Association of timing of gabapentinoid use with motor recovery after spinal cord injury

Abstract

Objective

To explore the hypothesis that earlier administration of acute gabapentinoids is beneficial to motor recovery after spinal cord injury in humans.

Methods

This is an observational study using a cohort from the European Multi-Centre Study about Spinal Cord Injury. Patient charts were reviewed to extract information regarding the administration and timing of gabapentinoid anticonvulsants. The primary outcome measure was motor scores, as measured by the International Standards for Neurological Classification of Spinal Cord Injury, collected longitudinally in the first year after injury. Sensory scores (light touch and pinprick) and functional measures (Spinal Cord Independence Measure) were secondary outcomes. Linear mixed effects regression models included a drug-by-time interaction to determine whether exposure to gabapentinoids altered recovery of muscle strength in the first year after injury.

Results

A total of 201 participants were included in the study and had a median age of 46 and baseline motor score of 50. Participants were mostly men (85%) with sensory and motor complete injuries (50%). Seventy individuals (35%) were administered gabapentinoids within the first 30 days after injury, and presented with similar demographics. In the longitudinal model, the administration of gabapentinoids within 30 days after injury was associated with improved motor recovery when compared to those who did not receive gabapentinoids during this time (3.69 additional motor points from 4 to 48 weeks after injury; p = 0.03). This effect size increased as administration occurred earlier after injury (i.e., a benefit of 4.68 points when administered within 5 days).

Conclusions

This retrospective, observational study provided evidence of the beneficial effect of gabapentinoid anticonvulsants on motor recovery after spinal cord injury. More critically, it highlighted a potential time dependence, suggesting that earlier intervention is associated with better outcomes.

Classification of evidence:

This study provides Class IV evidence that gabapentinoids improve motor recovery for individuals with acute spinal cord injury.

Gabapentinoid anticonvulsants (i.e., pregabalin and gabapentin) represent the front line pharmacologic interventions for the management of neuropathic pain.^1,2 In cases arising from damage to the spinal cord, administration often occurs early after injury (e.g., within 1 month).^3,4 Within this acute time frame, these medications have the potential to modify the protection or repair of the spinal cord by antagonizing voltage-gated calcium channels in the CNS.^5–7

In support of these modifying effects, recent observational studies in humans have identified an association between the early administration of gabapentinoids (but not other anticonvulsants) and enhanced neurologic recovery.^3,4,8 A notable limitation of previous studies was that anticonvulsant use was self-reported. This is problematic, particularly in acute care settings, as self-reported use of medications is often inaccurate.⁹ Moreover, previous analyses had limited information on timing of administration, which is critical to determine whether the effect of gabapentinoids is time-dependent.

To further examine the relationship between gabapentinoids and neurologic recovery, we performed a single-center observational study that abstracted medication usage from patient charts. The fundamental goal was to elucidate the effect of gabapentinoids on neurologic recovery after spinal cord injury using well-defined criteria on usage and timing of administration. Based on previous studies, we hypothesized that motor recovery would be greater in those administered gabapentinoids in the early stages of injury compared to those administered late or not at all.

Methods

Research question and hypothesis

Our research question was to assess whether gabapentinoid use is associated with motor recovery following spinal cord injury, with the hypothesis that early administration will be beneficial.

Study design and data source

The design was an observational cohort study, which provides Class IV evidence for the above research question. The cohort was defined based on participation in the European Multi-Centre Study about Spinal Cord Injury (EMSCI) between 2007 and 2016. The chart review was performed at a participating center (i.e., Trauma Centre Murnau, Germany). This chart review extracted information on specific anticonvulsant use, including the type of anticonvulsant, the first date of administration, the maximum dose received (and date), and the duration of intake. Patient charts were then linked with neurologic and functional outcomes from the original EMSCI database to create an observational cohort. The EMSCI database was queried for the following individual patient information: date of injury, sensory and motor scores from the International Standards for the Neurological Classification of Spinal Cord Injury (ISNCSCI),¹⁰ Spinal Cord Independence Measure 3 (SCIM 3) scores, the American Spinal Cord Injury Association Impairment Scale (AIS) injury severity grades, and level of injury (including upper cervical [C1–C4], lower cervical [C5–C8], thoracic [T1–T12], and lumbar [L1–L5]).

Standard protocol approvals, registrations, and patient consents

The participating EMSCI site obtained approval by the local research ethics board, and individuals consented to have their data entered into the database. Further details on the EMSCI database can be found elsewhere (emsci.org, ClinicalTrials.gov Identifier NCT01571531). Approval for this study (secondary analysis) was received from an institutional ethical standards committee on human experimentation at the University of British Columbia.

