Gabapentin in patients with the pruritus of cholestasis: A double-blind, randomized, placebo-controlled trial


  • This trial has been registered at, and its registration number is 00058890.

  • Potential conflict of interest: Nothing to report.


Pruritus is defined as the second order of nociception, the first being pain; thus, there is a rationale to study gabapentin, a drug that increases the threshold to experience nociception. The aim of this double-blind, randomized, placebo-controlled trial was to study the effect of gabapentin on the perception of pruritus and its behavioral manifestation, scratching, in cholestasis. The participants were 16 women with chronic liver disease and chronic pruritus. Hourly scratching activity (HSA) was continuously recorded for up to 48 hours at baseline and on treatment for at least 4 weeks in an inpatient setting. The perception of pruritus was assessed by interviews and by a visual analog score (VAS) of pruritus recorded every hour while patients were awake. Patients were randomized to the study drug (gabapentin or placebo) at a starting dose of 300 mg orally per day in divided doses to a maximum of 2,400 mg or until relief from pruritus. Gabapentin was associated with an increase in mean HSA, in contrast to the placebo, which was associated with a decrease. The mean VAS decreased significantly among those taking the placebo and in some patients on gabapentin. In conclusion, gabapentin did not provide a significant therapeutic advantage over the placebo; in fact, it was associated with an increase in the perception of pruritus and in HSA in some patients. (HEPATOLOGY 2006;44:1317–1323.)

Evidence suggests that the pruritus of cholestasis is, in part, mediated by neurotransmitters/neuromodulators, including endogenous opioids.1 The idea of altered neurotransmission in the mediation of the pruritus of cholestasis has led to the study of drugs with neuropsychiatric effects for the treatment of this type of pruritus.2 As pruritus is a nociceptive stimulus,3 there is a rationale to study gabapentin, a drug that increases the threshold of nociception.4, 5 Furthermore, gabapentin was reported to be effective in the treatment of brachioradial pruritus.6 These reports were the basis for the present study of the effect of gabapentin on the pruritus of cholestasis.

Patients and Methods

The study was approved by the institutional review board of Columbia University College of Physicians and Surgeons and of the New York Presbyterian Hospital, where it was conducted. All patients signed a consent form prior to entry into the study. A physician with expertise in liver disease was the study monitor. The recruitment rate was less than anticipated; thus, after a 3-year period we stopped recruiting for the study.


Fifty-six patients with chronic liver disease and chronic pruritus were evaluated as potential candidates for the study. Twenty-six patients were found to be eligible. Of these, 9 did not wish to participate after the index visit or follow-up visits, and one did not return for further evaluation. Thus, 16 patients were enrolled in the study (Fig. 1).

Figure 1.

Potential candidates for participation in the study, reasons for exclusion from randomization, and subjects randomized.

The study group consisted of 16 women whose median age was 49 years (range 44-63 years). Nine patients (patients 1, 2, 4, 6, 9, 10, 12, 15, and 16) had primary biliary cirrhosis (PBC), one patient (patient 3) had primary sclerosing cholangitis (PSC), and 6 patients (patients 5, 7, 8, 11, 13, and 14) had chronic liver disease secondary to infection with the hepatitis C virus. The duration of the pruritus was 1-12 years, except in patient 11, for whom it was 4 months. All the patients reported it interfered with their quality of life. All patients had a serum hepatic panel consistent with cholestasis, with a median serum alkaline phosphatase activity of 397 U/L (range 102-853 U/L, normal 33-96 U/L).

The drugs used to treat the pruritus, including cholestyramine, antihistamines, naltrexone, rifampicin, sertraline, and ondansetron, alone or in combination, had not provided satisfactory relief. Two patients who had chronic hepatitis C (patients 5 and 14) were taking ursodeoxycholic acid without any relief of pruritus.

Study Design.

The study was a double-blind, randomized, placebo-controlled trial with a treatment phase of 4 weeks duration. The trial was designed to study patients aged 18-80 years with chronic pruritus secondary to liver disease. The exclusion criteria were history of hepatic encephalopathy, ascites, history of variceal bleeding, malignancy, inability to practice contraception, pregnancy, creatinine greater than 1.7 mg/dL, hemoglobin less than 10 g/dL, status of post–liver transplantation, and infection with the human immunodeficiency virus.

At baseline, the principal investigator took a complete history and performed a physical examination, and laboratory tests including a complete blood count, coagulation and comprehensive metabolic panels, and, when appropriate, a pregnancy test were obtained.

