A randomized, controlled crossover trial of ondansetron in patients with primary biliary cirrhosis and fatigue

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Fatigue is common in primary biliary cirrhosis (PBC). Altered central serotonergic neurotransmission may be involved in its pathogenesis. This multicenter, randomized, double-blind, placebo-controlled, crossover trial evaluated the efficacy of ondansetron, a selective 5-HT3 receptor subtype antagonist, for treating fatigue in PBC. A crossover design was chosen, allowing subjects to serve as their own controls—appropriate to evaluate fatigue, a subjective symptom. Sixty patients with clinically stable PBC, a Fatigue Severity Score (FSS) > 4, and no other identifiable cause for fatigue were enrolled. Subjects were randomized to receive ondansetron (4 mg) or placebo orally 3 times daily for 4 weeks (period 1). Subjects then crossed over, after a minimum 1-week washout period, for a further 4 weeks of ondansetron or placebo (period 2). Fatigue was measured at the beginning and end of each period by using the FSS and Fatigue Impact Scale (FIS). Six patients withdrew; the remaining 54 subjects had a mean baseline FSS of 5.55 (±0.1). Response to study medication in period 1 versus period 2 was not uniform; thus, it was necessary to analyze the trial periods separately. In period 1, there was no significant additional fatigue reduction on ondansetron over placebo. During period 2, FSS and FIS decreased significantly on ondansetron versus placebo (P = .001). However, period 2 results were invalidated because drug side effects unblinded subjects (constipation affected 63.0% of patients taking ondansetron, versus 13.3% on placebo). In conclusion, ondansetron administration did not confer clinically significant fatigue reduction when compared with placebo in our study population. (HEPATOLOGY 2005;41:1305–1312.)

Fatigue is common in primary biliary cirrhosis (PBC), affecting between 55% and 85% of patients.1–3 It diminishes quality of life and interferes with many activities of daily living.2, 4 No known treatment exists for fatigue in PBC. Previously investigated drugs, including ursodeoxycholic acid,5 cyclosporine,6 thalidomide,7 and antioxidants,8 were not effective in ameliorating this debilitating symptom.

The pathogenesis of fatigue in PBC is unclear. Fatigue in patients with PBC may be centrally mediated and not peripheral in origin.9 The mechanisms proposed for this centrally mediated fatigue include abnormal neuroendocrine function and altered serotonergic neurotransmission. These mechanisms have been demonstrated in a bile duct–ligated rat model simulating acute cholestasis10, 11 as well as in humans.12, 13 In the rat model, repeated administration of a serotonin-1a receptor agonist relieved fatigue as measured by activity scores during a swim tank test.11 In humans, administration of the serotonin reuptake inhibitor paroxetine resulted in decreased exercise endurance time in recreationally active young males.13

Questionnaire studies conducted in individuals with PBC have consistently found an association of fatigue with depression.1–3 Abnormalities in central serotonin transmission are thought to be key to the pathogenesis of depression.14, 15 Serotonin activates central corticotropin-releasing hormone–containing neurons, and these alterations in serotonin transmission could affect corticotropin releasing hormone levels, which in turn could contribute to a sense of fatigue.16

A case report described relief from the fatigue associated with chronic liver disease by long-term administration of ondansetron.17 Ondansetron is a serotonin receptor antagonist that is selective for the third serotonin receptor subtype. It is primarily metabolized in the liver, and although the clearance of ondansetron is reduced and the plasma half-life is increased in patients with severe hepatic insufficiency, no drug-related adverse events were observed in this patient who had chronic hepatitis C.17

In view of the aforementioned data on modulation of central serotonin neurotransmission to ameliorate fatigue in cholestasis, we hypothesized that ondansetron would be an effective treatment for fatigue in PBC. As such, we sought to study the efficacy and safety of orally administered ondansetron for this purpose.

Abbreviations

PBC, primary biliary cirrhosis; AMA, antimitochondrial antibody; FSS, Fatigue Severity Score; FIS, Fatigue Impact Scale; PAB, Performance Assessment Battery; HRSD, Hamilton Rating Scale for Depression; PSQI, Pittsburgh Sleep Quality Index.

Patients and Methods

Inclusion Criteria.

