Several explanations may be offered for the discrepancies in the literature. First, nonrandomized studies may be subject to selection bias, because RA itself, and its concomitant decreased mobility, may adversely affect bone mass [1]. Second, it has been documented that negative influences of glucocorticoid agents may be most pronounced in the first months of therapy [4], and studies evaluating bone loss in the chronic phase of treatment may thus underestimate the effects of these drugs on bone. Randomized, controlled studies are necessary to avoid selection bias. We report the results of the first randomized, placebo-controlled study of the initial effects of low-dose prednisone on bone mineral density in patients with active RA.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Patients were selected from a single university outpatient clinic, from April 1988 to April 1991. All patients with classical or definite RA [5] who started taking intramuscular gold salts were considered for participation in the trial. The decision to start intramuscular gold therapy was left to the rheumatologists in charge of the patients. Patients were admitted to the study only if they fulfilled at least three of five criteria: five or more joints that were tender or painful during motion; three or more swollen joints; an erythrocyte sedimentation rate more than 28 mm/h; morning stiffness for at least 60 minutes; and a hemoglobin level less than normal limits.

Patients with other diseases or medication, including calcium supplements and postmenopausal hormone replacement therapy, that might affect bone mass, were excluded from the study. Introduction of these medications during the study was not permitted. Premenopausal patients were allowed to continue taking oral contraceptive drugs. Female patients who were postmenopausal for fewer than 3 years, or who had irregular menstrual cycles, were also excluded from the study. All patients were white. Informed consent was obtained from all patients. The study was approved by the local Ethics Committee.

Study Design and Medication

All patients were treated with intramuscular aurothioglucose (Auromyose; 20% oily suspension, Noury Pharma, Oss, the Netherlands) according to a standard dosage regimen. All patients started with a test dose of 10 mg before the trial. They then continued with a dose of 50 mg weekly, for 20 weeks. Modification of this regimen was possible if side effects occurred. Aurothioglucose was discontinued after 20 weeks if it was not effective. If a satisfactory response occurred, the dose was decreased. Patients were monitored for side effects weekly during the first 8 weeks of treatment and once every 2 weeks thereafter.

The trial was a double-blind, placebo-controlled study for 20 weeks. Patients were followed for an additional 24 weeks after this period. All patients were randomly assigned to receive either prednisone or placebo treatment. No stratification or block randomization was used. All assessments were done without knowledge of the patients' treatment status. Prednisone was given in a dose of 10 mg once daily for 12 weeks. Thereafter, the dose was gradually reduced: 7.5 mg/d in weeks 13 and 14, 5.0 mg/d in weeks 15 and 16, 2.5 mg/d in weeks 17 and 18, and no prednisone in weeks 19 and 20. This resulted in a total dosage after 20 weeks of 1050 mg and a mean daily dose of 7.5 mg. All patients used nonsteroidal anti-inflammatory drugs throughout the study.

Assessments of Rheumatoid Arthritis

All clinical assessments were done by a single observer. Disease activity was measured by a composite index (disease activity score) consisting of the erythrocyte sedimentation rate (millimeters in first hour), the Ritchie articular index [6], the number of swollen joints, and a visual analogue scale for general health (range, 0 to 100 mm; a zero score indicating excellent health), which has been constructed and validated previously [7, 8]. Assessments were done at baseline; after 1, 2, and 4 weeks; and thereafter every 4 weeks.

Functional capacity was assessed at baseline and after 12, 20, and 44 weeks using a health-assessment questionnaire (range, 0 to 3; a zero score indicating no functional impairment) [9].

Assessments of Bone Mineral Density

Bone mineral density (BMD) was measured in the lumbar spine by dual-energy, quantitative computed tomography (effective radiation dose less than 300 microSievert) with a Somatom DR3 (Siemens AG, Erlangen, Germany). The details of this method have been published previously [10-12].

In short, patients were positioned with the lumbar spine over a calibration phantom, consisting of two stable plastics: a water-equivalent and a bone-equivalent standard containing 200 mg of hydroxyapatite per milliliter. Scans were made through the midvertebral level of lumbar vertebrae L2 to L4, with a slice thickness of 8 mm. Separate regions of interest were automatically defined for the trabecular bone and the anterior cortical bone of the vertebral body. The BMD was then calculated by comparison of the absorption in the region of interest with the absorption in the calibration phantom and was expressed in milligrams of hydroxyapatite per milliliter.

