Rituximab for the treatment of fatigue in primary biliary cholangitis (formerly primary biliary cirrhosis): a randomised controlled trial

Fatigue: the commonest symptom in primary biliary cirrhosis

Historically, the main focus in terms of clinical impact and its treatment in PBC has been progression of the disease to cirrhosis and, ultimately, death. In recent years, with the advent of large-scale representative patient cohorts and validated patient-reported outcome measures (PROMs), it has become widely accepted that there is another important aspect to patient impact: that is quality-of-life impairment related to systemic symptoms. 12,17,20–24 Although quality-of-life impairment can be significant in patients with decompensated cirrhotic disease, this state is very uncommon in the condition, meaning that the conventional complications of end-stage liver disease contribute relatively little to overall quality-of-life burden. 25,26 In contrast, the disutility associated with the non-stage-specific symptoms is substantial. 24 Data from the very large UK-PBC cohort have also demonstrated that the burden of impaired quality of life is age related, with younger patients experiencing the worst symptoms and the greatest quality-of-life burden. 27

The symptom reported most frequently by PBC patients is chronic fatigue, which is present in > 50% and which has an often striking impact on patients. 28–30 There are currently no effective treatments for fatigue in PBC, with international treatment guidelines recommending only supportive treatment. 31 Unsurprisingly, given its prevalence, impact and the lack of effective treatments, fatigue is frequently identified by PBC patients as the principal priority area for new therapy development. There has been a significant recent increase in the understanding of the potential mechanisms that might contribute to fatigue in PBC, and significant advances in the development of clinical tools that are able to quantify the impact of that fatigue and that represent key tools for the study of fatigue pathogenesis and, critically, treatment. The key advance in this regard has been the development of the PBC-40, a patient-derived, disease-specific, quality-of-life measure that has become the standard measure used in all clinical trials. 14,21,32

Fatigue severity in PBC is not associated with the stage of the disease and is frequently seen in its most severe form in patients with early disease and a seemingly very good prognosis from the point of view of risk of death from their liver disease. 12,28,33 This can result in a discrepancy between the physician’s and patient’s perception of how well the patient is doing, with the former seeing a level of liver injury suggesting a good prognosis and the latter focusing on their severe fatigue. 30 Furthermore, there is no evidence to suggest that either current first- (UDCA) or second-line therapy has any impact on fatigue severity. 12,14 As prognostic management gets better in PBC, so the contrast between length of life and quality-of-life improvement will become increasingly marked.

The issue of the role played by depression in the expression of fatigue in PBC is a controversial one, with many patients being clinically labelled as having depression simply because they are fatigued, making the association self-fulfilling. Although depressive problems can clearly occur in fatigued PBC patients (frequently as a consequence of the considerable limitation to lifestyle that they experience), and positive coping strategies are an important part of managing fatigue, overt depressive illness is actually a rare cause of fatigue in PBC. 34,35 In fact, in most fatigued PBC patients it is peripheral, activity-limiting fatigue that predominates. 36 Fatigue in PBC can be associated with significant fatigue disturbance (in particular daytime somnolence) and autonomic dysfunction (particularly in the form of baroreflex sensitivity loss). The extent to which these associations reflect causes or consequences of chronic fatigue is unclear at present. 1

The scale of the problem of fatigue in primary biliary cirrhosis

Population-based studies undertaken using the PBC-40, which contains a fully validated fatigue domain, have suggested that > 50% of PBC patients across the UK experience moderate or severe fatigue, with > 20% experiencing sustained severe fatigue (≈4000 patients nationally), with a profound impact on life quality. 12,24,27 Fewer than 20% of severely fatigued patients are able to work, with the majority of this group dependent on state benefits for income. Severe fatigue also causes social isolation in people who lack the energy to leave their house and interact with others, a particular issue among younger patients. 24 Severely fatigued PBC patients also frequently have problems with their ability to care for others (90% of PBC patients are female and they often have young children and/or dependent elder relatives themselves), with cases of Meals on Wheels being provided for the children of socially isolated, severely fatigued PBC patients being far from uncommon. The management of severe fatigue in PBC, which is currently largely supportive, is a major draw on NHS budgets (the trial investigators are currently following up > 150 severely fatigued patients with, typically, monthly clinic follow-up). 26,31 Severe fatigue is also a frequent driver for patients seeking liver transplantation, although the evidence to suggest that there is any benefit from transplantation for PBC fatigue is limited. 25,37 Treatment able to reduce the impact of severe fatigue in PBC would transform the lives of patients and their families, substantially reduce the costs of NHS treatment (the costs of follow-up and supportive management of fatigued patients and the costs of transplantation where the indication is advanced liver disease but the patient driver is fatigue) and, importantly, the costs to society of this socially isolated, economically unproductive and dependent group of people. 38

Bioenergetic abnormality is frequent in primary biliary cirrhosis and is associated with fatigue

Primary biliary cirrhosis patients’ descriptions of fatigue emphasise an inability to sustain physical activity and concepts related to ‘energy’ or ‘power’ depletion (a further group describe ‘brain fog’, which is associated with significant cognitive symptoms). Furthermore, on repeat testing, handgrip strength reduces rapidly in fatigued PBC patients and is associated with perceived fatigue severity. 36 More recently, using magnetic resonance (MR) spectroscopy approaches, we demonstrated (and replicated in a second cohort) the presence of significant peripheral muscle bioenergetic abnormality in PBC patients. 39,40 Abnormality is also present in cardiac muscle, suggesting the presence of a broadly based bioenergetic abnormality. 41 These studies have suggested the presence of abnormality in mitochondrial function (in particular, recovery response to the energetic burden of repeat exercise) and a tendency for excessive intramuscular acidosis during, and in the recovery from, exercise. Intramuscular acidosis in a well-recognised contributing factor to muscle fatigue in the general population. 42

