Early changes in blood-based joint tissue destruction biomarkers are predictive of response to tocilizumab in the LITHE study

The LITHE study has previously been described by Kremer et al., Smolen et al. and Bay-Jensen et al. [7,11,25] (clinicalTrials.gov identifier: NCT00106535). Briefly, the study is a 2-year phase III, multicentre, randomized, three-arm, placebo-controlled, parallel group trial in patients with moderate to severely active RA, who had inadequate responses to methotrexate (MTX). Patients were randomized 1:1:1 to one of three treatment groups: 4 mg/kg or 8 mg/kg TCZ, or placebo (PBO) in combination with a stable dosage of MTX (10–25 mg/week). TCZ and PBO were given intravenously every 4 weeks. Patients who failed to respond to treatment during the study, ie., they experienced <20 % improvement from baseline in the swollen joint count (SJC) and tender joint count (TJC) at week 16 or later, could receive blinded rescue therapy in a stepwise fashion between weeks 16 and 28 (see study description in paper by Kremer et al. [7,11,25]. The patients receiving rescue treatment after week 16 were described as early non-responders in this sub-analysis, whereas the patients who experienced at least a 20 % improvement from baseline in SJC and TJC were described as early responders. Table 1 provide a short summary of the patient characteristics.

The study was approved by the ethics committee at each participating institution (the list can be found at https://​clinicaltrials.​gov/​ct2/​show/​NCT00106535/​) and was conducted according to the principles of Good Clinical Practice and according to the Declaration of Helsinki. The sub-analysis of this study was done on anonymized data, and the statistical analysis plan was approved by the projects teams at Nordic Bioscience and Hoffmann-La Roche. All patients provided written informed consent allowing for retrospective analysis of the blood samples for assessment of biomarkers of joint tissue turnover. No steering committee was used for this study.

In the study protocol, the use of serum for prospective and retrospective exploratory biomarker analysis was included and scheduled to be collected from patients who provided informed written consent. The blood was collected in the morning after an overnight fast for >8 hours, at baseline, and weeks 4 and 16. All samples were aliquoted and stored frozen at a temperature below −70 °C until assayed.

The biomarkers were measured at three different sites using different aliquots. CRP was measured by the different central laboratories associated with the study sites. CRP was measured in all patients in the biomarker sub-study (n = 740). MMP3, osteocalcin (OC), CTX-I and C2M were measured by Synarc laboratory in Lyon France. Only 676 samples of the 740 originally included were available for measurement from the second aliquot and the volume was not sufficient to measure all biomarkers (e.g., n = 626 for measurement of C2M). C1M, C3M and CRPM were measured by Nordic Bioscience Laboratory in Roedovre, Denmark. Only 599 samples of the 740 originally included were available for measurement in the third aliquot and the volume was not sufficient to measure all biomarkers (e.g., n = 585 for measurement of C1M).

Serum levels of C1M [26], C2M [27], C3M [28] and CRPM [14] were measured, blinded to patient identity, by manual competitive ELISA. Briefly, for C1M and C2M, which measure MMP-degraded type I and II collagen fragments, respectively, Biotin-K-GSPGKDGVRG or biotin-KPPGRDGAAG (American peptide, Sunnyvale, CA, USA) were coated onto the streptavidin-pre-coated 96-well plate (Roche Diagnostics, Mannheim, Germany) and left for 30 minutes at 20 °C. After washing (PBS + 10 % tween 20), the calibrators, controls and undiluted serum samples were added, followed by peroxidase-conjugated monoclonal antibody NB104-4D3 or NB44-3C1, and incubated at 4 °C overnight. The sample-antibody mix was washed off the well plate and peroxidase reaction was visualized by 3,3′,5,5′-tetramethylbenzidine (TMB, Kem-En-tec, Taestrup, Denmark) at 20 °C and stopped with sulphuric acid after 15 minutes. For C3M, which measures the MMP-derived type III collagen neo-epitope, 96-well streptavidin-coated plates were coated with 0.4 ng/mL of KNGETGPQGP-biotin and left for 30 minutes at 20 °C. After washing, calibrators, controls and serum samples (diluted 1:1 in incubation buffer) were added, followed by peroxidase-conjugated antibody NB51-G12. The sample-antibody mix was incubated at 20 °C for 60 minutes. TMB was added after washing of the plates and incubated at 20 °C and stopped with sulphuric acid after 15 minutes. CRPM measurement followed the same procedure as C3M, however, applying a different peroxidase-conjugated antibody (NB94-1A7) and coater (KAFVFPKESDK-biotin). After sample/calibrator incubation, the wells were washed and incubated with 100 μL of TMB at 20 °C for 15 minutes, followed by the addition of 100 μL stop solution (sulphuric acid) in each well. The colorimetric reaction was measured at 450 nm with reference at 650 nm using the Softmax Pro®, version 5 software (Molecular Devices, Sunnyvale, CA, USA). The intra- and inter-assay coefficient of variation (CV) was <8 % and <10 %, respectively, for all assays described above.

Serum total MMP-3 was measured by a two-site ELISA using two polyclonal antibodies raised against human MMP-3 (Quantikine®, R&D systems, Abingdon, UK). The intra- and inter-assay CV was <10 %. Serum OC and CTX-I were measured by an automated multiplex assay (IMPACT bone chip, Roche Diagnostics, Mannheim, Germany) using the same antibodies as those employed in the corresponding single marker assays (Elecsys, Roche Diagnostics, Mannheim, Germany) [29].

No imputation was made on missing values, however, all data points were included resulting in different numbers of values being included in different analyses. Missing data points were observed for most variables, including clinical, demographic and biomarker parameters.

Summary statistics of general demographics, baseline RA characteristics, and baseline ACR criteria demographics (Table 1) included the number of patients and the mean and standard deviation. Changes in clinical scores and biomarkers from baseline to 4 weeks were investigated by the Wilcoxon paired rank sum test. The primary analysis (Figs. 1 and 2) was based on the mean percentage (+/− standard error of the mean (SEM)) relative to the baseline measurement of the biochemical markers.

The area under the curve (AUC) for the separation by the biomarkers between early responders and non-responders was tested by receiver-operator characteristic (ROC) curve analysis (Table 2). For markers with significant and consistent associations in the primary analysis, further associations between change in biomarker at baseline and week 4 were investigated using a dichotomous approach, either logistic regression or classification and regression analysis, in which patients were categorized according to designation of high and low levels, and responders and non-responders, using unpruned data. Odds ratios were calculated using these cutoff values.

All statistical analyses were performed using MedCalc version 12.3.0. Graphing was performed using Prism Graphpad version 5.03.