Measurements were made at 125 and 85 kilovolt (kVp), and three images were reconstructed from each scan: at 125 kVp, at 85 kVp, and a calcium-equivalent density image after processing by material decomposition techniques. Dual-energy quantitative computed tomography results were calculated from the calcium density image.

For duplicate measurements, the in-vivo coefficient of variation ranged between 6.1% for the trabecular bone and 5.2% for the cortical bone.

To exclude the possibility that new vertebral fractures caused changes in the BMD, lateral spine radiographs were obtained in all patients at baseline and after 44 weeks and were evaluated by an experienced radiologist.

Statistical Analysis

The change in BMD (expressed as percentage from baseline) was chosen as an outcome measurement. This variable followed a normal distribution sufficiently enough to allow the use of t-tests. Paired and unpaired t-tests were used for within- and between-group comparisons, respectively. Ninety-five percent confidence intervals (95% CI) were also calculated. The possible influence of an imbalance between the two groups with respect to menopausal state was evaluated using analysis of variance. Cumulative disease activity and cumulative functional capacity were defined as areas under the curve for the disease-activity scores and the health-assessment questionnaire scores, respectively. Differences between placebo and prednisone-treated patients were analyzed using the t-test. All analyses were done using intention-to-treat principles.

Results

Top Methods Results Discussion Author & Article Info References

During the recruitment period, 55 patients started aurothioglucose treatment. Three patients were inadvertently not reported by their treating physician. Nine patients refused to be included in the study for various reasons. One patient did not fulfill the required criteria for active disease. Two patients were excluded because of irregular menstrual cycles, indicating that menopause might be starting. The remaining 40 patients consented to participate in the study. The baseline BMD was similar in both groups. The mean trabecular BMD was 111 mg/mL in the prednisone group and was 111 mg/mL in the placebo group. The mean cortical BMD was 272 mg/mL in the prednisone-treated patients and was 276 mg/mL in the placebo-treated patients. Other baseline characteristics were also similar in the two treatment groups for most variables (Table 1). There were more postmenopausal women in the prednisone group, and the duration of postmenopause was longer in the placebo group. During the first period of the study, aurothioglucose was discontinued in nine patients (four in the prednisone and five in the placebo group) due to side effects. One of the five patients in the placebo group developed an acute, irreversible, and lethal pancytopenia. At the end of the first period of the trial, 16 patients in the prednisone group and 15 in the placebo group still used aurothioglucose. One patient in the prednisone group started methotrexate (7.5 mg weekly) after 18 weeks. Three patients in the prednisone group and four in the placebo group used no second-line antirheumatic treatment at the end of the first trial period. One patient in the placebo group with progressive disease refused further treatment with the trial medication after 4 weeks. She was then treated with prednisone, 10 mg/d, but assessments were carried out according to the protocol and data from this patient were included in the placebo group for all analyses (according to intention-to-treat principles).

During the second period of the trial, six patients who were originally assigned to the prednisone group, and who had all discontinued the trial medication according to the protocol, again started to use prednisone. Two patients from the placebo group used prednisone between weeks 20 and 44. For one patient, who was originally assigned to prednisone therapy, no BMD data were available at week 44.

The composite disease-activity score and its four constituents all showed a marked and rapid improvement of disease activity in the prednisone-treated patients. Cumulative disease activity, as expressed by the area under the curve for the disease-activity scores, in the initial 20 weeks of the trial, was lower in the prednisone-treated than in the placebo-treated patients: mean (SD), 71.5 (15.5) compared with 88.1 (18.2); P = 0.004. The prednisone-treated patients also showed an improvement in functional capacity. Changes in health-assessment questionnaire scores were less marked, and no statistical difference was noted between the two groups in the area under the curve for this variable in the initial 20 weeks of the trial: the mean (SD) was 13.5 (8.2) in patients receiving prednisone and was 10.9 (6.9) in patients receiving placebo (P > 0.2).

