Patients with AS (n
= 180) were recruited from six secondary care rheumatology centres in the UK (Bath, North Bristol, Cannock, Durham, Romford and Stoke-on-Trent). Patients invited to take part in the study were over 21 years of age, had AS according to the modified New York Criteria of 1984 [24
], and provided written informed consent according to the declaration of Helsinki. The Trent Research Ethics Committee and the six site-specific National Health Service trusts approved the multicentre study. The majority of patients had been on a stable dose of nonsteroidal anti-inflammatory drugs for at least the previous 3 months. Patients on anti-TNF therapy, systemic steroid use in the previous 3 months or bisphosphonates in the previous 12 months were excluded from the study. Other exclusions included current use of methotrexate, sulfasalazine, raloxifene, calcitonin, phenylbutazone or hormone replacement therapy.
The principal aim of this study was to evaluate the effect of bisphosphonate (alendronate) on global health (Bath ankylosing spondylitis global health (BAS-G)) in AS (Bisphosphonates in Ankylosing Spondylitis trial). Secondary aims were to determine whether there were changes in disease activity, function and bone status, and whether changes in biomarker levels were different between patients treated with or without alendronate. A preliminary report has been published, which indicated that alendronate had no significant effect on changes in CRP, cytokine and MMP levels [25
]. A further aim of the study was to investigate the relationship between circulating biomarker levels and measures of disease activity and function in AS patients on standard nonsteroidal anti-inflammatory drug therapy. Data collected from the baseline visit are reported here.
Baseline assessment included questionnaires to assess disease activity (Bath ankylosing spondylitis disease activity index (BASDAI)), function (Bath ankylosing spondylitis functional index (BASFI)) and well-being (BAS-G). The BASDAI is based on six questions related to fatigue, spinal pain, peripheral arthritis, enthesitis and morning stiffness (both severity and duration). The BASFI is a set of 10 questions that consider activities related to functional anatomy and the patients' ability to cope with everyday life. The BAS-G consists of two questions that assess the effect of the disease on the patient's well-being. A 10-cm visual analogue scale is used to answer questions to the BASDAI, BASFI and BAS-G. All measures are scored between 0 and 10, with higher values indicating worse disease activity, function or well-being.
Blood samples were taken for measurement of cytokines, MMPs and TIMPs, as well as a full blood count, urea and electrolyte levels, serum calcium and CRP. Information was also collected on smoking status, smoking duration, average number of cigarettes smoked per day and age of smoking cessation. Pack years were calculated (1 pack year = 20 cigarettes/day for 1 year) to provide a quantitative measure of smoking history. The effects of smoking intensity were assessed by categorising participants according to pack-year history, as in previous studies in AS [26
]: category 1, 0 pack years; category 2, 1 to 15 pack years; category 3, 16 to 30 pack years; and category 4, > 30 pack years.
Measurement of cytokines, MMPs and TIMPs
Sera were separated from bloods collected in plain Becton Dickinson Vacutainer® tubes (Becton Dickinson, Oxford, Oxfordshire, UK) at study entry. All sera were stored at -70°C until required. Measurement of the various cytokines, MMPs and TIMPs was performed using multiplex, bead-based (Luminex®) assays on a Bio-Plex™ 200 suspension array system (Bio-Rad Laboratories, Hemel Hempstead, Hertfordshire, UK). The levels of 30 cytokines (Human cytokine 30-plex panel; Life Technologies, Paisley, UK), five MMPs and four TIMPS (Fluorokine multi-analyte MMP and TIMP kits; R&D Systems Europe, Abingdon, UK) were measured in separate multiplex assays according to the manufacturers' instructions. The following biomarkers were measured: cytokines and cytokine receptors - IL-1β, IL-1 receptor antagonist, IL-2, IL-2R, IL-4, IL-5, IL-6, IL-7, IL-10, IL-12p70, IL-13, IL-15, IL-17, IFNα, IFNγ, TNFα, epidermal growth factor, basic fibroblast growth factor, VEGF, hepatocyte growth factor (HGF), granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor; chemokines - CXCL8, CXCL10, Eotaxin, macrophage inflammatory protein-1α (CCL3), macrophage inflammatory protein-1β (CCL4), monokine induced by gamma interferon (CXCL9), monocyte chemotactic protein-1(CCL2) and regulated upon activation, normal T-cell expressed, and secreted (CCL5); MMPs - MMP-1, MMP-2, MMP-3, MMP-8 and MMP-9; and TIMPs - TIMP-1, TIMP-2, TIMP-3 and TIMP-4. High and low control samples were used in each assay. To test the reproducibility of MMP measurements we also measured the same samples using standard ELISA kits for MMP-3 and MMP-8 (R&D Systems).
All data were tested for normality and the appropriate parametric or nonparametric tests were selected. Continuous data were expressed as mean ± standard deviation or median (interquartile range), as appropriate. Univariate correlations between biomarker levels and disease measures were carried out using Spearman's correlation. A multivariate variable selection procedure using the algorithm of McHenry [27
] was initially used to select cytokine, MMP or TIMP variables that showed the strongest association with each clinical assessment. Multiple regression analysis or multivariate logistic regression analysis were used to investigate the association between clinical measures and biomarker levels while adjusting for other possible confounders. Where appropriate, data transformation to normality (log or square-root transformation) was carried out before analysis. When cytokine levels were below the level of detection, we carried out imputation of the lowest standard for that particular cytokine [28
Principal component analysis
Principal component analysis (PCA) is an exploratory technique that reduces the dimensionality of a large number of variables to a smaller set of uncorrelated independent components, and allows the identification of combinations of variables that best explain the differences between observations. PCA allows the identification of patterns within the variables and expresses them in a way that highlights the similarities and differences between them. Principal components (PCs) were extracted using varimax rotation, with the factor selection based on an eigenvalue cutoff of 1.0. The PCs identified were used in multivariate analysis to look for associations with clinical measures.
Hierarchical cluster analysis
Biomarker levels were first converted to log2 and expressed relative to the normalised mean value. These measurements were used to generate heat maps using Genesis software (version 1.7.2; Alexander Sturn, Institute for Genomics and Bioinformatics, Graz University of Technology, Graz, Austria). The Genesis program uses a hierarchical clustering method that enables groups of variables with similar expression levels to be clustered together, as well as grouping together patient samples with similar expression patterns.
Statistical analyses were carried out using Number Cruncher Statistical Software package for Windows (Number Cruncher Statistical System 2000; NCSS Statistical Software, Kaysville, UT, USA). The significance level was set at P = 0.05.