Acute myeloid leukemia: 2012 update on diagnosis, risk stratification, and management


  • Elihu H. Estey

    Corresponding author
    1. Division of Hematology, University of Washington
    2. Clinical Research Division, Fred Hutchinson Cancer Research Center
    • c/o Seattle Cancer Care Alliance, 825 Eastlake Ave E, MS G3-200, Seattle, WA 98109-1023
    Search for more papers by this author

  • Conflict of interest: Nothing to report


Disease Overview: Acute myeloid leukemia (AML) results from accumulation of abnormal immature cells in the marrow. These cells interfere with normal hematopoiesis can escape into the blood and infiltrate lung and CNS. The most common cause of death is bone marrow failure. It is likely that many different mutations and/or epigenetic aberrations can produce the same disease, with these differences responsible for the very variable response to therapy, which is AML's principal clinical feature.

Diagnosis: This rests on demonstration that the marrow or blood has >20% blasts of myeloid lineage. Blast lineage is assessed by multiparameter flow cytometry with CD33 and CD13 being surface markers typically expressed by myeloid blasts. It should be realized that clinical/prognostic considerations, not the blast % per se, should be the main factor determining how a patient is treated.

Risk stratification: Two features determine risk: the probability of treatment-related mortality (TRM) and, more important, even in patients aged >75 with Zubrod performance status 1, the probability of resistance to standard therapy despite not incurring TRM. The chief predictor of resistance is cytogenetics with a monosomal karyotype (MK) denoting the disease is essentially incurable with standard therapy even if followed by a standard allogeneic transplant (HCT). The most common cytogenetic finding is a normal karyotype (NK) and those of such patients with an NPM1 mutation but no FLT3 internal tandem duplication (ITD), or with a CEBPA mutation, have a prognosis similar to that of patients with the most favorable cytogenetics [inv(16) or t(8;21)] (60–70% cure rate). In contrast, NK patients with a FLT3 ITD have only a 30–40% chance of cure even after HCT. Accordingly analyses of NPM1, FLT3, and CEBPA should be part of routine evaluation, much as is cytogenetics. Risk is best assessed considering several variables simultaneously rather than, for example, only age.

Risk-adapted therapy: Patients with inv(16) or t(8;21) or who are NPM1+/FLT3ITD− can receive standard therapy (daunorubicin + cytarabine) and should not receive HCT in first CR. It seems likely that use of a daily daunorubicin dose of 90 mg/m2 will further improve outcome in these patients. There appears no reason to use doses of cytarabine > 1 g/m2 (for example, bid × 6 days), as opposed to the more commonly used 3 g/m2. Patients with an unfavorable karyotype (particularly MK) are unlikely to benefit from standard therapy (even with dose escalation) and are thus prime candidates for clinical trials of new drugs or new approaches to HCT; the latter should be done in first CR. Patients with intermediate prognoses (for example, NK and NPM and FLT3ITD negative) should also receive HCT in first CR and can plausibly receive either investigational or standard induction therapy, with the same prognostic information about standard therapy leading one patient to choose the standard and another an investigational option. Am. J. Hematol. 87:90–99, 2012. © 2011 Wiley Periodicals, Inc.

Disease Overview

Acute myeloid leukemia (AML) is a disease of aging with a median age of 67 years at diagnosis [1]. The most commonly recognized antecedent is cytotoxic therapy for a solid tumor [therapy-related—(t-)AML]. Older patients with seemingly spontaneously arising (de novo) AML are more likely than younger de novo patients to have cytogenetic abnormalities identical to those seen in t-AML suggesting that cumulative exposure to environmental analogues of cytotoxic therapies predisposes to “de novo” AML in many older patients [2]; indeed the incidence of the AML characterized by other cytogenetic findings, such as t(15;17), t(8;21), or inv(16), is not age dependent. AML is also commonly associated with an antecedent hematologic disorder (AHD), most frequently myelodysplasia (MDS), less commonly a myeloproliferative, or plasma cell, neoplasm [3]. AML developing after an AHD is also associated with the cytogenetic abnormalities seen in t-AML, such as a “monosomal karyotype” (MK). Although older patients have poorer outcomes than younger patients, the age effect is largely explained by the associations between older age and (a) poor performance status, (b) comorbidities, (c) t-AML and/or AHD (collectively known as “secondary AML”), and most importantly, (d) specific cytogenetic abnormalities [4–5].

