AML risk stratification models utilizing ELN-2017 guidelines and additional prognostic factors: a SWOG report

Biomarker Research

Table 3 Multivariable models for CR, OS and RFS

Models	AUCs or C-statistics
Models	MNCs		VLBs
Complete Response (CR, AUC)	Discovery	Validation	Discovery	Validation
ELN₂₀₁₇	0.66	0.73	0.66	0.67
AGE + ELN₂₀₁₇	0.71	0.72	0.71	0.72
AGE + PS + GATA2 + BAALC (MNCs)	0.77^a	0.67	N/A	N/A
AGE + PS + MN1 + GATA2 (VLBs)	N/A	N/A	0.72^a	0.70
Overall Survival (OS, C Statistic)	Discovery	Validation	Discovery	Validation
ELN₂₀₁₇	0.60	0.68	0.59	0.68
AGE + ELN₂₀₁₇	0.72	0.70	0.71	0.71
AGE + ELN₂₀₁₇ + ERG1 + EVI1 + JAG1 + JAG1^aELN (MNCs)	0.73^a	0.69	N/A	N/A
AGE + ELN₂₀₁₇ + EVI1 + ERG1 + CCNA1 (VLBs)	N/A	N/A	0.71^a	0.73
Relapse-Free Survival (RFS, C Statistic)	Discovery	Validation	Discovery	Validation
ELN₂₀₁₇^b	0.60	0.54	0.61	0.54
AGE + ELN₂₀₁₇^b	0.67	0.63	0.66	0.62
AGE + ELN₂₀₁₇^b + EVI1 + CCNA1 (MNCs)	0.69^a	0.65	N/A	N/A
AGE + ELN₂₀₁₇^b + EVI1 + CCNA1 (VLBs)	N/A	N/A	0.72^a	0.65

^aAUCs or C-statistics for the integrated models come from cross-validation. 1/5 of the discovery data were used to build a model, and the remaining 4/5 of the data were fit to this model, resulting in five C-statistics or AUCs. The mean of the five is presented in the table
^bDue to small sample size, ELN₂₀₁₇ in the RFS models was categorized into adverse vs. not adverse (including intermediate, favorable, and unknown)

ISSN: 2050-7771