Prognostic Import of French-American-British (FAB) System As Embedded in 2008 Revision of World Health Organization Classification of AML: Review of SWOG Data

After many years of use, the French-American-British (FAB) system for classification of AML has been replaced by that of the World Health Organization (WHO), as revised in 2008 (Blood 2009;114: 937–51). While unlike FAB, WHO uses genetic and clinical information to specify AML subtypes, the FAB system remains embedded in the WHO category of “AML not otherwise specified” (NOS). Here we ask whether the subdivision by FAB of AML NOS is of prognostic significance.

FAB data were available on 970 newly-diagnosed patients treated on various SWOG protocols from 1986 to 2009. The protocols generally called for 3+7 type induction therapy. Exclusion of patients with recurrent genetic abnormalities, therapy-related AML (T-AML), or AML with myelodysplasia related changes (MDS-AML), as stipulated in WHO 2008 (NPM1 and CEBPA mutation status was unavailable), left 447 with AML NOS whose FAB type was recorded. We used multivariate analyses adjusting for age, sex, cytogenetic risk (SWOG criteria), pretreatment performance status, blood counts, and marrow blasts, to assess whether FAB was independently associated with outcome in these 447 patients and used areas under the receiver operator characteristic curve (AUC) to quantify the predictive ability of FAB, with AUC of 1.0 indicating perfect prediction and an AUC of 0.5 equivalent to a coin flip.

The 447 patients had median age of 58 years, CR rate 56% and median survival 17.2 months. Given the association between T-AML or MDS-AML and FAB types M6 or M7 (p=0.02 in this dataset), only 2% of the 447 patients were M6 or M7; 8% were M0, 27% M1, 26% M2, 25% M4, and 12% M5 (we excluded patients with M3). CR rates were 47% M0, 56% M1, 53% M2, 54% M4, 72% M5, and 45% M6 or M7. Survival probability was longest with M6/M7 followed by M5. The multivariate analyses showed the following for the effect of FAB (considered as M0, M1, M2, M4, M6/M7):

Results were similar regardless of whether we considered FAB as above or grouped as M0, M1/M2, M4/M5, M6/M7. In particular, FAB was not associated with any outcome and its removal from more complete adjusted models had little effect on the models' predictive ability.

These results question the utility of the FAB system as embedded in WHO 2008.

No relevant conflicts of interest to declare.