A Randomized Study of the Effects of Gabapentin on Postamputation Pain

Patients were recruited from the Department of Orthopedic Surgery at Aarhus University Hospital, Aarhus, Denmark. The enrolment period began on May 13, 2002, and lasted until May 24, 2005. Inclusion criteria were age older than 18 yr and lower limb amputation because of peripheral vascular disease. Exclusion criteria were ipsilateral reamputation; amputation of foot or toes only; dementia or inability to answer an in-depth questionnaire about pain; psychiatric disease; severe cardiac, pulmonary, or liver disease; severely impaired kidney function (creatinine clearance ≤ 30 ml/min); current history of alcohol or drug abuse; known allergy to gabapentin; and treatment with anticonvulsants or tricyclic antidepressants.

Written informed consent was obtained from all patients. The protocol was approved by the Regional Ethics Committee (No. 2001-0258), Aarhus, Denmark; the Danish National Board of Health (No. 2612-1806), Copenhagen, Denmark; and the Danish Data Protection Agency (No. 2011-41-1428), Copenhagen, Denmark. The study was conducted in accordance with the guidelines for Good Clinical Practice and monitored by the Unit of Good Clinical Practice, University of Aarhus, Aarhus, Denmark.

Patients were randomly assigned to receive gabapentin or placebo for 30 days after the amputation, using a computer-generated randomization list in block sizes of 8 and 10. The study drugs (gabapentin and placebo supplied by Pfizer-Pharmacia, Ballerup, Denmark, as identical-appearing capsules) were prepared by the hospital pharmacy in identical containers marked with the name of the project and consecutive patient numbers. During the study, the randomization list was held securely at the hospital pharmacy and released only after study completion. Patients received one capsule (300 mg gabapentin or placebo) on the first postoperative day, three capsules (900 mg) on days 2–4, four capsules (1,200 mg) on days 5 and 6, five capsules (1,500 mg) on days 7 and 8, six capsules (1,800 mg) on days 9 and 10, seven (2,100 mg) capsules on days 11 and 12, and eight capsules (2,400 mg) on days 13–30. Patients with a creatinine clearance ≥ 30 ml/min but ≤ 60 ml/min received a maximum dose of 1,200 mg. If patients experienced intolerable side effects before the maximum dose of 1,200/2,400 mg was reached, they were allowed to stay on a lower dose for the rest of the study period. Patients who did not tolerate a minimum dose of 900 mg were withdrawn from the study. Throughout the study, all patients received their daily dose of study medication in three divided doses.

Anesthesia was conducted according to the standards at our institution. If there was no contraindication, an epidural catheter was placed through the L2–L3 or L3–L4 interspaces immediately before the amputation. Based on the individual patient (i.e. , hemodynamic stability), the anesthetist decided whether the epidural pain treatment should be applied alone (in that case, a test dose of 2% lidocaine with 3 ml epinephrine was administered, followed by 5- to 15-ml bolus doses of 0.5% bupivacaine) or combined with spinal or general anesthesia.

Postoperatively, bupivacaine (0.25%, 3–8 ml/h) was given epidurally. The infusion was stopped 2 or 3 days after the amputation. If the stump pain became severe, the infusion was restarted and continued for another 1 or 2 days. All patients received 1 g paracetamol four times daily. Opioids were used to supplement the epidural pain treatment and for treating the pain after removal of the epidural catheter. Patients who did not receive an epidural catheter had either general or spinal anesthesia for the amputation and paracetamol and opioids for postoperative pain treatment. Nonsteroidal antiinflammatory drugs were not used routinely.

At inclusion in the study, the intensity of the preamputation pain (mean pain intensity during the last week before amputation) was recorded using a numeric rating scale (NRS) with 0 as “no pain” and 10 as “worst possible pain.” The duration of preamputation pain was also recorded (days, weeks, months, years).

