Effects of Naloxone Infusions in Patients with the Pruritus of Cholestasis: A Double-Blind, Randomized, Controlled Trial

Methods

The study protocol was approved by the Clinical Research Subpanel of the National Institute of Diabetes and Digestive and Kidney Diseases. All participating patients signed a written informed consent form.

Patients

Twenty-nine patients with pruritus and cholestasis associated with a cholestatic liver disease or advanced chronic hepatocellular disease were studied. Needle biopsy of the liver or endoscopy and retrograde cholangiography were done when clinically indicated. In each case, causes of pruritus other than cholestasis, (notably, other general medical and psychiatric disorders and pruritic skin diseases) were systematically excluded by taking a comprehensive medical history, doing a full physical examination, and obtaining results of laboratory tests and procedures relevant to each patient's medical status. In addition, the results of a careful evaluation by a dermatologist (MLT) had to be negative for a cause of pruritus other than cholestasis. This skin evaluation included examination of the skin under Wood's light when appropriate and photography of skin lesions. In addition, post-trial follow-up of the patients studied for 10 to 32 months did not reveal any cause of pruritus other than cholestasis.

Exclusion criteria for the trial were the presence of another condition known to be complicated by pruritus (for example, pregnancy, azotemia, thyroid dysfunction, iron deficiency, malignant diseases, and neurologic disorders) and the occurrence of a gastrointestinal hemorrhage, hepatic encephalopathy, or ascites during the previous 6 months. Diagnoses and demographic characteristics of the patients are shown in Table 1. The duration of pruritus was life-long for the patient with the Alagille syndrome, intermittent in the patients with benign recurrent intrahepatic cholestasis although present for at least 1 month before study entry (range, 1 to 12 months), and from 1 to 15 years in the other patients. Among the patients studied, the severity of the pruritus was considered intractable, with severe sleep deprivation, in 6 patients; a major hindrance to regular activities in 20; and tolerable in 3. Serum biochemical profiles on each patient were consistent with cholestasis (median alkaline phosphatase level, 9.67 µkat/L [range, 1.3 to 24.4 µkat/L; normal, 0.5 to 2.0 µkat/L]; median total bilirubin level, 22.3 mmol/L [range, 5.1 to 243 mmol/L; normal, 2.0 to 18 mmol/L]). One patient with chronic hepatitis C had a serum alkaline phosphatase level in the normal range at the time of study. That patient also had a serum bilirubin level of 22.2 mmol/L and fasting and postprandial serum bile acid concentrations that were threefold higher than the upper limit of normal in the absence of features of hepatic decompensation.

In none of the patients was the use of conventional medications for the pruritus of cholestasis associated with clinically acceptable relief from pruritus. Seventeen of the patients had been treated with cholestyramine, antihistamines, or phenobarbital in various combinations. Another patient, who could not tolerate cholestyramine, received antihistamines. One patient with primary biliary cirrhosis had been receiving ursodeoxycholic acid for treatment of cholestasis for 2 years with no apparent effect on pruritus. Therapy with ursodeoxycholic acid, which was not classified as an antipruritic medication, was not discontinued. The patient with the Alagille syndrome had been given rifampicin for pruritus. Therapy with this drug had been discontinued several months before her participation in the trial when the patient developed neuropathy.

Study Design

Therapy with all antipruritic medications was discontinued 5 days before the study began. During the trial, each patient received four consecutive 24-hour intravenous infusions in an arm vein, with no intervals between infusions. The nursing staff provided conventional maintenance of the intravenous cannula. Two of the infusions consisted of 5% dextrose/0.45% NaCl (placebo), and two contained naloxone (Narcan; DuPont, Manati, Puerto Rico) (0.2 µg/kg body weight per minute) in 5% dextrose/0.45% NaCl. The total volume infused during 24 hours was 0.5 L. Each naloxone infusion was preceded by an intravenous bolus injection of 0.4 mg of naloxone in 1 mL of normal saline, and each placebo infusion was preceded by an intravenous bolus injection of 1 mL of normal saline. The order of naloxone and placebo infusions was assigned by balanced randomization. The patients, nursing staff, and clinicians were blinded to the content of the infusions and the bolus injections. Only the statistician (DWA) and the designated pharmacy staff knew of the randomization code.

