Efficacy and Tolerability of Gabapentin in Adults with Sleep Disturbance in Medical Illness: A Systematic Review and Meta-analysis

Background and purpose The aim of this study was to systematically review the efficacy and tolerability of gabapentin in the treatment of sleep disturbance in patients with medical illness. Methods PubMed was searched for randomized, double-blinded, placebo-controlled trials that reported sleep changes during gabapentin treatment up to November 2015. Findings This review included 26 studies involving 4,684 participants. Except for Composite Endpoint 3 [standardized mean difference (SMD) = 0.09, 95% confidence interval (CI): −0.05–0.22] compared with the placebo group, the gabapentin group showed superior outcomes on our endpoints: Composite Endpoint 1 (SMD = 0.50, 95% CI: 0.28–0.71), Composite Endpoint 2 (SMD = −0.53, 95% CI: −0.77 to −0.30), Composite Endpoint 4 (SMD = −0.38, 95% CI: −0.58 to −0.19), Composite Endpoint 5 [risk ratio (RR) = 1.79, 95% CI: 1.24–2.58], and Composite Endpoint 6 (RR = 0.48, 95% CI: 0.32–0.72). However, the patients in the gabapentin group showed worse tolerance than those in the placebo group (RR = 1.38, 95% CI: 1.08–1.76). Implications This study is the first to systematically assess the clinical value of gabapentin for the treatment of sleep disorders. We found that regardless the type of sleep outcomes, gabapentin displayed stable treatment efficacy for sleep disturbance in patients with medical illness. However, when an average dose of approximately 1,800 mg/day was used, the risk of treatment discontinuation or drug withdrawal was relatively high. We recommend that further studies confirm these findings in patients with primary sleep disorders.

melatonin receptor agonists can all contribute to a certain extent (6,7), few of these treatments can either restore patients' normal sleep structure or completely cure sleep disorders. Gabapentin, an apha-2-delta voltage-gated calcium channel ligand (8) that is widely used for the treatment of epilepsy, neuropathic pain, and restless legs syndrome, can enhance slowwave sleep in both normal individuals (9) and epileptic patients (10,11) and can improve slow-wave sleep and sleep efficiency and reduce nighttime awakening in patients with primary sleep disorders (12). However, these findings have not been verified with randomized controlled trials. Clinical studies have revealed that gabapentin could improve the objective and subjective outcomes of sleep disturbance in patient with medical illness . Gabapentin Enacarbil (GEn) or XP13512 is a prodrug of gabapentin, used as an anticonvulsant and for pain relief in postherpetic neuralgia. This new formulation of gabapentin was designed for increased oral bioavailability over gabapentin. It provides reliable drug absorption and consistent bioavailability (16). Nevertheless, the results derived from these studies had certain inconsistencies and did not undergo any systematical evaluation. Through a systematic review of the use of gabapentin to treat restless legs syndrome, neuropathic pain, alcohol dependence, hot flashes in menopause, fibromyalgia, phantom limb pain, human immunodeficiency virus (HIV)associated sensory neuropathies, and bipolar disorder, this study attempted to evaluate the efficacy and tolerability of gabapentin for the treatment of sleep disturbance in patients with medical illness.

MaTerials anD MeThODs
This systematic review and meta-analysis were performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) (38). There are no ethical issues involved in our study because our data were based on published studies.

Data sources and search
PubMed was searched for all clinical trials related to the present research topic (up to November 8, 2015). The keywords selected from the Medical Subject Headings (MeSH) included intervention, study type, and endpoint event. The search range was "title/ abstract/keywords. " No language restrictions were applied. In addition, we screened the reference lists of all included trials to identify additional eligible studies. Detailed information regarding the search terms used in the literature search is provided in the Supplementary Material.

study selection
Eligible Trials (1) Participants: all included patients were 18 years or older and had/did not have a record of baseline sleep status; (2) intervention: the patients in the treatment group received gabapentin, gabapentin enacarbil, or XP13512 (Gabapentin), and the patients in the control group received placebos with a treatment duration of at least 7 days; (3) endpoints: all included trials reported sleep changes and treatment discontinuation or drug withdrawal events that were possibly or probably related to the study drugs; (4) study type: randomized, double-blinded, controlled trials were included.

