Address correspondence and reprint requests to Dr. Joan Montaner, Neurovascular Research Laboratory, Neurovascular Unit, Institut de Recerca, Hospital Vall d'Hebron, Pg Vall d'Hebron 119-129, 08035 Barcelona, Spain.
The neuroprotective actions of citicoline have been documented for experimental stroke therapy. We used a systematic review and meta-analysis to assess this evidence. From 64 identified studies using citicoline in stroke animal models, only those describing ischemic occlusive stroke and reporting data on infarct volume and/or neurological outcome were included (14 studies, 522 animals). Overall, the quality of the studies was modest (5, 4–6), while the absence of studies involving animals with co-morbidities, females, old animals or strain differences indicated that studies did not fulfill the STAIR recommendations. Weighted mean difference meta-analysis showed citicoline to reduce infarct volume by 27.8% [(19.9%, 35.6%); p <0.001]. In the stratified analysis, citicoline effect on reducing infarct volume was higher in proximal occlusive models of middle cerebral artery (MCA) compared with distal occlusion. Moreover, the efficacy was superior using multiple doses than single dose and when a co-treatment was administered compared with citicoline monotherapy, the only independent factor identified in the meta-regression. Citicoline improved neurological deficit by 20.2% [(6.8%, 33.7%); p =0.015], but only four studies including 176 animals reported these data. In conclusion, this meta-analysis provides evidence of citicoline efficacy in stroke animal models and shows the optimal neuroprotective profile and the missing experimental requirements before jumping into clinical trials.
Stroke remains one of the leading causes of mortality and disability in western countries. Nevertheless, no neuroprotective agent has demonstrated clear efficacy in phase III clinical trials that can be translated into clinical recommendations or guidelines (Gladstone et al. 2002). Regarding citicoline and based on the evidence shown by animal models, several studies have been conducted in human stroke. Although some of those studies have shown overall benefits (Clark et al. 1999, 2001), their results are broadly inconclusive.
Cytidine diphosphocholine (CDP choline or citicoline) is an essential intermediate in the synthesis of structural phospholipids of cell membranes. Moreover, the formation of this compound from phosphorylcholine is the rate-limiting step of this biosynthetic pathway (Adibhatla and Hatcher 2007, 2010). Phospholipids are essential constituents of cells, specifically cell membranes, and have a high turnover rate, which requires the continuous synthesis of these compounds to ensure the adequate function of cell membranes. Damaged cell membranes and impaired phospholipid metabolism have been implicated in the pathophysiology of cerebral ischemia (Goto et al. 1988). In vitro studies using nerve tissues have shown that hypoxia induces a time-dependent decrease in the synthesis of structural phospholipids (Cohen 1973), suggesting the relevance of the new phospholipid synthesis after stroke. As a result of these physiological and pathological effects, the administration of citicoline has been proposed as a neuroprotective therapy for human stroke (Secades 2002).
With this background, the potential neuroprotective role of citicoline has been evaluated in several animal ischemia models. In those, citicoline has shown to decrease infarct volume when is used alone (Aronowski et al. 1996; Schabitz et al. 1996; Adibhatla et al. 2005) or in combination with other thrombolytics (Andersen et al. 1999; Shuaib et al. 2000; Alonso de Lecinana et al. 2006) or neuroprotective drugs (Onal et al. 1997; Schabitz et al. 1999, Ataus et al. 2004; Sobrado et al. 2003). Regarding neurological function, however, data are uncertain. Several studies assessing global (Kakihana et al. 1988) and focal (Aronowski et al. 1996; Hurtado et al. 2007) cerebral ischemia models have reported an improvement on neurological outcome, while other studies could not found differences after citicoline treatment (Schabitz et al. 1996, 1999). However, no systematic review has explored the adherence of citicoline experimental research on stroke models to the Stroke Therapy Academic Industry Roundtable (STAIR) recommendations (STAIR 1999). The updated STAIR recommendations reinforce the previous suggestions defining dose response and time windows with both histological and functional outcomes in multiple animal species. The STAIR recommendations also include that after initial evaluations in young, healthy male animals, further studies should be performed in females, aged animals, and animals with comorbid conditions such as hypertension, diabetes, and hypercholesterolemia.
