Permutated blocked randomization (blocks of 4) with stratification for center, age (<50 years and >50 years), and sex, made up by a random digit generator, was used to allocate the patients to the intervention groups, consisting of either a high-intensity exercise program (RAPIT) or usual care physical therapy. An administrative assistant, who was not aware of the block size, allocated the interventions.
In March/April 1998, the patients randomized to the RAPIT group started participation in a supervised, twice-weekly, group exercise program of 1¼ hours each session. This exercise program was aimed at increasing and maintaining cardiovascular and muscle fitness (strength and endurance) and included both weight-bearing and impact-loading exercises. Overall, each session had 3 parts: bicycle training (20 minutes), exercise circuit (20 minutes), and sport or game (20 minutes). Each training component was preceded by a warm up that included impact-bearing exercises, such as stepping aside and stair walking, and ended with a cool down period.
Bicycle training started with continuous vigorous bicycling (cycling frequency 50–70 revolutions per minute) for 5 minutes. Within 6 months, the duration of vigorous bicycling was increased to 18 minutes. Bicycle load was based on 2 indicators: 1) heart rate during bicycling, and 2) rating of perceived exertion (RPE) (17). Heart rate during the bicycle training had to be between ∼70% and 90% of the predicted maximal heart rate. Prediction of maximal heart rate was calculated with the formula 220 − age. A rating of 4–5 on the RPE (range 0–10) was expected to reflect an effort with enough intensity to improve cardiovascular fitness (18, 19).
The exercise circuit consisted of 8–10 different exercises intended to improve muscle strength, aerobic fitness, muscle endurance, joint mobility, and activities of daily living (for example, walking, turning around in bed, getting up, lifting). Exercises were interspersed with rest. The proportions of exercise duration and rest duration changed from 90 seconds and 60 seconds, respectively, in the first weeks of the program to 90 seconds and 30 seconds, respectively, after 6 months. Each exercise was repeated ∼8–15 times. The exercises within the exercise circuit were changed every 8 weeks.
The sport or game component consisted of badminton, volleyball (with a softer ball, if necessary), indoor soccer (with a low-weight ball), basketball, relay games, or pat-catch games. If necessary, the program was adapted to individual limitations to achieve the same training goals. A group of supervisors from all participating centers met every 8 weeks to evaluate the actual training circuit and to develop a new circuit and to ensure the uniformity of the exercise program in all centers.
Patients assigned to the usual care group were treated by a physical therapist only if deemed necessary by their attending physician. In both groups, the physicians had free choice with respect to medical prescriptions and any other treatment strategies, including additional individual physical therapy.
The number of RAPIT group sessions attended was recorded and expressed as a percentage of the maximum possible number of attended sessions. Attendance at any group or individual physical therapy sessions other than those of the RAPIT group was recorded in both intervention groups.
The attending physicians were informed about the treatment allocation and the results of each BMD measurement, including a suggestion for treatment with bisphosphonates as indicated by the current Dutch guidelines. The guidelines recommend treatment with bisphosphonates for patients with osteopenia (T score less than or equal to −1.0) who are currently taking prednisone and/or for any patient with osteoporosis (T score less than or equal to −2.5).
At baseline, sociodemographic characteristics were registered, as well as postmenopausal status (yes/no), body mass index, disease duration, presence of rheumatoid factor (yes/no), ACR functional class (16), past use of DMARDs (number) and glucocorticoids (ever/never), and current use of medication.
At baseline and every 3 months thereafter, the participants reported on their use of medication in the past 3 months, and their disease activity and physical capacity were assessed. BMD and radiologic joint damage were assessed at baseline and at 12 and 24 months.
The use of medication, such as oral glucocorticoids, hormone replacement therapy, bisphosphonates, and vitamin D and/or calcium supplementation, was expressed as the number of months that the patients reported the use of the respective drug. Since the trial lasted 24 months, the duration of use of each drug could therefore vary from 0 to 24 months. The use of DMARDs at each followup visit was compared with the use at the previous visit and coded by a clinician (ZdJ) into 3 categories: 0 = change of DMARD or change in dosage of current DMARD due to decreased disease activity, 1 = no change in DMARD or dosage of current DMARD, and 2 = change of DMARD or change in dosage of current DMARD due to increased disease activity. The numbers were summed up over the whole study period and thus could, after 2 years, vary from 16 (maximum increase due to increasing disease activity) to 0 (maximum decrease due to decreasing disease activity during the 2 years).
Disease activity was assessed with the original Disease Activity Score with 4 variables (DAS4). The DAS4 is a compiled index based on the number of swollen joints (maximum 44), tender joint score (Ritchie Articular Index), Westergren erythrocyte sedimentation rate, and the patient's global assessment of disease activity measured on a visual analog scale ranging from 0 (asymptomatic) to 100 (worst possible disease activity) (20). The DAS4 ranges from 0 (no disease activity) to 10 (severe disease activity). To express the change in disease activity during the 2-year study period, the area under the curve (AUC) was calculated from the combined 3-month assessments of disease activity.
Physical capacity was determined by aerobic fitness and muscle strength. Aerobic fitness was measured by a standardized ergometer test, with results expressed in watts (21). Muscle strength of the knee extensors was measured by an isokinetic dynamometer at 60°/second, with results expressed in newtons (22).
