This study was embedded in a multicenter randomized clinical trial studying persons older than 18 years presenting with recent-onset arthritis, the so-called tREACH study (treatment in the Rotterdam early arthritis cohort) 20. The primary aim of this study is to establish the best treatment strategy for patients with early arthritis.

Patients were included if arthritis in at least one joint was observed by a rheumatologist, and complaints were present for less than 12 months. With a prediction model developed by Visser et al. 21, patients were stratified according to their risk of having persistent erosive disease after a follow-up period of 2 years (high, intermediate, and low probability). We studied 41 patients in the high-probability group. These patients were randomized to three different treatment strategies, all including GC, either oral GC (15 mg prednisone/day, two treatment arms) or intramuscular GC (single depot of methylprednisolone, 120 mg, or triamcinolone acetonide, 80 mg, one treatment arm). All tREACH patients were naïve to GCs and DMARDs (Figure 1).

After a minimum of 1 year of follow-up, the diagnosis of the patients was verified in medical documentation or, if necessary, in consultation with the treating rheumatologist.

In an independent cohort, 37 patients with established RA and active disease (FLARE study) were recruited. Active disease was defined as disease activity requiring GC therapy according to the treating rheumatologists 22. All patients received a single intramuscular depot of GC (methylprednisolone, 120 mg, or triamcinolone acetonide, 80 mg). None of the FLARE patients had used GC in the last 3 months and were taking stable DMARD therapy (Figure 1). As a control group, we studied healthy laboratory employees (n = 20). None of the controls was using a GC.

Of the 41 high-probability patients included via the tREACH study, 38 were ultimately diagnosed as having definite RA. In this group of early RA, two patients were lost to follow-up, leaving 36 patients for complete analysis. After randomization, oral GCs were prescribed to 22 patients, and 14 patients were given a single depot of intramuscular GC. In the FLARE study, two patients were lost to follow-up for logistic reasons. In 10 patients, only one of the assays could be performed due to limited amount of PBMCs. Ultimately, 32 patients could be evaluated for binding capacity of the GC receptor, and 32 patients for the bioassay (in 27 patients, both assays were performed). Patients lost to follow-up were included in the baseline analysis (two patients in each cohort).

In the tREACH cohort, in vitro GC sensitivity was assessed with the GC bioassays (for logistic reasons, only enough PBMCs were available for the GC bioassays). In patients participating in the FLARE study, in vitro GC sensitivity was assessed by both the GC bioassays and GC binding capacity.

The GC bioassays were performed as described previously 17. In short, peripheral blood was drawn in all patients before start of treatment by using Cell Preparation Tubes with Sodium Heparin (Becton Dickinson, Breda, The Netherlands), allowing isolation of PBMCs. Cells were resuspended in RPMI 1640 medium containing L-glutamine supplemented with penicillin (100 U/ml) and streptomycin (100 μg/ml) and 10% fetal calf serum (FCS) and precultured overnight in a 48-well plate (Costar, Amsterdam, The Netherlands, 5.0 × 10⁵ cells/well in duplicate, density of 4.0 × 10⁶/ml). A single batch of FCS was used throughout. Before use, this batch was analyzed for cortisol content, which was found to be below the detection limits.

Trypan blue staining revealed the viability of isolated cells to be greater than 95%. The next day, cells were incubated with dexamethasone 0, 0.33, 1, 3.3, 10, 33, 100, and 333 nM dexamethasone and stimulated with 10 μg/ml phytohemagglutinin (Sigma-Aldrich, Zwijndrecht, The Netherlands). After 4 hours in the incubator, total RNA of the cells was collected (Total RNA isolation Kit, Roche, Almere, The Netherlands). Reverse transcription was performed by using 100 ng total RNA per reaction. Quantitative real-time PCR analysis was carried out on a 7900HT Taqman machine (Applied Biosystems, Nieuwerkerk aan den IJssel, The Netherlands), according to the manufacturer's instructions. Data were analyzed by using the SDS 2.4 software (Applied Biosystems). GC-specific transactivation of the GILZ gene and transrepression of the IL-2 gene were measured while correcting for the housekeeping gene hypoxanthine phosphoribosyltransferase (HPRT) by using the ΔΔCT method; primers and probes were obtained from Biolegio, Nijmegen, The Netherlands (see Additional File 1 Table S1). Half-maximal effective concentration (EC₅₀) was calculated by using nonlinear regression in GraphPad Prism 5.0 and used as a read-out for in vitro GC sensitivity. The EC₅₀ values of GILZ and IL-2 in PBMCs were not significantly influenced by the cellular composition (percentages lymphocytes and monocytes) of the PBMCs (data not shown).

