Teriparatide for Joint Erosions in Rheumatoid Arthritis: The TERA Trial (TERA)

I. Introduction:

While generalized osteoporosis causes tremendous disability in patients with Rheumatoid Arthritis (RA) and occurs relatively frequently in such patients, there has been little research on treatments for osteoporosis in patients with RA. Not only are there important questions about the effects of teriparatide on Bone Mineral Density (BMD) in patients with RA, but little is known about how it might affect localized bone erosions or RA disease activity.

Recent data in a mouse model of RA suggest that intermittent para-thyroid hormone (PTH) in the setting of potent immunosuppressives may indeed heal bone erosions. This study showed an additive effect of PTH in addition to a biologic on erosion healing. To the best of our knowledge, this has yet to be demonstrated in humans. That is the primary aim of the proposed study.

II. Objectives and Hypotheses:

To assess the effects of teriparatide among a group of patients with RA and erosions, all using biologics, with respect to:

1. Joint erosion volume by 3-dimensional computed tomography (3D CT) scan;
2. Lumbar BMD by dual-energy X-ray absorptiometry (DXA);
3. Hip BMD by DXA; and
4. RA disease activity measured by the Disease Activity Score (DAS) and acute phase reactants.

The hypotheses to be tested include:

1. Joint erosion scores, measured by 3D CT scan, will be significantly improved at study completion in patients taking teriparatide.
2. Teriparatide will significantly increase BMD at all sites as measured by DXA.
3. RA disease activity measures will be stable during the study year.

III. Statistical Analyses:

The total erosion volume will be calculated for each hand/wrist and for each of the six sub-regions: radius, ulna, proximal carpals [scaphoid, lunate, triquetrum, and pisiform), distal carpals (capitate, hamate, trapezium, trapezoid, and the carpometacarpal (CMC) joints), metacarpophalangeal (MCP) joints, and proximal interphalangeal (PIP) joints. We will perform the hand level-analysis with the individual hands as our study units. In a pre-trial study, repositioning reproducibility was excellent at the hand-level (N = 10) as noted in Duryea et al (REF ). The average total erosion volume in a single hand/wrist was 428.1 mm^3. Average erosion volume was smallest at the ulna (6.4 mm^3) and largest in the distal carpals (144.8 mm^3). The intra-class correlation (ICC) values were excellent, ranging from 0.97 to 1.00. The root mean square stand deviation (RMSSD) was 31.2 mm^3 with a coefficient of variation (CoV) of 7.3%. The CoVs for the six measured hand regions ranged from 7.6% to 21.0%. Individual regions with increased erosion volume tended to have correspondingly larger RMSSD values, while the dependence of the CoV on the total volume was less pronounced.

The outcome for the primary analysis will compare the changes in erosion volume from baseline to follow-up for the subjects receiving teriparatide versus those not. If we find that the change in whole hand erosion volume is significantly better for teriparatide users than controls, then we will have met the primary outcome. In a secondary analysis, we will assess the change in erosion volume at each of the six anatomic sites. As these are secondary analyses, no correction for multiple testing will be pursued. Significance at the whole hand and anatomic site level will be based on a two-tail test of significance with a p-value < 0.05 considered statistically significant.

For all analyses, we will analyze the data at the level of the hand, adjusting for within subject correlation between the two hands using a generalized linear mixed model (GLMM). Baseline subject characteristics will be compared between the two groups using two sample t tests, Chi-square tests or non-parametric tests when applicable. Characteristics of interest include age, gender, duration of RA, serologic status [rheumatoid factor (RF) positive and anti-citrullinated protein antibody (ACPA) positive], use of oral corticosteroids, use of concomitant non-biologic disease-modifying antirheumatic drugs (DMARDs), baseline DAS score, baseline Total Sharp Score, and baseline health assessment questionnaire (HAQ) score. If any of these characteristics are found to be imbalanced across groups (p-value < 0.10), those variables will be introduced as possible covariates in the model. However, for this small study with 48 hands, the final model will include no more than 5 predictors.

Several exploratory subgroup analyses will be pursued. These include subgroups of patients based on/with:

• Corticosteroid users or not at baseline;
• Less than or greater than 2 years of RA at baseline;
• In RA remission, defined in three ways: as DAS28 ≤ 2.6, simple disease activity index (SDAI) ≤ 3.3, and boolean [swollen joint count (SJC), tender joint count (TJC), patient global and c-reactive protein (CRP) ≤ 1];
• high sensitivity CRP (hsCRP) < 3mg/L versus ≥ 3mg/L;
• Less than median # of erosions (based on Sharp Score) versus greater than;
• Less than median erosion volume at baseline versus greater than; and
• Depth of erosion: 25% deepest erosions versus other erosions.

The secondary analyses will all be considered exploratory and will follow the same analytic strategy as the primary analyses. Due to small sizes, unadjusted GLMM model will be used in sub-group analyses.

Analyses of secondary outcomes noted above will also be considered exploratory. They will also follow the same analytic strategy as the primary analyses.

If follow up is less than complete, we will estimate the effect under intention-to-treat. Several missing value techniques will be applied to impute the missing outcomes including last observation carried forward, single imputation (e.g. replacing missing values with the sample mean or median), and multiple imputation. Sensitivity analysis will compare these methods with complete case analysis.

IV. Sample Size Estimates:

We have used the estimates derived from the pre-trial replication study to estimate sample size for the trial. We considered analyses both at the level of the subject and at the level of the individual hand. Assessing individual hands provides twice as many observations (two hands for each patient), but these observations are not independent of one another. In our pilot analysis that evaluated 5 patients (10 hands), log-transformation was applied to total volumes to achieve an approximately normal distribution. The average value of log-transformed total volume was 5.32, with standard deviation of 1.34. Assuming that a reasonable ICC between two hands/wrists, on the same patient ranges from 0.20 to 0.50, we estimated the sample size required to achieve 80% power, given a range of differences with a significance level of 0.050 using a two-sided test (see Table 2) 17. To detect a moderate difference of 15%-25% between groups, the sample size needed to achieve adequate statistical power would range from 10 to 34 per study group.

We recruited 12 patients per group (24 hands), assuming an ICC of 0.5.