AML is thought to commonly arise in a very small population of quiescent stem cells recognized by their expression of CD34 and CD123, but not CD38 [6]. Most such cells lose their stem cell properties and proceed to proliferate, differentiate, and ultimately die. However, those remaining act as a reservoir maintaining the great bulk of the AML population. While remissions reflect reduction inthis bulk population, cure of AML probably requires management of the stem cell compartment; the high frequency of relapse reflects a failure to accomplish this. AML stem cells are less sensitive than the bulk population to therapeutic agents, such as cytarabine (ara-C), and the larger the stem cell compartment the worse the prognosis [7]. However, some AMLs, for example, those associated with t(15;17), t(8;21), or inv(16), probably arise in relatively more differentiated normal hematopoietic cells possibly explaining why they are more curable than AMLs of stem cell origin, which are often characterized by a MK [8]. These observations indicate that AML comes in several (probably many) forms, with management dependent on the specific type of AML.


Patients typically present with symptoms of anemia. If bleeding is prominent, it is crucial to immediately rule in/rule out acute promyelocytic leukemia (APL), particularly if the WBC is > 10,000/μl. Morphologic examination, using blood if it contains >1,000 blasts/μl), can be employed for this purpose and in particular to justify immediate institution of arsenic trioxide (ATO) + all-transretinoic acid (ATRA). The diagnosis of APL must however be confirmed by demonstrating the pathognomonic PML-RARα rearrangement and the resultant 15:17 translocation; this is typically done using fluorescent in situ hybridization (FISH) [9].

The diagnosis of non-APL AML requires >20% blasts of myeloid lineage in marrow or blood [World Health Organization (WHO) criteria] [10]. Multicolor flow cytometry (MFC) has largely replaced histochemical staining for establishment of blast lineage. Myeloid lineage antigens include CD33, CD13, CD117 (CKIT), CD14, CD64 CD41, and glycophorin A. If there are >1,000–2,000 blasts/μl blood specimens suffice for diagnostic and stratification purposes and obtaining a marrow only delays treatment [11].

The 20% blast criterion contrasts with the previous 30% [12], suggesting the arbitrary nature of any cutoff. Indeed, although patients with 10–19% blasts (“MDS” by WHO criteria) are often ineligible for “AML protocols,” the natural histories of AML and MDS with 10–20% blasts are very similar and therapeutic outcome may depend more on covariates such as cytogenetics, de novo vs. secondary AML, and age than on the distinction between “MDS” and “AML,” with prognostically unfavorable characteristics more common in MDS [13]. Thus, the 20% cut-off should serve only as an approximate guide to clinical decisions, with specific treatment resting on benefit/risk considerations. For example, the WHO regards patients with <20% blasts but with t(8;21) or inv(16) as AML because they respond so well to AML therapy. In contrast, frail patients who present with low WBC counts might appropriately be treated as MDS even if the blast count is >20%.

Risk Stratification

After the diagnosis is made, the first consideration is the need for emergency treatment, as based on high (>10,000 in APL, >50,000 in other types), or rapidly rising WBC, or symptoms suggestive of pulmonary of CNS infiltration. Although leukapheresis likely reduces the risk of leukostasis, or of tumor lysis once definitive therapy begins, it is of unclear benefit in removing blasts from organs, not obviously associated with improved outcomes [14], and often delays administration of chemotherapy. Although frequently given to reduce WBC, hydroxyurea has much less activity than ara-C [15] and is less likely to enter lung or CNS than high doses (1–3 g/m2) of the latter. Although investigational treatment protocols typically exclude patients given ara-C, rather than hydroxyurea to reduce a high WBC, ara-C should be used if there are CNS or pulmonary symptoms and if protocol treatment cannot be administered promptly.