In PBC, the length of time that significant acidosis is present in muscle and the rate of recovery from that acidosis are strongly associated with the severity of fatigue that patients experience. 43–45 The cumulative effect of a pro-acidotic effect and delays in recovery is well demonstrated using an integrative approach exploring the ‘area under the curve’ (AUC) for pH during exercise and in the recovery from exercise, which gives a measure of the scale of excess proton exposure in PBC and its association with fatigue (Figure 1). Recent studies using metabolomics approaches have shown lactate accumulation within tissues in PBC. 46 Our own unpublished study of anaerobic threshold in PBC patients (compared with age- and sex-matched patients with primary sclerosing cholangitis and sedentary control subjects), assessed using an incremental exercise approach, shows significantly lower anaerobic threshold values in PBC that cannot be accounted for by either deconditioning or by the presence of cholestatic liver disease per se. 47 These observations further support the presence of bioenergetic abnormality in PBC and critically do so using complementary experimental approaches. Taken together, these studies all point to excessive utilisation of anaerobic metabolism pathways in muscle in PBC [metabolism of pyruvate to lactate via the actions of lactate dehydrogenase rather than to acetyl coenzyme A via the actions of pyruvate dehydrogenase (PDH)]. This leads, we hypothesised, to excessive and inappropriate lactic acid accumulation in muscle in PBC and, in the absence of an effective adaptive response to it, fatigue. In this model, capacity to adapt to acid load is a physiological variant that is naturally protective from fatigue in some people and that can be induced by exercise therapy. The critical question is ‘why do PBC patients seemingly overutilise anaerobic metabolism pathways and how can we use that knowledge to treat their fatigue?’.

FIGURE 1.

Area under receiver operating characteristic (AUROC) changes in pH vs. recovery time following exercise. Blue is normal, green is non-fatigued PBC patients and black is fatigued PBC patients.

Primary biliary cirrhosis is characterised by autoimmune responses directed at bioenergetic enzymes

The obvious link between PBC as an autoimmune disease and PBC as a disease characterised by overutilisation of anaerobic pathways and associated fatigue is the characteristic autoantibody response in PBC, which is directed at PDH, the enzyme that controls entry into aerobic metabolism pathways. High-titre anti-PDH antibodies are present in > 95% of PBC patients and are able to fully block PDH function during in vitro assays. 48,49 The basis of this inhibitory function (which is so consistent that one of the assays for anti-PDH in PBC utilises inhibitory capacity as a readout) is the specific reactivity of anti-PDH antibodies with a lipoic acid cofactor present on the E2 and E3 binding protein components of PDH, to which the principal autoepitopes have been localised, which is essential for enzymatic function and which is blocked by antibody binding. 50

Targeting autoreactive responses in primary biliary cirrhosis as a potential therapy for fatigue

Three strands of evidence support a potential direct role for anti-PDH antibody responses in bioenergetic abnormality and fatigue in PBC.

1. The mitochondrial bioenergetic abnormality in PBC is directly associated with the level of serum anti-PDH antibody, with median anti-PDH titre being > 200-fold higher in patients with mitochondrial bioenergetic abnormality than in those without (9.9 × 10⁵ vs. 4.1 × 10³; p < 0.05). 39 At least part of the anti-PDH autoantibody response in PBC is directed at the lipoate component of PDH and is reactive with free lipoic acid in the circulation. 50 Given that free lipoic acid plays an important role in regulating PDH function (it activates PDH through inhibition of PDH-kinase), its consumption by autoantibody in PBC represents one potential mechanism for the bioenergetic effect of that antibody. 51

2. Patients who are anti-PDH antibody positive but who have normal liver function (‘pre-PBC’ patients) experience fatigue of the same severity as ‘classic’ PBC patients, suggesting that fatigue is a feature of the presence of anti-PDH rather than PBC per se. 52

3. Fatigue is typically not improved following liver transplantation in PBC. 38,53 Anti-PDH universally remains present at unchanged levels following liver transplantation in PBC.

A direct role for anti-PDH antibody in bioenergetic abnormality and fatigue was the justification for exploration of therapies able to deplete the B-cell compartment, exemplified by the anti-CD20 antibody rituximab (MabThera^®, Roche Products Ltd) as a treatment for fatigue in PBC.

Rituximab in primary biliary cirrhosis

A pilot study performed in Canada exploring the use of rituximab in PBC (in 13 patients) provided proof of concept, showing that the agent is safe and well tolerated in patients and is associated with a clinically significant reduction in fatigue. 54 Fatigue severity was assessed using the Fatigue Severity Scale (potential range 9–63 points), with a fall being seen from pre treatment (median 36 points, range 11–59 points) to post treatment (median 29 points, range 12–55 points). Taking into account the floor value for the Fatigue Severity Scale, this represents a median fall in fatigue severity over 6 months of 26%. This compares with our own case–control study of fatigue in PBC, which suggests that fatigue severity in age- and sex-matched normal controls is 30% lower than in PBC patients, suggesting the potential for rituximab therapy to return PBC patients to close to normal with regard to their perceived fatigue. 28 However, this pilot study did not attempt to explore the mechanism of the effect and, as it did not use severe fatigue as an inclusion criterion, the extent of possible improvement for such patients is unclear. Moreover, the study was not optimised for the study of fatigue (fatigue was a secondary outcome and only some of the patients who participated had fatigue, potentially underestimating the clinical effect). Other evaluations of rituximab in PBC have focused exclusively on disease severity as an end point. 55 Both studies showed a sustained reduction in anti-PDH antibody levels of all isotypes, supporting the concept that rituximab has a beneficial effect on fatigue through depletion of PDH-reactive antibody. In all studies in PBC to date, rituximab has been found to be well tolerated.

