Pregabalin and gabapentin for chronic low back pain without radiculopathy: a systematic review

Background  Chronic low back pain (CLBP) is a global health problem, and gabapentin and pregabalin are often used in the treatment of patients without associated radiculopathy or neuropathy. Therefore, determining their efficacy and safety is of enormous value. Objective To examine the efficacy and safety of using gabapentin and pregabalin for CLBP without radiculopathy or neuropathy. Methods  We performed a search on the CENTRAL, MEDLINE, EMBASE, LILACS, and Web of Science data bases for clinical trials, cohorts, and case-control studies that evaluated patients with CLBP without radiculopathy or neuropathy for at least eight weeks. The data were extracted and inserted into a previously-prepared Microsoft Excel spreadsheet; the outcomes were evaluated using the Cochrane RoB 2 tool, and the quality of evidence, using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system. Results  Of the 2,230 articles identified, only 5 were included, totaling 242 participants. In them, pregabalin was slightly less efficacious than amitriptyline, the combination of tramadol/acetaminophen, and celecoxib, and pregabalin added to celecoxib showed no benefit when compared to celecoxib alone (very low evidence for all). On the other hand, although one study with gabapentin did not support its use in a general sample of patients with low back pain, another found a reduction in the pain scale and improved mobility (moderate evidence). No serious adverse events were observed in any of the studies. Conclusion  Quality information to support the use of pregabalin or gabapentin in the treatment of CLBP without radiculopathy or neuropathy is lacking, although results may suggest gabapentin as a viable option. More data is needed to fill this current gap in knowledge.


INTRODUCTION
Low back pain, typically defined as pain below the costal margin and above the inferior gluteal folds, with or without leg pain, 1 is usually classified according to duration as acute (< 6 weeks), subacute (6 to 12 weeks), or chronic (> 12 weeks). 2 Extremely common in populations throughout the world and occurring in all age groups, from children to the elderly, [3][4][5] it is a global health issue that has been responsible for 60.1 million disability-adjusted life years (DALYs) in 2015, 6 and it is currently the leading cause of disability. 7 It is estimated that up to 84% of all adults have at least 1 episode at some point in their lives, and it is one of the most common reasons for a primary care visit. 8,9 Although rapid improvement in pain and disability and return to work is the norm within the first month, 10 symptoms may persist beyond 12 weeks in some people. 11 When this happens, the use of medications to provide symptomatic pain relief, enabling the patient to participate in active therapies and encouraging increased function and improved coping can be implemented.
Furthermore, low back pain can be classified as mechanical, radicular (neuropathic), or primarily nociplastic in nature, 12 and the prevalence of the neuropathic pain ranges from 16% to 55% in patients with chronic low back pain (CLBP). [13][14][15] Therefore, drugs that were originally antiepileptics and their derivatives, mainly gabapentin and pregabalin, have been used as an alternative to other more traditionally recommended drugs in the treatment of CLBPnon-steroidal anti-inflammatory drugs (NSAIDs), duloxetine, tramadol, among others -, which have several limitations, adverse effects, and risks that are well-known with the long-term use. [16][17][18][19][20][21] However, evidence proving the real efficacy and safety of gabapentin and pregabalin in the treatment of CLBP, especially in the absence of radiculopathy or neuropathy, is still limited, with mixed and often inconclusive results. In addition, there are frequent reports of adverse effects associated with these medications, which highlights the need for further studies and analyses of the real pros and cons of their use. [22][23][24][25] Therefore, the present study aims to evaluate gabapentin and pregabalin in terms of their efficacy and safety in the treatment of CLBP without radiculopathy or neuropathy, according to the results published so far in the medical literature, through a systematic review.

Palavras
Studies with pregnant women, with people in conditions eminently indicative of immediate surgical or interventional treatment, who had significant cognitive impairment, with low back or back pain caused by pathological entities (such as infections, neoplasms, metastases, osteoporosis, rheumatoid arthritis, fractures, or trauma) were excluded, as well as studies that were not published in full as articles (such as posters or conferences annals). In case of clinical trials, those whose protocols could not be found on international clinical trials databases were also excluded.