Vital signs were continuously recorded every hour throughout the study by an automated blood pressure monitor (Critikon Corp., Tampa, Florida). Any symptoms or abnormal signs occurring during the infusions, including those compatible with a reaction similar to an opiate withdrawal syndrome, were carefully assessed.

Assessment of Pruritus Perception and Scratching Activity

Patients were asked to record the severity of their pruritus by making a mark with pen or pencil on a visual analog scale every 4 hours while awake. The scale consisted of a 10-cm horizontal line; 0 cm (the beginning of the scale on the left side) represented no itching and 10 cm (the end of the scale on the right side), the worst itching ever. A visual analog score was the number of centimeters (to the nearest millimeter) between 0 cm and the point on the scale at which the patient made a mark [15, 17].

Scratching activity, independent of limb movements, was continuously recorded during the infusions. The monitoring system used consists of a vibration (scratch) transducer (Piezo Film Division, Altochem Corp., Philadelphia, Pennsylvania), an FM transmitter and receiver (Radio Shack [32-1221]), a custom-made signal processor, and a personal computer. The vibration transducer was taped to the middle fingernail of the dominant hand, and the fingernail was not cut during the study. The transducer consists of a rectangular piece of piezoelectric film, 1 cm² in area, 28µmthick and metallized on both sides with silver ink. It is connected by a thin cable to the transmitter box (2 × 7 × 6 cm), which is attached to the same arm by a Velcro cuff. The transducer converts the strain produced by vibrations of the fingernail as it traverses the skin in the act of scratching to an electrical voltage. The electrical signal is transmitted across the room, where it is received, processed, and logged by the personal computer as a scratching activity index in units of counts per unit time (30 seconds). Only signals above a preset threshold and within the frequency band shown by Fourier analysis to be associated with vibrations of the scratching fingernail (50 to 1000 Hz) were recorded [15, 16]. The validity of this index as an accurate quantitative measurement of scratching activity has been confirmed by independent observations and examination of videotapes [16].

Statistical Analysis

In this four-period crossover trial [18], each patient was assigned to receive, without interruption, four 24-hour infusions (two of naloxone and two of placebo), the order of which was chosen by balanced randomization from the six possible arrangements. All of the infusion bags and their contents were identical in appearance. Twenty-three patients received all four infusions, four received three, and two received one naloxone infusion and one placebo infusion. The balance achieved in the infusion assignment was as follows:

Infusion 1: 11 naloxone and 12 placebo

Infusion 2: 12 naloxone and 11 placebo

Infusion 3: 10 naloxone and 13 placebo

Infusion 4: 13 naloxone and 10 placebo

Data collected from each patient were reviewed and, if necessary, edited (DWA) to ensure that only observations made during corresponding diurnal intervals during placebo and naloxone infusions were compared in the analysis.

Whenever possible, the Statistical Analysis System library [19] was used in analyzing data. Analyses of variance [20] were done on data for mean hourly scratching activity and on data for mean daily visual analog score. Residual variation in hourly scratching activity was not normally distributed (P < 0.001; Shapiro-Wilk test) [21]. Accordingly, we used log₁₀ hourly scratching activity values in the analysis. Differences in log responses were transformed back to the original scale of measurement as ratios. Residual variation in visual analog scores tended to be normally distributed (P = 0.34; Shapiro-Wilk test) [21], which enabled us to use untransformed visual analog score data in a separate analysis of variance. The outcomes of the two analyses of variance were similar. The levels of significance of the three principal sources of variability were as follows: between-treatment means (naloxone compared with placebo), P < 0.01 for hourly scratching activity and P = 0.01 for visual analog score; among-patient means, P < 0.001 for both hourly scratching activity and visual analog score; and patient-treatment interaction, P > 0.5 for both hourly scratching activity and visual analog score.

Order effects, patient-order interactions, and drug-order interactions were not significant (P > 0.5). We sought evidence of carryover effects by comparing the outcome of a treatment given (for example, naloxone) when preceded by the alternative (placebo) with the outcome of the treatment when the preceding treatment was the same (that is, naloxone). If two identical sequences (for example, naloxone and placebo followed by naloxone and placebo) were available in a given patient, only the first was used in the analysis.