Inclusion criteria

• Patient had capacity to consent and provided written informed consent for participation in the study prior to any study-specific procedures.

• Moderate or severe fatigue as assessed using previously designated cut-off points of the PBC-40 fatigue domain (i.e. fatigue domain score of > 33).

• Presence of AMA (anti-PDH antibody) at a titre of > 1 : 40.

• Adequate haematological function (haemoglobin level of > 9 g/l), absolute neutrophil count of > 1.5 × 10⁹/l and platelet count of > 50 × 10⁹/l.

• Bilirubin level of ≤ 50 µmol.

• International normalised ratio ≤ 1.5.

• Child–Pugh score of < 7.

• Eastern Cooperative Oncology Group performance status of < 2.

• Adequate renal function: Cockcroft–Gault estimation of > 40 ml/minute.

• Women of childbearing potential had a negative pregnancy test prior to study entry and were using an adequate contraception method, which was continued for 3 months after completion of treatment. Acceptable forms of effective contraception included established use of oral, injected or implanted hormonal methods of contraception:

  • placement of an intrauterine device or intrauterine system

  • barrier methods of contraception – condom or occlusive cap (diaphragm or cervical/vault caps) with spermicidal foam/gel/film/cream/suppository

  • male sterilisation (with the appropriate post-vasectomy documentation of the absence of sperm in the ejaculate)

  • true abstinence – when this was in line with the preferred and usual lifestyle of the subject

  • aged ≥ 18 years.

Exclusion criteria

• Advanced or decompensated disease (variceal bleed, hepatic encephalopathy or ascites).

• History or presence of other concomitant liver diseases (including hepatitis due to hepatitis B virus or hepatitis C virus or evidence of chronic viraemia on baseline screening), primary sclerosing cholangitis or biopsy-proven non-alcoholic steatohepatitis).

• Average alcohol ingestion of > 21 units per week (for males) or > 14 units per week (for females).

• Chronic sepsis or intercurrent condition likely to predispose the patient to chronic sepsis during the study.

• Previous treatment with B-cell-depleting therapy.

• Previous history of aberrant response or intolerance to immunological agents.

• Presence of significant untreated intercurrent medical condition itself associated with fatigue.

• Presence of significant risk of depressive illness [Hospital Anxiety and Depression Scale (HADS) score indicating cosiness].

• Current statin therapy or statin use within 3 months of enrolment.

• Ongoing participation in other clinical trials or exposure to any investigational agent 4 weeks prior to baseline or within five or fewer half-lives of the investigational drug.

• Major surgery within 4 weeks of study entry.

• Vaccination within 4 weeks of study entry; patients requiring seasonal flu or travel vaccines will be required to wait a minimum of 4 weeks post vaccination to enrol in the study.

• Pregnant or lactating women.

• Psychiatric or other disorder likely to have an impact on informed consent.

• Patient was unable and/or unwilling to comply with treatment and study instructions.

• Any other medical condition that, in the opinion of the investigator, would interfere with safe completion of the study.

• Hypersensitivity to the active substance (rituximab) or to any of the excipients [sodium citrate, polysorbate 80, sodium chloride, sodium hydroxide, hydrochloric acid and water (for infusion)] or to murine proteins.

• Active, severe infections (e.g. tuberculosis, sepsis or opportunistic infections).

• Known human immunodeficiency virus infection.

• Clinical history of latent tuberculosis infection unless the patient had completed adequate antibiotic prophylaxis.

• Aspartate transaminase (AST)/alanine transaminase (ALT) levels four times above the upper limit of normal.

• Severe immunocompromised state.

• Severe heart failure (New York Heart Association class IV) or severe uncontrolled cardiac disease.

• Malignancy (other than basal cell carcinoma) within the last 10 years.

• Demyelinating disease.

• Previous participation in this study.

• Any contraindication to rituximab therapy not covered by other exclusions.

Withdrawal criteria

The study protocol required that the study drug be discontinued if:

• the participant developed elevated serum ALT/AST levels four times above normal limits for each local laboratory

• the participant decided they no longer wished to continue

• cessation of study drug was recommended by the investigator.

Participants had the right to withdraw from the study at any time for any reason and without giving a reason. The investigator also had the right to withdraw patients from the study drug in the event of intercurrent illness, adverse events (AEs), serious adverse events (SAEs), suspected unexpected serious adverse reactions (SUSARs), protocol violations, cure, administrative reasons or other reasons. If a patient decided to withdraw from the study, all efforts were made to report the reason for withdrawal as thoroughly as possible. If a patient withdrew from the study drug only, efforts were made to continue to obtain follow-up data (with the permission of the patient).

Participants who wished to withdraw from study medication were asked to confirm whether or not they were still willing to provide:

• study-specific data at follow-up visits 5–19 [https://njl-admin.nihr.ac.uk/document/download/2005806 (p. 33); accessed September 2017]

• end-of-study data as per visit 19, at the point of withdrawal

• questionnaire data collected as per routine clinical practice at annual follow-up visits.

If participants agreed to any of the above, they were asked to complete a confirmation of withdrawal form to document their decision.

Participants withdrawing from the study after receiving the second infusion were not replaced. In practice, all seven withdrawals from the study were replaced prior to study drug administration.

Setting

All screening, consent, treatment and follow-up assessments requiring patient attendance took place at the clinical research facility at the Royal Victoria Infirmary in Newcastle. Data recording took place in the same setting.

