Patient inclusion/exclusion criteria

Sixty patients were recruited from the outpatient rheumatology clinic of the Ghent University Hospital between May 2006 and January 2008. Patients were eligible for study if: (1) they were 18 years or older; (2) had hand osteoarthritis (meeting the American College of Rheumatology criteria)20 characterised by painful, inflammatory episodes of the interphalangeal joints; (3) presented with at least one interphalangeal finger joint in the ‘E’ phase as defined by Verbruggen and Veys7 on radiography; and (4) were willing to self-administer subcutaneous injections or allow a suitable person to perform this.

Patients were excluded from the study if they had received previous treatment with any investigational agent within 30 days (or five half lives of the product when longer). Previous treatment with chondroitin sulfate, glucosamine, avocado-soybean unsaponifiables, tetracyclines, corticosteroids or any immunomodulating drug with possible effects on pro-inflammatory cytokine metabolism within 90 days was another reason for exclusion. Patients with chronic inflammatory rheumatic disease (eg, RA, spondylarthropathy, psoriatic arthritis, gout, chondrocalcinosis or other autoimmune diseases) were excluded as well as underlying comorbidities, eg, uncontrolled diabetes, unstable ischaemic heart disease, congestive heart failure, active inflammatory bowel disease, recent stroke (within 3 months before screening), chronic leg ulcer, lymphoproliferative disorders, history of cancer, active hepatitis B or C infection, positive HIV status and persistent or recurrent infections requiring treatment with antibiotics within 30 days before enrolment.

Levels of rheumatoid factor, anti-citrullinated cyclic peptide, erythrocyte sedimentation rate and C-reactive protein were measured at screening. Screening for tuberculosis was done by purified protein derivate skin test and x-ray of the chest.

This investigator-initiated study was declared at www.ClinicalTrials.gov (EudraCT number 2006-000925-71) and complied with the principles of the Declaration of Helsinki. The study protocol was approved by the Ghent University Hospital Ethics Committee. All patients provided written informed consent.

Expected intergroup difference and sample size calculation

The sample size was calculated based on the results from previous studies7 21 22 and is applicable for the evolution between baseline and 1 year. The number of patients required for this study was 20 in both treatment arms. It can be estimated that after 1 year, 60% of patients in the placebo group and 10% of patients in the adalimumab group will have started the erosive phase in their ‘S’ or ‘J’ joints. Taking into account 20 patients in each subgroup, a power of 80% and an α-level of 5%, the present study should be able to detect a significant difference between the placebo and treatment groups when 50–55% of the treated patients have started the erosive phase in one or more non-erosive interphalangeal joints. Considering a drop-out rate comparable with the previous studies22 a total of 60 patients were included.

Randomisation and treatment

Patients received 40 mg adalimumab or placebo administered subcutaneously every 2 weeks for 52 weeks. Randomisation of the study medication was done in by the subsidising company and successive treatment allocation numbers were administered by the investigators following the order of inclusion.

Stable doses of analgesics and anti-inflammatory drugs were allowed as adjuvant therapy. In the case of latent tuberculosis or the presence of other risk factors for activation of latent tuberculosis, the patient was treated with tuberculostatic therapy (isoniazid treatment) at least 3 months before initiation of the trial and continued up to 9 months.

