Prediction gain by DcR3 over a model of traditional cardiovascular risk factors
| Multivariate Cox Hazards Model | HR (95% CI) | P Value | AICa | ΔAICb |
|---|---|---|---|---|
| Cardiovascular mortality | ||||
| cardiovascular risk factors | 335.5 | 0.0 | ||
| + VCAM-1 (1 SD unit) | 1.8 (0.7−4.4) | 0.15 | 346.7 | 11.2 |
| + albumin (1 g/L) | 0.9 (0.6−1.3) | 0.48 | 345.0 | 9.5 |
| + hs-CRP (1 SD unit) | 1.6 (0.9−2.5) | 0.05 | 341.4 | 5.9 |
| + IL-6 (1 SD unit) | 1.7 (1.1−2.7) | 0.01 | 339.3 | 3.8 |
| + DcR3 (1 SD unit) | 1.5 (1.1−2.4) | 0.03 | 341.5 | 6.0 |
| All-cause mortality | ||||
| cardiovascular risk factors | 872.5 | 0.0 | ||
| + VCAM-1 (1 SD unit) | 1.7 (1.0−3.1) | 0.03 | 889.6 | 17.1 |
| + albumin (1 g/dl) | 0.7 (0.5−0.9) | 0.01 | 882.7 | 10.2 |
| + hs-CRP (1 SD unit) | 1.3 (0.9−1.7) | 0.05 | 892.3 | 19.8 |
| + IL-6 (1 SD unit) | 1.6 (1.2−2.1) | 0.01 | 885.4 | 12.9 |
| + DcR3 (1 SD unit) | 1.5 (1.2−2.0) | <0.01 | 884.0 | 11.5 |
DcR3, decoy receptor 3; HR, hazard ratio; 95% CI, 95% confidence interval; AIC, Akaike Information Criterion; VCAM-1, vascular cell adhesion molecule-1; hs-CRP, high-sensitive C-reactive protein.
↵a AIC (15) for a given model is a function of its maximized log-likelihood and the number of estimable parameters (K): AIC = −2 (maximized log-likelihood) + 2K. In the baseline Cox proportional hazards model, there were six traditional cardiovascular risk factors, including age, sex, smoking status, diabetes, total cholesterol, and systolic BP. VCAM-1, albumin, hs-CRP, IL-6, or DcR3 was added subsequently in order to assess the independent predictive contribution by DcR3.
↵b ΔAIC is the difference between the AIC of baseline model and the AIC of VCAM-1, albumin, hs-CRP, IL-6, or DcR3, respectively. Therefore, the lower the ΔAIC, the higher prediction gain of pertinent biomarker.