In parallel, clinical studies in stroke patients have shown benefits in terms of infarct size reduction (Warach et al. 2000), 3-month disability and complete neurological recovery (Clark et al. 1999, 2001). In pooled data analysis (Davalos et al. 2002; Saver 2008), citicoline therapy was associated with higher rates of total recovery, death or disability at long-term follow-up. However, the recently published ICTUS Trial (Davalos et al. 2012) has failed to demonstrate efficacy in the treatment of moderate-to-severe acute ischemic stroke. These contradictory results suggest that there is a need to re-analyze the experimental evidence to get some clues on a new failure of neuroprotection strategies.
In this study, we conducted a systematic review and meta-analysis to identify qualitatively and quantitatively any evidence of citicoline as a neuroprotectant on stroke animal models. Such type of studies allows to identify possible influences modifying citicoline benefit on experimental stroke and, therefore, to select the optimal requirements for drug administration for jumping into future clinical trials.
Data sources and data extraction
A computerized literature search and hand searching of abstracts from scientific meetings (European stroke conferences from 1995 to 2011; American Academy of Neurology Annual Meetings from 2000 to 2011) were performed to find publications studying the effect of citicoline treatment on animals subjected to focal cerebral ischemia. For electronic search, Pub Med database was accessed (December 14, 2011) to find English and Spanish publications combining the following terms: ‘citicoline’ OR ‘CDP choline’ AND ‘cerebral ischemia’, ‘citicoline’ OR ‘CDP choline’ AND ‘brain ischemia’ OR ‘citicoline’ OR ‘CDP choline’ AND ‘stroke’, with no publication date limit but with ‘other animals’ species limit. Data extraction was achieved from the identified studies that described animal models of focal cerebral ischemia and in which data of citicoline effect on infarct size and/or on neurological outcome was reported. When data were not shown, authors were contacted to request additional information. Data presented graphically were obtained using digital ruler software (Universal desktop ruler).
Effect size of both infarct volume reduction and neurological improvement from citicoline treatment and control groups were calculated in our meta-analysis to assess the effect of citicoline. For this purpose, the total number of animals, mean and standard error of mean (SEM) or standard deviation (SD) of both infarct volume and neurological data were extracted from the studies. Other information regarding the method used for ischemia induction, type of ischemia (permanent or transitory), method of injury quantification, neurobehavioral scale, treatment dose, single or multiple doses, time of drug administration (minutes), time of assessment (hours), route of delivery, anesthetics, co-morbidity and co-treatments were included in our database.
Study quality was scored according 10 items (Macleod et al. 2004): (1) peer review publication, (2) control of temperature, (3) random allocation to group, (4) allocation concealment/blinded induction of ischemia, (5) blinded assessment of outcome, (6) use of anesthetic without significant intrinsic neuroprotective activity, (7) animal co-morbidities, (8) sample size calculation, (9) compliance welfare and (10) statement potential conflict of interest. Studies that were grateful to pharmaceutics or stated conflict of interest were included in the last category.
Extracted data were entered to the CAMARADES MS Access database. From each comparison, the effect size and standard error were calculated as a proportion of the control group, where these proportions were the reduction on infarct volume or the improvement on the neurological deficit of the treated group with respect to the control group.
The meta-analysis was performed using Weighted Mean Difference with random effects model to avoid heterogeneity (DerSimonian and Laird 1986). To assess the global effect for infarct volume and neurological outcome, one-sample t-test was done. A stratification of the analysis was also conducted to further examine the impact of several factors, such as the quality of the study, the administered dose (amount and repetitive vs. single dose), the use of combined therapies with citicoline and the type of ischemic model. Stratified meta-analysis was assessed partitioning heterogeneity by χ2 in the analysis after Bonferroni correction; p-values lower than 0.0038 were considered statistically significant.
