Skip to main content
. Author manuscript; available in PMC: 2020 Mar 1.
Published in final edited form as: Arthritis Rheumatol. 2019 Jan 24;71(3):451–459. doi: 10.1002/art.40727

Table 3:

Diagnostic accuracy of serum S100A8/A9 and S100A12 levels at the time of anti-TNF withdrawal regarding the prediction of flare within 30 days, 60 days, 90 days or 8 months.

S100A8/A9 (cut-off 690 ng/ml) S100A12 (cut-off 120 ng/ml)
30 days 60 days 90 days 8 months 30 days 60 days 90 days 8 months
Sensitivity 67% 38% 40% 66% 56% 54% 53% 33%
Specificity 42% 58% 58% 45% 70% 70% 71% 68%
Positive likelihood ratio 1.16 0.91 0.95 1.20 1.85 1.84 1.86 1.05
Youden index 0.09 −0.04 −0.02 0.11 0.25 0.24 0.25 0.02
Hazard ratio*(95% CI), P 0.68 (0.18–2.58), 0.57 0.83 (0.28–2.52), 0.75 0.89 (0.32–2.49), 0.82 0.72 (0.36–1.42), 0.34 2.83 (0.70; 11.49), 0.15 2.63 (0.82; 8.43), 0.10 2.59 (0.87; 7.67), 0.09 1.13 (0.55; 2.33), 0.74
AUC (95% CI) 0.51 (0.31; 0.70) 0.51 (0.33; 0.69) 0.51 (0.34; 0.67) 0.56 (0.44; 0.67) 0.64 (0.50; 0.77) 0.66 (0.54; 0.79) 0.64 (0.51; 0.77) 0.51 (0.39; 0.62)
*

hazard ratio for the occurrence of disease flare if level is above threshold

AUC, area under the receiver operating characteristic curve; CI, confidence interval

The Youden index is calculated as follows: sensitivity + specificity - 1