Inclusion and exclusion criteria

The EMSCI includes individuals with spinal cord injury caused by a traumatic event (including single event ischemia). Individuals are excluded if they had prior dementia or reduced learning disabilities, peripheral nerve lesions above the level of injuries or a prior polyneuropathy, or a severe craniocerebral injury. For the purpose of our study, individuals were also excluded if they were missing baseline measurement of injury severity (AIS grade) or level of injury (i.e., 4-week recordings if available and 1 week otherwise). Individuals with AIS E injuries or injuries caudal to T12 at baseline were excluded. Finally, individuals with fewer than 2 motor score assessments over the first year after injury were excluded in order to ensure there were sufficient data for longitudinal analyses.

Outcomes

Motor and sensory scores as defined by ISNCSCI were examined at 1, 4, 12, 24, and 48 weeks after injury.¹⁰ The primary outcome was total motor score (0–100). Secondary outcomes included the sensory measures pinprick score and light touch score (each 0–112) and the functional score from the SCIM 3 (0–100).¹¹ Outcomes were modeled using all available scores from baseline to 48 weeks after injury (i.e., after exposure), as done previously.⁴

Drug exposure

The extraction of anticonvulsant use through a review of patient charts was conducted by 2 research assistants and supervised and approved by L.G. Anticonvulsant classification was then verified using the WHO Collaborating Centre for Drug Statistics Methodology's (whocc.no/) Anatomic Therapeutic Chemical classification system (code: NO3). Administration of anticonvulsants was classified as having occurred early (within 30 after injury), or late/never, as previously examined.⁴ Additional time windows of administration also examined included 0–5, 6–15, 16–30, 31–60, and 61–120 days after injury, exclusively.

Statistics

All analyses were performed using RStudio statistical software, version 1.1.453.¹² The longitudinal progression of total motor scores was first visually examined using Trellis plots. A multivariable longitudinal analysis was conducted using a linear mixed effects regression model including random slopes and intercepts (R package: lme4). Lme4 uses general-purpose nonlinear optimizers to estimate the variance–covariance matrices of the random effects. These linear mixed effects regression models are useful to longitudinally fit data with varying starting points and patterns of recovery, as well as participants with missing data. The model was adjusted for baseline level and severity of injury (AIS). An AIS-by-time interaction term was also included to account for anticipated differences in recovery related to initial injury severity, with the expected outcome that more severe injuries will recover less over time.¹³ To account for the nonlinear pattern of neurologic recovery profile after spinal cord injury (i.e., an initial rapid improvement that then plateaus),¹⁴ time was modeled as a quadratic variable. Finally, a drug-by-time interaction term examined the effects of gabapentinoids on neurologic recovery. An analysis of variance (ANOVA) was then applied to statistically compare models (i.e., whether the reduction in the residual sum of squares was statistically significant). These analyses were conducted for the primary outcome measure (total motor score), and then repeated for sensory and functional outcomes.

Prior studies have shown that different methods to control for confounding yield variable effect estimates.¹⁵ Therefore, to determine whether the effect estimate was reliable across different methodologies, a matched cohort was created based on propensity scores. The cohort was matched using a 1:1 ratio (late/nonusers: early users) based on baseline AIS grades and level of injury using an optimal matching algorithm (R package: MatchIt). Motor scores were then modeled longitudinally using a mixed effects regression to examine the effect of the drug-by-time interaction in the matched cohort.

Following the initial modeling, timing of administration was assessed by redefining the drug exposure categories relative to time of injury, and each of these new exposures was run in a respective longitudinal model. These new exposure categories included gabapentinoids that had been initially administered 0–5, 6–15, 16–30, 31–60, and 61–120 days after injury, exclusively. p < 0.05 was considered statistically significant.

Data availability

Requests for data are evaluated on an individual basis, and anonymized data may be shared with qualified investigators.

Results

The chart review included 251 individuals who sustained a spinal cord injury. After applying our study-specific exclusion criteria, a total of 201 individuals remained. The cohort was primarily male, with a high proportion of motor and sensory complete injuries (AIS A) (table 1).

Table 1.

Cohort demographics

The anticonvulsants administered included gabapentin, pregabalin, and carbamazepine. Of the 201 individuals, 70 (35%) received gabapentinoids within the first 30 days after injury (i.e., early administration). Those in the early administration group did not significantly differ from the late/nonusers with regards to age at injury, sex, or baseline AIS grades. Late/nonusers tended to have a lower proportion of upper cervical injuries and a higher proportion of thoracic injuries (table 1).

