TABLE 4.

Comparison of AUC of predictive models for solid and micropapillary components

Models		Predictors	AUC
			Training (2017–2019)	p	Validation (2020)	p
Solid component	Model 1	Gender + tumor size	0.723(0.687–0.762)	Ref.	0.704 (0.661–0.751)	Ref.
	Model 2	Model 1+ CEA	0.771(0.726–0.812)	0.009	0.747 (0.706–0.782)	0.034
	Model 3	Model 2 + Ki‐67 + EGFR + tumor differentiation	0.962(0.948–0.979)	<0.001	0.942 (0.908–0.968)	<0.001
Micropapillary component	Model 1	Gender + tumor size	0.699(0.636–0.760)	Ref.	0.711 (0.675–0.750)	Ref.
	Model 2	Model 1 + CEA	0.746(0.703–0.788)	0.045	0.753 (0.722–0.784)	<0.001
	Model 3	Model 2 + Ki‐67 + tumor differentiation	0.837(0.790–0.881)	0.002	0.901 (0.874–0.942)	<0.001

P value of <0.05 was marked in bold.