Percentage changes with respect to baseline BMD are presented in Table 2. In the first 20 weeks, an 8.2% decrease was noted in the trabecular BMD within the prednisone group (P = 0.001). Little change was observed in the placebo group (P > 0.2), and a statistical difference was noted between the two groups (P = 0.003). The cortical BMD did not change in either treatment group (P 0.2). To correct for the imbalance in menopausal characteristics between the two treatment groups, an analysis of variance was done. For this purpose, patients were classified as men, premenopausal, postmenopausal for 7 years or less, or postmenopausal for more than 7 years. No effects of this classification on the change in trabecular BMD in the first 20 weeks were shown, and prednisone treatment remained associated with trabecular bone loss in this period (P = 0.01).

After discontinuation of trial medication, some patients started to use glucocorticoid agents between weeks 20 and 44 (6 patients originally assigned to receive prednisone and 2 patients originally assigned to receive placebo treatment). The mean BMD decreased between weeks 20 and 44 in these patients (9.7% in the trabecular bone and 7.0% in the cortical bone). Due to the small number of patients, no further analyses were done in this subgroup. The remaining patients, who did not use glucocorticoid agents between weeks 20 and 44, showed a 5.3% increase (P = 0.03) in trabecular BMD after discontinuing prednisone treatment. After discontinuing placebo treatment, little change was found (P > 0.2), and the changes in the two groups were again statistically different (P = 0.03). Cortical BMD did not change in the second period of the trial (P > 0.2). The change in mean trabecular lumbar BMD in those patients who did not use prednisone between weeks 20 and 44 is shown further in Figure 1. Individual changes in trabecular BMD during the study are shown in Figure 2.

View larger version (13K): [in this window] [in a new window]   Figure 1. Mean changes in lumbar, trabecular bone mineral density during and after low-dose prednisone treatment in a subset of patients who did not use prednisone between weeks 20 and 44. The change (mean ± SE) in bone mineral density was measured using dual-energy, quantitative computed tomography. The prednisone dose was 10 mg/d from week 0 to 12 and was tapered between weeks 12 and 20.

View larger version (29K): [in this window] [in a new window]   Figure 2. Changes in lumbar, trabecular bone mineral density in individual patients receiving low-dose prednisone or placebo between weeks 0 and 20. Patients who used prednisone in the second part of the study (between weeks 20 and 44) are indicated with a dashed line. = men; = premenopausal women; \#9671; = postmenopausal women.

No vertebral fractures were noted on lateral spine radiographs, at baseline, or after 44 weeks.

Discussion

Top Methods Results Discussion Author & Article Info References

The deleterious effects of glucocorticoid agents on bone are well recognized [13, 14]. However, it has been suggested that in patients with RA who frequently use relatively low doses of prednisone ( 10 mg/d), the favorable effects of these drugs on the inflammatory process and on physical activity may outweigh the negative effects on bone. Previous reports [1] on this subject have shown contradictory results. Flaws in study designs may partly explain the discrepancies in the literature. First, patients receiving prednisone will in general have a more severe disease than those not treated by glucocorticoid agents. Because patients with RA with more active disease and concomitant functional impairments are at risk for increased bone loss [1], this may introduce a selection bias in nonrandomized studies. Second, bone loss may be most severe in the first months of treatment [4]. In consequence, studies that evaluate bone loss some time after the initial period of treatment may underestimate the effects of glucocorticoid agents on bone. We report the results of the first randomized, placebo-controlled study that evaluates the initial effects of relatively low-dose prednisone therapy on BMD in patients with active RA.

In a period of 20 weeks, the mean trabecular BMD decreased by 8.2% in the patients receiving prednisone therapy (P = 0.001), despite an improvement in disease activity and functional capacity. In patients receiving placebo, no trabecular bone loss was seen (P > 0.2), and the two treatment groups were statistically different (P = 0.003). No changes were observed in the anterior cortical bone of the vertebral body, confirming the hypothesis of preferential loss in the trabecular bone [13, 14]. The presence of more postmenopausal women with a somewhat shorter duration of postmenopause in the prednisone group might partly explain the observed differences between the two treatment groups. However, because patients in the first years after menopause, who may be expected to loose bone more rapidly, were excluded from the study, we do not think that the imbalance has had a great effect on the results. The fact that prednisone therapy remained associated with rapid trabecular bone loss in the analysis of variance supports this view.