GC binding capacity was measured by using a whole-cell dexamethasone-binding assay, as described previously, with minor modifications 23. In brief, by using PBMCs from the same isolation procedure, incubation was started in a volume of 200 μl (0.5 to 2 × 10⁶ cells) containing [³H] dexamethasone at concentrations of 1 to 30 nM with and without a 400-fold excess of unlabeled dexamethasone reflecting nonspecific and total binding of [³H] dexamethasone, respectively. Two tubes without labeled dexamethasone were incubated under the same conditions for determination of cell number and viability at the end of the procedure. The PBMCs were incubated during 1 hour at 30°C in a shaking water bath. The incubation was stopped by the addition of 2 ml cold saline, followed by centrifugation and two washing steps. Finally, the PBMCs were resuspended in 250 μl saline. Radioactivity in 200 μl of this suspension was counted in a liquid scintillation counter. Specific binding was calculated by subtracting nonspecific binding from total binding. EC₅₀ values, receptor number, and ligand affinity (1/K_D) were calculated by using the nonlinear regression method (GraphPad Prism, version 5.0; La Jolla, CA, USA).

Trained research nurses examined patients before and after 2 weeks of standardized GC treatment. Disease Activity Score (DAS, 44 joints) was calculated according to the following formula: DAS = 0.54 × √RAI + 0.065 × SJC44 + 0.33 × ln(ESR) + 0.007 × GH (RAI, Ritchie Articular Index; SJC44, 44 swollen-joint count; ESR, erythrocyte sedimentation rate; GH, general health on a 100-mm scale). As primary outcome, the relative decrease in DAS (100 × ((DAS_baseline - DAS_{after 2 weeks})/DAS_baseline)) was used as an index for in vivo GC sensitivity. By using this continuous outcome variable, a floor effect in patients with relatively low disease activity was prevented. In addition, continuous variables represent the full information, in contrast to (arbitrary) categoric data (that is, response criteria). We chose a 2-week interval for follow-up in tREACH patients to minimize the influence of the disease-modifying effects of the other antirheumatic drugs on the DAS. A similar follow-up period was chosen in the FLARE study to make comparisons between the groups possible. During the study period, the dose of DMARD(s) already being used was not changed, and no additional anti-rheumatic therapy was started.

To explore further the effectiveness of GC therapy, the impact of GC treatment on performing activities of daily living was assessed by using the health assessment questionnaire disability index score (HAQ-DI). The HAQ-DI is a widely used and validated tool to quantify functional disability in RA 24 and comprises questions about different aspects of daily life. In particular, the minimal import difference (MID) in the HAQ-DI score is the smallest difference in HAQ-DI score that patients sense as a difference. In clinical trials, the MID in HAQ-DI improvement ranged from 0.22 to 0.24 1617. As a result, patients were classified as responder (HAQ-DI_baseline - HAQ-DI_2wks ≥ 0.25) or nonresponder (HAQ-DI_baseline - HAQ-DI_2wks < 0.25).

We measured blood pressure and body weight before and after 3 months of GC therapy in tREACH patients. Furthermore, glycosylated hemoglobin (HbA_1c) was measured at baseline and after 3 months in tREACH patients (HbA_1c analyzer, type Adams A1c HA-8160, Menarini Benelux).

Differences in continuous variables between the cohorts were tested by using analysis of variance (ANOVA). GILZ-EC₅₀ values were normally distributed (Kolmogorov-Smirnoff P > 0.20), whereas IL-2-EC₅₀ was square-root transformed, and the number of receptors and K_D were both natural logarithm transformed to normalize the data. Bonferroni post hoc tests were used to correct for multiple testing.

Pearson or Spearman correlation coefficients were used to describe the bivariate relations between in vitro parameters of GC sensitivity and DAS at baseline and relative decrease in DAS.

ANOVA analysis was applied to test for differences in in vitro parameters of GC sensitivity between HAQ responders and nonresponders. Paired t tests or Wilcoxon Signed Ranks Tests were used for analysis of alterations in DAS, HAQ-DI scores, blood pressure, body weight, and HbA_1c values.

To test for potential confounders, each of the individual in vitro parameters of GC sensitivity (that is, IL-2- and GILZ-EC₅₀, K_D, and number of GRs) and selected covariates were modeled by using linear regression (relative decrease in DAS as the dependent variable). These selected covariates included gender, age, and, based on potential synergistic immunomodulating properties with GCs, use of NSAIDs, number of DMARDs, and use of anti-TNF-α agents.

Orally and intramuscularly treated patients were analyzed separately because of nonequivalent cumulative dosages of GC (cumulative GC dosage: oral > intramuscular). We considered differences statistically significant if P ≤ 0.05 (two-sided).

All subjects signed informed consent, and the study was approved by the medical ethics committee of the Erasmus Medical Center.

