Skip to main content
. Author manuscript; available in PMC: 2016 Nov 1.
Published in final edited form as: Arthritis Rheumatol. 2015 Nov;67(11):2905–2915. doi: 10.1002/art.39279

Table 3.

Relationship of plasma lipid (PGE2), PBL inflammatory gene transcriptome and demographic traits with radiographic joint space narrowing (JSN) over 24 months in a multivariate analysis.

Combined Predictors AUC Estimate AUC 95% Confidence Interval p value p value adjusted
JSN>0mm vs. JSN≤ 0mm (N=72 vs.39) Age, Gender, BMI 0.5 0.39 0.62 0.473 0.473
Age, Gender, BMI, COX-2 0.65 0.55 0.76 0.002 0.017
Age, Gender, BMI, IL-1β 0.51 0.39 0.62 0.446 0.464
Age, Gender, BMI, TNFα 0.44 0.33 0.55 0.153 0.229
Age, Gender, BMI, COX-2, IL-1β 0.65 0.54 0.76 0.005 0.022
Age, Gender, BMI, COX-2, TNFα 0.61 0.5 0.72 0.027 0.072
Age, Gender, BMI, IL-1β, TNFα 0.46 0.35 0.58 0.272 0.295
Age, Gender, BMI, COX-2, IL-1β, TNFα 0.61 0.5 0.72 0.030 0.072
Age, Gender, BMI, COX-2, IL-1β, TNFα, PGE2 0.59 0.48 0.7 0.062 0.107
JSN>0.2mm vs. JSN≤ 0mm (N=64 vs. 39) Age, Gender, BMI 0.57 0.46 0.68 0.094 0.149
Age, Gender, BMI, COX-2 0.67 0.56 0.78 0.002 0.017
Age, Gender, BMI, IL-1β 0.6 0.48 0.71 0.047 0.089
Age, Gender, BMI, TNFα 0.54 0.42 0.65 0.257 0.295
Age, Gender, BMI, COX-2, IL-1β 0.67 0.56 0.77 0.001 0.017
Age, Gender, BMI, COX-2, TNFα 0.65 0.55 0.76 0.003 0.017
Age, Gender, BMI, IL-1β, TNFα 0.56 0.44 0.67 0.174 0.246
Age, Gender, BMI, COX-2, IL-1β, TNFα 0.66 0.54 0.77 0.003 0.018
Age, Gender, BMI, COX-2, IL-1β, TNFα, PGE2 0.63 0.52 0.74 0.008 0.032
JSN>0.5mm vs. JSN≤ 0mm (N=44 vs. 39) Age, Gender, BMI 0.55 0.43 0.68 0.211 0.262
Age, Gender, BMI, COX-2 0.62 0.49 0.74 0.032 0.072
Age, Gender, BMI, IL-1β 0.56 0.43 0.68 0.182 0.246
Age, Gender, BMI, TNFα 0.54 0.41 0.67 0.273 0.295
Age, Gender, BMI, COX-2, IL-1β 0.64 0.52 0.76 0.012 0.041
Age, Gender, BMI, COX-2, TNFα 0.61 0.48 0.73 0.048 0.089
Age, Gender, BMI, IL-1β, TNFα 0.55 0.42 0.68 0.2136 0.2621
Age, Gender, BMI, COX-2, IL-1β, TNFα 0.63 0.51 0.75 0.02 0.06
Age, Gender, BMI, COX-2, IL-1β, TNFα, PGE2 0.61 0.48 0.73 0.0496 0.0893

For measuring a biomarker’s predictivity for radiographic progression in medial knee OA based on threshold of JSN progression (>0.0, >0.2 and >0.5 mm/24 months) as outcomes, we used the ROC curve with multivariate predictive model support vector machines. The null model against which all other predictive models are compared and p-values are calculated is the random model (i.e., coin-flipping model). The random model has AUC of 0.5.