During the 1-month treatment period, the intensity of stump and phantom pain was recorded every evening (mean intensity during the past 24 h). Stump pain was defined as pain localized to the region of the stump. Phantom pain was defined as pain experienced in the missing part of the limb. Intensity of pain after 7, 14, 21, and 30 days was calculated as a mean of the previous seven daily pain scores; thus, values for median intensities of pain in the gabapentin and placebo group included decimal values. Five major postoperative interviews were performed after 7, 14, and 30 days and 3 and 6 months. At all five interviews, the stump and phantom pain was described using the McGill Pain Questionnaire.21If phantom pain was present, the pain frequency (constant, daily, or with daily intervals), number and duration of phantom pain attacks on days with pain, and intensity of pain (mild, moderate, severe) were recorded. After 3 and 6 months, patients were asked to record the intensity of stump and phantom pain (mean intensity during the last week before interview, 0–10). After 30 days, both the patient and the examiner were asked whether they thought that the patient had received gabapentin or placebo.

Sensory testing was conducted after 14 and 30 days, most often at the patient’s home. Testing was performed at the lateral aspect of the stump. If amputation was performed at the thigh or knee, the L2–L3 dermatome was examined (above m. vastus lateralis), and if amputation of the calf was performed, the L4 dermatome was examined. Both the amputation stump and the contralateral side were examined. The following measures were performed: allodynia to touch, thresholds to mechanical pressure, and temporal summation (“windup-like pain”). Allodynia was measured by gently stroking the skin with a brush, which was moved from outside the stump area toward the stump. Allodynia was considered to be present if a sensation of slight touch changed to a sensation of tenderness or pain. The pressure pain tolerance threshold was determined by using a handheld pressure algometer (Somedic, Horby, Sweden) with a circular probe area of 1 cm²and a pressure application rate of 150 mmHg/s. When pressure pain tolerance threshold was reached, the patient pressed a push button that immediately froze the digital display. Each value was determined three times. Windup-like pain was assessed by repeatedly tapping the skin with a stiff von Frey hair (665 units = 447 g) at a rate of 3 taps/s. The stimulation was discontinued after 60 s or earlier if it became unbearable. Windup-like pain was considered to be present if the evoked pain exceeded zero on the NRS (0–10).

The primary outcome measures were rates of phantom pain and intensity of stump and phantom pain at the end of the 30-day treatment period (calculated as a mean of the last 7 daily pain scores) and after 6 months. Secondary outcome measures were (1) frequency, (2) duration, and (3) intensity of phantom pain attacks (before unblinding, two of the investigators ranked all answers on phantom pain attacks according to severity on a 0–100 point scale, and thereafter, they reached a consensus about the individual points given); (4) descriptions of pain (McGill Pain Questionnaire); and (5) consumption of opioids at the end of the 30-day treatment period and after 6 months.

The natural history of phantom pain after amputation shows an incidence of approximately 70%,1,2and in most cases, pain is not severe. Because medical treatment may cause side effects, we decided that to be clinically relevant, the treatment should reduce the risk of phantom pain from 70% to 30% or reduce the intensity of stump and phantom pain by two points on the NRS (0–10). Phantom pain was considered to be present if greater than zero on the NRS (0–10). Before the start of the study, we estimated that a sample size of 18 patients in each group would be sufficient to detect such differences. A standard biostatistical formula was used (n = 2 ×ς²/(μ¹−μ²)²f(α, β), based on the premises α= 0.05, β= 0.2, ς²= 2.1).

Results are presented as median with range, 25% and 75% percentiles, or mean ± SD. Results regarding incidence of phantom pain are shown with 95% confidence intervals (CIs) for differences. The Mann–Whitney U test was used to analyze differences between the two treatment groups, the Wilcoxon signed-rank test was used for analysis of paired data, and the chi-square or Fisher exact test was used to analyze differences in dichotomous data.

Only patients who completed 1 week of treatment and tolerated a minimum dose of 900 mg were included in the data analysis. Patients who were withdrawn after 1 week were interviewed after 30 days and 3 and 6 months, if possible.

Forty-six patients were randomly assigned to treatment with gabapentin (n = 23) or placebo (n = 23) during the 3-yr enrolment period. The groups were matched in terms of baseline characteristics, although the gabapentin group tended to include fewer diabetic patients (P = 0.07; table 1). Five patients did not complete the first week of treatment; thus, results from 41 patients were included in data analysis. Another 7 patients were withdrawn from the study before the end of the 30-day treatment period, and 1 patient died after 3 months. No patients were withdrawn because they did not tolerate the minimum dose of 900 mg. Thirty-three patients completed the 6-month trial period: 15 received gabapentin and 18 received placebo (fig. 1).