Experimental intervention: rituximab therapy

The investigational medicinal product (IMP) used in the clinical trial was 1000 mg of intravenous (i.v.) rituximab (MabThera). This product was approved by the European Agency for the Evaluation of Medicinal Products (marketing authorisation number EU/1/98/067/002). The licence holder of MabThera is Roche Products Ltd. The product was supplied as vials containing 500 mg of rituximab at a concentration of 10 mg/ml. Supplies of rituximab were stored in a refrigerator at between 2 °C and 8 °C in a secure location in their original packaging in order to protect them from light. Supplies of rituximab were labelled as IMP in accordance with regulatory requirements. Storage and supply of the rituximab was delegated to the local pharmacy at the trial centre. Rituximab was released for use in the study once all the appropriate regulatory and governance approvals were in place. Patients randomised to receive rituximab therapy were given treatment at the infusion rates recommended for rheumatoid arthritis (RA) patients (i.e. for the first infusion at an initial rate of 50 mg/hour; after the first 30 minutes the rate could be increased in increments of 50 mg/hour every 30 minutes to a maximum rate of 400 mg/hour). Second doses were infused at an initial rate of 100 mg/hour and increased by 100 mg/hour at 30-minute intervals to a maximum of 400 mg/hour. Rituximab has a shelf life of 30 months. The prepared infusion solution of rituximab (MabThera) was physically and chemically stable for 24 hours at temperatures between 2 °C and 8 °C and subsequently for 12 hours at room temperature. Further details can be found in the MabThera Summary of Product Characteristics [https://njl-admin.nihr.ac.uk/document/download/2005806 (p. 56)].

Control intervention: placebo infusion

Patients randomised to receive placebo received conditioning in addition to the control saline infusion. The control infusion was delivered in a double-blind manner to participants using the same placebo that was used in our RA studies, under the supervision of a clinician (as per the clinical research facility protocol).

Conditioning

In line with recommendations for the administration of rituximab in other conditions, patients received a conditioning regimen prior to the infusions of study medication on days 1 and 15. Given that this was a double-blinded trial, both arms received the conditioning regimen. This conditioning regimen was administered 30 minutes prior to infusion of study medication and to all patients, to maintain the double blind. The conditioning regimen comprised 1 g of paracetamol orally, 10 mg of i.v. chlorpheniramine and 100 mg of i.v. methylprednisolone. The paracetamol, chlorpheniramine and methylprednisolone were sourced locally.

Adverse reactions during treatment administration

During the infusion, patients were clinically monitored for the onset of clinical features of infusion-related reactions (IRRs), the most common AEs in rituximab usage.

The protocol stated that patients who developed evidence of severe IRRs, especially severe dyspnoea, bronchospasm or hypoxia, would have the infusion interrupted immediately, with restart of the infusion (at not more than half the rate at which the reaction took place) not taking place until complete resolution of all symptoms and normalisation of laboratory values and chest radiographic findings. If the same severe adverse reactions (ARs) occurred for a second time, the treatment would be stopped. Such reactions were not observed in the study and this protocol was therefore not invoked.

Mild or moderate IRRs were addressed by halving the rate of infusion. The infusion rate was increased again on improvement of symptoms. Paracetamol and/or non-steroidal anti-inflammatory drugs were allowed for the symptomatic management of IRRs.

Concomitant medication

Concomitant therapies were managed as per the MabThera Summary of Product Characteristics. A complete listing of all concomitant medication received during the treatment phase was recorded in the case report form (CRF).

For patients who were receiving UDCA therapy, the dose of UDCA could not be changed during the patients’ 12-month participation in the study. No other disease-modifying agents could be introduced during the trial. Therapy aimed at reducing pruritus and its impact could be introduced if unavoidable at the discretion of the investigators; however, this proved unnecessary.

Live vaccines could not be administered during the study (see Exclusion criteria).

Primary outcome measure

The primary outcome variable was fatigue severity in PBC patients, assessed at 3 months using the fatigue domain of the PBC-40: a fully validated, psychometrically robust, disease-specific quality-of-life measure. 21

Secondary outcome measures

The following secondary outcome measures were prespecified:

• The PBC-40 domain scores covering itch, cognitive, emotional, social and other symptom domains at 3, 6, 9 and 12 months.

• Physical activity assessed using 7-day physical activity monitoring.

• Perceived fatigue severity assessed using a self-completion fatigue diary.

• Daytime somnolence [as assessed using the Epworth Sleepiness Scale (ESS)57], vasomotor autonomic symptoms [assessed using the Orthostatic Grading Scale (OGS)58] and functional status [assessed using Patient-Reported Outcomes Measurement Information System Health Assessment Questionnaire (PROMIS HAQ), incorporating the Cognitive Failure Questionnaire (COGFAIL)29]. Reduction in depressive and anxiety-related symptoms was assessed using the HADS. 59

• Serum anti-pyruvate dehydrogenase complex (PDC) antibody levels assessed using enzyme-linked immunosorbent assays (to confirm whether or not any clinical effect is directly related to antibody modulation). 60,61

• Peripheral muscle bioenergetic function on exercise (to confirm whether or not any clinical effect was directly related to effects on muscle bioenergetic function). Approaches used included MR spectroscopy assessment of muscle pH change during a fixed exercise protocol, time taken for recovery of pH to baseline following exercise, AUC for pH (a measure of total muscle pH exposure with exercise) and anaerobic threshold assessed using cardiopulmonary exercise testing (CPET). 39,40,47

Exploratory end points

• Biochemical parameters associated with severity of the liver disease in PBC (ALP), ALT, bilirubin and albumin.

• Serum immunoglobulins (Igs) [total Ig, immunoglobulin G (IgG) and immunoglobulin M (IgM)].

  During the course of the study, a continuous variable model for the prediction of outcomes in PBC was described and validated. This used routinely collected clinical data and could be incorporated into the protocol as exploratory end points. 62 Concern had also developed around second-line therapeutics in PBC and cardiovascular risk related to lipids. In the light of these changes, the following exploratory end point was added during the study:

  • blood lipid levels.

