Baseline Cohort Demographics
There were 251 patients enrolled in the GiACTA trial; one patient withdrew prior to randomization and was not included in the analyses. Thus, 250 patients were included in our analyses. The mean ± SD age was 69 ± 8 years and 187 (75%) were female (Table 1). The mean ± SD baseline BMI was 25.9 ± 4.7 kg/m2. A slight majority of patients (131, 52%) had relapsing disease at baseline; the remainder had newly diagnosed disease. Of the 250 patients included, seven (3%) were classified at baseline as being underweight; 115 (46%) as having normal weight; 85 (34%) as being overweight; and 43 (17%) as obese. The baseline prednisone dose was less than or equal to 30 mg daily in 129 patients (52%) and greater than 30 mg daily in 121 patients (48%).
Glucocorticoid Administration During the Trial
Over the course of the 52-week trial, patients were treated with a mean ± SD cumulative prednisone dose of 3077 ± 2577 mg. The greatest GC exposure occurred during the first 24 weeks of the trial, when patients received a mean of 1999 ± 1421 mg—approximately two thirds of their overall GC exposure. Between weeks 24 and 52, patients received a mean of 1035 ± 1448 mg—approximately one-third of their overall GC exposure. Patients who were randomized to tocilizumab received less prednisone than those randomized to prednisone-only (2191 vs. 4318 mg). Patients with relapsing and newly diagnosed disease received similar amounts of prednisone (3006 vs. 3149 mg, respectively). Patients who experienced disease flares received more prednisone than those who did not flare (4732 vs. 1712 mg).
BMI Change During the Study
The mean BMI increased from 25.9 ± 4.7 at baseline to 27.1 ± 4.9 at 24 weeks, and then remained stable at 27.1 ± 4.5 at 52 weeks (p = 0.01 for change from 0 to 24 weeks and p = 0.91 for change from 24 to 52 weeks) (Fig. 1A). BMI change was most notable over the first 24 weeks of the study, with patients gaining a mean of 1.1 ± 1.7 kg/m2 during weeks 0–24. BMI change from weeks 24 to 52 was 0.0 ± 1.3 kg/m2. Regarding BMI category changes, 41 (21%) had a decrease in BMI, 97 (49%) had an increase of 0 to 2 kg/m2, 54 (27%) had an increase of > 2 to 5 kg/m2, and five (3%) had an increase of > 5 kg/m2. Figure 1B demonstrates the change in BMI status for patients between 0 and 52 weeks. Between patients with relapsing versus newly diagnosed disease, more patients with relapsing disease experienced decreases in BMI, and fewer patients with relapsing disease experienced a > 2 to < 5 kg/m2 increase in BMI (22 vs. 32%) (p = 0.05 for differences across all categories) (Table 2). Patients randomized to tocilizumab had similar changes across BMI categories compared with those randomized to prednisone-only (p = 0.91 for differences across all categories) (Table 2).
Factors Associated with BMI Change on Mixed-Effects Univariable and Multivariable Analyses
In unadjusted analyses including all patients, the cumulative prednisone dose over 52 weeks was associated with an increase in BMI (ranging from β = 0.94 kg/m2 for 0–1 g exposure to β = 1.40 kg/m2 for ≥ 4 g exposure; p for trend < 0.001) as was relapsing disease (β = 1.57 kg/m2 compared to those with newly diagnosed disease; p = 0.01). Neither randomization to tocilizumab (β = 0.34 kg/m2 compared to prednisone-only; p = 0.56) nor flares (β = 0.14 kg/m2; p = 0.65) were associated with BMI change on univariable analysis (Table 3). In fully adjusted analyses, the cumulative prednisone dose over 52 weeks was associated with increase in BMI (ranging from β = 0.92 kg/m2 for 0–1 g exposure to β = 1.45 kg/m2 for ≥ 4 g exposure; p for trend < 0.001) (Table 3), relapsing disease was independently associated with a lower increase in BMI (β = − 0.42 kg/m2 compared to those with newly diagnosed disease; p = 0.002), and disease flares during the trial were associated with a numerically lower BMI increase (β = − 0.10 kg/m2 per flare; p = 0.15). Randomization to tocilizumab was not associated with BMI change (β = 0.11 kg/m2 compared to prednisone-only; p = 0.39). Results were generally similar in subgroup analyses of those with newly diagnosed and relapsing disease, with the exception of flares, which were associated with a decrease in BMI in patients with newly diagnosed disease (β = − 0.18 kg/m2 per flare; p = 0.03) (Table 4) but not in patients with relapsing disease (β = − 0.04 kg/m2 per flare; p = 0.73) (Table 5).
Due to the change in direction of the beta coefficients between the unadjusted and adjusted mixed-effects models for relapsing disease and flares, additional analyses were performed. Because patients with relapsing disease had a higher BMI at baseline, we sequentially added covariates in sensitivity analyses and noted that the direction of the association changed when baseline BMI was incorporated into the model. Similarly, since patients who flared received additional doses of prednisone, we sequentially added covariates to these models and found that adjusting for cumulative prednisone and baseline BMI were associated with changes in the direction of the association between flare and BMI.
BMI Change and Relapsing Disease at Baseline
Given that relapsing disease was associated with a smaller increase in BMI than newly diagnosed disease, we evaluated these disease type subgroups further. Over the 52-week trial, patients with relapsing disease at baseline had no significant change in BMI between 0 and 52 weeks (26.9 vs. 27.6 kg/m2; p = 0.27) whereas those with newly diagnosed disease had a significant increase (24.8 vs. 26.5 kg/m2; p < 0.01). Patients with relapsing disease had no significant change between 0 and 24 weeks or between 24 and 52 weeks (Fig. 2). The increase in BMI among patients with newly diagnosed disease was observed between 0 and 24 weeks but not between 24 and 52 weeks (Fig. 2).
BMI Change and Flares Across Treatment Arms Over 52 Weeks
Patients randomized to tocilizumab received far less prednisone compared with those who were randomized to prednisone-only (median of 1583 vs. 3607 mg; p < 0.001) and had fewer flares (median of 0 vs. 1; p < 0.001) but experienced similar changes in BMI (mean of 1.2 vs. 1.1 kg/m2; p = 0.31) (Table 6). Similar results were seen when assessing BMI change and cumulative glucocorticoid exposure at 24 weeks and when evaluating results in patients with either newly diagnosed or relapsing disease.
Analyses with Primary Outcome of Weight Change
In addition to BMI, we also analyzed weight change (kg) and found overall similar results. In the multivariable, fully adjusted mixed-effects model, cumulative prednisone exposure was independently associated with weight change with increasing estimates in increasing exposure categories. Relapsing disease and flares during the trial (particularly among newly diagnosed subjects) were both associated with numerically less weight gain, but neither reached statistical significance.