Development of an Individualized Prediction Model
Multivariable logistic regression analysis began with the following immune variables: CD3, CD8, CD56, CD68, PD-1, PD-L1, the distance between CD8 and CD56, CD8 and CD68, CD8 and PD-1, CD8 and PD-L1, CD56 and PD-1, CD56 and PD-L1, PD-1 and PDL-1. Immune signature was applied to develop a diagnostic model for pLN metastasis by using the two groups. An ROC curve is made for each meaningful marker and every cutoff value was emerged. Thereafter, these cutoff values are divided into two groups. At this point, we can make a diagnostic ROC curve and the AUC reach 0.843 (Figure 6A). The model that incorporated these above independent predictors was developed and presented as the nomogram (Figure 6B).
Internally Validation of the Prediction Model
With the nomogram, we then tested the remaining 40 percent of cases, and obtained a ROC curve with an AUC reaching 0.888. (Figure 6C).