Multivariate meta-regression was performed using STATA software (College Station, TX, USA), metareg function, to explore potential confounders among the variables (quality score, dose, co-treatment and type of ischemic model) in the analysis (Frantzias et al. 2011). Moreover, funnel plot was used to assess publication bias. Asymmetry was detected in the plot by visual inspection and when bias was observed, a new corrected estimate was calculated using ‘trim and fill’ by the Egger regression method.
Identification of papers
Electronic and hand searching identified a total of 64 articles potentially eligible only reading titles and abstracts. Of these, 34 were excluded from the study for the following reasons: 23 studies did not describe experiments reporting the focal cerebral ischemia model; six of them were written in Chinese and two in French; three were abstracts with no data obtainable. After that, 30 full articles were full-screened to decide whether appropriate to be included in the meta-analysis. Twelve of these did not report data on infarct volume or neurological deficit and therefore were excluded. Moreover, after screening, we detected two reviews and two studies not based on focal cerebral ischemia, which were also excluded (Fig. 1). Seven authors were contacted to request additional information, and just two of them provided the needed data. Finally, 14 studies were included to perform the meta-analysis (Table 1).
Table 1. List of articles included in the meta-analysis
The 14 identified studies produced a total of 44 comparisons involving 522 animals (242 and 280 in the control and in the citicoline-treated group, respectively) to assess citicoline effect on both infarct volume reduction and neurological outcome. The estimate effect for the infarct volume [32 comparisons involving 522 animals (242 in the control group and 280 in the citicoline group, respectively)] showed a 27.8% significant reduction [95% confidence interval (CI), (19.9%, 35.6%); p <0.001] in the citicoline-treated group with respect to the control group. Regarding the neurological deficit, a statistical improvement by 20.2% [(6.8%, 33.7%), p =0.015] was similarly detected. However, data from neurological outcome were only obtained from four studies involving 12 comparisons with a total of 176 animals, 72 of which were controls and 104 citicoline-treated animals (Fig. 2). Therefore, the effect of citicoline on neurological outcome was not assessed in the stratified meta-analysis because of the small data on each group.
Study characteristics according to the quality score are shown in Table 2. Overall, the median score was modest (5, interquartile range 4–6). No study scored low (0, 1 or 2) or high (9 or 10). Control of temperature and use of anesthetics without described neuroprotection were applied in 10 studies. Two studies used hypertensive animals and none of them described sample size calculation.
Table 2. Quality score
Author and Year
(1) Peer review publication, (2) control of temperature, (3) random allocation to group, (4) allocation concealment/blinded induction of ischemia, (5) blinded assessment of outcome, (6) use of anesthetic without significant intrinsic neuroprotective activity, (7) animal comorbidities, (8) sample size calculation, (9) compliance welfare, (10) statement potential conflict of interest.
When we further analyzed the effect of citicoline on infarct volume according to the type of ischemia (permanent or transient), the effect was present in both models. The stratified analysis showed a 25.4% (17.6%, 33.3%) infarct volume reduction when permanent ischemia was performed and a 30.2% (15.3%, 45.1%) reduction when transient ischemia was used (Fig. 3a). No differences between both models were found (χ2 = 1.58, df = 1, p =0.21).
Moreover, we sought to investigate citicoline-dose effect on infarct volume reduction. We observed a similar effectiveness when stratifying the analysis by low-dose (less than 250 mg) and high-dose (more than 250 mg) citicoline [27.4% (14.6%, 40.2%) vs. 27.4% (18.2%, 36.6%), respectively; Fig. 3b]; however, the heterogeneity between studies was significant (χ2 = 27.2, df = 1, p <0.0001). Conversely, multiple doses of citicoline led to higher infarct volume reduction compared with single dose [22.6% (14.1%, 31.1%) vs. 31.1% (18.8%, 43.5%) for χ2 = 69.9, df = 1, p <0.0001].