Overall, patients with early (tREACH cohort) and established RA (FLARE cohort) had higher mean EC₅₀ values in the IL-2 assay than did healthy controls (although not statistically significant in the FLARE cohort), indicating that RA patients needed a higher dosage of dexamethasone to suppress IL-2 mRNA expression in vitro. In contrast to this, similar EC₅₀ values were measured in the GILZ assay (Figure 2A, C). Patients participating in the FLARE study had a higher number of GRs compared with healthy controls, while having comparable affinity (1/K_D) of the receptor (Figure 2E, F). The percentage of monocytes was measured in subsets of FLARE patients and healthy controls and did not differ significantly (mean ± SD, 24.6 ± 9.2 in FLARE patients versus 20.9 ± 5.0 in healthy controls). Ligand affinity of monocytes and lymphocytes did not differ significantly. The number of glucocorticoid receptors per cell was about threefold higher in monocytes as compared with lymphocytes (data not shown). The maximum induction of GILZ and repression of IL-2 tended to be lower in the established RA cohort as compared with healthy controls (P = 0.068 and P = 0.101, respectively) (Figure 2B, D).

No correlations were found between the DAS and parameters of in vitro GC sensitivity. Of the variables used to calculate the DAS, a negative association was observed between the RAI and IL-2-EC₅₀ (ρ = -0.465; P = 0.005), but only in the early RA patients. No gender differences were noted at the mean level of the IL-2-EC₅₀ and GILZ-EC₅₀, number of GRs, or the affinity of the receptor.

HAQ-DI sum scores before start of treatment did not show any correlations with in vitro parameters of GC sensitivity. Male and female individuals did not differ significantly in HAQ-DI sum scores.

GILZ-EC₅₀ and IL-2-EC₅₀ were positively correlated, but only in the patients with early RA (ρ = 0.383; P = 0.028). In patients with established RA, the number of GRs was inversely correlated with GILZ-EC₅₀ and IL-2-EC₅₀ (ρ = -0.401; P = 0.042; and ρ = -0.462; P = 0.020 respectively). K_D was also inversely correlated with GILZ-EC₅₀ (ρ = -0.413; P = 0.032), but not with IL-2-EC₅₀. Finally, K_D and GR-number were correlated (ρ = 0.627; P < 0.001; see Additional File 2 Figure S2).

After 2 weeks of GC treatment, a significant decrease in disease activity was measured in both orally and intramuscularly treated patients (ΔDAS_oral = 0.92; P < 0.001; ΔDAS_{intramuscular} = 0.89; P < 0.001; and see Additional File 3 Table S3). The interquartile range in relative decrease in DAS was 22% and 43% in orally and intramuscularly treated patients, respectively, indicating greater variability in relative decrease in DAS in the intramuscularly treated group.

In patients treated with a single intramuscular depot of GC (all FLARE patients and a proportion of tREACH patients), a modest inverse relation was found between in vitro GC sensitivity as reflected by IL-2-EC₅₀ values and the percentage improvement in DAS after 2 weeks (P = 0.029; Figure 3A). Similarly, near-significance was reached for GILZ-EC₅₀ values and the relative decrease in DAS, but also only in intramuscularly treated patients (P = 0.054; Figure 3B). In addition, the number of GRs displayed a modest positive relation with the improvement in DAS in patients with intramuscular depots of GC, and a positive trend was observed for the K_D of the GRs (P = 0.008 and P = 0.070, respectively; Figure 3C, D).

With multiple regression, however, both the number of GRs and K_D of the receptor were significant factors contributing to the relative decrease in DAS. The negative association between IL-2-EC₅₀ and GILZ-EC₅₀ values and relative decrease in DAS persisted, although only near-significance was reached (Table 2).

Of note, in the subgroup of patients with evaluation of GC-binding capacity (FLARE study), age and use of NSAIDs were also independent predictors of improvement of disease activity after 2 weeks of GC treatment. Age and use of NSAIDs both had positive β coefficients, indicating a better response with older age and use of NSAIDs.

After 2 weeks of GC treatment, a significant decrease in HAQ-DI sum scores was measured (ΔHAQ-DI = -0.40; P < 0.001). However, 12 of 34 patients still had to be classified as nonresponders. Responders had lower EC₅₀ values of GILZ and higher numbers of GRs with higher K_D (Figure 4). IL-2-EC₅₀ values tended to be lower in responders.

The mean systolic and diastolic blood pressure of patients was reduced at their 3-month follow-up visit (systolic RR_baseline, 145.0 mm Hg; systolic RR_{3 months}, 134.4 mm Hg; P = 0.005; and diastolic RR_baseline, 87.4 mm Hg; diastolic RR_{3 months}, 82.6 mm Hg; P = 0.003). This decrease was observed in both orally and intramuscularly treated patients. Body mass index did not change significantly after 3 months (BMI_baseline, 26.9; BMI_{3 months}, 26.7).

At baseline, the mean HbA_1C was 5.51% (reference range, 4.5% to 6.0%). Three patients had HbA_1C values above the upper limit of the normal range. After 3 months, the mean HbA_1C was even somewhat lower (5.31%; P = 0.016). Nine patients had a higher HbA_1C, four patients had an equal percentage of HbA_1C, and 13 patients had an improvement. No relation was found between alterations in HbA_1C and blood pressure and in vitro GC sensitivity, as measured by the bioassay.