Table 4.
Comparison of different esophageal cancer risk prediction models
Model comparison | Difference of AUC | Za | P valuea | NRI | Zb | P valueb |
---|---|---|---|---|---|---|
Training set | ||||||
Genetic vs. non-genetic | 0.032(-0.028,0.091) | 1.034 | 0.301 | 0.041 | 0.755 | 0.450 |
Genetic vs. combined | 0.089(0.048,0.130) | 4.256 | < 0.001 | 0.122 | 3.234 | 0.001 |
Non-genetic vs. combined | 0.058(0.030,0.086) | 4.067 | < 0.001 | 0.082 | 2.082 | 0.037 |
Validation set | ||||||
Genetic vs. non-genetic | 0.004(-0.072,0.079) | 0.091 | 0.927 | -0.002 | 0.020 | 0.984 |
Genetic vs. combined | 0.060(0.007,0.113) | 2.223 | 0.026 | 0.075 | 1.214 | 0.225 |
Non-genetic vs. combined | 0.057(0.021,0.092) | 3.084 | 0.002 | 0.076 | 2.128 | 0.033 |
NOTE: The difference of AUC was analyzed using Delong’s test. NRI, net reclassification improvement.
a represents the z-statistic and P value from Delong’s test;
b represents the z-statistic and P value from NRI analysis