Table 2.
AUC (ROC Curve), IDI, and NRI for logarithmically transformed uNGAL creatinine corrected [log2(uNGALcc)] and pNGAL
| Variables per model | AUC (ROC curve) | P value | IDI score | P value | NRI score | P value |
|---|---|---|---|---|---|---|
| No variables | 0.5 | — | — | — | — | — |
| uNGALcc only (leukocyte-negative) | 0.714 | 0.027* | — | — | — | — |
| pNGAL only | 0.546 | 0.590† | — | — | — | — |
| Covariates only | 0.744 | — | — | — | — | — |
| Covariates only (leukocyte-negative) | 0.750 | — | — | — | — | — |
| uNGALcc + covariates (leukocyte-negative) | 0.849 | 0.157‡ | 0.157 | 0.016§ | 0.638 | 0.046‖ |
| pNGAL + covariates | 0.751 | 0.419¶ | −0.002 | 0.496 | −0.190 | 0.523 |
Established risk factors (clinical covariates) were BMI, HbA1c, and total daily insulin. Bold P values are statistically significant.
*AUC of “uNGALcc-only” model (leukocyte-negative) is significantly different to AUC = 0.5.
†AUC of “pNGAL-only” model is not significantly different to AUC = 0.5.
‡uNGALcc (leukocyte-negative) improved the AUC relative to AUC of 0.750 for a logistic model containing only the covariates, although this was not significant (P = 0.157).
¶Relative to AUC of 0.744 for a logistic model containing only the covariates.
§Statistically significant improvement (P = 0.016) in predicted risk of PE after the addition of uNGALcc to model with covariates only.
‖Statistically significant improvement (P = 0.046) in the reclassification of preeclampsia risk after the addition of uNGALcc to model with covariates only.