Table I.
n∗ | Median ADAMTS-4 ng/mL (IQR) | P-value† | Coefficient‡ (95% CI) | Concordancec (95% CI) | Nagelkerke'sR2 (95% CI) | P-value§ | |
---|---|---|---|---|---|---|---|
Univariable analysis | |||||||
Age | 170 | 0.08 (0.04–0.13) | 0.69 (0.62–0.78) | 0.10 (0.03–0.19) | <0.001 | ||
OA | 159 | <0.001 | 0.66 (0.59–0.73) | 0.11 (0.04–0.22) | <0.001 | ||
No | 118 | 0.0 (0–0.3) | 2.4 (1.2–3.6) | ||||
Yes | 41 | 1.3 (0–4.1) | |||||
Injury | 160 | 0.10 | 0.58 (0.52–0.65) | 0.03 (0.00–0.12) | 0.10 | ||
None | 32 | 0 (0–0) | |||||
Chondral | 95 | 0 (0–2.2) | 1.7 (0.2–3.3) | ||||
Other | 33 | 0 (0–2.8) | 1.4 (−0.4–3.3) | ||||
Effusion | 122 | <0.001‖ | 0.78 (0.69–0.86) | 0.31 (0.17–0.49) | <0.001 | ||
No | 71 | 0.0 (0.0–0.0) | |||||
Small | 28 | 0.0 (0.0–0.75) | 1.8 (0.4–3.2) | ||||
Moderate | 14 | 1.7 (0.2–3.4) | 3.6 (1.9–5.3) | ||||
Large | 9 | 8.5 (4.0–12.3) | 5.4 (3.4–7.3) | ||||
Inflammation | 144 | <0.001 | 0.68 (0.60–0.76) | 0.11 (0.03–0.24) | <0.001 | ||
No | 62 | 0.0 (0.0–0.0) | |||||
Yes | 82 | 0.2 (0.0–2.8) | 2.4 (1.1–3.6) | ||||
Multivariable analysis | |||||||
Full model | 104 | 0.88 (0.83–0.96) | 0.46 (0.36–0.66) | ||||
Age | 0.72¶ | 0.13¶ | 0.07 | ||||
OA | 0.65 | 0.10 | 0.64 | ||||
Injury | 0.54 | 0.01 | 0.86 | ||||
Effusion | 0.79 | 0.34 | <0.001 | ||||
Inflammation | 0.70 | 0.15 | 0.04 | ||||
Reduced model | 106 | 0.88 (0.83–0.95) | 0.46 (0.34–0.63) | ||||
Age | 0.05 (0.01–0.09) | 0.72¶ | 0.13¶ | 0.02 | |||
Effusion | 0.79 | 0.34 | <0.001 | ||||
No | |||||||
Small | 1.7 (0.4–3.0) | ||||||
Moderate | 3.3 (1.7–4.8) | ||||||
Large | 4.4 (2.6–6.2) | ||||||
Inflammation | 0.71 | 0.16 | 0.006 | ||||
No | |||||||
Yes | 1.8 (0.5–3.0) |
Notes (IQR: interquartile range; CI: confidence intervals).
Number of patients for each analysis, which vary due to data missingness.
P-values from univariate non-parametric statistical analyses.
Coefficients from censored regression.
P-values from censored regression.
P-value from Jonckheere trend test.
Concordance and Nagelkerke's R2 values for individual predictors in the full and reduced models are the univariable values when calculated using each model's dataset. They are therefore NOT partial concordance or R2 coefficients.