If emergency treatment is unnecessary, information should be gathered to assess prognosis with standard therapy, most commonly 3 days of an anthracycline and 7 of ara-C (3 + 7) [16]. If prognosis is unsatisfactory, investigational treatment should be advised. It is tempting to begin all patients on 3 + 7 before prognostic information becomes available, while relying on an allogeneic transplant (HCT) once patients found to have a poor prognosis after induction therapy begins achieve CR. However, patients, who although in CR by conventional criteria, have minimal residual disease (MRD) at time of HCT are much more likely to subsequently relapse [17]. Since MRD at HCT reflects inadequacy of pre HCT chemotherapy, consideration should be given to investigational induction therapy in poor prognosis patients. In cooperative group studies, rates of treatment-related mortality (TRM) are <15–20% in patients, even those aged >75, with performance status 1 [4], and hence the major cause of failure to enter CR is resistance to therapy (similarly the risk of relapse far outweighs that of death in CR even in patients in their 70s [18]). Nonetheless the risk of TRM, not to mention morbidity, does not appear to justify use of 3 + 7 to patients with high probabilities of resistance (no CR) to it.

Predictors of Sensitivity/Resistance to Standard Therapy

Cytogenetics: These can be classified as follows (Fig. 1):

Figure 1.

Survival in newly diagnosed AML according to pre-treatment cytogenetics [19].


Collectively known as core-binding factor (CBF) abnormalities, inversion 16 [inv(16)] or t(8;21) are associated with CR rates >90% and 3-year event-free survival rates of 60%, with relapse rates falling precipitously past 3 years for all types of AML [20]. Nonetheless, inv(16) and t(8;21) differ in that subsequent CRs after relapse are more common in the former [21]. However in patients age > 65 [22], inv(16) patients age > 35–40 [23], or t(8;21) patients with initial WBC >5,000–10,000 [24], 3-year success rates are closer to 20–30%. The presence of CKIT mutations also seems to affect outcome in CBF AML as noted below.


Multiple cooperative groups have identified monosomies of chromosomes 5 or 7, deletions of the long arm of chromosome 5, abnormalities of the long arm of chromosome 3, and “complex abnormalities” involving at least three distinct aberrations as comprising a high-risk group in younger and older patients [25]. However more recent findings strongly suggest that “monosomal karyoype” (MK) should be the primary criterion for “worst prognosis cytogenetics” [19, 26]. The deleterious effect of complex karyotypes largely reflects associations with MK, defined as at least two autosomal monosomies, or one autosomal monosomy in conjunction with a structural abnormality; indeed noncomplex MK does as poorly as complex MK. In various SWOG studies CR rates following 3 + 7 in patients with MK were: age 31–40 (27%) (13 patients), age 41–50 (14%) (23 patients), age 51–60 (24%) (43 patients), and age > 60 (13%) (90 patients) [19]. Even in patients age < 60 median survival among MK+ patients is only about 6 months [26].

Unfavorable without MK.

This most commonly includes patients with monosomies of chromosomes 5 or 7, deletions of the long arm of chromosome 5, abnormalities of the long arm of chromosome 3. Their CR rates even in younger patients rarely exceed 60–65%.


Some restrict this category to patients with a normal karyotype (NK) while others also include patients with abnormalities other than those in categories (a)-(c). NK is the most frequent cytogenetic finding and has the most variable prognosis with standard therapy, ranging from cure to never achieving CR.

Molecular markers

Given the extremely heterogeneous outcome of NK AML testing for molecular abnormalities is of most use in this subset [27].

Established markers.

NPM1 and FLT3 ITD.

Approximately 50% of NK patients will have an NPM1 mutation and 1/3 an internal tandem duplication (ITD) in the FLT3 gene. Such patients tend to present with high WBC. On average patients with mutated NPM1 but without FLT3 ITD have prognoses similar to those seen in CBF AML [28]. In contrast patients with a FLT3 ITD typically have prognoses resembling those in the “unfavorable without MK” group [28]. However, the effect of a FLT3 ITD likely depends on the presence of other mutations (for example, an NPM1 mutation lessens the effect of a FLT3ITD) or the amount of the mutant ITD protein (allelic burden). For example, data from the MRC in the United Kingdom indicate that the difference in cumulative incidence of relapse between patients with allelic burdens >50% and patients with allelic burdens 1–49% is similar to the difference between patients in the 1–49% group and patients without an ITD [29].