Data extraction
Using a unified form, two investigators independently extracted the data and created the data spreadsheet, which were then cross-checked to ensure data accuracy. Disagreements were resolved by consensus. The extracted data mainly included the six composite endpoints and treatment discontinuation or drug withdrawal events that were possibly or probably related to the study drugs.

endpoint Definitions
Because of the diversity of outcomes reported in the included trials, only a limited number of trials provided data that could be pooled for each meta-analysis. To reach a sufficient statistical level, we introduced the concept of "composite endpoint" to pool the data related to sleep outcomes with similar significance and a consistent direction.
Based on the treatment outcomes and relevant data provided by the original trials, seven composite endpoints were analyzed for evaluation. Composite Endpoints 1-6 were used to evaluate the efficacy of gabapentin, and Composite Endpoint 7 was used to evaluate treatment discontinuation or drug withdrawal events that were possibly or probably associated with gabapentin. Composite Endpoints 1-4 indicated sleep improvement after treatment. Specifically, Composite Endpoint 1 represented the net increase in the evaluation indices provided in the trials in which the index values increased, but the baseline values were not provided. Composite Endpoint 2 represented the net decrease of evaluation indices provided in the trials in which the index values decreased but the baseline values were not provided. Composite Endpoint 3 and Composite Endpoint 4 represented     the posttreatment values of the evaluation indices provided in the trials in which the index values increased and the trials in which the index values decreased (none of these trials provided the baseline values), respectively. Composite Endpoint 5 (Excellent, 0 or Good) represented the sleep outcomes that received the highest grades in the survey, e.g., the overall quality of sleep was evaluated as "Excellent, " or the ability to function was evaluated as "Good, " or the number of nighttime awakenings caused by RLS symptoms was 0, or the number of hours awake per night because of RLS symptoms was 0 in the past week. Composite Endpoint 6 (Poor, ≥3, ≥5, or 7) represented the sleep outcomes that were graded the lowest in the survey, e.g., the overall quality of sleep was evaluated as "Poor, " or the ability to function was evaluated as "Poor, " or the number of nighttime awakenings caused by RLS symptoms was ≥5, or the number of hours awake per night because of RLS symptoms was ≥3, or the number of nights with RLS symptoms was 7 in the past week.

Quality assessment
Two investigators evaluated the methodological quality of all included trials according to the Cochrane Collaboration's tool for assessing bias [the Reviewer's Handbook (39)].

grading of recommendations assessment, Development, and evaluation (graDe) classification
Based on the GRADE study group criteria (20), we graded the evidence quality for all of the endpoints.

Data synthesis and analysis
Based on the formula and endpoint definition, the values of the same endpoints in each trial were pooled first and then the data from different trials were pooled together for analysis. The standardized mean difference (SMD) and risk ratio (RR) were used to assess the abovementioned endpoints. Prior to the metaanalysis of each endpoint, statistical heterogeneity across the various trials was tested using Chi-square test. A P-value greater than the nominal level of 0.10 and I 2 ≤40% indicated a lack of heterogeneity across trials, allowing for the use of a fixed-effects model; otherwise, a random-effects model was used. The inverse variance method was used for continuous variables, and the Mantel-Haenszel method was used for dichotomous variables. In addition, a sensitivity analysis was conducted by removing each trial one at a time, and the publication bias was evaluated using the Egger test. SPSS Predictive Analytics Software version 18.0 (SPSS, Inc., Chicago, IL, USA) was used for the Chi-square tests, and Stata Statistical Software version SE 12.0 (Stata Corp. LP, College Station, TX, USA) was used for all other analyses.