Quality-of-life and symptom assessment tools

Fatigue severity for the primary end point was assessed using the PBC-40 fatigue domain. This is a validated, disease-specific and patient-derived quality-of-life measure. The fatigue domain score ranges from 11 to 55 (11 items scored 1–5), with higher values denoting worse fatigue.

Other symptom severity was assessed by the relevant domain of the PBC-40 (itch, cognitive, social, emotional and other symptoms). Functional status was assessed using the PROMIS HAQ to assess function. The PROMIS HAQ measures the functional and physical ability of the participants (covering, dressing, arising, eating, walking, hygiene, reach, grip and activity). The score is on a 0–100 scale with higher scores indicating worse functional ability. Anxiety and depression were assessed by HADS score (range 0–42). Daytime somnolence was assessed using the ESS (range 0–24), vasomotor autonomic symptoms using the OGS (range 0–20) and cognitive functionality using COGFAIL (range 0–100), where higher scores indicate worse outcome for all cases.

Fatigue diaries

Participant-held diaries were used to gather qualitative information on symptoms and functional ability. The diaries measured fatigue using a scale of 1–6 (1 = no fatigue and 6 = extreme fatigue). Participants were asked to complete the diaries six times during the study. They completed the diaries for a period of 1 week during the first week of each month at baseline and 1, 3, 6, 9 and 12 months. Participants returned the diaries at their final visit.

Magnetic resonance spectroscopy

Magnetic resonance data relating to muscle energetics and pH were acquired at baseline and at 12 weeks using a 3-T Intera Achieva scanner (Philips, Best and NL; Philips Medical System, Andover, MA, USA). The steps used for acquisition and analysis have been described, in detail, in published protocols. 39,40 In summary, this involved controlled plantarflexion using a purpose-built exercise apparatus developed for operation within the MR imaging scanner. Participants performed two 180-second bouts of plantarflexion contractions at 25% and then at 35% of maximal voluntary contraction, with each bout preceded by 60 seconds of rest and followed by 390 seconds of recovery. Phosphorous spectra were collected at 10-second intervals. The minimum pH seen in the exercise and recovery period, the time required post exercise for pH to return to within 0.01 units of baseline levels (calculated as the sum for each individual for the three bouts to form a total pH recovery time) and the mean AUC for pH for the three exercise episodes, which reflected total acid exposure, were calculated.

Anaerobic threshold

Anaerobic threshold was assessed using conventional CPET. Participants cycled on a stationary ergometer (Lode BV, Groningen, the Netherlands) at between 60 and 70 revolutions per minute. The test was terminated either voluntarily by the participant or when they were unable to maintain a pedal frequency of 60 revolutions per minute. Expired air was collected at rest and during exercise using a breathing mask and analysed online using a gas analysis system (MetaLyzer II, CORTEX Biophysiks GmbH, Walter-Kohn, Germany). Anaerobic threshold was determined using the computerised v-slope method at baseline and at the 12-week follow-up.

Physical activity levels

Participants completed physical activity monitoring using wrist-worn tri-axial GENEActiv accelerometers (Activinsights Ltd, Cambridgeshire, UK). The accelerometer was worn continuously on the right wrist for a period of 7 days in free-living conditions. Raw accelerometer data were processed in R studio (www.cran.r-project.org; accessed February 2017) using R-package GGIR version 1.2.8 (The R Foundation for Statistical Computing, Vienna, Austria). Patients were included in the analysis if they had worn the monitor for a minimum period of 5 days (with at least one of these days at the weekend). Only days with at least 22 hours of valid data were retained for further analysis. The first and last days of the raw accelerometer measurement were excluded as they were influenced by the monitor distribution and collection procedure.

Following data processing and the exclusion of patients who did not meet the wear-time criteria, the average magnitude of wrist acceleration was calculated via metric Euclidean Norm Minus One (ENMO) as previously described. 63 The output from metric ENMO is in mg (1 mg = 0.001 g = 0.001 × 9.8 m/s² = 0.001 × gravity). In addition, the average acceleration (mg) during the most active 5-hour period of each day was assessed pre versus post intervention.

Immunology

Quantification and phenotyping of total B-cell populations and B-cell subsets was undertaken using a standard fluorescence-activated cell analysis-based approach. Total B-cell levels in peripheral blood were evaluated using a direct immunofluorescence reagent (Fast Immune™ CD19/CD69/CD45, BD Biosciences, San Jose, CA, USA).

Anti-PDH antibody total and individual isotype levels were studied on day 0 and at the primary end point (12 weeks after therapy), using a well-established enzyme-linked immunosorbent assay developed within our research group. 61

Biochemistry and haematology

Biochemical (liver biochemistry, electrolytes and lipids) and haematological (platelet count) assessments were undertaken using the routine clinical laboratories at the research centre. These are quality-assured clinical parameters. The UK-PBC risk score was calculated using the published formula. 62 The values generated are percentage risks of death from liver-related causes or need for liver transplant at 10 years.

Data Monitoring and Ethics Committee

An independent DMEC was appointed, the terms of reference of which were governed by a signed DMEC charter. It consisted of an independent chairperson, an independent expert clinician and an independent statistician and was convened to undertake independent review. The purpose of this committee was to monitor efficacy and safety end points. Only the independent DMEC members and the unblinded statistician had access to unblinded study data. The committee met eight times during the study.

Trial Steering Committee

A TSC was established to provide overall supervision of the study, the terms of reference of which were governed by a signed TSC charter. The TSC consisted of an independent chairperson, an independent clinician, an independent consumer representative, the chief investigator/PI, senior trial manager, trial manager and trial statistician. Representatives of the sponsor were invited to all TSC meetings. The committee met eight times during the study.

