in vivo visualization); anesthetic agent used; dose of αCGRP and time of administration from SAH; reporting and method of randomization; reporting and method of blinding; animal welfare guideline statement; statement of sample size calculation; whether there was a statement of potential conflicts of interest and the use of animals with comorbidities. Study quality was assessed using the CAMARADES 10-item quality checklist (42). One point was awarded for each of (1) publication in a peer-reviewed journal, (2) statement of control temperature, (3) randomized intervention allocation, (4) intervention allocation concealment, (5) blinded assessment of outcome, (6) avoidance of anesthetics with marked intrinsic neuroprotective properties (ketamine), (7) statement of a priori sample size collection, (8) statement of compliance with regulatory requirements, (9) conflicts of interest statement, and (10) use of animals with comorbidities.

For meta-analysis, we recorded arterial diameter for intervention and control groups as a percentage from baseline (mean values and a measure of variance with the number of animals per group). Where a single control group was used for multiple treatment groups, we adjusted the size of the control group entered into the meta-analysis by dividing the size of the control group by the number of treatment groups served (43).

Two of the nine publications did not report quantitative data in their text, only presenting graphed data (26, 32). Mean values with SEM were estimated from the graphs of these studies using Universal Desktop Ruler for Windows.

We performed meta-analysis using normalized mean difference with a random effects model (43). We used univariate meta-regression to explore associations of animal species, sex, strain, and quality issues. We used meta-regression of transformed data using a three-component cubic spline to assess dose–response. Where multiple experiments were performed in the same publication we treated these as separate studies (43). All in vivo experiments investigating the effect of αCGRP in animal models of SAH where outcome was reported as a change in arterial diameter were included. We also included experiments which examined in vivo SAH and αCGRP administration with postmortem in vitro measurements of artery diameter. Where one study administered different doses of αCGRP to separate groups of animals but had a shared control group we treated these as separate experiments and divided the number of controls between the experiments as per Vesterinen et al. (43). Results are presented as the mean ± SE unless otherwise stated. Statistical analysis was performed with STATA 14 (StataCorp LP).

The total number of animals included in analysis was 251, and the median number of animals used per experiment was 9 [interquartile range (IQR) 6–14]. Twelve experiments used rabbits (n = 156, 62% of all animals). Four experiments used New Zealand White, and eight experiments used Japanese White. Five experiments used mongrel dogs (n = 45, 18% of all animals). Two experiments used rats (n = 40, 16% of all animals). One used Wistar rats, and the other used Sprague Dawley rats. The remaining experiment used Macaca fascicularis (n = 10, 4% of all animals). The median number of study quality checklist items scored was 4 (IQR 2–6). No studies used animals with comorbidities; reported a statement of potential conflicts of interest or stated an a priori sample size calculation. 40% of studies reported control of body temperature. 20% described a randomized treatment allocation, and 45% reported allocation concealment. Half of the experiments used a blinded assessment of outcome, and half used an anesthetic agent other than ketamine. All studies were published in peer-reviewed journals, and 70% reported compliance with local animal welfare guidelines. In 16/20 (80%) of the experiments, SAH was induced by autologous blood injection by either single or double injection methods, the remainder were induced with blood clot placement. Further study characteristics are presented in Tables 1 and 2.

All 21 publications reporting in vivo and in vitro experiments demonstrated a dilation of cerebral arteries after αCGRP administration. Of the 20 eligible in vivo experiments (taken from the nine eligible publications) included in meta-analysis, there was a 40.8 ± 8.2% increase in cerebral vessel diameter in the αCGRP group compared with controls (p < 0.0005, 95% CI 23.7–57.9, I² 96%, Figure 2). There was also a significant dose–response to αCGRP in the 10 experiments, which administered a single dose into the cerebroventricular system (Figure 3).

The effect size tended to be lower in studies that reported randomization, blinded assessment of outcome, blinded induction of SAH, and use of an anesthetic agent without intrinsic neuroprotective properties. However, none of these observations reached statistical significance. There was also a trend toward lower effect size in studies reporting compliance with more quality checklist items. This ranged from 57.3 ± 10.7% (p < 0.05) from experiments with a quality score (QS) of 1 to 28.1 ± 9.1% (p < 0.01) from experiments with a QS of 6 (see Figure 4). There was no statistically significant difference in treatment effect between species.