To be eligible for the study, patients were required to be between the ages of 18 and 70 years. A diagnosis of PBC was required, defined as elevation in serum alkaline phosphatase with antimitochondrial antibody (AMA) positivity or compatible liver histology. Each subject was required to score ≥4 on the Fatigue Severity Score survey (FSS), a scoring system that has been externally validated for use in patients with PBC.1, 18 This FSS cutoff score was chosen as a value that would correlate well with verbally reported significant fatigue in PBC patients.1 The FSS survey was administered to each subject at a screening clinic visit and repeated on a second visit at least 2 weeks later, to verify that the fatigue severity was stable before enrollment in the study.

Exclusion criteria were any concomitant medical condition that would cause fatigue, such as untreated hypothyroidism, renal failure, anemia (hemoglobin < 90 g/L), or excess alcohol consumption (more than 14 and 7 standard drinks per week for men and women, respectively). Medications that could cause fatigue (antidepressants, sedatives, antihistamines) or interact with ondansetron metabolism (cytochrome P450 3A, 1A2 or 2D6 inhibitors or inducers) were not permitted. The use of beta blockers was permitted, but patients were stratified on the basis of beta blocker use. Pregnancy, breast-feeding, or intention to conceive were grounds for exclusion. Patients with signs of hepatic decompensation in the past 6 months, including gastrointestinal hemorrhage, hepatic encephalopathy, ascites, International Normalized Ratio > 1.3, or serum albumin < 3.5 g/L were excluded.

Trial Design.

The study was conducted in a randomized, double-blind, and placebo-controlled manner, in a crossover design. With this design, patients acted as their own controls because they received active medication in one trial period, and crossed over to receive placebo in the other trial period. The study was conducted in three Canadian centers (Toronto, Montreal, Calgary), and received approval by the local human subject review board at each site. Written informed consent was obtained from each subject before enrollment in the trial.

Each of the 2 trial periods was 4 weeks in duration, with a minimum of 1 week washout between periods. Randomization was conducted using a series of coded envelopes, made from a computer-generated random list. By selection of the next available envelope, each patient was randomized at enrollment to receive either placebo in period 1 and ondansetron in period 2 (group I), or vice versa (group II).

Ondansetron was administered in single 4-mg oral disintegrating tablets, 3 times daily. This reduced dose of ondansetron was selected to offset potential decreased hepatic drug clearance in our study population with mild to moderate liver disease. Subjects taking placebo medication received single orally disintegrating tablets identical in appearance to the ondansetron medication, 3 times daily. Patients and investigators were blinded to treatment, and the randomization code was not broken until the study was complete.

Outcome Measurements.

The primary measure of outcome was reduction in fatigue as quantified by the FSS. Fatigue severity was also measured using the Fatigue Impact Scale (FIS),19 which has cognitive, social, and physical dimensions. The FIS has been previously validated for use in patient populations with PBC2, 3, 20 and has been shown to have good reproducibility for fatigue measurement.19

The Walter-Reed Performance Assessment Battery (PAB)21 was used to measure fatigue on a more objective basis. The PAB has been used for fatigue quantification in studies in which fatigue is induced by means such as sleep deprivation, fragmentation, or disruption of circadian rhythms.22 It is administered by a computer program that records measures of sleepiness (Stanford Sleepiness Scale) as well as response speed and accuracy in 5 different cognitive tests requiring mental manipulation and attention (choice reaction time, logical reasoning, serial addition and subtraction, spatial orientation [Mannikin], and conflict resolution [Stroop]). The Stanford Sleepiness Scale quantifies a patient's current subjective feeling of sleepiness on a 7-point scale (higher score means more sleepiness). Neurocognitive performance on the 5 cognitive tests is quantified by the number of correct question responses per minute. To obtain comparable and reliable results on the Walter-Reed PAB, subjects must have prior understanding of how the cognitive tests are performed. Accordingly, patients were permitted one instructional session to practice the cognitive tasks in the PAB before test scores during the trial were recorded.