Mutations in this gene occur in 10% of patients with NK AML [28]. It too confers a favorable prognosis, particularly in patients with double mutations and without a FLT3 ITD. The profound effect of NPM1, FLT3 ITD, and CEBPA status on outcome in NK AML (Fig. 2) suggests that assessment of these mutations should be routinely done in newly diagnosed AML. For example, comparison of patients with and without sibling donors for myeloablative HCT suggests that patients in the NPM+/FLT3 ITD− or the CEBPA + subtypes do not benefit from HCT, while patients with other genotypes do (Fig. 3) [28]. Similarly, the 30% 3-year survival rate following standard therapy in patients aged 70 or above with NPM1 mutations (Fig. 4) [30] suggests that some of these patients might be less enthusiastic about participation in a clinical trial than other similarly aged patients.

Figure 2.

Relapse-free, and overall, survival in newly diagnosed AML according to genotype [28]. [Color figure can be viewed in the online issue, which is available at]

Figure 3.

Relapse-free, and overall, survival according to genotype and whether patients had donors for myeloablative HCT [28]. [Color figure can be viewed in the online issue, which is available at]

Figure 4.

(A) Disease-free (that is relapse-free) and (B) overall survival of patients age ≥ 60 years with cytogenetically normal de novo AML according to NPM1 mutation status. (C) Same as (A) but including only patents age ≥ 70. (D) Same as (B) but including only patents age ≥ 70 [30]. [Color figure can be viewed in the online issue, which is available at]


Approximately 20–30% of patients with t(8;21) or inv(16) have a CKIT mutation. The cumulative incidence of relapse at 5 years has been estimated as 30–35% for patients with either t(8;21) or inv(16) but without a CKIT mutation vs. 70% for t(8;21) patients with a CKIT mutation and 80% for inv(16) patients with a mutation in exon 17 of CKIT [31]. However because CKIT mutations are do not appear to affect outcome in the more common subtypes of AML (for example, NK AML) it is probably practical to test for these mutations only once the patient is known to have inv(16) or t(8; 21).

Other markers.

Currently, 85% of patients with NK AML have a mutation [27], although the prognostic significance of some of these is unclear. However, routine testing for the following may become advisable in the near future.


Mutations in DNA methyltransferase 3a [DNMT3a] occur predominantly in patients with intermediate prognosis cytogenetics, including 25–35% of those with an NK [32, 33]. DNMT mutations are associated with mutations in NPM1 and ITDs of FLT3. They seem to convey poorer prognosis in NK AML, with this effect limited to patients who are not NPM+/FLT3 ITD negative [33].


Mutations in isocitrate dehydrogenase 1 (IDH1) or 2 (IDH2) also have been found chiefly in patients with intermediate prognosis cytogenetics, with 2/3s–3/4s patients having an NK. As with DNMT3a, the effect of mutations in IDH1 has been reported to be context dependent. Thus, IDH1 mutations are independently associated with lower relapse rates in patients with, but higher relapse rates in patients without, FLT3 ITDs [34]. IDH2 illustrates the potential influence of mutation site. While an IDH2 mutation at R140 has been found to confer on patients with an otherwise intermediate prognosis a probability of relapse similar to that found in patients with inv(16) or t(8;21), a mutation at R172 confers a probability of relapse resembling that seen in patients with unfavorable cytogenetics [35].


The TET2 gene provides yet another example of a context dependent mutation. In particular, TET2 mutations, which like the mutations described above, are most commonly found in NK AML, are independently associated with inferior rates of CR, relapse-free survival, event-free survival, and survival but only in those NK patients who are NPM+/FLT3− or CEBPA mutated [36].

Gene and protein expression.

The prognostic influence of some mutations, for example in the WT1 gene, has differed in different series [37, 38]. While such disparities may reflect the presence/absence of other mutations or differences in treatment, they may also reflect whether the mutation results in expression of an abnormal protein. This possibility has prompted profiling of whole genome [39] and micro-RNA [40, 41] expression DNA methylation [42, 43] (as well as expression studies of specific genes (BAALC and ERG [27]), and studies of protein expression [44]. Several reports indicate that these studies potentially add prognostic information independent of that provided by analyses of NPM, FLT3, and CEBPA [39–43].