To determine if depression existed at baseline, a psychiatrist (I.H.G.) evaluated the patients prior to their receiving the study drug, using the Hamilton Depression Rating Scale, a measure of the severity of current depression,7 and the Structured Clinical Interview Questionnaire (SCID) for DSM IV, Axis I Disorders, an interview measure for the diagnosis of depression and anxiety syndromes.8 A dermatologist (D.E.) examined the patients to exclude dermatological causes of the pruritus.

Drugs prescribed for the treatment of pruritus were discontinued 5 days before the collection of baseline data and were not restarted during the study, as was done in previous studies.9–12 Patients who took antihistamines to sleep were kept on those drugs.

The data were collected in the inpatient and outpatient units of the General Clinical Research Center (GCRC) of New York Presbyterian Hospital. Each patient had 2 hospitalizations, at baseline and while on treatment. The patients were contacted by telephone by the principal investigator at weeks 1 and 3 after randomization to assess how they were tolerating the drug and to adjust the dose, and they were evaluated in person at week 2 with an exam and blood work.

Evaluation of the Perception of Pruritus and Scratching Activity.

The perception of pruritus was assessed every hour while the patients were awake with a 10-cm visual analog scale from which a visual analog score was derived, as previously described.9–12

A scratching activity monitoring system specifically designed to record scratching behavior independent of gross body movement13 and successfully utilized in several previous clinical trials was used.9–12 Continuous recording of scratching activity was performed over one or two 24-hour periods (depending on the patient's availability), beginning on the morning of day 2 of admission and ending on the morning of day 4 during both the baseline and on-treatment hospitalizations.

After the baseline HSA and VASs were collected, subjects were randomized to receive the placebo or the active drug according to a randomization code generated and kept at the research pharmacy. After at least 4 weeks on the study drug subjects were admitted to the GCRC for data collection, after which the code was broken. On completion of the treatment period, patients who had been randomized to the placebo were offered gabapentin for 4 weeks to provide them with possible relief; patients who had been randomized to gabapentin and who reported relief received a 1-week supply of the drug, to be continued by the referring physician, as per protocol.

Each patient documented on a form the number of capsules taken and any effects associated with their intake. The forms were collected and reviewed on follow-up appointments.

Study Drug.

Gabapentin was obtained from the manufacturer (Parke Davis Pharmaceuticals, Ltd., Vega Baja, Puerto Rico) and packaged at the research pharmacy in 100- and 300-mg maroon capsules, like those of the placebo. Under code, the study drug was started at 100 mg by mouth 3 times a day for 3 days, to be increased, if necessary and in the absence of side effects, by 300 mg every 3 days to a maximum of 2,400 mg daily in divided doses. Patients 1-8 received a starting dose of the study medication of 300 mg once a day; however, because of reported vomiting by 2 patients, one of whom was a study participant (patient 8), after a dose of 300 mg of gabapentin, the protocol was amended to a starting dose of 100 mg, as described previously.

Statistical Methods.

Exploratory descriptive statistics and graphics were analyzed according to whether patients were randomized to gabapentin (i.e., the gabapentin group) or to placebo (i.e., the placebo group), with those in each group further assessed both pre- and on-treatment, making 4 categories. Mean differences in measurements were determined by subtracting each on-treatment value from each pretreatment value for each subject and calculating the mean of all the differences. A positive mean difference indicates that, on average, the pretreatment mean HSA and the pretreatment mean VAS were greater than the corresponding on-treatment values. Paired t tests were computed for each subject. One-sided P values were calculated because we hypothesized a decrease in HSA, particularly for subjects taking gabapentin. For the HSA the significance level of each test was Bonferroni-adjusted at 0.05/11 = 0.004. Correlations between pre- and on-treatment HSA and VAS scores for the same patient were examined using a standard correlation test; a correlation coefficient of zero indicated changes in the values of one variable did not affect the values of the other variable.

The data contained imputed values as replacements for missing observations; a nearest-neighbor hot deck imputation method was used to estimate these values.14, 15 In hot deck imputation, missing data are replaced by values taken from complete records of subjects with characteristics similar to the subject for whom data are missing.14, 15 Two mixed-effects models, one with HSA as the dependent variable and one with VAS as the dependent variable, were run on the imputed data using the PROC MIXED procedure of SAS V8.2 (SAS Institute, Cary, NC). These results were calculated using data collected in the first 24 hours, as these were available from most subjects. The fixed effects were time of measurement, treatment group (i.e., patients randomized to gabapentin or to placebo), treatment phase (i.e., pretreatment or on-treatment), and an interaction term of treatment group and treatment phase. Subjects were included as nested random effects because they represented a random sample from a population of patients with chronic pruritus secondary to liver disease, and the effects due to the subjects per se were dependent on the effect due to the treatment group. The data were log-transformed to stabilize the variance among groups and to make the distribution of the HSA measurement normal. In the mixed-effects model differences were considered significant at P less than .05.