Because naloxone infusions were associated with highly significant overall effects on both hourly scratching activity and visual analog score, the question arose whether the separate effects of the drug on the two response variables were associated in individual patients. We tested this hypothesis by determining a nonparametric correlation coefficient between the two types of response (differences between placebo and naloxone in log₁₀ hourly scratching activity compared with the corresponding differences in visual analog scores).

Among the 29 patients studied, 11 had complete 96-hour records of scratching activity. For each set of data from these 11 patients, we constructed a periodogram to detect the presence of periodicity in the data. We used the Fisher method [22] to determine the level of significance of the dominant rhythm in the time series.

The content of some of the infusions administered during the study was predicted in a blinded fashion by the recipient (35 infusions) and by one of the investigators (NVB) (46 infusions). The proportion of correct calls made by each patient on whom data were available were averaged and compared to the corresponding averages for the physician by the Yates mean score test [23].

Results

Side Effects of Naloxone Infusions

One patient who received four infusions developed symptoms and signs similar to those of opiate withdrawal during infusions 1 and 3. When the randomization code was broken, these infusions were identified as those containing naloxone. The first reaction was characterized by an intense feeling of anxiety, restlessness, and a diastolic blood pressure of 100 mm Hg; it subsided within 2 to 3 hours after the infusion was stopped. A similar but milder reaction occurred during the subsequent naloxone infusion. In addition, three patients associated some infusions, later identified as those containing naloxone, with ill-defined anxiety. The frequency of symptoms of a nonspecific nature was similar during naloxone infusions (34%) and placebo infusions (24%).

Predictions of the Content of Infusions

The average of the proportions of correct calls was 0.69 of those made by the patients (95% CI, 0.41 to 0.77) and 0.65 for the clinician (CI, 0.50 to 0.81). The difference between these two mean values, 0.06, was not significant (P > 0.2); moreover, neither average significantly exceeded 0.5, the expected value if the calls had been made at random. These findings suggest that the blinding was effective.

Perception of Pruritus

Thirty-eight percent of the patients reported on their perception of itch during one or more infusions, which enabled us to score the change in this perception ( −2,-1, 0, 1, or 2). For each patient, the mean of the scores recorded during placebo infusions was subtracted from the corresponding mean for naloxone infusions. The results of this analysis indicated that the perception of pruritus tended to be lower during naloxone infusions than during placebo infusions (chi-square sign test, 4.0; P < 0.05).

Visual Analog Scores

We averaged visual analog scores recorded by the patient during each infusion and averaged the resulting daily means according to treatment. Treatment means (naloxone compared with placebo) for each patient are plotted in Figure 1 (inset). There was a highly significant correlation between the response while the patient received naloxone and the response while the patient received placebo (Spearman r = 0.70; P < 0.01) [24]. We therefore used a paired-sample analysis and calculated differences between responses during placebo infusion and those during naloxone infusion for the individual patients (Figure 1); these differences ranged from −3.05 to 1.05, with an overall average difference of −0.582(CI, −0.988 to −0.176).This overall difference was highly significant (t = 2.94; P < 0.01). The average visual analog score for patients was 2.24 during naloxone infusions (range, 0.40 to 5.15; n = 28) and 2.86 during placebo infusions (range, 0.40 to 5.8; n = 29). The difference in the mean visual analog score (the score during naloxone administration minus the score during placebo administration) was similar for patients with and those without primary biliary cirrhosis (P = 0.43). No relation between biochemical indices of cholestasis (serum bilirubin and alkaline phosphatase levels) and the visual analog score difference was found (the Spearman rank correlation coefficient and the slope of linear regression analysis were nearly zero).