Search methods
We identified studies through advanced searches on the Cochrane Central Register of Controlled

Data collection and analysis
One reviewer extracted and gathered the search results and excluded clearly ineligible studies based on title and abstract. After that, the full articles of all remaining studies were retrieved. After reading these articles in full, those clearly ineligible were excluded. The remaining studies were reanalyzed by two reviewers, and only then were they excluded or included in the final composition of the review. We resolved any disagreements by consensus among the review authors. One of the reviewers manually extracted and inserted the data into a spreadsheet prepared by consensus by the reviewers using the Microsoft Excel (Microsoft Corp., Redmond, WA, United States) software.
One of the reviewers assessed the risk of bias of all included studies using version 2 of the Revised Cochrane risk-of-bias tool for randomized trials (RoB 2), 26 described in ►Appendix 6 (►Supplementary Material). We classified each of the criteria as "low risk", "some concerns", and "high risk". For the criteria classified as "some concerns", we did not contact the trial authors for further information.
Efficacy data were examined according to previouslyestablished outcome measures (►Appendix 7 -►Supplementary Material). Any serious adverse events were mentioned in separately from the less serious ones. As for the other outcomes, such as any other pain-related outcome indicating some improvement, they were assessed using the Descriptor Differential Scale (DDS) and a standard numerical rating scale (0 ¼ "no pain", 10 ¼ "worst imaginable pain"). Adverse events were measured by the proportion of participants who experienced them.
We did not assess clinical heterogeneity for any of the clinical trials included, as they were very different from the start, both in terms of intervention and comparator, and in relation to the general population studied. Because of this too, only a meta-analysis of the proportion of adverse events experienced comparing gabapentin versus placebo between two studies could be carried out.
Among our outcomes, we used dichotomous data of known usefulness. 27 We would only perform a meta-analysis if there were at least two studies with sufficiently similar participants, interventions, comparisons, and measurement of outcomes. Otherwise, we would describe the results of comparable clinical trials in the review text. To assess and synthesize the quality of evidence for each result, we used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions, 28 based on the following domains: limitations of design, inconsistency of results, indirectness, imprecision, and other factors (such as publication bias). Finally, we developed "Summary of Findings" tables to present the certainty (or quality) of the body of evidence.

Description of the studies
We identified 2,230 potential articles through the primary electronic search and manual review of the research protocols found in it (►Figure 1). The main reasons for exclusion were because some articles did not contain a population of patients with low back pain without radiculopathy/neuropathy, not even after the division into groups (such as with and without radiculopathy/neuropathy), or they did not provide enough information about the presence of radiculopathy/ neuropathy. Finally, less frequent were the study protocols in which there were no publications or dissemination of data until the end of our search, on August 20, 2022. All articles or protocols contained at least one alternative title in a language understandable to the authors. After the selection by title, all articles also contained at least one alternative abstract in an understandable language. Finally, all articles read in full were written in understandable language and, after a selection, five were included. The sample size of these studies ranged from 30 to 200 randomized participants, with a total of 445 participants.
However, because only one of the studies contained a pure sample of interest to us (McCleane, 2000 29 ), that is, the other studies also contained participants who did not fit our selection criteria (such as patients with radiculopathy or associated neuropathy), only a portion of the population of these other studies was included. In this case, the sample size of the five trials ranged from 20 to 93, with a total of 242 participants. All of these studies were randomized controlled clinical trials and had significant particularities, as can be seen in the ►Table 1.
Among the outcomes previously established for the present research (►Appendix 7 ►Supplementary Material), only those described as follows were reported in the selected studies and were applied to the present review. The primary outcome, participant-reported reduction in pain intensity of 50% or more, was measured through the Visual Analog Scale (VAS). The assessment of functional improvement was measured through the Oswestry Disability Index (ODI) and Roland Morris Disability Questionnaire (RMDQ).