Regarding the effect of combined therapies with citicoline, we observed that animals that were co-treated with other drugs showed a greater increase in infarct volume reduction by 40.2% (27%, 53%) with respect to non-co-treated animals [24% (16%, 31.6%); χ2 = 60.2, df = 1, p <0.0001, Fig. 3c].
As only two studies (Hurtado et al. 2005; Han et al. 2011) used citicoline in a pre-treatment protocol, we were not able to analyze whether administration of citicoline as a pre-treatment was more effective than post-treatment.
The stratified analysis performed to investigate whether applying random allocation to treated or control groups modifies the estimate of the infarct volume reduction, showed a slight higher effect in studies that stated random allocation [29.6% (20.7%, 38.5%)] from those that did not [26.3% (14.6%, 38.1%), χ2 =4.1, df = 1, p =0.044].
With regard to the method used for ischemia induction, intraluminal filament and autologous embolism triggered a similar effect on infarct volume reduction [31.4% (21.5%, 41.4%) and 38.4% (20.7%, 56.1%), respectively]. However, the effect was reduced [19% (9.2%, 28.9%); χ2 =60.2, df = 2, p <0.0001] when surgery was performed to occlude the MCA distally using a surgical clip.
Meta-regression was conducted to further explore meta-analysis heterogeneity. The analysis identified only one factor to account for 17.25% of between-study variance (τ2 = 196.3, adjusted r2 = 0.173). Combined therapy with citicoline increases the efficacy by 16.15% (−0.15%, 32.45%) compared with those studies that used citicoline alone (p =0.052).
Finally, we sought to identify whether the effect of small studies may contribute to publication bias in citicoline analysis. Funnel plot evidenced asymmetry (Egger regression p <0.001; Fig. 4). The effect on infarct volume reduction when it was adjusted by publication bias with an estimated number of 40 comparisons was 19.17% (10.62%, 27.72%).
Studies assessing the neuroprotective effect of citicoline in animal stroke models have shown positive results in terms of infarct size reduction. As those studies have been conducted in different models of focal cerebral ischemia, type of ischemia, citicoline dose, single or multiple doses, or simultaneous use of other neuroprotective or thrombolytic therapies, we conducted a meta-analysis for a better understanding and translation of experimental evidence of neuroprotection to future clinical trials. With this translational objective, focal ischemia animal models were chosen as the pathophysiological model closer to human stroke. The analysis of the experimental evidence of citicoline effect in global ischemia models and its comparison with focal ischemic models, although out of the translational objectives of our meta-analysis, could be of great interest for future studies.
The meta-analysis results reinforce the evidence for a neuroprotective role of citicoline therapy in experimental stroke. These results provide strong evidence about the effect of citicoline in brain infarct volume reduction. For neurological outcome, our analysis shows consistent benefit of citicoline therapy, although limited data were available and the number of experiments reporting this end point is reduced (Aronowski et al. 1996; Schabitz et al. 1996, 1999; Hurtado et al. 2007), making these findings less robust. Further evidence is required in this area to assess that reduction in infarct volume leads to neurological improvement in experimental stroke models.
Although the quality of the 14 studies included in this meta-analysis is not optimal, there is not a wide dispersion regarding quality score between the studies that may limit the generalizability of the results. The lowest scores were extracted from studies with only abstracts available, not containing all the information needed to appropriately score those studies. Although inclusion of abstracts might be a limitation to the qualitative analysis and its publication is not exposed to the same rigor of review than full studies, those were included to collect the maximum number of studies reporting infarct volume or neurological outcome.
One of the factors that may be of interest in the design of future studies is the ischemia induction method. In our meta-analysis, no differences were found between proximal occlusion models (intraluminar and embolic models). However, the effect was reduced in distal occlusion models. Those data suggest that the effects of citicoline might be more effective in large strokes, as has been suggested in clinical studies (Clark et al. 1999; Davalos et al. 2002), where the benefit of citicoline has been found to be greater in patients with moderate-to-severe strokes (baseline NIHSS score of 8 or higher).