Eight patients (gabapentin, n = 5; placebo, n = 3) had a creatinine clearance ≥ 30 ml/min but ≤ 60 ml/min and were only allowed to receive a maximum dose of 1,200 mg. The 41 patients who completed the first week of treatment received a median dose of 2,100 mg (range, 900–2,400 mg) gabapentin (n = 21) or a median dose of 8 capsules containing placebo (equivalent to 2,400 mg gabapentin; range, 4–8 capsules, equivalent to 1,200–2,400 mg gabapentin) (n = 20).

Side effects (nausea, stomachache, fatigue, confusion, nightmares, itching, ataxia [all transient but in most cases with a possible relation to study medication]) were reported by 17 patients; 9 received gabapentin and 8 received placebo. Four patients were withdrawn from the study (gabapentin, n = 2; placebo, n = 2), and the dose of study medication was either reduced or temporarily stopped in 7 patients (gabapentin, n = 5; placebo, n = 2) because of side effects.

The incidence of phantom pain among 41 patients who completed at least 1 week of treatment with the study drug was 55.0% versus  52.6% (risk difference, 2.4%; 95% CI, −28.9 to 33.7%; n = 39; P = 0.88; 30 days) and 58.8% versus  50.0% (risk difference, 8.8%; 95% CI, −23.3 to 40.9%; n = 37; P = 0.59; 6 months) (gabapentin vs.  placebo) (fig. 2). Median intensities of phantom pain were 1.5 (range, 0–9.0) versus  1.2 (range, 0–6.6) (n = 33; P = 0.60) and 1.0 (range, 0–6.0) versus  0.5 (range, 0–5.0) (n = 37; P = 0.77), and median intensities of stump pain were 0.85 (range, 0–8.2) versus  1.0 (range, 0–5.4) (n = 33; P = 0.68) and 0 (range, 0–8.0) versus  0 (range, 0–5.0) (n = 37; P = 0.58) after 30 days and 6 months, respectively (gabapentin vs.  placebo) (fig. 3). Rank scores on the severity of phantom pain attacks are shown in figure 4. Median rank scores of phantom pain were 20 (range, 0–90) versus  40 (range, 0–100) (n = 39; P = 0.49) and 10 (range, 0–60) versus  7.5 (range, 0–95) (n = 37; P = 0.56) after 30 days and 6 months, respectively (gabapentin vs.  placebo). There were no differences with regard to rate of phantom pain, intensity of stump and phantom pain, or rank scores of phantom pain between the two treatment groups at any of the other postoperative interviews.

Also, descriptions of pain using the McGill Pain Questionnaire revealed no difference between the gabapentin and the placebo group. After 30 days, the median Number of Words Chosen/Pain Rating Index was 3.5 (range, 0–16)/5.4 (range, 0–66) versus  1.0 (range, 0–12)/1.0 (range, 0–45) (n = 39; P = 0.78/0.76) for phantom pain and 1.0 (range, 0–13)/0.9 (range, 0–54) versus  3.0 (range, 0–10)/4.4 (range, 0–32) (n = 39; P = 0.37/0.29) for stump pain (gabapentin vs.  placebo). After 6 months, the median Number of Words Chosen/Pain Rating Index was 2.5 (range, 0–13)/2.2 (range, 0–35) versus  3.0 (range, 0–14)/6.0 (range, 0–36) (n = 37; P = 0.90/0.93) for phantom pain and 1.0 (range, 0–15)/1.2 (range, 0–40) versus  0 (range, 0–11)/0 (range, 0–26) (n = 37; P = 0.38/0.58) for stump pain (gabapentin vs.  placebo). The consumption of opioids (converted to morphine equivalents22) is shown in table 2; there were no significant between-group differences, but there was a trend for a lower opioid use in the gabapentin group after 30 days.

A subpopulation analysis of patients who did experience phantom pain showed no difference between the two treatment groups. After 1 month, 21:39 (53.8%) patients had phantom pain (11 belonged to the gabapentin group and 10 belonged to the placebo group), and after 6 months, 20:37 (54.1%) had phantom pain (10 belonged to the gabapentin group and 10 belonged to the placebo group). Median intensities of phantom pain were 3.0 (range, 1.0–9.0) versus  2.8 (range, 0.6–6.6) (n = 39; P = 0.99) and 2.0 (range, 1.0–6.0) versus  2.0 (range, 1.0–5.0) (n = 37; P = 0.73) after 30 days and 6 months, respectively (gabapentin vs.  placebo).