Early administration of gabapentinoids was associated with improved motor recovery. The 2 groups showed different motor recovery profiles, as indicated by the drug-by-time interaction term in the longitudinal analysis (table 2). The early administration of gabapentinoids was associated with an overall benefit of 3.69 motor points in the 4–48 weeks after injury (95% confidence interval [CI] 0.35–7.04). ANOVA confirmed that including the drug-by-time interaction term significantly improved the model fit (χ²(1) = 4.75, p = 0.03). A drug-by-time² interaction did not significantly improve fit and thus was excluded.

Table 2.

Longitudinal model of total motor scores from 4 weeks to 48 weeks in the entire and matched cohorts

Of the secondary outcomes, early gabapentinoid administration was significantly associated with greater recovery of light touch scores (over 6 points, 95% CI 1.51–12.37, table 3). Early administration had no significant effect on pinprick or SCIM scores, with overall estimated benefits of 2.51 (95% CI −3.92 to 8.98) and 0.56 points (95% CI −6.14 to 7.36), respectively (table 3).

Table 3.

Longitudinal models of secondary outcomes from 4 weeks to 48 weeks

Similar results were found in the longitudinal model of the propensity score matched cohort (n = 140, 70 early users), in which the 2 groups did not significantly differ in distribution of AIS grades or neurologic levels of injury. This analysis revealed a larger effect size (over 6 motor points improvement from 4 to 48 weeks after injury, 95% CI −0.19 to 12.91) for the drug-by-time interaction term (table 2). Figure 1 depicts the individual recovery profiles and longitudinal model trajectories alongside the predicted change scores of the early vs late/never gabapentinoid users from the matched cohort.

Figure 1. Recovery in early vs late/never gabapentinoid users in the matched cohort.

Individual trajectories and predicted changes in the recovery of early and late/never gabapentinoid users from the matched cohort by (A) total motor score (MS), (B) light touch score (LT), and (C) spinal cord independence measure (SCIM).

Longitudinal motor scores were then modeled using the redefined exposures for gabapentinoids. The times included 0–5 (n = 14 users), 6–15 (n = 32), 16–30 (n = 24), 31–60 (n = 17), and 61–120 (n = 13) days after injury. Though limited in sample size, the effect sizes decreased as administration occurred later after injury (table 4, figure 2). For example, the drug exposure interaction term for 0–5 days after injury indicated the greatest improvement in total motor score from 4 to 48 weeks after injury (4.68 points, 95% CI −0.95 to 10.31).

Table 4.

Comparison of drug-by-time term in longitudinal models of total motor scores from 4 weeks to 48 weeks using varying timing of exposure

Figure 2. Results of the longitudinal modeling of motor scores by exposure.

The effect sizes of the longitudinal models using mutually exclusive time points of administration of gabapentinoids (0–5, 6–15, 16–30, 31–60, and 61–120 days after injury) in the total cohort.

Discussion

The current analysis revealed that early exposure to gabapentinoids was associated with a beneficial effect on the recovery of sensorimotor function after spinal cord injury. In using medication details extracted directly from patient charts, we found higher rates of gabapentinoid exposure as compared to self-report.⁴ Further analyses on timing indicated that the largest effect occurred with administration in the first 5 days of injury. This is consistent with the concept of a window of opportunity to promote protection or repair in the injured spinal cord. According to the Oxford Center of Evidence-Based Medicine, this study provided Class IV evidence of their benefit on neurologic function after acute spinal cord injury.¹⁶

As in previous studies,^3,4 we observed a beneficial effect of gabapentinoids administered within 30 days of injury on the recovery of muscle strength. This benefit increased marginally if administered within 5 days of injury. The magnitude of effect is similar to that previously reported using data from multiple centers participating in EMSCI.^3,4 In general, a strength of multicenter cohort studies (as is often used in large-scale clinical trials) is the inclusion of larger and more heterogeneous sample sizes, which increases generalizability. However, single-center studies, as used here, offer a number of distinct advantages. Collecting detailed data such as drug administration is more feasible when dealing with a single center, and there is decreased variability introduced by different acute care and rehabilitative practices.