Risk of bias in the studies included
The assessment of the risk of bias is presented in ►Figure 2.  30 In terms of 'allocation', the five studies selected 29-33 reported a randomization procedure. Of these, one 31 did not provide clear information on allocation sequence concealment, two 29,30 adequately described treatment allocation concealment, and two 32,33 were not explicit in the description of treatment allocation concealment, although we can infer that there was. In the five studies, any baseline differences observed among the intervention groups appear to be by chance.
Two 29,30 of the studies reported blinding of the patients, caregivers, and outcome assessors, and two 32,33 reported blinding of the outcome evaluators. The latter 33 presented an incongruity when stating in the introduction that it was a single-blinded study, while in the methods section the authors 33 stated that it was a double-blinded study. One of the studies 31 did not blind the patients, caregivers or outcome assessors.
In terms of incomplete outcome data, for the continuous outcomes, with the availability of data from 95% of the participants, the dropout rate was considered small. For the dichotomous results, the dropout rate was considered small when data from at least 80% of the participants were available. When the dropouts were justified, such as in case of adverse events, with the description of the group they were in, we considered those studies less prone to bias. Three of the studies 29,32,33 reported small dropout rates; the other two studies 30,31 reported dropout rates higher than 20%, but only these two studies performed an intention-to-treat analysis (ITT).
The studies showed differences in baseline characteristics and time of outcome assessment. The mean age was much higher in the study by Sakai et al., 32 (72.5 years) than in the others (41.5 to 56.04 years), and the proportion of male patients was higher in Atkinson et al. 30 (78.7% versus 44.44% to 66.67%). Furthermore, the follow-up varied from 4 32,33 to 14 31 weeks. On the other hand, all studies at least avoided cointerventions (or advised that they should only be performed when necessary, not regularly). In one study, 30 maintenance of stable complementary NSAID therapy was allowed. In another study, 29 the patients were allowed to remain on a stable dose of NSAIDs and to continue the use of a compound analgesic based on paracetamol and codeine as rescue analgesia. However, they were asked to remain using the same compound preparation and to take it only when needed, not regularly.
None of the studies showed significant conflicts of interest. We did not create funnel plots to assess potential   publication biases due to the small number of studies included.

Gabapentin compared to placebo
The study by McCleane, 29 the only one that contained a pure sample of participants of our interest, obtained a reduction in the mean pain 0-10 verbal numeric rating scale only during the use of gabapentin: from 7.10 (95% confidence interval [95%CI]: 6.26-7.94) to 6.39 (95%CI: 5.39-7.39); p < 0.05-moderate degree of evidence (GRADE). One of the participants withdrew due to the side effects of gabapentin, and adverse events were reported in both groups, however, with a significant higher number among those using gabapentin (9 out of 30) than among those taking placebo (  with no significant difference between them (2.2 versus 1.6; p ¼ 0.253). Importantly, these results from Atkinson et al. 30 referred to a mixed population of patients with radiating (46 patients) and non-irradiated (62 patients) pain, but the reduction in intensity was similar among these participants, both within and between treatment arms (none of the pvalues from the mixed-model analysis was significant

Pregabalin compared to tramadol/acetaminophen
For the subgroup of patients without neuropathic pain, Sakai et al. 32 obtained significant pain improvement in the VAS at 4 weeks in both groups (p < 0.05 for both)very low degree of evidence (GRADE). However, in the tramadol/acetaminophen group, this improvement could already be observed after 2 weeks (p < 0.05), that is, the effect in the pregabalin group took longer to be observedvery low degree of evidence (GRADE). As for the functional improvement measured by the RMDQ, no significant improvement was noted for the pregabalin group, whereas, for the tramadol/acetaminophen group, a significant improvement was observed after two weeks of administrationvery low level of evidence (GRADE). Adverse events were reported in both groups, with a significantly higher number in the group that was using tramadol/acetaminophen (risk ratio: 0.59 [95%CI: 0.35-0.99]; p ¼ 0.04).  Abbreviations: CI, confidence interval; GRADE, Grading of Recommendations Assessment, Development and Evaluation. Notes: GRADE Working Group grades of evidence: high certaintywe are very confident that the true effect lies close to that of the estimate of the effect; moderate certaintywe are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different; low certaintyour confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect; very low certaintywe have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. a Lowered a level because of severe inaccuracy due to very few events in the study. A total of 4 out of the 42 recruited patients discontinued the treatment due to adverse events (epigastralgia and/or nausea), with one taking pregabalin monotherapy, one taking celecoxib monotherapy, and two taking pregabalin plus celecoxib.

