Main findings
This multicentre, randomised, double-blind, placebo-controlled trial showed that in women with CPP and no obvious pelvic pathology gabapentin was no more effective than placebo in reducing pain. The incidence of side effects and SAEs was higher in the gabapentin group than in the placebo group. Gabapentin appears to exert an effect on the ACC in women with CPP. Those taking gabapentin show a significant reduction in ACC responses to punctate stimuli after treatment, which in turn was associated with improvements in pain interference scores.
Strengths
To the best of our knowledge, this study is the only randomised placebo-controlled clinical trial to report on the treatment of CPP with gabapentin. The robust study design, which included blinding to treatment allocation of both participants and investigators, ensured internal validity, enabling the results to be interpreted with confidence. Randomisation was concealed via a computer-generated allocation sequence and achieved balanced groups with respect to pain symptoms during menstruation, psychological functioning and concomitant hormone use, all potentially prognostic for reported pain.
Chronic pelvic pain can fluctuate or follow the menstrual cycle; therefore, eliciting a pain score at a single time point is unlikely to capture the effect of gabapentin or reflect the women’s experience of pain. Instead, we sought a pain score weekly over a 4-week period, asking participants to rate both worst and average pain for the preceding week, and defined a minimum number of responses to create a valid outcome. Although it is preferable to have a single primary outcome, a survey of our patient involvement group found that worst pain and average pain were equally important to women. We, therefore, chose to use dual primary outcomes and considered both worst and average pain scores. These outcomes were considered separately and an improvement in one (or both) would conclude that gabapentin was efficacious. All of the outcome data in the GaPP2 study were subjective or participant-reported outcomes (rather than laboratory measurements), but the study was blinded, which reduced the risk of incurring assessor bias.
We calculated the sample size based on a recognised minimally important difference for chronic pain of 1 point on a 0–10 NRS, and used a SD from a comparable pilot study. We applied appropriate adjustments to account for the dual primary outcome in both the sample size calculation and the analysis. We recruited the target number of women and missing outcome data were as anticipated, with follow-up rate for the dual primary outcome of 80% of women.
A high proportion of women reported taking at least half of the study drug doses throughout the trial. The dose of gabapentin that participants received was based on individual adjustment of the dose by the participants themselves, which reflected their perception of pain relief and side effects. Adjustments were made in accordance with existing dosing recommendations, up to a dose of 2700 mg per day, and the final median doses ranged from 1200 mg per day to 1800 mg per day during the treatment phase. This is in line with the current NeuPSIG [neuropathic pain special interest group of IASP (International Association for the Study of Pain)] recommendations for the treatment of neuropathic pain,91 where gabapentin at a dose of 1200–3600 mg in three divided doses is a first-line treatment.
The fact that a mechanistic substudy was included is a strength of this study as it allows us to begin to disentangle the treatment response from the placebo response, in addition to exploring the mechanism by which gabapentin itself might work. fMRI provides a sensitive, objective outcome measure and, thus, allows a smaller sample size to be used than when patient-reported outcomes alone are used. All imaging was performed at the same site on the same MRI scanner and, thus, no additional variation with regard to data collection systems needs to be accounted for in the analysis. Furthermore, the evoked stimulus investigated was a punctate probe, which delivers a fixed force. This is less subject to variation than, for example, thermal stimuli, for which differences in methods of fixation to the skin or equipment calibration can produce variation in the actual stimulus delivered. It is relatively unusual for cohorts in such studies to be all women and, therefore, these findings may be of relevance to other chronic pain conditions (that are more prevalent in women), given how little is known about the mechanism of action and predictors of response to gabapentin.
Limitations
There are some limitations of our study that should be considered.
Twenty per cent of participants failed to provide pain scores at the end of treatment. Our analytical approach involved imputation of missing responses using a recognised method, but still makes an assumption about missing data being missing at random. Any deviation from this assumption could give rise to different results. The sensitivity analysis was almost identical to the observed data comparison and the CIs for both did not reach the minimally important clinical difference, so it is unlikely that a meaningful treatment effect was missed because of missing data.