Study monitoring

Monitoring of study conduct and data collected was performed by a combination of central review and site monitoring visits to ensure that the study was conducted in accordance with GCP, the protocol and clinical trial regulations. Study site monitoring was undertaken by the trial manager. The main areas of focus included consent, SAEs, essential documents in the investigator site file and drug accountability and management.

Site monitoring included the following:

• All original consent forms were reviewed as part of the study file (the presence of a copy in the patient hospital notes was confirmed for 20% of participants).

• All original consent forms were compared against the study participant identification list.

• All reported SAEs were verified against treatment notes/medical records (source data verification).

• Essential documents were accessible in the investigator site file.

• Source data verification of primary end point data and eligibility data was undertaken for 20% of participants entered in the study.

• Drug accountability and management was checked; an unblinded monitor conducted pharmacy monitoring visits.

Central monitoring included the following:

• All applications for study authorisations and submissions of progress/safety reports were reviewed for accuracy and completeness, prior to submission.

• All documentation essential for study initiation was reviewed prior to site authorisation.

Sample size

The study was planned to detect a mean change in PBC-40 fatigue domain score of 5 units at 3 months’ follow-up (equating to an average of a 0.5-point change per question, a difference in PBC-40 score demonstrated in our population-based studies to be associated with significantly higher levels of social function). 24 Previous studies had shown a standard deviation (SD) of 8 units and a correlation of 0.6 between baseline and follow-up, so, using a power of 90% and a 5% significance level, this required outcome data from 35 participants per arm. A total of 78 participants (39 per arm) was planned to be recruited and randomised, assuming 10% attrition at 3-month follow-up. 21

However, as recruitment was slower than expected, even after expanding recruitment beyond the north-east region of England, a revised sample size calculation was undertaken. Consequently, trial recruitment was extended by 6 months and the power of the trial was reduced to 80%, with a revised target sample size (with other estimates of parameters unchanged) of 58 participants (29 per arm). These changes to the design of the trial were agreed by the funder and submitted as a substantial amendment, which was accepted by both the Research Ethics Committee (REC) and MHRA during October 2015.

Statistical analysis

A complete statistical analysis plan was finalised and signed before datalock and unblinding (see Report Supplementary Material 1). It provides full details of statistical analyses, variables and outcomes.

All analyses were performed on the intention-to-treat (ITT) population. A pragmatic ITT approach was used in which patient outcomes were reported at their visit nearest the scheduled appointment date. Baseline characteristics of the study population were summarised separately within each randomised group.

The primary analysis was the PBC-40 fatigue domain. Descriptive statistics were reported at each time point (baseline and 3, 6, 9 and 12 months) in both arms. PBC-40 fatigue domain scores at 3 months were compared in the intervention and placebo groups using multiple linear regression with adjustment for baseline PBC-40 fatigue score, age in years, UK-PBC risk model score at 10 years and patient location (managed by the Newcastle Clinics for Research and Service in Themed Assessment centre for at least 1 year or not) at baseline. The results were reported as an adjusted difference in means with a 95% confidence interval (CI). Bootstrap estimation was used throughout. The time course of the comparison between intervention and control groups over the 12-month follow-up period using the time points above was assessed for PBC-40 fatigue domain using repeated measures analysis of variance (ANOVA).

The above analyses for PBC-40 fatigue score were repeated for secondary outcomes, namely:

• other PBC-40 domain scores (itch, cognitive, social, emotional and other symptoms)

• clinical symptom and functional capability scales (ESS, OGS, PROMIS HAQ, COGFAIL, HADS and fatigue diary)

• metric ENMO activity monitoring (daily average and best 5 hours)

• bioenergetics outcomes (minimum pH post exercise, pH recovery half-time, pH fall with exercise, AUC for pH and anaerobic threshold)

• biological outcomes [serum immunoglobin level (IgG and IgM)].

Euclidian Norm Minus One, bioenergetics and biological outcomes were collected only at baseline and 3 months, so no repeated measures analyses were performed.

Descriptive analyses were reported for other secondary outcomes listed below [additional repeated measures (ANOVA up to 12 months) were carried out for bilirubin and serum ALP outcomes]:

• full blood count (haemoglobin concentration, white blood cell count and platelet count)

• liver function test [prothrombin time, bilirubin level, ALP level, ALT level, AST level, albumin level, GGT level, activated partial thromboplastin time and C-reactive protein (CRP) concentration]

• lipid profile [low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL) and triglyceride]

• urea and electrolytes (U&E) (urea, creatinine, sodium and potassium concentrations)

• anti-mitochondrial antibody titre (expressed as number and percentage) and anti-PDC antibody level

• CD19 B-cell depletion.

Change in serum ALP level (at 3 and 12 months from baseline) was also measured and we classified a drop in level of > 15% or normalisation (within normal range) as being clinically significant. The number (%) of clinically significant results in each arm was reported.

Mechanistic analyses

Scatterplots and correlations between the change in PBC-40 fatigue domain score over 3 months and change in five bioenergetics outcomes were produced (see Figure 4).

All statistical analyses were performed using the statistical software package Stata^® (version 14.1; StataCorp LP, College Station, TX, USA).

Protocol specifications

For purposes of the specific study protocol:

• All non-SARs were recorded at visits 2–19.

• Any SAEs were recorded throughout the duration of the trial until 12 months after trial therapy was stopped

• The SAEs excluded any pre-planned hospitalisations (e.g. elective surgery) that were not associated with clinical deterioration.

• The SAEs excluded routine treatment or monitoring of the studied indication that was not associated with any deterioration in condition.

• The SAEs excluded fatigue (primary outcome measure, already documented and monitored within study).

Adverse reactions during treatment administration

During the infusion, patients were monitored clinically for the onset of clinical features of IRRs (the most common AEs in rituximab usage). Patients who developed evidence of severe IRRs, especially severe dyspnoea, bronchospasm or hypoxia, would have had the reaction managed as outlined in Adverse reactions during treatment administration. In practice, this did not occur.

