Validation and update of a multivariable prediction model for the identification and management of patients at risk for hepatocellular carcinoma

Table. 2 Characteristics and performance of the original and updated ASAP diagnostic prediction models

Characteristics	ASAP model¹	Recalibration-in-the-large	Recalibration	Model revision
Calibration parameters
Intercept	0	− 3.243	− 3.577	–
Slope	1	1	1.192	–
Model coefficients²
Intercept	− 7.57711770	− 10.82011770	− 12.60892	− 12.95521
b_age	0.04666357	0.04666357	0.0556229754	0.06178
b_sex	− 0.57611693	− 0.57611693	− 0.6867313806	− 0.7449
b_AFP	0.42243533	0.42243533	0.5035429134	0.55643
b_PIVKA-II	1.10518910	1.10518910	1.3173854072	1.28268
Model performance
Calibration intercept	− 3.243	− 0.000	− 0.000	− 0.000
Calibration slope	1.192	1.192	1	1
Residual deviance	851.25	212.552	208.489	207.962
df	1012	1011	1010	1007
LRT chi-square p-value³	< 0.001	0.332	0.913	–
Emax (95% CI)	0.526 (0.483–0.557)	0.053 (0.042–0.159)	0.087 (0.049–0.198)	0.082 (0.043–0.185)
Eavg (95% CI)	0.246 (0.230–0.262)	0.009 (0.007–0.018)	0.006 (0.004–0.015)	0.005 (0.004–0.015)
C-statistic (95% CI)	0.982 (0.970–0.990)	0.982 (0.972–0.991)	0.982 (0.971–0.991)	0.982 (0.971–0.991)
Brier score	0.127107476	0.028182937	0.0280551	0.02789345
AIC	851.25	214.55	212.49	217.96

AFP alpha-fetoprotein, AIC akaike information criterion, CI confidence interval, df degrees of freedom, Eavg average absolute difference in the predicted and calibrated probabilities, Emax maximum absolute difference in the predicted and calibrated probabilities, LRT likelihood ratio test, PIVKA-II protein induced by vitamin K absence or antagonist-II
¹The model coefficients are derived from the original ASAP model reported by Yang et al [8]
²b_age, b_sex, b_AFP, and b_PIVKA-II are the regression coefficients of age, sex, AFP, and PIVKA-II, respectively
³Comparison with the model revision using the methods described by Vergouwe et al [15]

ISSN: 1559-0275