There was a potential placebo response observed in the trial. Although we acknowledge that placebo responses are observed universally in almost all placebo-controlled randomised clinical trials,92 the response observed in this trial is very relevant because of the side-effect profile of gabapentin and its potential addictive properties.29 However, the trial was not designed to investigate this placebo response.
The rate of adherence to the trial regimen (women reported taking at least half of the study drug doses throughout the trial) was high; however, this was not validated against an objective method, such as pill counting.
The limitations of the mechanistic substudy were the small sample size, as not all women returned for their follow-up scan despite the best efforts of the research team. Moreover, by chance, the baseline data between the two groups were different for a number of variables. However, this has been accounted for as much as possible in the analysis strategies.
Despite recruitment from many hospitals across the UK, the study participants are overwhelmingly white women, which limited the generalisability of our study. CPP and dysmenorrhoea are commonly reported across the globe,3 but barriers to seeking medically and culturally appropriate care may exist,93 which is compounded by the well-documented under-representation in clinical trial research among black, Asian and minority ethic populations in the UK.94
Interpretation of findings
In conclusion, our results show that gabapentin did not relieve pain or improve physical and psychological function in women with CPP, as compared with placebo, over a course of 16 weeks. Gabapentin was associated with higher rates of side effects than placebo.
Data from a recent review65 showed that the number needed to treat to be 6.6 (95% CI 5.0 to 10) to achieve at least 50% pain-intensity reduction in painful diabetic neuropathy (1331 patients), and 6.7 (95% CI 5.4 to 8.7) to achieve at least 50% pain-intensity reduction in postherpetic neuralgia (2260 patients). The NeuPSIG review91 suggests that across all neuropathic pain conditions the number needed to treat is 6.3 (95% CI 5.0 to 8.3), for a 50% reduction in pain intensity with an associated number needed to harm of 25.6 (95% CI 15.3 to 78.6). The lack of treatment effect of gabapentin in women with CPP may reflect differences in the aetiology of neuropathic pain, suggesting that recommendations from guidelines on neuropathic pain may not apply to women with CPP.13,91 Alternatively, it may be that women gain less benefit but are more susceptible to harm from gabapentin. It is not possible to extract information on sexual dichotomies in responses from any of the existing systematic reviews.
Gabapentin was associated with higher rates of side effects than placebo in the trial (e.g. dizziness, drowsiness and visual disturbances), which is consistent with other published studies.21,95 Another recent meta-analysis of all trials for postherpetic neuralgia and painful diabetic neuropathy95 showed that, compared with placebo, gabapentin was associated with more drowsiness (14% for gabapentin vs. 5% for placebo; p < 0.001), dizziness (19% for gabapentin vs. 7% for placebo; p < 0.001), peripheral oedema (7% for gabapentin vs. 2% for placebo; p < 0.001) and gait disturbance or ataxia (14% for gabapentin vs. 3% for placebo; p < 0.001).
Although more women in the placebo group were able to correctly guess their allocation at the end of the treatment period (78/106 placebo, 54/111 gabapentin), their use of rescue medication was similar. It cannot be concluded that women who perceived that they were taking placebo compensated by increasing their analgesic use and, thus, negated any effect of gabapentin.
Public and patient involvement
We have been supported throughout the project by the charity Pelvic Pain Support Network (PPSN) and, in particular, its chief executive officer. Public and patient involvement was crucial in improving the acceptability of the GaPP2 trial and promoting engagement of gynaecologists. We engaged with PPSN throughout, improving our understanding of the opinions and uncertainty surrounding treatments for CPP, and the anecdotal evidence of the use of gabapentin that women have provided here. A survey disseminated by PPSN helped us to decide to include worst and average pain scores. We also discussed the other outcome questionnaires to be included in the trial. Members of the PPSN commented on patient-facing materials to ensure that they were clear and comprehensive. The PPSN promoted participating trial centres with their contact details on the website. We will engage with PPSN regarding the dissemination of our findings, providing a Plain English summary of the findings and the uncertainties around the evidence we have discussed here. This will be distributed via PPSN’s website and on their social media channels. Any future research groups taking forward the research recommendations from this project would benefit from engaging with PPSN.