DISCUSSION
The present review contains information from 5 studies totaling 242 participants. In the comparison between gabapentin and placebo, McCleane 29 reported a subtle reduction in the score on the pain scale only with the use of gabapentin, while Atkinson et al. 30 observed a decrease in pain in both groups, with no significant difference between them. Both studies demonstrated a greater presence of adverse events with the use of gabapentin than with placebo. Pregabalin, in turn, was compared with amitriptyline, tramadol/acetaminophen, celecoxib, pregabalin/celecoxib, and, finally, pregabalin/celecoxib was compared with celecoxib. In these comparisons, in no case was the pregabalin monotherapy superior to its comparator for pain relief, sometimes being inferior. 31,33 As for safety, there was no significant difference between the compared groups, with the exception of those in the study by Sakai et al., 32 in which patients using pregabalin reported fewer adverse events than those submitted to the coadministration of tramadol/acetaminophen.

Overall completeness and applicability of the evidence
Only the study by McCleane 29 study entirely composed of patients of interest to the present reviewindividuals with nociceptive/mechanopostural CLBP, without radiculopathy or neuropathy -, which demonstrates the difficulty of finding studies with samples exclusively composed by this particular group. In the other studies, it was necessary to extract data from subgroups of the total set of patients, which significantly compromises the quality of the results for the purposes of the present review. Regarding our main outcomes, none of the studies reported the occurrence of any serious adverse event. Other outcomes, such as improvement in pain in the VAS and functional improvement by the ODI or RMDQ, were partially covered by the included articles; however, as highlighted, the difficulty in finding research entirely on patients of interest to us compromises the applicability of the results. Moreover, the fact that we chose not to examine the grey literature may have led to a higher risk of non-reporting bias (such as non-publication bias).

Quality of the evidence
Although in general the studies included were not of very low methodological quality or high risk of bias, the quality of their evidence was greatly affected by several factors. The Notes: GRADE Working Group grades of evidence: high certaintywe are very confident that the true effect lies close to that of the estimate of the effect; moderate certaintywe are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different; low certaintyour confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect; very low certaintywe have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. a Lowered two levels because of very serious limitations of the study, due to some concerns about the risk of bias in the study by not blinding the allocation; b lowered a level due to severe inaccuracy, because the sample size criterion was not met in the per-protocol analysis; c lowered a level due to "Indirectness" of the severe evidence, as only part of the population was of interest to us.
need to extract data from only a portion of the population, the very small sample size, and the risk of bias due to the nonblinding of the participants or the uncertainty regarding this process severely affected the quality of the evidence. This resulted in the fact that, finally, we maintained only two outcomes in one of the studies with moderate quality of evidence, with the other results being classified as of very low quality of evidence.