Four studies reported an effect on neurological outcome after αCGRP administration (17, 21, 22, 35). The standardized mean difference was 1.31 (95% CI −0.49 to 3.12, Q 40.5, n = 65 animals) in favor of αCGRP. Tian et al. reported neurological outcome based on a comprehensive scoring system (0–48, 0 = best score, 48 = worst score) measured three times daily and based upon the assessment of four functions, which has been used elsewhere (35, 44). The mean neurological outcome on day 7 for the αCGRP group was significantly better than for the control group (10.67 ± 1.16 versus 22.33 ± 2.08, respectively, p < 0.001). Imaizumi et al. assessed neurological outcome based on food intake and a slope tolerance test on days 2, 3, and 4 post-SAH (21). No significant difference was found between the αCGRP and control groups for either assessment. Inoue et al. reported no significant difference in food intake, observable hemiparesis, consciousness disturbance or response to stimulation between the αCGRP and control groups (22). Ahmad et al. reported neurological outcome from grade I (normal) to grade III (unable to stand and presented abnormal posture) in addition to performing a slope tolerance test. Two rabbits in the control group were grade II (slow in response but able to walk) and III, respectively, all other rabbits were normal, and there was no statistical difference between the groups in the slope tolerance test. In all studies where it was measured, food intake was decreased after SAH, but there was no significant difference between the αCGRP and control groups. Inoue et al. noted a significant decrease in weight in their αCGRP group compared with the control group at day 14, but note no other adverse effects and were unable to explain this change in weight.

Eight publications reported physiological parameters that might be associated with adverse events. There was no significant difference in systemic arterial pressures or arterial blood gas results between the aCGRP or control groups. Imaizumi et al. found that all animals tended to have an increased respiratory rate for approximately 6 h after intrathecal injection of either αCGRP or vehicle and demonstrated a high blood pH and low pCO₂, but again no difference between the groups (21). Both Nozaki et al. (16) and Toshima et al. (36) demonstrated a decrease in mean arterial blood pressure when αCGRP was administered intravenously, which was not seen by intrathecal administration. The study by Toshima et al. demonstrates a marked decrease in mean arterial blood pressure following intravenous administration of αCGRP which is not seen with intracisternal administration (~70 versus ~40 mmHg at 30 min after αCGRP administration).

There were no female-only experiments and the majority of experiments in this systematic review used the single hemorrhage model of SAH (60%). The other forms were double injection and clot placement. Animals rarely develop a vessel narrowing-related ischemic neurological deficit from any of these methods. Megyesi et al. note that this is probably because animal brains have a plentiful collateral blood supply (54). While this is probably irrelevant for measuring the effect of αCGRP on vessel diameter, it becomes more problematic when trying to assess neurological outcomes. Another translational problem arises from the times of administration of αCGRP and assessment of neurobehavioral outcomes. In humans, DCI is said to occur most commonly between days 3 and 10 and cerebral vessel narrowing is maximal between days 6 and 10 after ictus (12). The experiments assessed in this review administered αCGRP before and up to 3 days after SAH and assessed the response within hours to one week after administration (Table 1). The authors also note that the best model of vasospasm seems to be the primate model in which a blood clot is placed around a large cerebral vessel (54). Only one of the studies we analyzed uses this model. More recently, Titova et al. in their systematic review state that dog models of SAH are considered superior and that the ability of murine models to reflect human vasospasm is disputed (55).

One of the challenges of translating these animal data and methods to human trials is the invasive nature of the intrathecal route. Previous human trials with intravenous administration of αCGRP have used a continuous infusion owing to the short half-life of αCGPR in the systemic circulation (approximately 7–10 min) (56). However, when administered into the cerebrospinal fluid the effects of αCGRP have continued for 4–6 h (21, 26). Therefore, a continuous infusion of αCGRP into the cerebral spinal fluid may not be necessary. Furthermore, Toyoda et al. (37) and Sun et al. (32) demonstrate novel approaches to administering αCGRP, one via gene transfer and the other by intranasal delivery. If intraventricular administration of αCGRP in humans ameliorates cerebral vessel narrowing and avoids the adverse effects seen with intravenous delivery, both gene transfer and intranasal delivery offer potential alternatives without the difficulties associated with an intraventricular drain.