Serum biomarkers and changes in clinical/MRI evidence of golimumab-treated patients with ankylosing spondylitis: results of the randomized, placebo-controlled GO-RAISE study

Patients and study design

The details of the GO-RAISE study have been described previously [3, 4]. The protocol was reviewed and approved by the institutional review board or independent ethics committee at each site of this multicenter trial. (See Acknowledgements section below for further details.) All patients provided written informed consent. The study was conducted at 57 sites in the United States, Canada, Europe, and Asia. The MRI substudy was conducted at ten sites with the capability and willingness to participate. All patients enrolled at participating substudy sites were included in the MRI substudy.

GO-RAISE patients were adults with definite AS (for ≥3 months) according to modified New York criteria, a Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) score ≥4 (0- to 10-point scale), a total back pain score ≥4 on a visual analogue scale (0- to 10-cm scale), and an inadequate response to current or previous nonsteroidal anti-inflammatory drugs (NSAIDs). Patients were randomly assigned in a 1:1.8:1.8 ratio to receive subcutaneous injections every 4 weeks of placebo, golimumab 50 mg, or golimumab 100 mg. Randomization was stratified by investigational site and screening C-reactive protein (CRP) concentration. Patients were allowed to continue stable doses of methotrexate, sulfasalazine, hydroxychloroquine, corticosteroids, and NSAIDs throughout study participation.

At week 16, patients who achieved <20% improvement from baseline in both the total back pain and morning stiffness scores entered into double-blinded early escape, such that patients in the placebo group received golimumab 50 mg, patients in the golimumab 50 mg group received 100 mg, and patients in the golimumab 100-mg group had no change to their study therapy. The placebo-controlled portion of the study ended at week 24, following which all patients who had been receiving placebo injections crossed over to receive golimumab 50 mg. Injections continued to be administered subcutaneously every 4 weeks through week 252, and the last study assessments for the period covered by the current analyses were at week 208.

Biomarker evaluations

Serum samples collected at weeks 0, 4, and 14 of the GO-RAISE trial were tested by Rules Based Medicine (Austin, TX, USA) and Pacific Biomarkers (Seattle, WA, USA) for selected markers using Luminex (Luminex, Austin, TX, USA) and enzyme-linked immunoassay platforms, respectively. The extensive list of candidate biomarkers implicated from previous studies (Additional file 1: Table S1) [12] was selected for correlation analyses with imaging and clinical endpoints, including leptin, interleukin (IL)-6, tissue inhibitor of metalloproteinase (TIMP)-1, complement component 3 (C3), intracellular adhesion molecule (ICAM)-1, CRP, haptoglobin, and serum amyloid-P.

Disease activity and progression evaluations

Clinical assessments of disease activity included the BASDAI [13] and the ASDAS employing CRP [14] determined at weeks 0, 14, and 104. MRI was employed to assess inflammation and structural changes in the spine. Serial spine 1.5-T MRI scans of the cervical, thoracic, and lumbar spine regions in the sagittal plane were acquired at baseline, week 14, and week 104 [5]. Two qualified readers who were blinded to treatment information, patient identity, and chronology of the images independently scored each sequence using modified versions of the ankylosing spondylitis spine magnetic resonance imaging score for activity (ASspiMRI-a) and the Ankylosing spondylitis spine magnetic resonance imaging score for chronicity (ASspiMRI-c) [15, 16]. ASspiMRI-a scoring of each vertebral unit (VU) was previously described [5]. For the modified ASspiMRI-c score, each VU was scored as follows: 0 (normal, no lesions), 1 (minor fatty degeneration), 2 (much fatty degeneration), 3 (one or two syndesmophytes), 4 (more than two syndesmophytes), 5 (vertebral bridging with <25% of disc length and/or bridging of one side), or 6 (vertebral fusion with >25% of disc length and/or bridging of both sides). If a fatty lesion was present in a VU scored 3–6, it was noted. Lateral view radiographs of the cervical and lumbar spine regions were obtained at weeks 0, 104, and 208 and scored by two trained readers blinded to treatment group and chronology using the modified Stokes Ankylosing Spondylitis Spine Score (mSASSS) method [17], as described previously [18].

Data analysis

All analyses of MRI data collected through week 104 employed observed data; missing data were not imputed. Of the unique serum biomarkers tested in this trial, 73 had <80% of values below the lower limit of quantitation at baseline and are included in the analyses described herein (Additional file 1: Table S1).