All clinical outcome assessments were done by 4 research physical therapists who were blinded to the treatment allocation and who were trained thoroughly both before the trial and after 1 year. During the study, the majority of the participants were assessed by the same assessor. A reproducibility study in 19 successive patients was performed at the last study visit and 2 weeks afterward by the same assessor, which yielded intrarater intraclass correlation coefficients (ICCs) for aerobic fitness, muscle strength, swollen joint count, and Ritchie Articular Index of 0.97, 0.98, 0.83, and 0.92, respectively.
Functional ability was assessed with the Health Assessment Questionnaire (HAQ) (23, 24). The HAQ score was assessed at baseline and every 6 months thereafter. The total HAQ score ranges from 0 (no functional limitations) to 3 (serious functional limitations). The change score can thus vary from −3 (maximal improvement) to +3 (maximal deterioration).
BMD measurements of the hip (the total hip region) and of the lumbar spine (involving L1–L4) were carried out using a dual-photon X-ray absorptiometer (DXA), with results expressed in gm/cm2. The scanning and standard quality procedures were followed. All measurements in each patient were performed using the same DXA. The in vitro reproducibility, expressed as the coefficient of variation, was 3.7% and 2.6% for the hip and 0.8% and 0.3% for the spine by Hologic QDR-2000 (Amsterdam, The Netherlands) and Hologic QDR-4500 (Leiden and The Hague, The Netherlands), respectively.
Radiologic damage of the small joints (the hands and feet) was assessed using the method devised by Larsen and modified by Scott et al (25). The following joints were assessed: the 10 proximal interphalangeal joints, the 10 metacarpophalangeal joints, the wrists (scored as one unit and multiplied by 5), the second to fifth metatarsophalangeal joints, and the first interphalangeal joints. According to this method, the Larsen score of the small joints varies from 0 (no joint space narrowing, no erosions) to 200 (maximum possible damage to the joints) and is a sum of both joint space narrowing and erosions of the hands and feet. All radiographs were scored by a single observer (AC) who was given no information about the chronology, patient's identity, and group allocation. The ICC for radiologic damage of the small joints, based on repeated readings of the 15 radiographs of the small joints by a single observer, was 0.97.
Radiologic damage of the large joints (the shoulders, elbows, hips, knees, ankles, and subtalar joints) was scored independently by 2 experienced readers (HK and ZdJ) also using the Larsen method (26). The Larsen score of the large joints ranges from 0 (no joint space narrowing, no erosions) to 60 (maximal possible damage to the large joints). The Larsen score presented is the mean of the scores of the 2 readers. The ICC for radiologic damage of the large joints, based on all readings of the 2 readers, was 0.95.
The target sample size was based on an estimated mean ± SD loss of BMD in RA patients of 4.12 ± 6.25% over 2 years (27). To demonstrate a complete inhibition of bone loss by the high-intensity exercise when compared with usual care, and based on 90% power to detect a significant difference (2-sided P = 0.05), a sample size of 48 patients would be required in each study group. To compensate for an expected dropout rate of ∼20%, we needed to enroll at least 58 patients in each study group. However, the RAPIT study was primarily designed with the HAQ score as a primary outcome variable. To detect a difference of 0.20 in the change of HAQ score (14), the target sample size was calculated at 150 patients per group. Thus, a final cohort of 300 patients was enrolled.
Final analyses, which were based on intention to treat as initially assigned, were performed on the data from the 281 patients who completed the study (designated “completers”). At all times, measures with a Gaussian distribution are expressed as the mean and SD, and measures with a non-Gaussian distribution are the median and interquartile range (IQR). Bone mass is expressed as BMD (in gm/cm2) and is also presented as T and Z scores. Osteoporosis was defined as a T score of less than or equal to −2.5 SD (28). Reduced bone mass was defined as a Z score of less than or equal to −1 SD.
To compare the effectiveness of the treatment over the total period of 2 years, BMD was further analyzed with a mixed-effects analysis of variance (ANOVA) model, with patient number as a random factor and treatment, time, and the treatment-by-time interaction as fixed effects. The effect analysis was performed after correction for the baseline differences.
The patient cohort was also analyzed according to individual bone loss. The individual changes in BMD values from baseline to followup (followup minus baseline) were divided in tertiles. The patients with the most BMD loss (lowest tertile) were designated as fast losers, while the patients with the least BMD loss or increased BMD (highest tertile) were gainers. The patients with slow loss (middle tertile) were designated as slow losers. Bivariate analyses (Mann-Whitney U test, Pearson's chi-square test) were used for comparisons between 2 groups, while Kruskal-Wallis test and Pearson's chi-square test were used for comparison between more than 2 groups.
Multivariate analyses were used to explore the relationships between the dependent variable, change in BMD at the total hip site, and the independent variables, the clinical characteristics that can be modified by treatment. The individual changes in muscle strength and aerobic fitness were dichotomized. Patients showing a decrease in muscle strength or aerobic fitness were designated as muscle strength or aerobic fitness losers, while patients showing an increase were gainers. The analyses were performed as a polytomous logistic regression model with BMD changes categorized in tertiles. Based on the results of the multivariate analysis and supposed clinical relevance, the variables were successively removed from the multiple regression model in a backward manner. All analyses were performed using SPSS program, version 10.0 (Chicago, IL). P values less than or equal to 0.05 were considered statistically significant.