Skip to main content

Table 2 Model improvement with urine NGAL added to the clinical risk models for the endpoints

From: Urine NGAL as a biomarker for septic AKI: a critical appraisal of clinical utility—data from the observational FINNAKI study

 

AKI

Severe AKI (KDIGO 2–3)

RRT

90-day mortality

Goodness of fita (clinical risk model)

0.406

0.400

0.973

0.365

Goodness of fita (new model including uNGAL)

0.395

0.338

0.749

0.990

Events (n)

217

134

46

136

Nonevents (n)

267

350

438

348

AUC

 uNGAL alone

0.690 (0.647 to 0.731)

0.728 (0.686 to 0.767)

0.769 (0.729 to 0.806)

0.600 (0.555 to 0.644)

 Clinical risk model

0.717 (0.670 to 0.764)

0.759 (0.710 to 0.809)

0.724 (0.643 to 0.805)

0.797 (0.754 to 0.840)

 New risk model (uNGAL added)

0.749 (0.704 to 0.794)

0.799 (0.755 to 0.843)

0.824 (0.761 to 0.886)

0.804 (0.762 to 0.846)

 Difference, p (clinical model vs new model)

0.017

0.011

0.005

0.27

Category-free NRI (%)

 cfNRIevents

1.38 (− 12 to 14.77)

22.39 (6.11 to 38.67)

47.83 (21.91 to 73.74)

− 4.41 (− 21.06 to 12.23)

 cfNRInonevents

48.69 (38.25 to 59.13)

49.14 (39.99 to 58.29)

46.58 (38.54 to 54.62)

34.48 (24.67 to 44.3)

 cfNRI

50.07 (33.01 to 67.13)

71.53 (52.88 to 90.19)

94.4 (67.16 to 121.65)

30.07 (10.71 to 49.43)

IDI and summary statistics

 IDIevents

0.0248 (0.0099 to 0.0398)

0.0398 (0.0185 to 0.061)

0.0615 (0.0361 to 0.0868)

0.0115 (− 0.0003 to 0.0233)

 IDInonevents

0.0202 (0.0092 to 0.0312)

0.0152 (0.0049 to 0.0256)

0.0065 (− 0.0007 to 0.0136)

0.0045 (− 0.0008 to 0.0099)

 IDI

0.045 (0.0264 to 0.0637)

0.055 (0.0317 to 0.0782)

0.0679 (0.0413 to 0.0945)

0.016 (0.0032 to 0.0289)

 ISold

0.5263 (0.5006 to 0.5519)

0.408 (0.3711 to 0.4449)

0.1947 (0.1344 to 0.255)

0.4509 (0.4112 to 0.4905)

 ISnew

0.5511 (0.5226 to 0.5796)

0.4476 (0.4083 to 0.487)

0.2557 (0.1917 to 0.3197)

0.4623 (0.4214 to 0.5032)

 IPold

0.3849 (0.3651 to 0.4048)

0.2267 (0.2111 to 0.2422)

0.0846 (0.0781 to 0.0911)

0.2146 (0.1968 to 0.2325)

 IPnew

0.3648 (0.3425 to 0.3871)

0.2115 (0.1937 to 0.2293)

0.0782 (0.0693 to 0.0871)

0.2102 (0.192 to 0.2283)

  1. 95% confidence intervals are shown in parentheses. aA Hosmer–Lemeshow goodness of fit was used to test calibration of the models. “New” refers to the classification model that includes the new biomarker and “old” refers to the classification model that does not
  2. cfNRI = cfNRIevents + cfNRInonevents; IS, integrated sensitivity; IP, integrated 1-specificity; IDI = (ISnew – ISold) – (IPnew – IPold)