Development of a Prognostic Scoring System for Secondary Acute Myeloid Leukemia (sAML) Arising from Myelodysplastic Syndromes (MDS), Myeloproliferative Disorders (MPD), or Therapy-Related AML (t-AML).

Background: Patients (pts) with sAML are evaluated together in large AML trials regardless of whether their sAML arises from antecedent MDS vs. from MPD vs. t-AML. Prognostic factors and outcomes may differ among these subgroups, and a prognostic scoring system would be helpful in determining which patients would benefit from induction chemotherapy.

Methods: We conducted a retrospective review of all pts with newly diagnosed, pathologically-confirmed AML at Cleveland Clinic between 1997 and 2007 to identify sAML pts treated with cytarabine-based induction chemotherapy. Data on known prognostic factors (age, white blood cell count (WBC) at diagnosis, cytogenetic risk groups (as defined by CALGB 8461), and AML etiology) were collected as baseline characteristics and controlled for in stepwise multivariable analyses. Complete response (CR) and overall survival (OS) were analyzed. A prognostic scoring system for OS was developed based on the number of poor prognostic features present, derived from significant multivariable factors. Pts received 1 point for adverse cytogenetics, 1 point for having 1-10% peripheral blasts, and 1 point for AML arising from MDS or MPD. Pts with 0 points were favorable, 1 point intermediate, 2 or more points unfavorable.

Results: Of 584 AML pts identified, 361 were treated with remission induction therapy, of whom 90 had AML arising from MDS, MPD, or t-AML. Thirty-nine (43%) had antecedent MDS, 21 (23%) an MPD, and 30 (33%) had t-AML, and 47% were female. Pts with AML arising from MDS were older at AML diagnosis (median of 67 years) vs. from MPD (61 years) and t-AML (60 years) (p=.02) but a shorter time from antecedent diagnosis/event (7 months, vs. 47 and 37 months, respectively (p<0.001)). Cytogenetic risk categories were favorable in 9 pts (10%), intermediate in 38 (42%), adverse in 27 (30%), and unknown in 16 (18%) and were similar among groups (p=.28). Median WBC at AML diagnosis was also lower for pts with AML from MDS (3.7k/uL, vs. 9.9 k/uL for AML from MPD and 8.9 k/uL for t-AML, p=.04), as were peripheral blasts (11%, vs. 29% for AML from MPD and 23% for t-AML, p=.01). The overall CR rate was 51% and was qualitatively lower for pts with AML from MDS (38%) than from MPD (62%) or t-AML (60%), but not significantly (p=0.11). The overall reinduction rate was 21% and did not differ among groups (p=.76). Median OS was significantly longer for pts with t-AML (15 months) vs. AML from MDS (8 months) or from MPD (11 months, p=.02). Available SNP karyotyping data on a subset of pts (6 from MDS, 1 from MPD, 4 with t-AML) did not reveal any shared or cryptic lesions that would distinguish responders from non-responders. In multivariable analyses, secondary AML etiology remained non-predictive of CR but was an independent prognostic factor for OS. Pts with t-AML had improved OS compared to AML from MDS (p=.01) or from MPD (p=.08). Favorable- or intermediate-risk cytogenetics, compared to adverse, were significant predictors of CR (p<.001) and OS (p<.001). Age <60 years (compared to ³60 years) was a significant predictor of CR (p=.02), but not OS. Patients with 1–10% peripheral blasts had worse OS vs. 0% or >10% blasts (p=.01). Using the prognostic scoring system described above, favorable pts (n=16) had a median OS of 40 months vs. 12 months for intermediate pts (n=33) and 4 months for unfavorable patients (n=27, p<.001).

Conclusions: Among sAML patients undergoing induction chemotherapy, pts with t-AML have a longer OS than AML from MDS or from MPD. Other important predictors of CR or OS were cytogenetics, age, and peripheral blast %. The sAML prognostic scoring system distinguished patients with improved survival from induction chemotherapy.