This was a three-part (Parts A, B and C), multicentre study to investigate the safety, tolerability, efficacy, pharmacokinetics and pharmacodynamics of intravenous (IV) GSK315234 in Parts A and B and subcutaneous (SC) GSK315234 in Part C in patients with RA. Parts A and B were randomised, double-blind, placebo-controlled, Bayesian adaptive dose-finding studies to investigate the effect of single (Part A) and three repeat (Part B) IV infusions of GSK315234 in patients with active RA on a background of MTX. Part C was a single dose, randomised, single-blind, placebo-controlled study of SC administered GSK315234 in patients with active RA on a background of MTX.
Patients between 18 and 75 years of age, who fulfill 1987 American College of Rheumatology (ACR) classification criteria of RA were recruited . They must have had active disease: Disease Activity Score 28 (DAS28) of >4.2 at screening and a pre-dose C-reactive protein (CRP) level of ≥0.5 mg/dl or an erythrocyte sedimentation rate (ESR) level ≥28 mm/hour at screening and pre-dose. Patients should have received at least three months of MTX and have been on a stable dose (up to 25 mg/week) for at least eight weeks prior to screening and be willing to remain on this dose throughout the study. Concomitant sulfasalazine or anti-malarial was permissible if it was taken in addition to MTX, and the dose was stable for at least four weeks for sulfasalazine and three months for anti-malarial prior to screening. Other DMARDs must have been withdrawn for more than one month prior to screening. Other oral anti-rheumatic therapies, such as non-steroidal anti-inflammatory drugs (NSAIDs), COX-2 inhibitors, oral glucocorticoids, were permitted providing the dose is ≤10mg/day of prednisolone (or equivalent) and stable for at least four weeks prior to screening and remains unchanged through the study. Patients must use acceptable contraception during the course of the study. Patients were excluded if they had received prior biologic therapy, have active infection, previous exposure or past infection caused by Mycobacterium tuberculosis, positive Hepatitis B surface antigen or Hepatitis C antibody result at screening, history of HIV or other immunodeficiency disease, history of malignancy, except for adequately treated non-invasive cancers of the skin (basal or squamous cell) or carcinoma in situ of the uterine cervix, positive pregnancy test, elevated liver function tests on more than one occasion: transaminases or alkaline phosphatase >3 times the upper limit of normal (ULN) or total bilirubin >1.5 times ULN, or any significant medical conditions. Patients with haemoglobin (Hb) <10 g/dl or platelet count <150 × 109/l were also excluded.
The study was approved by Institution Review Boards (Serbia: Institute of Rhematology Belgrade, Institute for Prevention and Treatment and Rehabilitation of Rheumatic and Cardiovascular Diseases Belgrade; Ukraine: Zaporizhzhya State Medical University, Keiv National Medical University, National Scientific Center, Institute Kiev, Keiv National Medical University, Donetsk State Medical University, Lviv National Medical University; New Zealand: Waikato Hospital, Hamilton, Christchurch Clinical Studies Trust Christchurch, P3 Research Wellington; Russia: State Research Institute Novosibirsk, Yaroslavi Regional Clinical Hospital, Yaroslavi; Smolensk Regional Clinical Hospital, Moscow Regional Research Clinical Institute Rhematology Institute Moscow; Australia: Nucleus Network Melbourne, Austin Center for Clinical Studies Heidelberg, Princess Alexandra Hospital Wooloongabba) and Charing Cross Research Ethics Committee, UK and registered (ClinicalTrials.gov no: NCT00674635). All patients gave signed informed consent. The study was conducted in accordance with the guiding principles of the Declaration of Helsinki.
Dosing and study design
In Part A, initially, six cohorts of eight patients each were enrolled (Cohorts 1 through 6). After the first interim analysis, an additional two cohorts of patients were enrolled (Cohort 7 and Cohort 8). Eligible patients within each cohort were randomized to GSK315234 (n = 6) or placebo (n = 2). A starting dose of 0.03 mg/kg GSK315234 was identified, and a Bayesian adaptive dose-finding algorithm based on a measure of clinical response (weighted mean DAS28) on Day 14 post-dose was used to identify subsequent doses that provided 90% of maximal benefit based on trial simulation of the Bayesian adaptive pharmacokinetics and pharmacodynamics (PK/PD) design. Patients in Cohorts 1 through 6 received 0.03 mg/kg, 0.3 mg/kg, 3 mg/kg (2 cohorts of patients were enrolled at this dose level), 10 mg/kg and 30 mg/kg of GSK315234; doses were administered in a dose rising fashion. Cohorts 2 through 6 were dosed a minimum of three weeks after dosing of the last patient in the previous cohort. Cohorts 7 and 8 enrolled simultaneously, and patients received 10 mg/kg or 20 mg/kg GSK315234.