Because severity of fatigue can be affected by depression or sleep quality, during the trial we measured depression using the Hamilton Rating Scale for Depression (HRSD)23 and sleep quality using the Pittsburgh Sleep Quality Index (PSQI).24 Both of these measures have been previously used in the study of patients with PBC.1

All outcomes were measured at the beginning and end of each trial period. Self-reported compliance (number of pills taken per day) was recorded in a daily diary. Compliance data were corroborated by collecting unused tablets at the end of each study period.

Safety.

A physical examination and laboratory tests were conducted at screening, as well as at the beginning and end of each trial period. In addition, patient interviews were conducted and relevant laboratory tests were performed mid-way through each trial period. Patients were contacted by telephone during weeks they did not visit the study clinic, to evaluate compliance and side effects. Minor and major adverse events were recorded, and medication dose was adjusted if necessary. Co-interventions resulting from new or discontinued medicines were monitored at each study clinic visit. Initiation on any contraindicated medication was cause for withdrawal from the study.

Statistical Analysis.

Sample size was determined at 60 subjects, using power calculation based on 30% expected reduction in FSS on ondansetron versus placebo, with a type I error <5%, power of 80%, and attrition rate of no more than 20%.

Study data were entered into a Microsoft Excel database and transferred to PC SAS (SAS Institute, Cary, NC, release 8.02) for further analyses. For both group I (placebo followed by ondansetron) and group II (ondansetron followed by placebo), means and their corresponding SEM were computed to describe the distributions of demographic, clinical, and patient characteristics at the initial screening visit. Distributions of categorical demographic variables were described for each group, using percentages. The extent of random imbalances in comparisons of the 2 groups at screening was described with P values according to the chi-square test for dichotomous demographic variables, and t tests or Wilcoxon rank sum tests for all the other variables.

Means and SEM were also used in computing baseline distributions of the corresponding variables at the beginning of each treatment period, as well as distributions of the corresponding changes from baseline for each treatment period.

The primary method for comparing the effects of placebo versus ondansetron for the response variables (FSS, FIS, PSQI, HRSD, PAB) was analysis of covariance for the change during period 1 (i.e., baseline vs. end of period 1), using baseline values as covariates (covariables). There were also 2 types of major analyses: one applied analysis of covariance to changes during period 2, with the corresponding values at period 1 baseline and period 2 baseline as the covariables; the other applied repeated-measures analysis of covariance to changes during both periods done through generalized estimating equation where appropriate.

The prevalence of study drug–related side effects such as constipation and headache, as well as the use of medication or study drug dose reduction for relief of such side effects were compared for the 2 treatments by an extension of the McNemar test.

Enrollment.

Recruitment was conducted between September 2000 and January 2002. Of the 253 patients screened, 193 were ineligible based on exclusion criteria (Fig. 1). Sixty patients from the 3 centers were enrolled and randomized (28 to group I, and 32 to group II). Six patients withdrew or were removed from the study—2 because of depression, 2 for intractable or intolerable constipation, 1 patient who refused to complete the required number of visits, and 1 patient who after randomization was found to be taking amitriptyline. A total of 54 patients remained and were included in the final data analysis. The analysis was not on an intention-to-treat basis, because there were only 2 time points (baseline and end) for each trial period.

Figure 1.

Study recruitment, enrollment, and randomization.

Demographics.

At randomization, patients in group I and group II were similar in age, sex, weight and height, number of months since diagnosis of PBC, Mayo risk score, serum alkaline phosphatase, total bilirubin, INR, AMA positivity, and screening FSS/FIS scores (Table 1). The mean FSS at screening was 5.55 (±0.1).

Table 1. Patient Demographics and Characteristics at Screening in Two Groups of Patients Randomized to Receive Placebo or Ondansetron First
VariableGroup I (n = 28)Group II (n = 32)P
  1. NOTE. Except where otherwise indicated, the values are mean ± SEM.