Dermatological Evaluation.

None of the patients had a dermatological disease associated with pruritus.

Psychiatric Evaluation.

Data on the full psychiatric evaluation were available for 13 of the 16 patients. One patient (patient 11) did not have an evaluation performed because she was moving to another state, and the admission could not be coordinated with the psychiatric exam, 1 patient (patient 5) dropped out of the study prior to the exam, and the exam of another patient (patient 4) was incomplete and not helpful, as only data from the SCID were available.

The results of the full 17-item Hamilton scale, which includes items intrinsic to medical conditions (i.e. fatigue, sleep, and concern about health), showed that 8 patients (patients 2, 3, 7, 10, 12, 13, 15, and 16 [62%]) scored in the range of mild depression, 3 (patients 1, 9, and 14 [23%]) in the range of moderate depression, and 2 (patients 6 and 8 [15%]) in the range of no to minimal depression (0-7); however, the scores of the latter group were 6-7, which might be considered minimal rather than absent depression. When the items relating to medical conditions were omitted, 7 patients (patient 1, 2, 7, 9, 12, 13, and 14 [54%]) scored in the range of mild depression, none in the range of moderate depression, and 6 (patients 3, 6, 8, 10, 15, and 16 [46%]) in the range of no to minimal depression. Four subjects remained in the mildly depressed range despite receiving antidepressant treatment (patient 9 received doxepin; patient 12, Lexapro; patient 13, Lamictal, Seroquel, and Ambien; and patient 14, Celexa), whereas one subject (patient 16) who went from moderate depression to none was on one antidepressant (Zoloft) and previously had been on another one (Celexa), suggesting she was considered to be depressed at one point.

On the SCID, one patient (patient 14) had major depressive disorder with atypical features. The remainder of the patients were diagnosed with a mood disorder because of a general medical condition, 8 with depressive features (patients 2, 3, 6, 8, 10, 12, 15, and 16), and 4 with major depression-like episodes (patients 1, 7, 9, and 13).

The results of the psychiatric evaluations suggested that liver disease and pruritus might have contributed to the depressive symptomatology of the subjects.

Side Effects Associated with Gabapentin and Placebo.

On gabapentin, patient 1 experienced a rise in her serum bilirubin at the end of the study period, interpreted as resulting from progression of her disease (PBC), which was confirmed by histology after she had a liver transplantation several months later. On gabapentin, patient 2 experienced fatigue, patient 4 dizziness, patient 7 worsening symptoms of carpal tunnel syndrome, patient 8 vomiting, and patient 16 dizziness on increasing dose and a fluctuating rise in serum creatinine. Gabapentin was discontinued in these patients. On placebo, patient 3 developed fatigue and leukopenia, and patient 7 developed symptoms of carpal tunnel syndrome.

Effect of Study Drug on Pruritus of Cholestasis.

The results and the subjects' oral reports of their perception of pruritus after 4 weeks of treatment are given in Table 1. Patients 4 and 7, who took gabapentin after completing treatment with placebo, did not continue on gabapentin, and patients 3, 11, 10, and 15 remained on gabapentin; patient 3 underwent liver transplantation several months later because of progression of her disease (PSC). Patient 12, who had stage IV PBC, underwent liver transplantation, as hepatocellular carcinoma was suspected from screening.

Table 1. Effect of Gabapentin or Placebo on the Perception of Pruritus as Reported Orally and on HSA and VAS
Patient No.Oral ReportHSA Mean DifferenceVAS Mean Difference
  • Abbreviations: HAS, hourly scratching activity; VAS, visual analog score.

  • *

    Based on 10 measurements. All others based on 22-24 measurements.

  • P values in parentheses; statistical significance defined as P = .004; a negative number indicates that the treatment measurement was higher (i.e., worse) than that at baseline.