Integration of cytogenetic, molecular genetic, and clinical data

The European Leukemia Network (ELN) categorizes patients into one of four groups depending on cytogenetic and molecular genetic (NPM1, FLT3, and CEBPA) status (Table I) [45]. An unpublished supplement to the ELN system is depicted in Table II. This recognizes five groups and differs from the ELN in that it (a) recognizes the unique significance of MK, (b) weighs FLT3 ITDs somewhat more unfavorably than the ELN, and (c) uses CKIT mutation status to discriminate among patients with “favorable” cytogenetics [inv(16) or t(8;21)]. While cytogenetics and molecular genetic data provide the most important prognostic information, clinical covariates should also be considered.

Table I. European Leukemia Net (ELN) Prognostic System [45]
Genetic groupSubsets
Favorablet(8;21)(q22;q22); RUNX1-RUNX1T1
 inv(16)(p13.1q22) or t(16;16)(p13.1;q22); CBFB-MYH11
 Mutated NPM1 without FLT3-ITD (NK)
 Mutated CEBPA (NK)
Intermediate IMutated NPM1 and FLT3-ITD (NK)
 Wild-type NPM1 and FLT3-ITD (NK)
 Wild-type NPM1 without FLT3-ITD (NK)
Intermediate IIt(9;11)(p22;q23); MLLT3-MLL
 Cytogenetic abnormalities other than favorable or adverse‡
Adverseinv(3)(q21q26.2) or t(3;3)(q21;q26.2); RPN1-EVI1
 t(6;9)(p23;q34); DEK-NUP214
 t(v;11)(v;q23); MLL rearranged
 −5 or del(5q); −7; abnl(17p); complex karyotype
Table II. Author's Modification of ELN System
Prognostic groupSubsets
  • a

    Patients with inv(16) or 6t(8;21) and CKIT mutations should be considered as belonging to the intermediate 1 group as should patients with t(8;21) and a WBC index > 20 and patients age > 65 and possibly patients with inv(16) age > 35.

 inv(16)a or t(16;16)
 NK with mutated NPM1 and no FLT3 ITD
 NK with double-mutated CEBPA (NK)
Intermediate 1NK with wild-type NPM1 and no FLT3-ITD
 Cytogenetic abnormalities other than best or unfavorable
Intermediate 2FLT3-ITD
Intermediate 3Unfavorable cytogenetics without MK

t- AML.

Many of the studies explicating on the role of various mutations have been done in patients with de novo AML. However data from the German AML Study Group [46] indicate that, after accounting for cytogenetics, NPM and FLT3 ITS status, patients with therapy-related AML have increased rates of relapse and shorter survival than patients without such a history. Shorter survival in younger patients (age < 60) chiefly reflects higher probabilities ofdeath in CR, particularly after HCT, perhaps owing to cumulative toxicity of cancer treatment but primarily reflects higher rates of relapse in older patients.

Antecedent hematologic disorder.

Although not studied in as much detail as t-AML an AHD has also been shown to be adversely affect outcome independent of cytogenetic status [13].

WBC count.

Increasing circulating blast count confers a poorer prognosis in patients with t(8;21) [24] (and particularly in patients with t(15;17) [9].

Age and performance status.

Much of the relation between age and therapeutic resistance (relapse or no CR despite not incurring TRM) reflects age's association with t-AML, an AHD or unfavorable cytogenetics [4, 5]. Similarly, much of age's influence on early death reflects an association between age and performance status (PS) [4, 47]. Of these, PS is the most important. For example, patients aged >75 but with PS 1 have lower early death rates than patients aged 66–75 with PS 2 (Table III) [4]. It is likely that some of age's effect on early death will also be due to an association with various comorbidities (for example, obesity and diseases of heart, lung, liver, and kidney).

Table III. Frequency of Early Death (Within 30 days of Start of 3 + 7 Therapy) According to Age and Performance Status (PS) in Southwest Oncology Group (SWOG) Studies [4]
 Age < 56Age 56–65Age 66–75Age > 75
PS 03/129 (2%)8/72 (11%)9/73 (12%)2/14 (14%)
PS 16/180 (3%)6/11Fo2 (5%)20/126 (16%)7/40 (18%)
PS21/46 (2%)6/34 (18%)16/52 (31%)7/14 (50%)
PS30/97/24 (29%)9/19 (47%)9/11 (82%)

Given these associations, it is not surprising that the prognosis of older patients (typically age > 60–65) is very variable following 3 + 7. Various prognostic systems have been developed for older patients [48–50], although these typically do not account for NPM, FLT3, and CEBPA status, Although it is well known that AML clinical trials enroll only a small minority of older patients with AML who likely have better prognoses than patients who are not enrolled, many of these prognostic systems emanate from countries [48–50] where a larger government role in health care makes it likely that the patients reported in studies are more representative than would be the case in the United States.