  2. Abbreviations: PBC, primary biliary cirrhosis; AMA, antimitochondrial antibody; ALP, alkaline phosphatase.

Demographic measures   
 Age, years52.4 ± 2.354.7 ± 2.1.467
 Female, %92.590.729
 Weight, kg72.9 ± 3.776.7 ± 3.7.485
 Height, cm160.1 ± 1.6165.4 ± 1.6.415
 Duration of PBC, months51.1 ± 8.354.3 ± 8.6.883
 Mayo Risk score4.4 ± 0.24.3 ± 0.1.525
 Patients taking beta blocker3 (10.7%)3 (9.4%).863
Laboratory investigations   
 AMA positivity85.2%89.3%.526
  (IF or ELISA), %   
 ALP (U/L) (N ≤ 110)208.6 ± 30.8149.3 ± 13.5.209
 Bilirubin, total (μmol/L, N ≤ 22)21.7 ± 3.414.6 ± 1.5.094
 INR1.01 ± 0.021.00 ± 0.01.904
Fatigue measurements   
 Fatigue Severity Score (FSS)5.6 ± 0.15.5 ± 0.1.626
 Fatigue Impact Scale (FIS)   
  Cognitive dimension21.6 ± 1.921.2 ± 2.07.935
  Physical dimension22.6 ± 1.622.4 ± 1.5.964
  Social dimension35.4 ± 3.136.3 ± 3.5.796

Results

Variance in FSS between administrations of the questionnaire before taking any medication, that is, between screening and baseline of period 1, was minimal (FSS varied by only ±0.06 units; 95% confidence interval = 0.19).

Although analysis of combined period 1 and 2 data indicated that the primary outcome, FSS, decreased significantly on ondansetron versus placebo (0.48 units, P = .025, Table 2), the response to study medication in period 1 versus period 2 was not uniform (Fig. 2). For this reason, the results of each trial period were analyzed separately. During period 1, subjects taking ondansetron and subjects taking placebo exhibited a decrease in their FSS and FIS (Fig 2; Table 3). However, there was no significant difference in the fatigue decrease measured on ondansetron versus placebo, by FSS or FIS (Fig. 2; Table 2). In contrast, in period 2 there was a significant difference in FSS between subjects. Those taking ondansetron exhibited a mean FSS decrease of 0.7 units, whereas subjects taking placebo showed a mean FSS increase of 0.2 units (2-sided P = .001; Fig. 2; Tables 2 and 3). Similar to the findings with FSS, all 3 dimensions of the FIS decreased significantly on ondansetron versus placebo in study period 2 (Tables 2 and 3).

Table 2. Estimated Difference Between Ondansetron and Placebo During the First and Second Trial Period, and During Both Trial Periods
VariablePeriod 1*Period 2Both PeriodsTreatment × Period P#
Estimate ± SEMPEstimate ± SEMPEstimate ± SEMP
  • Abbreviations: FSS, fatigue severity score; PSQI, Pittsburgh Sleep Quality Index.

  • *

    Results from analysis of covariance are shown for differences between end and beginning of period 1, with baseline period 1 as a covariable.

  • Results from analysis of covariance are shown for differences between end and beginning of period 2, with baseline period 1 and baseline period 2 as covariables.

  • Results from repeated-measures analysis of covariance are shown for the combined data for change in periods 1 and 2. The model included components for period, treatment, and baseline.

  • #

    From repeated-measures analysis of covariance with the model described in footnote ‡, expanded to include treatment × period interaction (or carryover effects).

Primary outcome measure       
 FSS0.06 ± 0.29.813−1.03 ± 0.29.001−0.48 ± 0.21.025.015
Other measures       
 Fatigue Impact Scale (FIS)       
  Cognitive dimension0.81 ± 1.91.670−6.02 ± 1.68.001−2.53 ± 1.30.054.008
  Physical dimension−0.68 ± 1.70.691−6.08 ± 1.85.002−3.35 ± 1.29.011.025
  Social dimension−0.44 ± 3.15.887−10.76 ± 2.7.001−5.67 ± 2.11.008.012
 PSQI Score−0.30 ± 0.81.709−1.08 ± 0.79.179−0.94 ± 0.58.111.198
 Hamilton Depression Score0.08 ± 0.68.901−0.86 ± 0.63.174−0.51 ± 0.47.281.214
Figure 2.

Change in Fatigue Severity Score (FSS) over periods 1 and 2.