Patients randomized to gabapentin   
1No relief−15.26 (0.361)−0.618 (0.002)
14No relief−122.15 (0.002)−0.629 (0.033)
16Initial relief/no relief in the last days of treatment8.2 (0.382)*−0.844 (0.005)
2Relief at times75.29 (0.027)1.132 (0.077)
6Some relief209.42 (0.019)2.461 (0.000)
10Relief35.46 (0.011)0.628 (0.098)
12ReliefTechnical failure (pre)−1.745 (0.012)
Patients randomized to placebo   
3Relief most of the time33.22 (0.000)3.202 (0.000)
4Relief42.53 (0.025)5.282 (0.000)
7No relief initially/relief in the last days of treatment47.10 (0.000)2.778 (0.001)
11Relief14.05 (0.085)−0.285 (0.717)
15Relief11.69 (0.183)1.683 (0.034)
13No reliefTechnical failure (pre)1.229 (0.013)

Three of the 16 patients dropped out of the study—one (patient 5) prior to randomization because she did not wish to remain hospitalized for baseline evaluation, one (patient 8) on day 2 because of side effects, and one (patient 9), who had been randomized to placebo, because of persistent severe pruritus within 2 weeks of randomization.

All the patients who completed the 4-week study under code went up to the dose of 2,400 mg per day except patients 15 and 16, who took 1,200 and 1,500 mg daily, respectively.

Results of Statistical Analysis.

The graphs for baseline HSA (Fig. 2A-B) show that scratching activity of the patients randomized to gabapentin (Fig. 2A) was higher than that of the patients randomized to placebo (Fig. 2B). The scratching activity of the patients on gabapentin did not decrease (Fig. 3A); in contrast, the scratching activity of the patients on placebo was near zero (Fig. 3B). The group randomized to gabapentin, however, seemed to have higher HSA at baseline than the group randomized to placebo, which might have given the placebo treatment some advantage. These observations are supported by the results of the statistical tests reported in Table 1.

Figure 2.

Mean baseline hourly scratching activity (HSA) in patients subsequently randomized to gabapentin (A) and to placebo (B).

Figure 3.

Mean hourly scratching activity (HSA) in patients randomized to gabapentin (A) and to placebo (B) after at least 4 weeks on treatment.

Correlations between HSA and VAS scores at baseline and on treatment varied from −0.23 to 0.97, suggesting the VAS correlated poorly with HSA. In addition, there was no relationship between baseline and on-treatment HSA or between baseline and on-treatment VAS, as correlations ranged from −0.3 to 0.7 for both variables.

Analysis of the imputed log-transformed HSA data showed the treatment group and the interaction of the treatment group with the treatment phase were significant (P = .013 and .026, respectively); however, the effects of the subjects and of the treatment phase per se were not significant (P = .2 and .16). This implies a detectable difference in HSA measurements between those subjects on gabapentin and those on placebo without consideration of the phase (i.e., baseline or on treatment) at which the measurement was taken. This result may reflect the difference between the treatment groups in the baseline measurements of the subjects (Fig. 2A-B); however, the difference was also affected by the treatment phase (i.e., pre- or on-treatment) during which the measurement was taken. Table 2 gives the means with standard deviations of the non-transformed HSA for the 4 groups: baseline prior to randomization to gabapentin and on gabapentin, and baseline prior to randomization to placebo and on placebo; these data helped to determine the direction of the effect. It can be appreciated that the significance of the interaction effect came from 2 sources: a large increase in the mean HSA associated with treatment with gabapentin and a large decrease in the mean HSA associated with treatment with placebo.

Table 2. Nontransformed Mean Hourly Scratching Activity* by Treatment Phase
GroupHourly Scratching Activity [mean (SD)]
  • *

    Hourly scratching activity is expressed in arbitrary units.

Pre gabapentin52.94 (89.35)
On gabapentin88.50 (125.09)
Pre placebo57.22 (99.52)
On placebo15.34 (41.23)


Our findings indicate that gabapentin does not have a therapeutic advantage over placebo for patients with the pruritus of cholestasis, some of whom were refractory to drugs used to treat this symptom; in fact, this drug appears to worsen pruritus. In contrast, treatment with placebo was associated with relief from pruritus for most patients by all measures used to assess this symptom.

The correlations between baseline HSA and VAS were low, suggesting that the VAS of the pruritus of cholestasis is an unreliable predictor of HSA, in general, as was found in a previous study.9 We also found that the time at which HSA and VAS was recorded influenced the results; accordingly, a VAS of pruritus obtained at random may not be interpretable.

The most stunning results of this study can be visually appreciated from Figs. 2 and 3. It is clear that patients on placebo (Fig. 3B) scratched less than those on gabapentin (Fig. 3A), indicating that the placebo effect resulted not only in subjective improvement, but also in a change in behavior. In contrast, some patients on gabapentin reported no relief of pruritus and scratched more and had a higher mean VAS, some significantly so, than at baseline. A question arises: did gabapentin interfere with the placebo effect? In trying to address this question, current ideas on the placebo response should be discussed.