The finding of an 8.2% decrease in mean BMD after 20 weeks, induced by low-dose prednisone, was quite remarkable. Similar magnitudes of bone loss have been documented after oophorectomy [15] and during immobilization [16]. Marked decreases in BMD were also found in patients treated with higher glucocorticoid doses. Julian and colleagues [17] recently reported a mean decrease of 6.8% in lumbar BMD, as assessed by dual-photon absorptiometry, 6 months after renal transplantation in 20 patients treated with a three times higher cumulative prednisone dose. LoCascio and colleagues [4] studied patients with various primary diseases and found a mean decrease in trabecular bone volume (iliac crest biopsies) of 34%, 5 to 7 months after the start of prednisone (daily dose, 10 to 25 mg). Comparison of our results with these previous reports seems to indicate that the degree of the initial, glucocorticoid-induced bone loss may not be closely related to glucocorticoid dose. In fact, Julian and colleagues [17] observed no correlation between the degree of bone loss and the cumulative dose of glucocorticoid agents. In contrast, an increase in glucocorticoid-induced osteoporosis with longer durations of therapy and higher cumulative doses has been observed in a cross-sectional study involving patients receiving long-term therapy [18].

Dual-energy, quantitative computed tomography has a relatively low precision, which suggests caution when interpreting the data of individual patients presented in Figure 2. Sixteen of 20 patients in the prednisone group seemed to lose bone. Eleven of 19 placebo-treated patients also had a decrease in trabecular BMD, although to a lesser degree than the prednisone-treated patients. Some patients, mainly in the placebo group, seemed to gain bone. No definite statements can, of course, be made, but it is noteworthy that gold salts have been shown to decrease bone resorption in in-vitro experiments [19], and they may increase metacarpal bone mass [20]. For clinical practice, it would be useful to know which patients are most susceptible to the deleterious effects of glucocorticoid agents. Inspection of the data in Figure 2 and the results of the analysis of variance do not suggest an important effect of sex and menopausal state. However, a study involving much greater numbers of patients and preferably using a more precise method of bone mass measurement is needed to answer this type of question.

Our data clearly show that relatively low doses of glucocorticoid agents can cause a considerable trabecular bone loss from the lumbar spine in the initial months of therapy. One should be cautious, however, not to overinterpret these findings. First, the study was done in patients with active inflammatory disease, in whom disorders of bone metabolism have been described [1]. In consequence, these patients may be more susceptible to the deleterious effects of glucocorticoid agents than patients with other primary diseases or with less active RA. Second, all patients started intramuscular gold treatment. It cannot be completely ruled out that the observed decrease in trabecular BMD is the result of an interaction between prednisone and gold salts. However, no data currently exist to support this hypothesis. Third, we have only investigated the initial effects of glucocorticoid agents on lumbar BMD, and further prospective studies are needed to clarify the long-term effects on bone mass and fracture risk. Finally, one may ask whether a lower dose than the one we used will also induce axial bone loss. Intramuscular administration of glucocorticoid agents has been used successfully, in a dose of 120 mg of methylprednisolone once every month [21], and the effects of this regimen on bone have not been evaluated but could be less marked.

The potential reversibility of glucocorticoid-induced bone loss is an important issue, but data on this subject are scarce. Our data suggest that prednisone-induced bone loss can be reversible after discontinuing treatment, when duration of therapy is limited to a short period. Reversibility of osteopenia has also been suggested after surgical cure of endogenous Cushing syndrome [22] and in a few young patients with sarcoidosis, who had been treated for a longer period with higher doses of prednisone [23].

Despite a general concern for the potential side effects, and a lack of data showing their long-term efficacy, low-dose glucocorticoid treatment is widely used in patients with RA [24]. Our data show that glucocorticoid agents can cause rapid and marked bone loss, even when relatively low doses are used, and that this effect may be reversible after discontinuing therapy. Therefore we suggest that the use of prednisone should be limited as much as possible to short periods of time, until randomized, controlled trials have shown the efficacy and safety of long-term, low-dose glucocorticoid treatment in patients with RA. The data also stress the need for an effective strategy to prevent glucocorticoid- induced bone loss. The marked decrease in bone mass in the initial months of therapy suggests that prevention should be considered from the beginning and throughout glucocorticoid treatment. Currently, no such preventive strategy is generally accepted, but some studies have suggested positive effects of calcium, vitamin D, bisphosphonate agents, estrogenic and androgenic hormones, and calcitonin in the prevention and or treatment of glucocorticoid-induced osteopenia [13, 14].

Abbreviations

BMD: bone mineral density

RA: rheumatoid arthritis