Table 3. Baseline Values and Changes After 4 Weeks of Treatment in Outcome Variables at the End of Each Study Period
VariableGroup I (n = 24)Group II (n = 30)
Period 1—PlaceboPeriod 2—OndansetronPeriod 1—OndansetronPeriod 2—Placebo
BaselineChange at 4 weeksBaselineChange at 4 weeksBaselineChange at 4 weeksBaselineChange at 4 weeks
  1. NOTE. Values are mean ± SEM.

  2. Abbreviations: FSS, fatigue severity score; PSQI, Pittsburgh Sleep Quality Index.

Primary outcome measure        
 FSS5.6 ± 0.1−0.6 ± 0.15.03 ± 0.2−0.7 ± 0.25.6 ± 0.1−0.6 ± 0.25.09 ± 0.20.2 ± 0.15
Other measures        
 Fatigue Impact Scale (FIS)        
  Cognitive dimension20.7 ± 1.9−3.8 ± 1.518.5 ± 2.01−4.7 ± 1.521.9 ± 2.1−3.3 ± 1.320.4 ± 2.20.8 ± 1.1
  Physical dimension21.6 ± 1.5−4.7 ± 1.819.8 ± 1.7−5.96 ± 2.223.1 ± 1.5−4.4 ± 1.720.3 ± 1.70.9 ± 1.6
  Social dimension33.4 ± 2.9−6.5 ± 2.629.7 ± 2.9−8.3 ± 2.338.1 ± 3.5−8.03 ± 1.934 ± 3.31.7 ± 1.7
 PSQI Score9.5 ± 1.06−0.2 ± 0.89.7 ± 0.8−1.8 ± 0.811.1 ± 0.7−1.03 ± 0.510.03 ± 0.8−0.2 ± 0.5
 Hamilton Depression Score5.6 ± 0.7−1.7 ± 0.53.9 ± 0.7−0.8 ± 0.66.2 ± 0.8−2.1 ± 0.54.3 ± 0.70.2 ± 0.4

There were no significant differences in sleep quality (PSQI) or depression (HRSD) scores on ondansetron versus placebo (Tables 2 and 3). No significant difference was found in any outcome measure among subjects taking beta blockers versus those not taking beta blockers.

Neurocognitive performance, as measured by the Walter-Reed PAB, improved in all subjects during period 1, but was relatively unchanged in period 2. Improvement in sleepiness and cognitive performance among patients taking ondansetron versus those taking placebo did not reach statistical significance (Table 4).

Table 4. Estimated Difference Between Ondansetron and Placebo—Walter-Reed Performance Assessment Battery
VariableBoth Periods*Treatment × Period P
Estimate ± SEMP
  • *

    Results from the GEE/ANCOVA method for the combined data for week 4–baseline 1 and week 9–baseline 2. The model included components for arm, treatment, and baseline.

  • From the results of the GEE/ANCOVA method (* expanded to include treatment × period interaction [or carryover effects].

Sleepiness Scale−0.05 ± 0.06.397.792
Throughput—number of correct responses per minute   
 Choice RT−1.3067 ± 0.2241.563.658
 Logic0.9272 ± 0.6248.134.055
 Serial−1.1391 ± 1.4327.431.423
 Spatial (Manikin)1.4963 ± 1.2379.235.465
 Stroop−0.2152 ± 2.1318.920.958

Side Effects.

Two major side effects were significantly associated with ondansetron administration—constipation and headache (Table 5).

Table 5. Side Effects of Study Medication
 Ondansetron—Frequency, n (%)Placebo—Frequency, n (%)
Period 1Period 2Period 1Period 2
Constipation27 (90.0)17 (63.0)6 (22.2)4 (13.3)
Headache11 (36.7)9 (33.3)9 (33.3)6 (20)

Constipation was very common among patients taking ondansetron, but not among patients taking placebo. Constipation was seen in 90% of subjects taking active medication in period 1, and 63% of subjects taking active medication in period 2 (versus only 22% and 13% of patients taking placebo, respectively). Of subjects taking ondansetron who were constipated, 63% in period 1 and 48% in period 2 required medication for relief of constipation; similar proportions of constipated subjects taking placebo required medication. A total of 4 subjects, all of whom were taking active medication at the time, required a dose reduction in study medication to assist in relief of constipation.

