Skip to main content
. 2018 Jan 4;18:1. doi: 10.1186/s12874-017-0458-6

Table 2.

Comparison of the patient-level true positive rate (TPR(·,τ1,τ2)) when the threshold for each screening algorithm was chosen such that the screening-level false positive rate is 10%, i.e FPR(·,τ1)=0.1. In each definition, the choice of the parameters τ1 and τ2 varies

Results from validation cohort
Screening algorithm A1 B1 C1 D1 A2 B2 C2 D2
AFP only 0.5753 0.5672 0.5442 0.5388 0.4019 0.4099 0.3564 0.3361
AFP+Lab+ ΔAFP 0.6137 0.6119 0.5907 0.5809 0.4766 0.4820 0.4158 0.3770
PEB: AFP 0.6055 0.6045 0.6023 0.6364 0.4579 0.4955 0.4653 0.4891
Number of HCC cases 365 402 430 451 107 222 303 366

A1: τ1=6 months and τ2=0, B1: τ1=12 months and τ2=0, C1: τ1=24 months and τ2=0, D1: τ1 is the maximum follow-up time and τ2=0. A2: τ1=6 months and τ2=3 months, B2: τ1=12 months and τ2=3 months, C2: τ1=24 months and τ2=3 months, D2: τ1 is the maximum follow-up time and τ2=3 months. AFP+Lab+ ΔAFP: updated laboratory-based algorithm, PEB: AFP: parametric empirical Bayes algorithm applied to AFP