Management of “Favorable” Patients (Fig. 1, “Best” in Table II)

Given the data depicted in figure and Tables I and II, these are prime candidates for standard therapy containing an anthracycline and ara-C (3 + 7). However, strong consideration should be given to using increased doses of daunorubicin during induction and ara-C once in CR. ECOG randomized patients aged < 60 to standard doses of ara-C + either daunorubicin at 45 or 90 mg/m2 daily × 3 during induction and showed higher CR rates (70% vs. 57%) and longer survival (medians of 24 vs. 16 months) for patients given the higher dose with this effect limited to patients with favorable or intermediate cytogenetics [51]. The HOVON performed a similar study in older patients and found that in patients aged 60–65 the higher dose produced higher rates of CR (73% vs. 51%), EFS, and survival (38% vs. 23% at 2 years); benefit was most apparent in patients with inv(16) or t(8;21) [52]. Fourteen years ago CALGB showed that favorable (i.e., already anthracyline and ara-C-sensitive) patients benefited most from increased doses of ara-C, leading to widespread use of ara-C at 3 g/m2 twice daily on days 1, 3, and 5 for several cycles as post-CR therapy [53]. However, data from recent randomized studies suggest that a dose of 1 g/m2 bid × 6 days is sufficient [54]. Whether the benefit of dose escalation also applies to NPM+/FLT3 ITD− AML is not known, but the principle that dose escalation is most beneficial to patients already sensitive to standard doses seems likely to apply.

Generally speaking the risk of relapse in favorable patients is not high enough to justify the risk (both short and very long term) of HCT in first CR. This is illustrated in Fig. 3 for NPM+/FLT3ITD− patients and in Table IV for patients with inv(16) or t(8;21). Table IV is derived from a meta-analysis of trials comparing myeloablative HCT (primarily from sibling donors) with continued chemotherapy (or autologous transplant) in patients in first CR [55]. To avoid bias, the authors compared patients with and without donors, rather than patients who were or were not transplanted; 70–80% of patients with donors were transplanted. Although donor–no donor comparison is not free of problems [56], particularly when use is made of unrelated donors, generally if the donor group does better so would patients who actually receive HCT.

Table IV. Effect of HCT in First CR on Relapse-Free Survival (RFS) and Survival (OS) [55]
Cytogenetic risk  Hazard rate (95% CI)
Patients in Donor groupPatients in No donor groupRFSOS
  1. Hazard rate < 1.0 favors HCT over chemotherapy or autologous transplant. Cytogenetic risk assessed by SWOG/ECOG or MRC criteria with results not materially affected by criteria used.

Favorable1883591.06 (0.80–1.421.07 (0.83–1.38)
Intermediate86416350.76 (0.68–0.85)0.83 (0.74–0.93)
Unfavorable2263660.69 (0.57–0.84)0.73 (0.59–0.90)

As noted above the prognosis of favorable patients is far from uniform. Specifically the prognosis of those inv(16) or t(8;21) patients with CKIT mutations [31], WBC > 10,000 (for t[8;21]) [24], or age > 60 [22] [>35 for inv(16)] [23]) is such that they might be more willing to consider HCT (reduced intensity for older patients) or drugs such as desatinib that inhibit CKIT and that are being evaluated in clinical trials in patients with inv(16) or t(8;21).

Management of Patients with MK or Unfavorable Karyotypes without MK (Fig. 1, “Worst” and “Intermediate 3” in Table II)

Whereas prognosis with standard therapy should prompt its use, albeit with dose escalation, in the typical favorable risk patient, prognosis with standard therapy should prompt use of investigational therapy in almost all patients with unfavorable cytogenetics, particularly with MK. As noted in the second paragraph in the “Risk Stratification” section, strong consideration should be given to beginning such therapy during induction.