resUlTs search results and Trial characteristics
Ninety-eight records were identified through database searches and were screened by reading titles, abstracts, and part of main text. After irrelevant papers, observational studies, duplicates, and trials that used non-placebo control drugs were excluded, 26 papers (13-19, 21-37, 40, 41) met the inclusion criteria. The included publications comprised eight RLS-related trials, eight neuropathic pain-related trials, and three alcohol dependencerelated trials, two trials involving hot flashes in menopause, one fibromyalgia-related trial, one trial involving phantom limb pain, one trial involving HIV-associated sensory neuropathies, and one bipolar disorder-related trial. Among the included studies, six trials were included only for systematic review and 20 trials were included for meta-analysis.
The included 26 trials involved 4,684 patients. The average follow-up length was 11.07 weeks/per patient, and the total followup time was 997.23 patient-years. The average age of 83.50% of the patients was 55.45 (±13.45) years. Among 96.50% of patients, males accounted for 42.73%; among 90.67% of patients, the average length of disease course was at least 6.23 (±9.76) years. The initial dose of gabapentin was 300 or 600 mg/day; after the dose-increasing phase, the minimum dose was 600 mg/day and the maximum dose was 3,600 mg/day, with an average dose of 1,793.92 mg/day. Figure 1 presents the screening process used in the study, Table 1 lists the main characteristics of all included trials.

Tolerability
All of the trials reported mild-to-moderate adverse effects. The moderate adverse effects occurred primarily during the dose-increasing phase and significantly decreased in frequency afterward. Drowsiness, dizziness, and weakness were the most frequently reported effects. These discomforts were tolerable for the majority of patients but resulted in drug withdrawal in a portion of patients. A meta-analysis of 20 trials (7, 14-17, 19, 21, 22, 25-28, 32-37) showed that for adverse events that were possibly or probably related to the study drug and could lead to treatment discontinuation and drug withdrawal, the gabapentin group had a 1.45-times higher risk than the placebo group (RR = 1.38; 95% CI: 1.08-1.76; Figure 3); For adverse events that were possibly or probably related to the study drug and could lead to treatment discontinuation and drug withdrawal, the incidences in the gabapentin group and the placebo group were 8.19 and 5.37% (P < 0.001), respectively. Sixteen trials (14, 17, 19, 21, 22, 25-29, 31-34, 36) reported serious adverse effects. However, other than one case of headache (34), one case of serious dizziness and drowsiness (21), and one case of vision disturbance (19), no serious adverse effects were associated with the use of gabapentin.
No serious adverse events associated with the use of placebos were found.

graDe classification
For the GRADE classifications of evidence quality, the high, moderate, low, and extremely low were 0, 3, 3, and 0, respectively ( Table 3).

sensitivity analysis
The sensitivity analysis indicated that, for Composite Endpoint 1, the removal of any one trial led to a lower limit of the CI of SMD that was higher than 0; for Composite Endpoint 2 and Composite Endpoint 4, the removal of any one trial led to an upper limit of the CI of SMD that was lower than 0; for Composite Endpoint 5 and Composite Endpoint 7, the removal of any one trial led to a lower limit of the CI of the RR that was higher than 1; for Composite Endpoint 6, the removal of any one trial led to the lower limit of the CI of the RR that was lower than 1 ( Figures  S3-S8 in Supplementary Material). The above results suggest that the results for these endpoints were robust and had a low sensitivity.

Publication Bias
The P values of all endpoints derived from the Egger test were greater than 0.05, indicating there was no publication bias ( Table 4).