Part B was a randomized, double-blind, placebo-controlled, repeat dose study based on changes in DAS28 and PK in Part A. Prior to administration of the first dose, eligible patients (n = 54) were randomized in a 2:1 ratio to receive GSK315234 (n = 37) or placebo (n = 17). For each patient, doses were administered approximately four weeks apart.
In Parts A and B, GSK315234 or placebo was administered by slow IV infusion over two hours.
Part C was a randomized, single-blind, placebo-controlled, single SC dose study. Eligible patients (n = 17) were randomized on a 3:1 basis to GSK315234A (n = 12) or placebo (n = 5). One patient in the placebo arm was randomized and dosed but was withdrawn as the DAS28 score was lower than 4.2 at pre-dose on day 1. Patients were administered 500 mg of GSK315234 or matching placebo as five SC injections of 1 mL each (multiple injections were needed to administer the full dose). SC injections were administered on the abdomen, rotating sites around the umbilicus.
A central randomization schedule generated using the GSK Randall system was used in all parts of the study. There is no stratification of sites or countries. GSK315234 or placebo was administered in a blinded fashion so that both patients and investigators remained blinded to treatment allocation.
The primary efficacy endpoint was the DAS28  response rate on Day 28 in Part A and Day 56 in Parts B and C. Secondary endpoints were ACR  and European League Against Rheumatism (EULAR) response rates , together with Outcome Measure in Rheumatology (OMERACT) core component measures : tender/painful count; swollen joint count; Patient’s Assessment of Arthritis Pain (100 mm visual analogue scale (VAS)); Patient’s and Physician’s Global Assessments of Arthritis (100 mm VAS); Health Assessment Questionnaire-Disability Index (HAQ-DI)  and Multi-dimensional Assessment of Fatigue (MAF). Laboratory efficacy measures included CRP and ESR.
Safety assessments including adverse events (AEs), vital signs, electrocardiograms and clinical laboratory tests (haematology, biochemistry and urinalysis) were carried out at each study visit.
Pharmacokinetics and pharmacodynamics
PD biomarkers after single and repeat IV doses included, but were not limited to, serum OSM and GSK315234A-OSM complexes. Immunogenicity was measured by human anti-GSK315234A antibodies.
Sample size estimation and sensitivity
In Part A, the use of a non-linear mixed effects procedure required simulation techniques to estimate power for a given sample size and expected magnitude of effect. Trial simulations of the Bayesian adaptive PK/PD design using a nonlinear PK/PD model were conducted using typical parameter estimates that six cohorts of eight patients each would provide power in excess of 95% to detect a PK/PD maximum effect value of 66% inhibition from baseline. The probability of a false positive under the null hypothesis was approximately 5%. When the number of cohorts was increased to eight, the power increased marginally accompanied by a steeper cumulative distribution curve. When the magnitude of response was 33%, the power reduced to 80%. If the response to GSK315234 had a slower onset but similar sized response to adalimumab, the power for the PK/PD analysis was 80%. Overall, a sample size of 48 (six cohorts) or 64 (eight cohorts) would provide at least 80% power assuming a similar response to adalimumab.
In Part B, a maximum of 54 patients was planned for enrollment. A treatment difference of 0.95 between the selected dose and placebo in DAS28 scores 56 days post dose could be detected with approximately 90% power based on preliminary estimates of between subject variability of DAS28 scores seen in the interim analysis of Part A. This assumes a standard deviation of 1.15 in the GSK315234 dose group and 1.25 in the placebo dose group, a two-sided test and an overall alpha of 5%.
In Part C, no statistical techniques were used to determine the sample size.
Statistical analysis plan
A repeated measure analysis using a mixed effect model was used, including treatment, visit, and treatment by visit interaction as fixed effects and patient as a random effect to analyse the primary efficacy endpoint (mean change from baseline in DAS28 scores at Day 28). Other effects such as baseline, baseline by visit, country, gender, age and baseline OSM level were fitted into the model when deemed necessary.
A Bayesian normal dynamic linear model was applied to DAS28 data of evaluable patients to estimate the dose–response relationship at Day 28, Day 56, and Day 84 in Part A.
Descriptive statistical methods were used to summarize all primary and secondary efficacy variables (absolute value and change from baseline). Significance tests were carried out at the 2-sided 5% level. Changes from baseline in tender joint count, swollen joint count, patient’s pain assessment, patient’s and physician’s global assessment, CRP, ESR, HAQ-DI and MAF were compared between treatment groups. If the assumption of normality was not satisfied for these data, then the data were transformed prior to analysis.
Weighted mean DAS28 score
The weighted DAS28 was defined by the area under the curve (AUC) divided by the number of days. The weighted mean DAS28 was calculated and summary statistics are presented for Parts A, B and C. The difference in the mean of the two treatment groups is presented together with the corresponding 2-sided 95% confidence interval and the 2-sided P-value.
For EULAR, ACR20, ACR50 and ACR70 response rates, the Cochran Mantel Haenszel (CMH) test was used for comparing the responder rates at each visit for GSK315234A versus placebo stratified by subgroup factors.