What this therapy should entail is far from clear. Higher dose daunorubicin does not appear beneficial in such patients [51, 52], recalling findings from CALGB that whether patients were randomized to post-CR ara-C doses of 3 g/m2 BID on days 1, 3, and 5, 400 mg/m2 daily × 5 by continuous infusion (CI) or 100 mg/m2 daily × 5 by CI had no effect on outcome [54]. The rapid entry into trials of many new drugs makes virtually any list of “new drugs in trial” incomplete and outdated. However, some of the more commonly investigated drugs have included the following:

Nucleoside analogs

The Polish Acute Leukemia Group has reported a randomized study showing that addition of cladribine to 3 + 7 (daunorubcin at 60 mg/m2) produced a higher rate of CR after a single course despite similar or less toxicity [57]. This has sparked interest in clofarabine, which is structurally related to cladribine and fludarabine but more active, at tolerable doses than the latter. Burnett et al. found that, after adjusting for other covariates, CR and survival rates in 106 patients (median age 71) considered unfit for 3 + 7 and thus given clofarabine were higher than when similarly unfit patients received low-dose ara-C (LDAC) and comparable with those observed when fitter older patients received 3 + 7-like therapy [58]. Of note, CR rates with clofarabine were similar in patients with unfavorable and intermediate cytogenetics (44% vs. 52%). In 70 relatively fit patients (median age 71) randomized to clofarabine or clofarabine + LDAC, the combination produced superior CR (63% vs. 31%) and survival rates, but survival remained short (median 11 months), even with the combination [59]. Further information about clofarbine's role in newly diagnosed AML will likely be forthcoming from an ongoing ECOG study randomizing older patients to 3 + 7 or clofarabine, albeit without ara-C and from MRC trials comparing daunorubicin + clofarabine with daunorubicin +ara-C for both induction and “consolidation” therapy.

Hypomethylating agents (HAs): azacitidine and decitabine

These drugs were first investigated in patients with MDS some of whom had 21–30% blasts and were thus reclassified as AML. In the azacitidine trial [60] physicians first declared a preference for supportive care only, LDAC, or 3 + 7 in a given patient. Patients were then randomized to the selected conventional care regimen or azacitidine. Among 113 AML patients (median age 70), median survivals were 24.5 months (azacitidine) and 16.0 months (conventional care) and were 12 months (azacitidine) and 5 months (conventional care) in the patients with unfavorable cytogenetics. Too few patients received LDAC or 3 + 7 to permit robust comparisons with azacitidine. However, unlike LDAC or 3 + 7, achievement of CR with azacitidine did not seem a precondition for longer survival.

Results of a trial randomizing patients aged ≥ 65 between decitabine at 20 mg/m2 daily × 5 and physician's choice of LDAC or supportive care only were presented at the 2011 meeting of the American Society of Clinical Oncology [61]. An “updated unplanned analysis” performed after 92% of patients had died showed median survival of 7.7 months for decitabine and 5.0 months for physicians' choice (nominal P = 0.03) The previously planned “final analysis” done after 82% of patients had died showed the same median survivals but P = 0.10.

It remains unclear whether response to HAs correlates with hypomethylation and, in particular, with re-expression of silenced genes. Such a correlation might encourage further studies combining HAs and histone deacetylase inhibitors, which appear to cooperate with HAs in inducing re-expression of silenced genes.

Given the survival times noted above, I do not believe that, at least as used alone, clofarabine, azacitidine, or decitabine obviously differ from standard therapy. Hence I find it reasonable to recommend clinical trials, rather than these drugs, for patients with unfavorable cytogenetics. Such a trial might, for example, include use of azacitidine or decitabine in combination with other agents. In general, however, it is very difficult to know which, if any, of several available trials would be most beneficial for a given patient. Thus, I generally recommend that patients participate in the most logistically feasible trial. An obvious task for the future is to move beyond the current focus on results that likely reflect an average of more and less sensitive patients and to identify patients who are more likely to respond to a given therapy.

