Randomised controlled trial of gabapentin in Complex Regional Pain Syndrome type 1 [ISRCTN84121379]

This study complied with the Declaration of Helsinki regarding investigations in humans after approval of the protocol by the Institutional Review Board of the University Hospital Maastricht, the Netherlands. Patients were recruited from a database with patients who, in recent years, had been diagnosed with complex regional pain syndrome type I in our hospital. All patients had been treated in our pain management and research center (dept. of Anesthesiology, University Hospital Maastricht, The Netherlands) and had received sympathetic blocks[28], mannitol infusions[29, 30], and transcutaneous neuromodulation[31]. All participating patients fulfilled the IASP criteria[32] for the diagnosis of CRPS type I and were included if they were between 18 – 75 years old and had a score for pain > 3, as rated on a visual analog score (VAS), where 0 is no pain and 10 is the worst pain imaginable. Apart from IASP criteria, all patients had functional loss and pain outside the original traumatized area. Patients were excluded in case of a possibility of health risk or confounding by other diseases of syndromes, like e.g., pregnancy, known kidney and/or severe liver disease, another (2nd) chronic pain syndrome, known nerve damage in the affected area, active infection or diabetes mellitus. Patients were participating in 8-week periods from 19-11-1998 until 2-12-1999. Gabapentin was not registered as a drug in the Netherlands before or during the trial. After the trial the producing pharmaceutical company supplied gabapentin for compassionate use if indicated.

Since our patient population consisted of chronic CRPS I patients with a multiple years' duration of pain complaints refractory to various treatments, we assumed that their pain complaints would be more or less stable. We therefore undertook a double blind, randomized crossover study. Randomization of patients took place after baseline measurements and written informed consent. The assignment scheme was generated by our hospital pharmacy from a table of random numbers. The closed envelopes containing the assignments were prenumbered and kept at the pharmacy. The first treatment group received gabapentin, followed by a washout period and placebo treatment. The second treatment group received placebo treatment, which was followed by a washout period and gabapentin treatment. Each medication period lasted three weeks separated by the two-weeks washout period. Medication was stocked and delivered to the patient at the hospital pharmacy. Both the gabapentin capsules and the identical placebo capsules were delivered immediately before the start of the two medication periods. Left over medication was recollected and counted.

The gabapentin (GBP) dose was slowly increased to reduce adverse side effects:

600 mg's GBP AN once a day on day 1–2

600 mg's GBP b.i.d. on day 3–4

600 mg's GBP t.i.d on day 5–21

Placebo dose was identically titrated. Patients were allowed to take their usual analgesics and were told preferably not to change the usual dose.

The patients were reevaluated at the hospital three weeks (T1), five weeks (T2) and eight weeks (T3) after randomization. During the trial, the patient noted her/his pain rate of the past 24 hr (VAS) and the use of additional analgesics in a diary. During each hospital visit the following assessments were done:

A blinded independent investigator (STvdB) did sensibility, allodynia and range of motion tests (see above). A physician (AvdV), who examined each patient, did all the other measurements throughout the trial. Patient, investigator and physician were unaware of the treatment received. We tested blinding by questioning physician and participants after each medication period.

The statistical analysis of VAS-scores was determined per patient using estimating medication and period effect through linear regression analyses. Possible relationship of patient characteristics and outcome was tested by Pearson R's test. Mann-Whitney analyses were used for monofilament sensitivity on log-transformed data. Three point scales and seven point scales were dichotomized and like the SCL-90-R, NPS and CRPS-Brief Inventory questionnaires intra-individual paired tested (McNemar, t-test, Bonferroni-Holm corrected for multiple tests). Student t-tests and regression analysis were used to test treatment effect, which is calculated in crossover studies as ((AT1-AT0)-(AT3-AT2))/2+((BT3-BT2))-BT1-BT0)/2, where A represents data of placebo starters and B data of gabapentin starters both before (T0,T2) and after (T1,T3) treatments[39]. Blinding was tested with Chi-square analyses. Possible related factors to therapy effect were analyzed with forward stepwise logistic regression. Data analyses required complete data sets. Patients who were not completing one or two treatments were excluded for analyses. We tested two-tailed, with α = 0.05 as a level of significance (Excel 2000, SPSS 10.0 for Windows).