Headache was also a common side effect, but was not seen significantly more often with use of ondansetron than with use of placebo (Table 5). Approximately 15% of patients on ondansetron or placebo required medication for relief of headache. Three subjects required a dose reduction in study medication to assist relief of headache, all of whom were taking ondansetron at the time.

Other side effects, such as dizziness or agitation, were uncommon and not seen more frequently on ondansetron versus placebo. One patient required a reduction in dose of study medication because of presyncope and drowsiness, and was taking placebo at the time.

There were no clinically significant changes in any of the biochemical laboratory tests evaluated during the trial.

Compliance.

Compliance was measured by count of remaining tablets at the end of periods 1 and 2. Mean compliance among all subjects was 96%. Eight subjects had compliance of less than 85% in either period 1 or 2. During the period of low compliance, 5 of the subjects were taking ondansetron, and 2 were taking placebo. One patient had compliance of less than 85% in both study periods.

Discussion

The analysis of results from periods 1 and 2 combined seems to indicate that the administration of ondansetron ameliorates fatigue in patients with PBC, as measured by the FSS and FIS questionnaires. However, this conclusion becomes much less convincing when the trial periods are examined separately, which is necessary because fatigue reduction attributable to study medication use is not uniform in periods 1 and 2 (Fig. 2). In period 1, all subjects exhibited a reduction in fatigue, but there was no significant difference between the response of those on ondansetron and those on placebo. This demonstrates a strong placebo effect in period 1. During this period, all patients may have believed they were on active medication, and accordingly may have responded with lower subjective impressions of fatigue on their questionnaires. Thus, the apparent fatigue reduction benefit of ondansetron over placebo in this study originates entirely from the data in period 2. However, the data from this period are invalidated by subject unblinding because of side effects from the active medication, as well as confounding effects from the crossover trial design.

Ondansetron-induced constipation is a likely cause of subject unblinding in period 2. Constipation is an expected side effect of ondansetron, because blockade of 5-HT3 receptors in the gut can result in decreased motility. In typical regimens of ondansetron employed for chemotherapy-induced nausea (8-32 mg/d for less than a week), constipation arises in less than 15% of patients.25 During our study, a regimen of ondansetron 12 mg daily for 28 days resulted in a surprising 90% and 63% of patients complaining of constipation (in periods 1 and 2, respectively), compared with 22% and 13% of those on placebo (Table 5). The high prevalence of constipation in our subjects may be attributable to delayed drug metabolism (because all subjects had liver disease, albeit compensated), as well as the protracted course of medication used. Furthermore, our subjects were informed during trial enrollment that constipation may be a side effect of study medication use. A short list of potential side effects told to a patient before receiving a medication is known to increase patient vigilance for adverse symptoms, and thus increase the risk of unblinding from such side effects.26, 27 In our study, increased patient vigilance for constipation as a drug side effect, combined with constipation that correlated strongly with active medication use, resulted in significant potential for side effect–mediated unblinding. Unblinding was evident in period 2, by which time patients could observe a change in constipation between the study periods and determine which treatment they were taking in the latter period. Unblinded subjects may have perceived their fatigue differently, resulting in over- or under-reporting of fatigue on study questionnaires, and altered motivation to concentrate and achieve optimal scores during neurocognitive testing.

Period 2 data are also invalidated by confounding effects from the crossover trial design. Though a crossover trial benefits from advantages such as patient enrollment efficiency and increased statistical power, confounding factors may be introduced, such as a carryover effect or sequence effect. Using analysis of covariance, we found a treatment period interaction for FSS (P = .015) and the 3 dimensions of the FIS (all P ≤ .025; Table 2). This result indicates a statistically significant confounding effect from the crossover trial design. A carryover effect is evident because patients in both randomization groups did not return to their original baseline FSS after period 1 was complete and period 2 began (Fig. 2). This emphasizes the importance of placebo effect in period 1; although the drug was likely completely metabolized during the washout period, the placebo effect may have taken longer to resolve to baseline. As well, we believe a sequence effect occurred because of unblinding from constipation caused by ondansetron, because the change in constipation between period 1 and 2 may have allowed patients to determine which treatment they were taking by period 2.