Potential biases in the review process
We tried to avoid bias in the review process by conducting a comprehensive search without language restrictions, developing a comprehensive search strategy to identify all available evidence to answer our research question. However, a double full review by two reviewers was only performed after the initial exclusion of clearly ineligible articles by one of the authors. This represents a limitation, as it may increase the risk of human error in this selection. Furthermore, only one of the reviewers performed the data extraction and assessed the risk of bias of the included studies, which also represents a limitation.
Aiming to expand the scope of the primary search, allowing for the inclusion of studies with less objective definitions than the current ones for chronic pain, [34][35][36] we considered CLBP or back pain as pain for at least two months. While this may theoretically have limited the generalizability of our findings, 4 out of the 5 included studies defined CLBP as pain lasting longer than 12 weeks, and only 1 (McCleane 29 ) did not define it clearly. We have covered, in addition to publications on the subject, registered trial protocols. Furthermore, to ensure compliance with the revision primarily proposed, before starting the searches, we submitted our research project to the Research Project Management System (Sistema Gerenciador de Projeto de Pesquisa, SGPP, in Portuguese) of Hospital Israelita Albert Einstein, which was developed in accordance with the Lean Six Sigma requirements of the Process Improvement Program. Searches were not carried out in the grey literature, considering the generally lower methodological quality of these studies.
Finally, we endeavored to conduct a systematic review that followed the guidelines published and provided by the Cochrane Handbook for Systematic Reviews of Interventions. 28

Agreements and disagreements with other studies or reviews
No study, except one 29 of those included in the present review, had a sample entirely composed of the population of interest to us, which limits the comparison with previous studies. However, our results were very similar to those found in another review 22 that evaluated the use of gabapentin and pregabalin for CLBP regardless of the neuropathic or radicular component. In its results, pregabalin was slightly less effective than other analgesics, such as amitriptyline, celecoxib, or tramadol/acetaminophen, and pregabalin used as adjuvant therapy (added to other medicationsto celecoxib, in the case of the present review) did not show Notes: GRADE Working Group grades of evidence: high certaintywe are very confident that the true effect lies close to that of the estimate of the effect; moderate certaintywe are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different; low certaintyour confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect; very low certaintywe have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. a Lowered a level because of severe study limitations due to some concerns about the risk of bias from non-blinding of patients; b lowered a level due to severe inaccuracy, because the sample size criterion was not met in the per-protocol analysis; c lowered a level due to "indirectness" of the severe evidence, as only part of the population was of interest to us.
benefits ether. However, unlike our findings, the gabapentin group experienced no significant reduction in pain compared to the placebo group (mean difference ¼ 0.22 units; 95%CI: -0.51-0.07; p ¼ 0.14). In fact, one 31 of the studies we included did not support the use of gabapentin for a general sample of low back pain with and without pain radiating to the legs, but another, 29 only with patients without radicular pain or neuropathy, found a subtle pain reduction in the 0-10 verbal numeric rating scale during gabapentin use (of 7.10 to 6.39; p < 0.05), as well as an improvement in mobility (from 4.65 to 5.46; p < 0.01), with moderate quality of evidence (GRADE) for both results. Three main reasons may explain the differences in the findings of the other review: 29 1) the present review did not consider one of the studies included in this other review because the patients had associated leg pain; 2) we used only the population portion of the study by Atkinson et al. 30 with pain confined to the low back; and 3) the aforementioned review converted all study outcomes for pain relief expressed as continuous scores into a common 0-10 numerical rating scale.
In conclusion, the present review showed that there is still no quality information to support the use of pregabalin or gabapentin for the treatment of nociceptive/mechanopostural CLBP without radiculopathy or neuropathy, although the results suggest that the gabapentin may be a viable option. This corroborates the need for further data to fill the current gap in knowledge regarding this very relevant question.
Authors' Contributions RTT: conceptualization, data curation, formal analysis, methodology, resources, visualization, writingoriginal draft, and writingreview and editing; AWP: conceptualization, formal analysis, methodology, project administration, supervision, validation, visualization, writingoriginal draft, and  Notes: GRADE Working Group grades of evidence: high certaintywe are very confident that the true effect lies close to that of the estimate of the effect; moderate certaintywe are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different; low certaintyour confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect; very low certaintywe have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. a lowered a level because of severe study limitations due to some concerns about risk of blinding bias; b lowered a level because of severe inaccuracy due to very few events in the study; c lowered a level due to "indirectness" of the severe evidence, as only part of the population was of interest to us. writingreview and editing; CACO: conceptualization, data curation, formal analysis, methodology, project administration, supervision, validation, visualization, writingoriginal draft, and writingreview and editing.