Olfactory Ensheathing Cell Transplantation in Experimental Spinal Cord Injury: Effect size and Reporting Bias of 62 Experimental Treatments: A Systematic Review and Meta-Analysis

Abstract

Olfactory ensheathing cell (OEC) transplantation is a candidate cellular treatment approach for human spinal cord injury (SCI) due to their unique regenerative potential and autologous origin. The objective of this study was, through a meta-epidemiologic approach, (i) to assess the efficacy of OEC transplantation on locomotor recovery after traumatic experimental SCI and (ii) to estimate the likelihood of reporting bias and/or missing data. A study protocol was finalized before data collection. Embedded into a systematic review and meta-analysis, we conducted a literature research of databases including PubMed, EMBASE, and ISI Web of Science from 1949/01 to 2014/10 with no language restrictions, screened by two independent investigators. Studies were included if they assessed neurobehavioral improvement after traumatic experimental SCI, administrated no combined interventions, and reported the number of animals in the treatment and control group. Individual effect sizes were pooled using a random effects model. Details regarding the study design were extracted and impact of these on locomotor outcome was assessed by meta-regression. Missing data (reporting bias) was determined by Egger regression and Funnel-plotting. The primary study outcome assessed was improvement in locomotor function at the final time point of measurement. We included 49 studies (62 experiments, 1,164 animals) in the final analysis. The overall improvement in locomotor function after OEC transplantation, measured using the Basso, Beattie, and Bresnahan (BBB) score, was 20.3% (95% CI 17.8-29.5). One missing study was imputed by trim and fill analysis, suggesting only slight publication bias and reducing the overall effect to a 19.2% improvement of locomotor activity. Dose-response ratio supports neurobiological plausibility. Studies were assessed using a 9-point item quality score, resulting in a median score of 5 (interquartile range [IQR] 3-5). In conclusion, OEC transplantation exerts considerable beneficial effects on neurobehavioral recovery after traumatic experimental SCI. Publication bias was minimal and affirms the translational potential of efficacy, but safety cannot be adequately assessed. The data justify OECs as a cellular substrate to develop and optimize minimally invasive and safe cellular transplantation paradigms for the lesioned spinal cord embedded into state-of-the-art Phase I/II clinical trial design studies for human SCI.

Fig 1. Autologous olfactory ensheathing cell (OEC) transplantation.

OECs are extracted from a. the olfactory bulb or b. the olfactory mucosa (left side). After cell propagation, OECs are transplanted on the injured spinal cord (right side). OECs are transplanted either to the lesion core or rostral/caudal parenchymal areas juxtaposed to the lesion site after acute, subacute, or beginning chronic time frames after SCI in varying concentrations and dosages. With their unique ability to reorganise the glial scar and guide regenerating axons from the peripheral to the central nervous system (reviewed recently [6]), OECs are being regarded as a promising cellular approach for SCI treatment. Intramedullary OEC transplantation in proximity to the injured spinal cord site may facilitate regenerating fibres to overcome molecular axonal outgrowth inhibitors surrounding the forming scar and promote functional outcome improvement. There is heterogeneity in morphology, antigen expression, and function in OECs derived from olfactory mucosa versus olfactory bulb [15,16].

Fig 2. Study selection.

(A) Interventional preclinical OEC transplantation studies analyzing effects on locomotor recovery after SCI. (B) Effect size in percent according to rank; black dots represent studies applying the Basso, Beattie, and Bresnahan (BBB) score for neurobehavioral outcome assessment, and white dots indicate studies using other scores. The vertical error bars represent the 95% CI of individual studies, and the horizontal gray bar represents 95% CI of all analyzed studies (overall effect size 23.6 [95% CI 17.8–29.5]). (C) Meta-regression demonstrating the effect size for studies reporting motor outcome by open field BBB testing is comparatively lower when compared with other scores (e.g., Tarlov scale).

Fig 3. Differential effects of OEC transplantation paradigms on locomotor outcome indicated by the BBB-score.

(A) Effect of the origin of transplant (olfactory bulb versus olfactory mucosa derived), cell dispersion (suspension versus tissue block), (B) experimental SCI model, (C) time of OEC application (acute, subacute, beginning chronic lesion milieu), (D) additional surgical scar resection before OEC transplantation, (E) localization of injections (rostral/caudal parenchyma versus lesion core), (F) number of injections per OEC transplantation, (G) total volume of OEC suspension per animal, (H) fractionated volume per injection, (I) number of transplanted cells, and (J) concentration of OEC on the effect size. Meta-regression identifies each aspect of study design significantly associated with different observed efficacies. Vertical error bars represent the 95% CI, and the horizontal gray bar represents the 95% CI of all analyzed studies. The width of the columns depicts the log of the number of used animals in this group.

Fig 4. Assessment of publication bias.

(A) Funnel plot displays the precision plotted against the effect size. When considering the Funnel plot without any imputation from the trim and fill method (black dots only), the analysed 62 experiments indicate an overall effect size (ES) of 20.3%. When applying the trim and fill method to detect possible missing experiments based on the funnel plot’s asymmetry, the overall effect size results in a reduced effect size of 19.2%. One imputation was done, indicating a possibly missing experiment (red dot). (B) Egger regression illustrates the precision (one divided by the standard error of the mean) plotted against effect size divided by the standard error of the mean. Egger regression line does not intersect the origin, indicating possibly underlying publication bias. (C) “Traffic light” colored funnel plot adds two layers of information in the context of study quality. The circle sizes represent the number of animals in the individual experiment and the dot colors reflect the individual points on the quality score. The “traffic light” funnel plot demonstrates that studies with higher quality scores (≥5; green dots) relate closer to the corrected calculated effect size, although the stratification did not account for significance heterogeneity (Quality score ≥5: 18.7 [95%CI 10.6–26.7]; <5: 22.1 [10.3–33.9]). However, the “traffic light” funnel plot also identified a cluster of studies reporting high study quality likely to further correct the observed effect size below the adjusted effect size of the trim-and-filled funnel plot.