Skip to main content

Table 4 A comparison of the performance of conventional biomarkers vs. tenascin-C to predict the escalation of glucocorticoids in patients with SLE

From: Serum tenascin-C discriminates patients with active SLE from inactive patients and healthy controls and predicts the need to escalate immunosuppressive therapy: a cohort study

Cox model Variable
Continuous AIC C3 C4 Anti-dsDNA Anti-nucleosome Tenascin
 AIC   165.02 163.52 162.67 161.87 160.40
 C3 165.02 1.000 0.220 0.125 0.076 0.032
 C4 163.52 0.220 1.000 0.357 0.199 0.078
 Anti-dsDNA 162.67 0.125 0.357 1.000 0.370 0.132
 Anti-nucleosome antibodies 161.87 0.076 0.199 0.370 1.000 0.227
 Tenascin 160.40 0.032 0.078 0.132 0.227 1.000
Categorical   Low C3 Low C4 Anti-dsDNA+ Anti-nucleosome+ Tenascin+
 AIC   165.45 164.44 163.21 160.25 160.06
 Low C3 165.45 1.000 0.314 0.134 0.023 0.020
 Low C4 164.44 0.314 1.000 0.268 0.041 0.036
 Anti-dsDNA antibodies+ 163.21 0.134 0.268 1.000 0.085 0.076
 Anti-nucleosome antibodies+ 160.25 0.023 0.041 0.085 1.000 0.667
 Tenascin+ 160.06 0.020 0.036 0.076 0.667 1.000
  1. anti-dsDNA anti–double-stranded DNA
  2. This table compares univariate Cox regression models to predict the escalation of therapy by glucocorticoids based on the Akaike information criterion (AIC). The lower the value of the AIC, the better the fit of the model. Predictors are treated in the upper part of the table as continuous variables and in the lower part of the table as categorical variables. Paired comparisons of the quality of each model are illustrated by their respective p values. The thresholds for conventional biomarkers were reference values, and a cutoff value of 654 ng/ml for tenascin-C was generated using receiver operating characteristic curve analysis to find the optimal discriminatory threshold to identify patients who would require escalation of glucocorticoids. Boldface type indicates statistically significant values