Recording and reporting of serious adverse events or reactions

All AEs were reported. Depending on the nature of the event, the reporting procedures below were followed (or would have been followed if relevant).

Serious adverse events and reactions (including suspected unexpected serious adverse reactions)

All SAEs, SARs and SUSARs during drug treatment were reported to the chief investigator within 24 hours of the site being notified of its occurrence. The initial report was made by via the secure fax to e-mail reporting system (Soho66, Sunderland UK), which generates a copy for the chief investigator, trial manager, senior trial manager and sponsor. In the case of incomplete information at the time of initial reporting, all appropriate information was provided as follow-up as soon as this became available. The relationship between the SAE and the treatment was assessed by the investigator at the site, as was the expected or unexpected nature of any SARs.

The MHRA and main REC were to be notified by the chief investigator or NCTU (on behalf of the sponsor) of all SUSARs occurring during the study in accordance with the following time lines: fatal and life-threatening within 7 days of notification, and non-life-threatening within 15 days. No SUSARs, however, occurred.

Baseline data

The baseline clinical data are outlined in Tables 1 and 2. The study population was predominantly female, in keeping with the established demographic profile of PBC. 12 The distribution of UK-PBC risk score data at baseline suggested a low overall risk, in keeping with the established observation that fatigue severity in PBC is unrelated to the severity of underlying disease. Participants were predominantly of white ethnicity and most were not current smokers. More than two-thirds of participants had been managed at Newcastle Clinics for Research and Service in Themed Assessment centre for at least 1 year. The median age was ≈54 years and typical body mass index scores were relatively high, indicating that there were some potentially overweight participants in each arm. Alcohol consumption was modest in both arms and well below the recommended weekly intake. The distributions were well balanced between the intervention and control arms, with the exception of the proportion of UDCA responders and UK-PBC risk score at 10 years prediction.

Primary outcome

Data for the primary outcome, fatigue severity assessed at 3 months using the PBC-40 fatigue domain, are summarised in Table 3. The primary comparison was at 3 months, in which the mean PBC-40 fatigue scores were 36.2 in the rituximab arm and 38.1 in the placebo arm. This represents a decrease in score from baseline of around 5 units in each arm. The unadjusted mean PBC-40 fatigue score was 1.9 units lower in the rituximab arm at 3 months, but when the distribution of covariates was taken into account, the adjusted mean difference was –0.9 (95% CI –4.6 to 3.1). Although the 95% CIs are wide, they do not extend beyond the clinically important difference of 5.

There was little difference in mean fatigue scores between trial arms at any of the time points. The time course of the comparison between the intervention and control arms over the 12-month follow-up period was assessed for the primary outcome (PBC-40 fatigue domain) using repeated measures ANOVA at the time points of baseline and 3, 6, 9 and 12 months. There was no significant difference between the two trial arms (F = 1.81; p = 0.18) or in the interaction between arms over time (F = 0.41; p = 0.80).

Secondary outcomes

The prespecified secondary outcome data are presented in the following sections by category.

Bioenergetics

We used a combination of MR spectroscopy of muscle and anaerobic threshold assessment using CPET to calculate changes in pH and recovery time post exercise (Table 10).

TABLE 10 - Bioenergetics outcomes at baseline and 3 months by trial arm

N/A, not applicable.

Unadjusted comparison of means.

95% CI for adjusted mean difference with adjustment for baseline measurement. AUC and recovery time analyses were additionally adjusted for age in years at randomisation and recovery time also included predicted UK-PBC risk score at 10 years in final model.

Bioenergetics outcome, time point	Trial arm								Difference at 3 months [adjusted difference in means (rituximab – placebo) (95% CI)b]
	Rituximab				Placebo
	n	Meana (SD)	Median (IQR)	Range	n	Meana (SD)	Median (IQR)	Range
pH resting
Baseline	18	7.024 (0.023)	7.01 (7.009–7.040)	6.991–7.074	23	7.015 (0.022)	7.01 (6.996–7.032)	6.977–7.053	N/A
3 months	19	7.023 (0.019)	7.022 (7.012–7.037)	6.993–7.063	24	7.017 (0.021)	7.017 (7.007–7.029)	6.970–7.058
Minimum pH post exercise
Baseline	18	6.75 (0.40)	6.93 (6.67–6.99)	5.73–7.04	23	6.84 (0.36)	6.98 (6.80–7.00)	5.28–7.01	–0.1 (–0.29 to 0.04)
3 months	19	6.81 (0.32)	6.97 (6.64–7.01)	5.82–7.04	24	6.90 (0.11)	6.95 (6.81–6.98)	6.61–7.07
Recovery time (seconds)
Baseline	18	24.6 (32.0)	11.5 (0–40)	0–120	23	60.8 (96.9)	20 (0–70)	0–310	13.5 (–16.9 to 35.0)
3 months	19	63.7 (78.6)	50 (0–80)	0–290	24	52.1 (86.3)	20 (0–50)	0–320
pH fall with exercise
Baseline	18	0.28 (0.39)	0.09 (0.03–0.40)	0.01–1.27	24	0.1 (0.36)	0.04 (0.02–0.21)	0–1.76	0.10 (–0.04 to 0.30)
3 months	19	0.21 (0.31)	0.05 (0.02–0.38)	0.004–1.19	24	0.12 (0.11)	0.09 (0.02–0.20)	0.01–0.40
AUC for pH
Baseline	18	59.60 (82.7)	19.96 (5.08–84.31)	2.41–261.7	24	41.76 (84.2)	8.95 (4.17–48.7)	0–405.8	26.9 (–11.6 to 75.74)
3 months	19	56.45 (84.47)	11.76 (3.42–112.7)	0.66–308.99	24	31.51 (38.03)	20.64 (3.39–43.06)	0.93–153.4
Anaerobic threshold
Baseline	28	11.32 (2.70)	11.5 (9.5–13)	7–18	28	10.89 (2.20)	11 (9–13)	7–15	1.41 (0.03 to 2.80)
3 months	23	12.70 (2.88)	13 (10–15)	8–19	22	10.86 (2.49)	10 (9–13)	7–17