After randomization 58 patients were enrolled, with a mean age of 44.0 (range 24–75) resulting in 29 patients in the gabapentin-placebo arm and 29 patients in the placebo-gabapentin arm; 49 patients completed the gabapentin period, 50 patients the placebo period, 46 patients completed both periods and were used for further within-patient paired analysis (Fig. 1). Twelve patients discontinued treatment of which 6 during the placebo treatment, 2 during washout and 4 during GBP treatment. Three of these four GBP users discontinued due to side effects (Fig. 1). Between randomization and start of (placebo) medication one patient withdrew after rereading the information letter about possible side effects. These patients were excluded from analysis, since intra-individual testing was necessary for most of the data-analyses. Patients, who could not be used for analysis, did not differ in their characteristics from the total group nor comparing between the two arms of treatment (Tables 1 and 2). When comparing the placebo-GBP arm and GBP-placebo arm on sexes, age and pain level before period 1 or 2, duration of illness, SCL-90-R score, we could not find a difference between the two arms (Tables 1 and 2). SCL-90-R score revealed increased values on any subscale comparing to standard norms, indicating personal distress (Table 2). We found relative higher scores on somatic and sleeping complaints. The SCL-90-R scores were identical to control chronic pain patients (N = 143), besides higher score on sleeping complaints (T. Forouzanfar, data not published). Trial medication was returned and counted afterwards, but revealed no lack of compliance in any patient.

After each medication period both patient and physician were asked about their ideas concerning study medication in the past period. The treating physician guessed the used medication correctly more often after both phases than can be explained by coincidence (p = 0.000). Blinding for patients was sufficient in the first phase, but not anymore after the second phase (p = 0.2 versus p = 0.000).

Each participant was tested throughout the study on response to mechanical stimuli with von Frey filaments. The CRPS patients had sensory deficits at baseline. Application of smaller filaments was not felt in multiple skin areas. We found, with Mann-Whitney analyses, that this sensory deficit was significantly reversed in gabapentin users in comparison to placebo users (p = 0.027). This difference was found in patients with upper and lower extremity CRPS, but was still significant in the subgroup of lower extremity CRPS (p = 0.011) as seen in table 4.

Mechanical allodynia to static and dynamic stimuli (soft touch and brush) was measured by a mean of 11-point scales (range 0–10) of 9 areas of the hand/feet corresponding to cutaneous nerve branches. We found no effect of gabapentin on allodynia in comparison to placebo.

No difference was found on the parameters edema, discoloration, range of motion of wrist/ankle and fingers/toes between placebo and GBP. 10 patients out of 45 improved in relative skin temperature during placebo use compared to 18 patients out of 45 in gabapentin, which is two sided tested not significantly different (McNemar analysis, p = 0.096).

The reported function improvement was, with 10 positive responders during GBP versus 7 positive responders during placebo, not significantly different (N = 46) between the two treatments. The SCL-90 showed no significantly better scores during gabapentin treatment. CRPS-BI showed improvement of sleep between placebo treatment and gabapentin treatment., but this effect disappeared after Bonferroni-Holm correction.

Dizziness, somnolence and lethargy were significantly more often reported during gabapentin use than during placebo. Before washout 95 % of patients (n = 21) reported side effects during gabapentin use versus 58 % in placebo treatment (n = 14). After washout this was respectively 63% (n = 15) in GBP and 32% (n = 7) in placebo use. For more details on side effects see table 6. Since a high incidence of side effects can produce a stronger placebo effect, we analyzed the possible correlation between side effects and pain relief. There was a small relation between perceived side effects and pain relief in placebo users in period 2 (p = 0.04, Pearson's R value is 0.4), but, whether in period 1 or period 2, no relationship was found during the use of gabapentin (p = 0.2 in period 1, P = 0.4 in period 2).