DiscUssiOn
This study revealed that without consideration of the type of sleep outcomes, gabapentin was significantly superior to placebos for the treatment for sleep disorders secondary to RLS, neuropathic pain, alcohol dependence, hot flashes in menopause, fibromyalgia, phantom limb pain, HIV-associated sensory neuropathies, and bipolar disorder. However, with an average dose of approximately 1,800 mg/day, gabapentin had a higher risk of treatment discontinuation and drug withdrawal compared with placebo. The above conclusion was drawn from an extensive summary of trials involving various primary diseases. Only a small portion of these trials reported the baseline sleep status (14, 18, 21, 26, 27,36,37,41), and none of these trials reported the sleep status prior to the disease. Because it was impossible to distinguish absolutely true, partially true, and false sleep disturbance, we could not exclude the contribution of false sleep disturbance to the final treatment efficacy in patients with medical illness. However, it is worth noting that more than 90% of the patients in these trials had an average disease course of 6.23 (±9.76) years.
In terms of the psychological aspects of insomnia, the intention to fall sleep often becomes a driving factor of sleep difficulty (47) and worries about being sleepless often cause early awakening or anxiety (48), particularly among patients who are prone to excessive worry or over thinking. Without timely correction, one episode of sleep difficulty can easily induce a second episode in patients with related psychological traits, and as a result, ongoing sleep difficulties ultimately lead to a chronic sleep disorder. Some researchers believe that the initiating event does not significantly affect the progression of chronic sleep disorders (49) and that chronic sleep disorders are not closely associated with primary disease and thus do not improve with the improvement of the primary disease. In other words, during the chronic course of the abovementioned primary diseases, false sleep disturbance might have transformed into true or partially true sleep disturbance in patients with medical illness for the majority of the sample pool. Thus, we believe the existence of false sleep disturbance in medical illness would not significantly affect the results of the efficacy analysis, and the improvement of sleep disorders can be attributed to the efficacy of gabapentin treatment. The following experimental evidence supports this deduction: gabapentin can shorten sleep latency (36), reduce awakenings (12,26,35,36), reduce fast-wave sleep (23), enhance slow-wave sleep (9-12, 18, 36), prolong the total sleep time (18,23,36), increase sleep efficiency (12,18,36), and improve the quality of sleep (17,23,35,36). In fact, because of its sedative effect in various diseases, gabapentin has been clinically used as a hypnotic (48). Nevertheless, its efficacy for primary sleep disorders remains to be verified by randomized controlled trials, and the optimal dosage that is effective and tolerable in most patients needs to be identified. It is necessary to emphasize that despite its insignificant impact on the progression of sleep disorders, the initial sleep difficulty can induce the recurrence of disease (49). In other words, the complete cure of sleep disorders requires a complete removal of the initiating stimulus. Therefore, the use of gabapentin in the abovementioned diseases can "kill two birds with one stone. " Moreover, it is worth noting that pooled statistics were used with the basic premise of analyzing the efficacy of gabapentin. In this study, we introduced the concept of "composite endpoints" to pool sleep-outcome data that had similar significance and consistent direction. In a broad sense, this research method is in accordance with the basic principle of meta-analysis (39).

research significance
Through a systematic review and meta-analysis, this study for the first time systematically evaluated the clinical value of gabapentin for the treatment of sleep disorders. Used as a starting point, this study could inspire more researchers to conduct in-depth research on this topic.

study limitations
Because of the difficulty of distinguishing false sleep disturbance from true ones in patients with medical illness, we were unable to exclude their contribution to the treatment efficacy. In addition, because of the limitations of the original trials, we were unable to conduct a meta-analysis of individual sleep outcomes and analyses related to treatment dose and timing or patient gender.
cOnclUsiOn This is the first study to systematically evaluate the clinical value of gabapentin for the treatment of sleep disorders. Regardless the type of sleep outcomes, gabapentin showed stable efficacy in the treatment for sleep disturbance in patients with medical illness with a relatively high risk of treatment discontinuation and drug withdrawal when used at an average dose of approximately 1,800 mg/day. Because the adverse events often occurred during the dose-increasing phase, and the dose was high, reducing the dose-increasing speed and lowering the dosage of gabapentin might reduce the risk. In addition, it would be ideal if our conclusions could be further verified in patients with primary sleep disorders.

aUThOr cOnTriBUTiOns
All authors contributed equally to this work.