In keeping with previous reports,47 anaerobic threshold at baseline was low in the PBC patient group. The mean values rose in the rituximab arm from baseline to 3-month follow-up, with no change in the placebo arm. The anaerobic threshold score at 3 months was higher in the rituximab arm than in the placebo arm, which was also the case after adjustment for baseline values (adjusted difference 1.41, 95% CI 0.03 to 2.80).

In keeping with our previous reported findings, although muscle pH values at rest were within the normal range, the minimum pH following the specific exercise task was highly variable among the PBC patients, with substantial acidosis seen in some patients. Minimum pH in muscle seen following exercise was slightly higher at 3 months than at baseline, but this was seen in both trial arms. No reduction in the time taken to recover to baseline pH after exercise and no reduction in the AUC for pH (a factor combining the degree of acidosis and the length of time taken to recover to the baseline level and an estimate therefore of degree of muscle intracellular acid exposure) were seen in either trial arm. Comparison of bioenergetics outcomes between trial arms shows that the adjusted differences in means at 3 months were all small with relatively narrow 95% CIs for the pH parameters; however, the 95% CI was wide for the comparison of AUC.

Plots of the changes in bioenergetics outcomes against changes in fatigue scores are shown in Figure 4 and the correlations are given in Table 11. The changes were frequently close to zero, with the exception of the anaerobic threshold: there were more non-zero changes in the rituximab arm, but these were not consistently in one direction. The correlation coefficients were all close to zero. There was little evidence that improved fatigue scores are associated with better bioenergetics outcomes. However, sample sizes in each arm were small and correlations of combined arms are based on just 41–45 participants, depending on outcome measure.

FIGURE 4.

Change in muscle bioenergetics outcomes compared with change in fatigue score. (a) Change in minimum pH post exercise; (b) change in AUC; (c) change in recovery time; (d) change in pH post exercise; and (e) change in anaerobic threshold.

TABLE 11 - Correlations (Pearson’s r) between changes in PBC-40 fatigue domain score and changes in bioenergetics outcomes by trial arm

Change in PBC-40 fatigue score, correlation with:	Trial arm						Overall
	Rituximab			Placebo
	n	Correlation	95% CI	n	Correlation	95% CI	n	Correlation	95% CI
Minimum pH post exercise	18	0.034	–0.440 to 0.493	23	–0.062	–0.463 to 0.359	41	–0.013	–0.319 to 0.296
pH fall with exercise	18	–0.044	–0.501 to 0.432	24	0.065	–0.347 to 0.457	42	0.011	–0.294 to 0.314
Recovery time (seconds)	18	–0.443	–0.754 to 0.030	23	–0.059	–0.460 to 0.362	41	–0.194	–0.474 to 0.121
AUC for pH	18	–0.082	–0.529 to 0.400	24	0.075	–0.338 to 0.465	42	–0.007	–0.310 to 0.297
Anaerobic threshold	23	–0.119	–0.507 to 0.308	22	0.416	–0.006 to 0.713	45	0.117	–0.183 to 0.396

Impact on biochemical disease parameters

The B-cell-depleting therapy (RITuximab) as a treatment for fatigue in Primary Biliary Cirrhosis (RITPBC) trial was not designed or powered to explore the impact of the drug on liver injury. The cohort of treated patients, however, exceeds the total world experience of PBC patients treated with biological agents to date. Furthermore, because the primary target was fatigue, a non-stage-associated symptom, treated patients were typically early in the disease course. This contrasts with other trials of biological agents, which have typically been performed in patients with advanced/aggressive disease. The RITPBC trial therefore presents a unique insight into the impact of rituximab therapy in early disease, which could inform future trials of disease-modifying therapy. It is for this reason that the biochemical results are reported in Tables 12–15.

Baseline biochemical parameters were typical of early-stage PBC, with median ALP values of 136 U/l and 131 U/l in the rituximab and placebo groups, respectively. Subsequent to the design of the trial, definitions have been developed for the risk status for PBC patients. These include dichotomous scoring systems such as the Toronto Criteria and continuous variable scoring systems such as the UK-PBC risk score. 4,62 Over the 12-month period of assessment in the placebo group, a typical PBC natural history pattern was seen, with small increases in the ALT and bilirubin levels (the latter of which were within the normal range). There were no significant differences between trial arms or in the interaction between arms over time in repeated measure ANOVA up to 12 months for bilirubin, but there were significant differences between trial arms (F = 6.17; p = 0.016) and in the interaction between arms over time (F = 2.91; p = 0.023) for ALP outcome (see Tables 12 and 13).

Median ALP level fell from 136 U/l to 106 U/l over 3 months in the rituximab arm, whereas the median ALP level rose from 131 U/l to 137 U/l in the placebo arm. Ninety-three per cent of patients in the rituximab arm had a normal ALP level at 3 months, compared with 65% in the placebo arm (Table 16). All parameters progressively returned to baseline levels by 12 months of follow-up.

There were no notable differences between the rituximab and placebo arms for any lipid parameters, creatinine concentration or serum electrolyte levels or any change from baseline measurements up to 12 months (see Tables 14 and 15).

Severe adverse events

For SAEs, we collected information on the number of patients in each trial arm who experienced an event and the number of patients assessed (see Table 17).