Outcomes

Primary endpoints were the control of structural damage on radiography. Therefore, posteroanterior radiographs of the hands were obtained at baseline and after 26 and 52 weeks. All interphalangeal joints (distal interphalangeal 2–5 and proximal interphalangeal 2–5) of both hands were scored except the interphalangeal joints of the thumb because of decreased visibility and reproducibility to evaluate on posteroanterior views. A total of 960 interphalangeal joints was scored. In order to score the erosive progression of the interphalangeal joints on the radiographs, two validated scoring systems were used as co-primary endpoints. The first scoring system was based on the consecutive anatomical phases recognised during progressive disease in erosive hand osteoarthritis.7 Non-erosive osteoarthritis joints (‘S’ or stationary osteoarthritis) have been shown to enter the ‘J’ phase when the joint space disappears, followed by the ‘E’ phase when manifest erosive changes occur. Next, the affected interphalangeal joint show signs of repair and the ‘R’ (remodelling) phase ensues. A fusion (‘F’), as an extreme form of joint remodelling, is seen extremely rarely. Characteristic features and changes are illustrated in figure 1. Anatomical phases of all interphalangeal joints were characterised and changes in phase were recorded by two readers. The percentage of absolute agreement between the readers was 93.6% (κ=0.92). Both readers showed intrareader reliabilities with a percentage of absolute agreement of over 95.9% (κ=0.95). In case of dispute, agreement was sought by discussing the radiographs. The primary endpoint was a reduction in the numbers of ‘N’, ‘S’ or ‘J’ joints that become ‘E’ within 1 year at the joint and at the patient level.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Clinical appearance of erosive osteoarthritis of the interphalangeal finger joints is illustrated in (A) where erosive progression and subsequent remodelling caused functional inability of the second, third and fifth digit, and in (B) where almost all interphalangeal joints of both hands were affected and disabled. Remodelling of the interphalangeal joints in (B) after the erosive process is shown on the corresponding radiograph (C). Typically, metacarpophalangeal joints are spared in this condition. (D and E) Erosive progression and subsequent remodelling in two different proximal interphalangeal (PIP) joints is shown in a series of radiographic images taken at 6-month intervals. PIP joints in which the synovial joint space has disappeared (‘J’ phase) pass through the erosive ‘E’ phase and eventually remodel (‘R’).

The second scoring system was the quantitative Ghent University score system (GUSS).23 Erosive progression and signs of repair or remodelling were scored by indicating the proportion of normal subchondral bone, subchondral plate and joint space on a 11-point rating scale (range 0–100 with intervals of 10 units). The three scores were summed to obtain the total GUSS score that ranged from 0 (severe joint destruction) to 300 (no damage) for each interphalangeal finger joint. ‘E’ phase interphalangeal joints thus have, on average, lower GUSS scores than ‘J’ or ‘R’ joints.

Two readers (RW, GV) scored the films independently blinded for order, treatment assignment and clinical response. Interreader and intrareader intraclass coefficients of correlation scores rated excellent for the total GUSS scores: 0.86 and 0.97, respectively. The smallest detectable change,24 as calculated on the mean scores of the two readers was reduced to 18 units after training during routine in-house training sessions. The mean scores of the two readers were used for the GUSS readings.

Secondary endpoints consisted of effects on clinical variables. Clinical assessment was performed by the same experienced examiner, blinded for treatment assignment, at screening, baseline and after 6, 12, 26, 39 and 52 weeks. This included the evaluation of pain on palpation and the search for palpable swelling as a result of synovial thickening and intra-articular effusion25 in 16 interphalangeal finger joints. The grip strength of both hands was measured using a hand grip dynamometer (My-Gripper; Yamasa, Tokei, Japan). The maximum of three attempts was recorded for the dominant and non-dominant hand. Level of pain, morning stiffness and function was registered by the self-reported AUSCAN questionnaire.26 The occurrence of any adverse events was questioned at every visit. Blood samples were obtained at baseline and at 6 and 12 months for routine safety tests.

Statistical analysis

Descriptive analyses were performed by the calculation of means, mean differences and proportions. t Tests and χ² tests were used to explore differences between groups for continuous and dichotomous variables, respectively. For the evaluation of joints, the classic methods neglect the fact that treatment is given on a patient level. Therefore, additional generalised estimating equation (GEE) modelling with a logit link function was used to calculate OR and differences of means. An additional level (time) was added to the GEE models for longitudinal analysis.

A multi-step method was used to evaluate subgroups of patients that might benefit from anti-TNFα therapy: (1) potential predictors of erosive progression were evaluated by univariate analysis; (2) variables that showed a 50% difference in odds (OR >1.5 or OR <0.667) were added in one multivariate GEE model; (3) significant predictors from that model were evaluated by interaction terms with therapy (placebo or adalimumab) in further regression analyses. For these subgroup analyses, disease progression was assessed using the anatomical phase scoring system7 and the GUSS23 as follows: for the evaluation of classic calculated OR, a new erosive joint was defined if a joint with a baseline ‘N’, ‘S’ or ‘J’ phase evolved to an ‘E’ phase after 1 year. In the GEE models, ‘E’ joints were compared with all other phases at the two different time points. For the evaluation of the GUSS, differences between baseline were calculated and the different scores of these joints showing no evolution were imputed with zero.

Concerning the descriptive analyses and the analysis of differences, the last observation carried forward method was used in the case of missing values (one patient (placebo) missed the 12-month visit). For the calculation of inference by GEE modelling, no imputation was used. All statistics were calculated using SPSS 16.