Scratching Activity

We calculated the daily geometric mean of hourly scratching activity using a method analogous to that used for the visual analog score data. Treatment means for each patient are plotted in Figure 2 (inset); there was a highly significant correlation between the response during naloxone infusion and that during placebo infusion (r = 0.86; P < 0.01). The distribution of each set of means was markedly skewed to the right (note the use of the logarithmic scale in the inset of Figure 2), which led us to assess responses in terms of ratios rather than differences. We did a paired-sample analysis based on the ratio of the hourly scratching activity during naloxone infusions to that during placebo (Figure 2); these ratios varied from 0.15 to 1.25, with an overall mean of 0.727 (CI, 0.612 to 0.842). The overall difference was significantly (P < 0.001) less than 1.0, the expected value if naloxone had no beneficial effect on scratching activity. The geometric mean hourly scratching activity during naloxone infusions was 74.8 (range, 6.5 to 323.9), and the corresponding mean for placebo infusion was 108.9 (range, 41.0 to 354.4). In only five patients were mean values for hourly scratching activity during naloxone infusions higher than the corresponding means during placebo infusions; no clinical or serum biochemical features enabled this subgroup of patients (Figure 2) to be distinguished from the others. The mean ratio of scratching activity during naloxone infusions to scratching activity during placebo infusions was similar for patients with and those without primary biliary cirrhosis (P > 0.50). We found no relation between biochemical indices of cholestasis (serum bilirubin and alkaline phosphatase levels) and hourly scratching activity response (the Spearman rank correlation coefficient and slope of linear regression analysis were nearly zero).

Relation between the Effects of Naloxone on Visual Analog Score and Scratching Activity

During placebo administration, the recorded values of hourly scratching activity and the visual analog score in individual patients seemed to be associated (Figure 3, top) (r = 0.34; P = 0.07). Although during naloxone infusions patients mostly showed improvement in both hourly scratching activity and visual analog score, a subgroup of five patients experienced relatively greater improvement in scratching activity than the other subgroup (Figure 3, bottom). Thus, hourly scratching activity in the subgroup of five patients decreased by 64% compared with only 16% in the other subgroup, whereas the visual analog score decreased equally (63%) in both subgroups. The subgroup of five patients comprised three patients with hepatitis C, one of the two patients with benign recurrent intrahepatic cholestasis, and one patient with primary biliary cirrhosis who had shown a similar response in the previous pilot study [15]. Within each subgroup, the association between visual analog score and hourly scratching activity was highly significant (see the legend of Figure 3).

Carryover Effects

Data relating to possible carryover effects are shown in Table 2. No significant evidence suggested either a naloxone or a placebo carryover effect in either the hourly scratching activity or visual analog score data; however, the visual analog score when patients received naloxone after receiving placebo was suggestively lower than the score when the patients received two consecutive course of naloxone (P = 0.1). The lack of evidence of significant carryover effects, in addition to the known short plasma half-life of naloxone in men (about 90 minutes) and the results of our pilot trial [15], provide justification for not incorporating washout periods in the design of this trial.

Circadian Rhythm of Scratching Activity

Of the 11 patients for whom complete 96-hour data were available after editing criteria had been applied, 7 showed a significant 24-hour rhythm in their scratching activity (Figure 4). Maximum scratching activity usually occurred between 1200 and 1800 h, and the nadir of activity usually occurred during hours of sleep. The presence of a circadian rhythm in scratching activity, as in other physiologic functions [25], suggests that scratching activity is under internal control. The significant downward trend of the data in Figure 4 is consistent with the sequence of treatments administered (in the patient whose activity rhythm is shown in Figure 4, the infusion sequence was placebo, placebo, naloxone, naloxone).

Discussion

Our findings indicate that in patients with the pruritus of cholestasis, intravenous infusions of naloxone are associated with significant reductions in a visual analog score of the perception of pruritus and in scratching activity. Thus, the inference that naloxone ameliorates the pruritus of cholestasis is supported by both subjective and objective criteria. The results imply that further clinical testing of the role of opiate antagonists in the management of the pruritus of cholestasis is warranted.