The placebo effect results from at least 2 types of changes, subjective and physiological (i.e., changes in neurotransmission).16 In this context, studies that used positron emission tomography demonstrated that the mechanism mediating the placebo response in Parkinson's disease is associated with a release of dopamine in response to the activation of the nigostriatal dopamine system by the placebo intervention.17 The association of a placebo intervention with dopamine release raises another question: can dopamine exert an antipruritic effect? The answer is yes.

Evidence suggests that the pruritus of cholestasis is mediated, at least in part, by increased central opioidergic tone, which is the reason this type of pruritus is treated with opiate antagonists.9–12, 18–21 It has been shown that the pruritus resulting from the central administration of morphine (i.e., pruritus secondary to increased opioidergic tone), which may be analogous to the pruritus of cholestasis, can be relieved effectively by droperidol,22, 23 a short-acting neuroleptic agent with affinity for the D2 dopamine receptor.24 Initial treatment with neuroleptics, which may be analogous to the acute effect of droperidol administration, is associated with activation of dopamine neurons and enhanced dopamine release.24 Accordingly, if droperidol decreases central, morphine-induced pruritus, endogenous dopamine release may decrease central, opioid-mediated pruritus (i.e., as it is hypothesized to occur in cholestasis). Thus, activation of the nigrostriatal system (i.e., dopamine release) by a placebo intervention may mediate a decrease in pruritus by signals transmitted via the neural networks that connect the basal ganglia, which release dopamine, with the cerebral cortex,25, 26 where itch stimuli and the urge to scratch are represented.27–29

Having considered the role of dopamine in the placebo effect, it does appear that gabapentin could have interfered with the placebo effect. Animal studies30, 31 have documented that gabapentin blocks dopamine release from the basal ganglia (i.e., the striatum) and from the nucleus accumbens, preventing dopamine-mediated behavior.31 Accordingly, it seems plausible that gabapentin, by virtue of its ability to block dopamine release, may have interfered with dopamine-mediated relief of pruritus triggered by a placebo intervention. It is evident that in this study gabapentin was not a placebo intervention; however, it is the expectation of relief and, likely, the degree of that expectation that appear to mediate, at least in part, the release of dopamine that occurs with a placebo intervention.17, 32 Thus, patients with pruritus expected relief, but this expectation, which mediates the placebo effect and would have resulted in dopamine release and hence an antipruritic effect, was blocked by gabapentin. This is our hypothesis. The dichotomous effect of gabapentin in this study can be explained by the differences among the patients in the threshold for activation of neurotransmission and in the degree of expectation.17, 32 A high expectation of relief may have resulted in a strong stimulus for dopamine release, which may not have been completely blocked by gabapentin at the doses used in this study, allowing enough dopamine to mediate neurotransmission with the result of some relief from the pruritus. In contrast, a low expectation of relief may have resulted in a weak stimulus for dopamine release, which could have been blocked by the gabapentin at the doses used, preventing dopamine-mediated relief of pruritus. This interpretation suggests that patients' expectation is an important question to address in studies of pruritus; unfortunately, information on the expectations of relief by the patients in our study was not sought a priori.

The small number of patients was a limitation of this study; however, the strong placebo effect observed has focused our attention on the antipruritic effects of placebo in the pruritus of cholestasis and emphasizes that any intervention on pruritus may be difficult to interpret without some objective measure of change. As documented by the psychiatric evaluation, the pruritus of cholestasis has a negative impact on the emotional life of patients, and it can be an indication for liver transplantation even in the absence of liver failure.33, 34 These facts stress the importance of exploring why this symptom continues to escape understanding, contributing to the lack of consistently effective and specific treatments.

In summary, in this study gabapentin was not associated with a significant amelioration of the pruritus secondary to liver disease; in fact, on average, it worsened the pruritus and the scratching behavior. Several scientific questions have emerged from this experience, including the role of dopamine neurotransmission in the treatment of pruritus and on the neurophysiologic basis for the placebo effect on the pruritus of cholestasis.


The authors thank Glen Gillen, MPA, OTR, Sabrina Salvant, MS, MPH, Jane Bear-Lehman, PhD, OTR, FAOTA, Jacqueline Brown, MS, OTR/L, CHT, and Batsheva Meisels, OTR/L, CHT, from Occupational Therapy for making the orthotic devices, the staff of the Research Pharmacy and the GCRC for facilitating the conduct of the study, Erica Torres for her administrative work, and Howard J. Worman, MD, for monitoring the study. Dr. Bergasa is grateful to Thomas Talbot, Engineer, for his continued support and advice on the use of the scratching activity monitoring system.