A total of 4940 analyses were performed to determine Spearman correlation coefficients describing relationships between serum biomarkers at weeks 0, 4, and 14 and subsequent efficacy endpoints, including ASDAS, BASDAI, ASspiMRI-a, ASspiMRI-c (all determined at weeks 0, 14, and 104), and mSASSS (determined at weeks 0, 104, and 208). Correlations at baseline were assessed on the basis of all patients, whereas those at follow-up time points were assessed according to randomized treatment groups (placebo versus golimumab). For all postbaseline assessments, changes from baseline in biomarkers were used. Only correlations found to be statistically significant after Bonferroni multiplicity-of-testing adjustment are reported herein.

Relationships between four biomarker levels (IL-6, leptin, TIMP-1, and C3) and changes from baseline to week 104 in mSASSS were also assessed using a generalized linear model. In this model, mSASSS scores were adjusted for VUs with fat indicated via ASspiMRI-c score at baseline, because the presence of fatty lesions on MRI images is a known contributor to radiographic (mSASSS) progression [19, 20] and because these markers affect adipocytes.

Logistic regression analyses were also conducted to assess the influence of baseline and earlier (week 14) changes in biomarker levels on the development of new fatty lesions on a patient level (yes or no, i.e., not detected by MRI at baseline but detected at week 14 or week 104). In such analyses, all patients were combined, with no adjustment for treatment group or multiplicity of testing. The resulting odds ratios and 95% confidence intervals were determined.

Patient analysis

One hundred thirty-nine patients had serum biomarker data, and 98 of these patients at 10 sites with MRI capability participated in the GO-RAISE MRI substudy. The vast majority of randomized patients (299 [84.0%] of 356) had pre- and posttreatment data obtained for the GO-RAISE radiographic substudy. Of these, up to 139 patients had sera collected for biomarker evaluations, and up to 125 underwent spine MRI scans. As reported previously, the demographic and baseline characteristics for the MRI substudy patients were generally consistent with those of the overall GO-RAISE patient population [5].

Correlations between baseline biomarker levels and baseline disease activity

Correlations between baseline biomarker levels and subsequent changes in measures of disease activity, radiographic progression, and spinal inflammation

Results of a separate generalized linear model used to evaluate potential interactions between biomarkers and baseline patient characteristics show that IL-6, leptin, TIMP-1, and C3 influenced mSASSS progression independently of demographic factors of age, sex, and human leukocyte antigen B27 positivity (data not shown).

Correlations between posttreatment biomarker levels and posttreatment measures of disease activity and spine inflammation

Reductions in C3 observed from baseline to week 4 (r _s = 0.55, p = 0.001) and week 14 (r _s = 0.49, p = 0.040) significantly correlated with improvement from baseline to week 14 in the BASDAI score. Similarly, improvement from baseline to week 4 in serum IL-6 (r _s = 0.61, p = 0.022) and C3 (r _s = 0.72, p = 0.005) levels, as well as improvement from baseline to week 14 in serum IL-6 levels (r _s = 0.59, p = 0.043), correlated significantly with improvement in ASspiMRI-a scores from baseline to week 14 of golimumab treatment (Table 2). All other correlations tested between posttreatment biomarker levels and posttreatment measures of disease activity or spinal inflammation yielded insignificant relationships.

Limitations

A limitation of the present study is the large number of analytes, which precludes analyses such as principal component analysis or supervised cluster analysis. To guard against identifying a false-positive relationship between the large number of analytes assessed and disease activity outcomes, we employed the Bonferroni correction in the statistical modeling to account for multiplicity of testing. The crossover to active anti-TNF agent after a relatively short placebo period complicates interpretation of “placebo group” imaging data at weeks 104 and 208. These analyses are also limited by the fact that golimumab doses were combined and not assessed individually; however, given that no dose-response relationship was observed between golimumab 50 mg and 100 mg when analyzed by randomized groups at week 104 [4], the doses were combined to maximize the patient numbers in the golimumab arm. Despite this approach, there remained few patients on whom to base correlation analyses for radiographic progression and development of fatty lesions. This is an unresolved issue and should be the focus of future research in this field. Additionally, IL-6 levels can be influenced by therapeutic corticosteroid use via suppression of endogenous cortisol levels [24, 25] and also by visceral fat secretion [26]; our analyses did not adjust for these potential confounders in assessing relationships between IL-6 and MRI-detected spinal inflammation and clinical measures of disease activity.