Also, a separate analysis of data from the 34 patients who were treated with gabapentin or placebo for 30 days (gabapentin, n = 16; placebo, n = 18) did not reveal any difference between the two treatment groups. The incidence of phantom pain (gabapentin vs.  placebo) was 62.5% versus  55.5% (risk difference, 7%; 95% CI, −25.9 to 37.9%; n = 34; P = 0.68; 30 days) and 60% versus  50.0% (risk difference, 10.0%; 95% CI, −23.9 to 33.9%; n = 33; P = 0.57; 6 months), respectively.

Almost all (37:41) patients who completed 1 week of treatment with gabapentin or placebo had epidural pain treatment for postamputation stump pain (gabapentin, n = 19; placebo, n = 18). The median duration of postoperative treatment was 94.5 h (range, 40–152 h) in the placebo group but only 72.8 h (range, 21–206 h) in the gabapentin group (P = 0.01; fig. 5).

Thirty-five patients underwent sensory examination (gabapentin, n = 18; placebo, n = 17). Six patients declined to participate, either because of pain or because they did not want to remove stump dressings. Four patients reported allodynia (gabapentin, n = 1; placebo, n = 3) and 11 reported windup-like pain (gabapentin, n = 5; placebo, n = 6) after 14 days and/or 30 days. The median pressure pain tolerance thresholds were lower at the stump compared with the contralateral side in almost all amputees after 14 days (1,553 mmHg [range, 488–2,528 mmHg]vs.  2,033 mmHg [range, 878–2,933 mmHg]; P = 0.00) and 30 days (1,440 mmHg [range, 495–3,473 mmHg]vs.  2,130 mmHg [range, 743–4,583 mmHg]; P = 0.00). Pressure pain tolerance thresholds at the stump side were similar in the gabapentin group and the placebo group after 14 days (1,410 mmHg [range, 518–2,340 mmHg]vs.  1,500 mmHg [range, 488–2,528 mmHg]; P = 0.94) and 30 days (1,658 mmHg [range, 698–3,473 mmHg]vs.  1,463 mmHg [range, 495–2,955 mmHg]; P = 0.98).

After 30 days of treatment, 10:39 patients correctly identified the treatment given for the following reasons: gabapentin: effect of treatment (n = 4), side effects (n = 1); placebo: no effect of treatment (n = 5). Six patients gave wrong answers, and 23 did not know.

In 5 cases, the investigator correctly identified the treatment given: gabapentin: side effects (n = 3); placebo: lack of effect (n = 2). In 8 cases, the answers were incorrect, and in 26 cases the treatment could not be identified.

In this prospective, randomized study, gabapentin, administered at a median dose of 2,100 mg for 30 days after amputation, did not reduce stump and phantom pain. Also, gabapentin did not reduce elements of central sensitization. Therefore, our findings suggest that treatment with gabapentin in the early postoperative period has no short-term (30 days) or long-term (6 months) effect on postamputation pain. However, at least three key issues need consideration in this negative clinical trial.

First, the sample size of the study was relatively small, which might explain the negative outcome. Before the start of the study, we decided that only a moderate or large effect of treatment would be clinically relevant. The sample size in the current study was of sufficient power to detect a 40% reduction in the incidence of phantom pain or a two-point reduction of pain intensity on the NRS. Patients who undergo amputation because of peripheral vascular disease are often on multidrug regimens because of coexisting cardiovascular, pulmonal, or endocrine diseases, and further medication should be avoided unless a substantial effect of treatment is documented. Also, although the incidence of phantom pain is high immediately after amputation, the intensity and frequency of pain attacks diminish over time, and only 5–10% of patients continue to have severe pain.