The fact that combined therapy of citicoline with other neuroprotective (Onal et al. 1997; Schabitz et al. 1999, Ataus et al. 2004; Sobrado et al. 2003; Onal et al. 2000) or thrombolytic agents (Shuaib et al. 2000) was associated with higher efficacy seems reasonable, as ischemia triggers a multitude of pathophysiological events, and disruption of the ischemic cascade events at different levels is likely to be more effective than at a single point (Fisher 1997). This design has been proposed as a new approach for clinical trials in human stroke (De Keyser et al. 1999). Perhaps, the combination with reperfusion strategies may be of interest for future trials. As the use of citicoline as a co-treatment seems important for efficacy, identification in future experimental research of those ideal combinations merits further efforts.
To extrapolate the findings from these pre-clinical results to clinical studies is difficult. Just two of the 32 comparisons (6%) included in our meta-analysis reported a negative effect on infarct size reduction or no improvement on neurological behavior. This effect is usually caused by the ‘file drawer problem’ (Sena et al. 2010), and therefore, citicoline effect may be overstated in this meta-analysis. To detect this problem, funnel plot was assessed and the effect was corrected by the estimated number of comparisons to avoid publication bias.
Another limitation to jump into clinical studies highlighted by this meta-analysis is the few number of studies using animals with co-morbidities, which is the typical situation in human stroke. Only two of the studies include the use hypertensive rats (Aronowski et al. 1996; Adibhatla et al. 2005) and none of them used animals with other co-morbidities such as diabetes, dyslipidemia or aged animals. Moreover, no studies were conducted on females. Neither studies exploring the efficacy of citicoline in experimental stroke have been conducted in other rodent strains such as mice or larger species such as primates. Therefore, not all the steps recommended by the Stroke Therapy Academic Industry Roundtable recommendations (STAIR 1999) for the evaluation of pre-clinical data with neuroprotective drugs have been fulfilled.
In conclusion, this systematic review and meta-analysis provides evidence of the neuroprotective role of citicoline, while this evidence is stronger for infarct volume reduction and limited for neurological outcome. Factors that showed to be important for translation into human studies are multiple-dose administration, large infarct size, and/or neurological deficit. There is a lack of trials exploring the combined efficacy of citicoline with other neuroprotective drugs that might be investigated in future human studies.
The recently published ICTUS trial (Bolland et al. 2009; Davalos et al. 2012) has evaluated the benefit of citicoline therapy administered in the first 24 h after stroke onset. The investigators conclude that under the circumstances of the study, citicoline is not efficacious in the treatment of moderate-to-severe acute ischemic stroke. According to our meta-analysis, the trial design is still far from the optimal neuroprotective profile for jumping into a stroke clinical trial; favorable points in its design include multiple-dose administration (1 g/12 h i.v. during 3 days and then orally until complete 6 weeks of treatment), selection of patients with moderate-to-severe stroke (NIHSS score > 7), and the possibility of combination with thrombolytics. However, other critical aspects are those related with the selection of the dose as the meta-analysis shows efficacy at lower doses also, and the inclusion time window (< 24 h) as only one pre-clinical study (Hurtado et al. 2007) has tested this time window while most of the studies administered the drug in less than 1.5 h after ischemia onset. After this failure, the fulfillment of the experimental gaps identified in this systematic review might help in the future translation of this information into the clinical practice.
Neurovascular Research Laboratory takes part into the Spanish stroke research network (RENEVAS, RD06/0026/0010) and the European Stroke Network (EUSTROKE 7FP Health F2-08-202213) and receives grants PI080481 and EC07/90195 to conduct research on stroke models. A.R. is supported by a Miguel Servet senior research contract (CP09/00265) and M.C. by a predoctoral fellowship (FI 10/00508) from the Carlos III Health Institute.
The authors thank Malcolm Macleod and Emily Sena, from CAMARADES group, for their advice and for helping the setup of the database and to Julio Secades for kindly supplying some of the articles or congress abstracts used in the meta-analysis.