An adequate evaluation of the effect of naloxone on the pruritus of cholestasis requires not only a double-blind, randomized study design but also the application of an appropriate method for objectively quantitating drug effect. Because pruritus, or itch, is defined as the need to scratch, it is an inherently subjective perception that cannot be directly quantitated. Thus, although generating visual analog scores has the appeal of providing numeric data, such scores remain intrinsically subjective and are influenced by interindividual variations in the ability of patients to integrate the perception of itch over time and to translate this perception into a visuospatial score that is subject to individual differences in its conceptualization [26]. These problems have been overcome by recognizing that scratching activity, which is the behavioral manifestation of pruritus, can be accurately and objectively quantitated independently of limb movements. The scratching activity index we used shows a diurnal variation, which suggests that this index reflects a biological phenomenon. The validity of this index in providing quantitative data on scratching activity has been previously established in clinical trials [15, 16].

Naloxone is an extensively characterized opiate antagonist [6]. The drug's major limitation is that it must be given by frequent injections or by continuous infusion in order to achieve a sustained opiate antagonist effect [6]. The infusion rate of naloxone chosen for our study was based on a previous dose-finding study and experience in a pilot trial [15] and is about four times the dose that has been reported to reverse morphine anesthesia [27]. In the pilot study, we observed no side effects from naloxone infusions [15]. However, in the current study, mild neuropsychiatric disturbances, which occurred in four patients, probably represented naloxone-induced reactions similar to opiate withdrawal that were analogous to the much more florid reactions precipitated in cholestatic patients by the oral administration (5 mg) of the opioid antagonist nalmefene [10]. Differences in the frequency and intensity of this opiate antagonist-induced reaction in cholestatic patients may be related to dose, drug potency, and opioid receptor subtype affinity [11]. The occurrence of this reaction implies that opioid-mediated neurotransmission or neuromodulation is increased in cholestasis.

Previous studies in which subjective criteria were used to assess pruritus have suggested that both naloxone [13, 14] and nalmefene [10] may ameliorate the pruritus of cholestasis. The results of the current study indicate that by each of two criteria, naloxone induced a beneficial effect in patients with the pruritus of cholestasis. Specifically, both the mean hourly scratching activity and mean visual analog score were significantly lower during naloxone infusions than during placebo infusions. The few instances in which hourly scratching activity was greater during naloxone infusions than during placebo infusions probably represent only treatment failures rather than naloxone-mediated increases in scratching activity. Overall, thus, the data indicate that naloxone infusions decrease the perception and the behavioral manifestation (scratching activity) of pruritus in cholestatic patients.

In cholestatic patients, naloxone presumably competes with endogenous opioid ligands for occupancy of opioid receptors and consequently would block the biological effects of the ligands. Because opiate agonists are known to induce pruritus [5-9], amelioration of pruritus in cholestatic patients by naloxone suggests that the pruritus of cholestasis is mediated, at least partially, by opioid agonist ligands. There are several potential reasons why naloxone did not completely abolish scratching activity in our study: The dose may have been insufficient, particularly because endogenous opioids rather than exogenous opiates were being antagonized [28]; a carryover of the chronic scratching habit into periods of naloxone infusion may have occurred [29] (however, no carryover effects were found); and, last, a component of the syndrome may not be mediated by opioid receptors. An opioid agonist-mediated component of the pruritus of cholestasis may be associated with changes in the availability of opioid receptor ligands at opioid receptors or modulation of the status of opioid receptor subtypes [11]. However, the relation of such changes in the opioid system (in the brain) to the accumulation of endogenous opioids in cholestatic plasma [10, 30, 31] or the synthesis of opioid peptides by the cholestatic liver [32] has not yet been defined.

Our findings indicate that an opiate antagonist can ameliorate the manifestations of pruritus in patients with cholestatic liver diseases. However, although administration of naloxone infusions may have a place in the emergency treatment of intractable pruritus associated with cholestasis, the properties of naloxone (such as its opioid receptor subtype specificity, rate of metabolism, opioid receptor antagonist potency, and oral bioavailability) indicate that this drug is not an appropriate opiate antagonist for the long-term management of this clinical problem. Nevertheless, long-term amelioration of the pruritus of cholestasis may be achieved by administering an opiate antagonist that is more slowly metabolized, more potent, and more bioavailable when given orally than naloxone. Nalmefene is an example of such a drug [33, 34], and preliminary experience with its use for this purpose is encouraging [10, 35, 36].