In addition to motor scores, we also observed a beneficial effect of early gabapentinoid administration on sensory (light touch) scores. This is in contrast to our previous study, where we observed no such effect.⁴ Total motor scores have the highest level of interrater reliability, followed by light touch and then pinprick scores.^17,18 Benefits for light touch may indicate decreased variability due to the single-center nature of this study, revealing subtle effects masked in previous multicenter studies. Interestingly, there was no significant effect on the functional measure (i.e., SCIM). This suggests that gabapentinoid-associated motor and sensory improvements are not yet reaching a clinically meaningful threshold.¹⁹ In these early stages of investigation, this is perhaps not surprising, given that the majority of individuals were exposed to gabapentinoids outside of the optimal time window (i.e., <5 days). Nevertheless, it may also suggest that, in the search for advanced treatments for spinal cord injury, multiple therapies providing marginal neurologic benefits must be combined to achieve an overall clinically significant benefit.²⁰

Gabapentinoids provide an exciting opportunity for drug repurposing in spinal cord injury because they are already administered in the acute phase of spinal cord injury, and thus have an established safety profile. For this reason, repurposing could prevent many of the costly and time-consuming hurdles of early-phase clinical trials. Unlike our previous observational studies, which examined the general administration of anticonvulsants,^3,4 this study focused solely on the effects of gabapentinoids. In addition, the previous studies included self-reported drug use at a specific time point (i.e., 4 weeks after injury), whereas in conducting a chart review, this study accurately identified all individuals administered anticonvulsants. This permitted a close examination on the exact timing of administration, which highlighted, for the first time, that early administration was important to neurologic recovery.

The increase in effect size closer to time of injury (i.e., within 5 days) strengthens the position that gabapentinoids are directly affecting neurologic recovery, and is important for several reasons. Early administration is believed to be critical for both neuroprotective and regenerative strategies after spinal cord injury. Indeed, the 30-day window defining early administration is longer than hypothesized for either neuroprotection (hours to days) or neuroregeneration (days to weeks).²¹ Furthermore, previous clinical trials have emphasized the need to administer pharmacologic interventions very early, within hours of injury.^22–24 This presents serious ethical concerns and is a major barrier in terms of establishing a valid baseline assessment of injury severity.²⁵ A 5-day window mitigates a number of issues, and may be an important consideration for a future trial design.^26,27

Although this study provides advantages over those previously published,^4,28 there are limitations. Firstly, as in all observational cohorts, the individuals receiving gabapentinoids were not randomized to treatment, and thus some unknown or unavailable confounder may be driving the observed effects on neurologic function. Confounding by indication is a potential concern, in that the beneficial effects of gabapentinoids could be attributable to the underlying reason for administration (i.e., neuropathic pain). Information regarding neuropathic pain status was unavailable in this cohort; however, a prior study addressed the presence and progression of neuropathic pain as a potential confounder.⁴ Furthermore, a recent observational study found that early (within 30 days) administration of nongabapentinoid anticonvulsants was not associated with any effect on neurologic recovery. These results indicate the potential specificity of gabapentinoids for a benefit, highlighting that other drugs of similar indications have no effect on neurologic recovery.⁸ Secondly, specific dosages were not collected as part of this review. While limiting in terms of examining a dose–response effect, gabapentinoids were administered according to established standards for managing neuropathic pain, which likely decreased the potential variability in dosing. Finally, the effect of very early administration (i.e., within 5 days) is based on low numbers (n = 14), and the results found in these analyses cannot be extrapolated beyond the observed time frame (i.e., 48 weeks). Indeed, the quadratic function included in our analysis, while providing reasonable fit, would yield inaccurate predictions of outcomes beyond 1 year.

The overall benefit of gabapentinoids on motor recovery has persisted in several studies, is unique to gabapentinoid anticonvulsants, and has strengthened with improved methodology.^3,4,8 Based on our results, gabapentinoids represent a potential candidate for drug repurposing in spinal cord injury.

Glossary

AIS

American Spinal Cord Injury Association Impairment Scale

ANOVA

analysis of variance

CI

confidence interval

EMSCI

European Multi-Centre Study about Spinal Cord Injury

ISNCSCI

International Standards for the Neurological Classification of Spinal Cord Injury

SCIM

Spinal Cord Independence Measure

Appendix. Authors

Study funding

This study was supported by Wings for Life, the International Foundation for Research in Paraplegia, the Craig H. Neilsen Foundation, the Canada Research Chair Program, and the Canadian Institutes of Health Research (ERA-NET). Catherine Jutzeler was supported by postdoctoral funding from the Craig H. Neilsen Foundation.

Disclosure

The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.