Second, the lack of effect may be related to the slow titration schedule of a low dose of gabapentin. We selected the dose of gabapentin on the basis of other studies showing an effect of similar doses on phantom limb pain20and various neuropathic pain states.23–25The dose of gabapentin was increased gradually to avoid side effects. In a recent randomized, crossover study, Smith et al.  26did not show any effect of gabapentin at a dose of 3,600 mg administered to 24 patients with chronic phantom pain. In the current study, the duration of epidural pain treatment was shorter in the gabapentin group, and consumption of opioids tended to be lower after 30 days, suggesting that the administered dose of gabapentin was sufficient to reduce immediate postoperative pain.

Third, the intensity of postamputation pain was low in both treatment groups, making it more difficult to detect a between-group difference. Almost all patients had epidural analgesia for postoperative pain treatment. Some early studies, many with design flaws such as small sample sizes, no or insufficient randomization, and nonblinded assessment of pain, prompted the use of perioperative epidurals to prevent phantom pain.8–10Thus, epidural analgesia may have reduced pain in the early postoperative period and “washed out” differences between the gabapentin and placebo group. Nikolajsen et al.  11conducted a prospective, randomized, blinded, placebo-controlled study and found no effect of such treatment, so is it not likely that the epidural treatment had any long-term effect on phantom pain. Also, some patients still reported very high intensities of pain, and the incidence of phantom pain observed in the current study was similar to the natural course of phantom pain.3 

A true preemptive study approach is difficult in patients undergoing amputation for peripheral vascular disease because the decision about amputation is usually not made until a few days before or even on the same day as the amputation. For ethical reasons, patients need some time to consider their participation in a clinical trial. Therefore, we chose a pragmatic approach and started the treatment as soon after the amputation as possible.

Despite careful randomization, the gabapentin group tended to include fewer diabetic patients. In a previous study including patients undergoing amputation for peripheral vascular disease, diabetic patients had a lower intensity of preamputation pain.2Preamputation pain increases the risk of phantom pain, and therefore, the observed skewed distribution of diabetic patients in the current study would disfavor a treatment effect of gabapentin. However, the median intensity of preamputation pain among patients with diabetes was 7 (range 1–10), which is similar to the intensity of pain among patients without diabetes. The duration of preamputation pain was longer in the gabapentin group. This may also disfavor treatment effect of gabapentin, but a well-controlled prospective study has shown that phantom pain is related to the intensity but not the duration of preamputation pain.2 

Patients came into surgery with high levels of preoperative pain. Experimental studies have shown that afferent noxious input induces a central sensitization of second-order neurons in the dorsal horn. Such central sensitization has several features, including spontaneous activity, increased response to afferent input, and expansion of peripheral receptive fields (for review, see Woolf and Salter27). It is possible that central sensitization induced by preamputation pain made patients in the current study more resistant to postoperative treatment with gabapentin, and we cannot exclude that a similar treatment approach could have a pain-reducing effect on chronic pain after other types of surgery.

Phantom pain is a complex pain condition. Abnormal sensory phenomena, e.g. , the sensation of voluntary movement of phantom fingers, telescoping (gradual shrinkage of the phantom limb), and changes in phantom sensations induced by the use of an artificial prosthesis, suggest that cerebral structures are also involved.28Actually, a series of studies have shown extensive reorganization of the primary somatosensory cortex and a correlation between the extent of reorganization and the amount of pain.29,30These changes may be unique to the cortex or may be, at least in part, the consequence of alterations at the level of the thalamus, brain stem, or spinal cord. It is possible that such extensive neuronal changes may be difficult to prevent or modulate by a single approach such as treatment with gabapentin.

Despite certain limitations, the current study suggests that gabapentin, started the first day after amputation and continued for 30 days, does not reduce postamputation stump or phantom pain. Therefore, our findings are in agreement with a study by Fassoulaki et al.  31They found that treatment with mexiletine and gabapentin started on the day before breast surgery and continued for 10 days reduced early postoperative pain, but there was no effect on chronic pain after 3 months. In contrast, another study with a similar design found that perioperative administration of venlafaxine reduced postmastectomy pain 6 months after surgery.32Further studies are required, perhaps with multimodal and longer-lasting treatments, to assess their effect on chronic pain after surgery, including postamputation pain.

The authors thank Lotte Larsen, R.N., and Kristina Esbensen, R.N. (Department of Orthopedic Surgery, Aarhus University Hospital, Aarhus, Denmark), for their help in recruiting patients.