One might ask why a crossover design was chosen to conduct this trial, if such confounding factors could be predicted. In fact, a crossover design was deliberately chosen for this study, as it was the best possible method to study a very subjective symptom, namely, that of fatigue. A crossover trial, unlike a traditional randomized trial, allows patients to serve as their own controls, minimizing between-subject variability, which can be considerable when measuring a subjective symptom such as fatigue.

Sound statistical methodology dictates that we must discard data from the second trial period in a crossover design study that has evidence of sequence or carryover effect.28, 29 On this basis, we must judge the trial results based solely on the findings from period 1. There was no significant difference in the reduction of fatigue on ondansetron versus placebo in period 1, due to a strong placebo effect. Therefore, we must conclude that ondansetron did not have a fatigue-reduction benefit greater than the considerable placebo effect we observed in our population of patients with PBC.

An important caveat about the fatigue measured by the FSS and FIS surveys in this study must be considered. Although these questionnaires have been validated to quantify fatigue in study populations of patients with PBC, they are only a surrogate measure of true fatigue. By nature, fatigue is subjective, difficult to quantify, subject to variable perception by the same subject over time, and of variable meaning and significance between subjects. Therefore, correlating one unit on the FSS scale with a reproducible, specific amount of fatigue that is of equal significance to any given patient in a study population at any point in time is difficult. Thus, in this study, the inherent difficulty in measuring a symptom as subjective and ill-defined as fatigue, particularly with questionnaires that are surrogate measures for this symptom, must temper any interpretations of clinical significance from fatigue changes that were measured.

Depression and impaired sleep quality are associated with the presence of fatigue in patients with PBC.1 During this trial, we measured depression (by the HRSD) and sleep quality (by the PSQI), and neither changed significantly on ondansetron versus placebo. Moreover, the Sleepiness Scale in the neurocognitive battery did not change. Because no change in these parameters was seen during the trial, depression and sleep quality were not confounding factors affecting the amount of fatigue measured by FSS or FIS in this study.

With regard to neurocognitive testing, patients appeared to have insufficient practice in performing the Walter-Reed PAB cognitive tasks to a level of stable performance before evaluation during the trial. Improvement in overall cognitive test performance was observed during period 1, but relatively stable performance was present in period 2. The substantial gains in neurocognitive performance seen in period 1 are likely attributable to task learning, regardless of treatment received. However, improvement in performance for 2 of the most difficult cognitive tasks, logical reasoning and spatial orientation, was 3 times greater for those subjects receiving ondansetron versus placebo in period 1. Once subjects were more significantly practiced in period 2, these differences were not evident. Because learning new cognitive skills can be impaired by fatigue, these data suggest that ondansetron may facilitate the learning of new, highly complex cognitive tasks. This notion must be considered a hypothesis only, because the study was not powered to prospectively assess neurocognitive performance measurements.

In conclusion, whereas analysis of the trial data as a whole suggests that in our study population ondansetron ameliorated fatigue associated with PBC, closer inspection shows that ondansetron actually provided no significant fatigue reduction benefit. Fatigue reduction associated with active medication was observed solely in period 2, the data from which are invalid because of unblinding from ondansetron-mediated constipation, as well as a statistically significant treatment-period interaction indicating carry-over and sequence effects from the crossover trial design (P = .015). However, the data from period 1 can be interpreted in a similar fashion to a traditional randomized controlled trial, regardless of any confounding effects in period 2. In period 1, ondansetron was no better than placebo for amelioration of fatigue. A small ondansetron-mediated reduction of fatigue might have been masked in period 1 by a large placebo effect. However, if such an effect was present, it was not large enough to be clinically significant, especially given both the considerable financial cost and observed side effects of the drug.

Although confounding effects invalidated the data from period 2, the data in period 1 provide good evidence that ondansetron is not effective for clinically significant fatigue reduction in our study population of patients with PBC. Because previous studies provide convincing evidence that fatigue in cholestatic disease may be centrally mediated, it is still possible that pharmacological agents targeting the central nervous system may help to ameliorate fatigue in PBC. However, future trials testing this hypothesis would be best conducted using an agent with fewer side effects.

Acknowledgements

Glaxo-Wellcome provided the study medication for these studies but had no involvement in trial design/execution, manuscript writing, or review.

Ancillary