Page 54 - SPONDYLOARTHRITIS IN COLOMBIA -

Outcome measurement
The outcome of interest in the current analyses was the request of SI-MRI or HLA-B27 by the rheumatologists. Information of this outcome was retrieved from the clinical file with a description of the test result as recorded in the medical chart. If information about SI-MRI or HLA-B27 was found, it was considered that this specific test had been requested by the rheumatologist (who decided to ask these test). Sacroiliitis on MRI was defined according to the local experienced musculoskeletal radiologist’s interpretation of the presence of active inflammatory changes in the SI-joint 20. According to usual clinical practice, no central reading was done. The HLA-B27 test was performed locally in the laboratory of immunology.
Statistical analysis
Descriptive statistics were used to calculate mean (± SD) for continuous data and percentages for categorical data. Explanatory variables were “SI-MRI-ordered” (yes vs no) and “HLA-B27-ordered” (yes vs no).
Age at symptom onset, age at assessment, symptoms at presentation and disease duration had less than 20% of missing data. BASFI, BASDAI, ASDAS-CRP, ASDAS-ESR, CRP and ESR had more than 20% of missing data. All other explanatory variables had complete data. In case of missing data,
imputation was performed by using the median or mode from the variable’s distribution as appropriate, but only variables with complete data in at least 80% of the cohort were included in the multivariable analysis (see below).
Patient´s characteristics were examined per subgroups of SI-MRI ordered (yes vs no) or HLA-B27 ordered (yes vs no) and analysed by univariable analysis (either by chi-square or by t-test if appropriate) followed by logistic regression analysis. Variables with a p-value <0.05 in univariable
52