Hematopoietic cell transplantation (HCT)

The results in Table IV suggest that HCT is useful in first CR in patients with unfavorable cytogenetics. However, the median age of patients in the meta-analysis was only about 40, and most had sibling donors, raising questions as to whether the same would apply in older patients or in patients with matched unrelated donors. The past 20 years have also seen the development of reduced intensity conditioning (RIC) regimens [62]. These reduce toxicity and are associated with 100-day death rates of 10–15% depending on comorbidities, but permit engraftment and subsequent development of T-cell mediated graft-versus-AML effects and allow patients in their 70s or with significant comorbidities to receive HCT. Results of large trials of RIC-HCT [63] as well as donor–no donor [64] analyses suggest that RIC-HCT is superior to chemotherapy in many older patients (Fig. 5). Furthermore, modern typing techniques have allowed matched unrelated donors to be identified for most patients without sibling donors [65]. Although a true comparison of matched sibling and matched unrelated donor HCT would require randomization of patients with sibling donors to receive transplant from the sibling or a matched unrelated donor such a trial is not feasible. With this constraint and bearing in mind the increased difficulties with donor–no donor analyses in the unrelated setting, data suggest equivalent results with matched sibling and matched unrelated HCT and that if HCT from a (living) matched unrelated donor is not feasible, double cord blood transplants are (at least) their equivalent [66–69].

Figure 5.

(A) Overall survival and (B) relapse-free survival following HCT according to cytogenetic risk [63]. [Color figure can be viewed in the online issue, which is available at]

However, several points are worthy of mention. First, donor–no donor analyses or even the more sophisticated Mantel–Byar methodology [70] cannot substitute for a trial randomizing patients with donors between immediate HCT and HCT delayed until there is increasing evidence of MRD. Second, the same covariates that predict poor outcome after chemotherapy (in particular cytogenetics and MRD after consolidation therapy) do the same after HCT, suggesting that chemotherapy and HCT are not as qualitatively different as often believed [17, 63]. Third, it is clear that immediate mortality after HCT is decreasing [71]; however, patients who are cured of AML after HCT have a 30% decrease in life expectancy, consequent to the effects of immunosuppression and development of second cancers [72]. Most importantly, neither chemotherapy nor HCT are satisfactory for patients with unfavorable cytogenetics. Assume that those of such patients who enter CR have a life expectancy of 12 months, consequent to relapse, without HCT. If, as Table IV indicates HCT reduces risk of death to 0.73 in such patients life expectancy is still only 12/0.73 = 16–17 months. Indeed patients over age 60 who had MK and received HCT in CR1 had only a 6% survival rate at 4 years [73].

Regardless of whether patients receive ablative or RIC-HCT the major cause of death remains relapse [63]. Trials of new HCT “conditioning regimens” are in progress, involving for example clofarabine or anti-CD45 antibodies labeled with 131I [74]. It seems likely that the future will see increasing use of immunotherapy [75] or “chemotherapy” to reduce post HCT relapse. Trials of azacitidine and the FLT3 inhibitor quizartinib (AC220, see below) are ongoing for the latter purpose [76]. In general new agents might be found more effective if tested in patients in CR [77] rather than with active disease, as is currently done.

Management of Patients with FLT3 ITD (Intermediate 2 in Table II)

Many of the same considerations apply here as in the preceding group. However, the use of FLT3 inhibitors is worthy of discussion. Midostaurin (formerly PKC412) and lestaurtinib (formerly CEP701) inhibit multiple kinases, among them FLT3, whereas sorafenib and, particularly AC220, are more specific for FLT3 and much more potent FLT3 inhibitors. Each of the four drugs inhibits wild-type FLT3, but more effectively inhibits FLT3 ITDs and has been studied essentially exclusively in ITD positive patients. Midostaurin and lestaurtinib produced minor responses in relapsed disease and are being studied in trials randomizing untreated patients to 3 + 7 ± the FLT3 inhibitor. A similar trial in 224 relapsed patients found no difference in CR or survival between patients given chemotherapy ± lestaurtinib [78]. However, the ability of patients' serum to inhibit ITD + cell lines correlated with clinical response, suggesting that response rates might be higher with a more potent inhibitor, such as AC220. This drug has considerable more single agent activity in relapsed patients than midostaurin or lestaurtinib [79] and is being combined with chemotherapy in untreated patients. Sorafenib, the only one of these drugs currently available commercially, has also been combined with chemotherapy [80], although a randomized trial giving chemotherapy ± sorafenib suggests that the probability of great benefits in ITD+ disease is unlikely [81]. An issue to consider in combining chemotherapy with FLT3 inhibitors is the ability of the former to increase serum concentrations of FLT3 ligand, which antagonizes the effect of the FLT3 inhibitor [82]. Cytotoxicity as a result of FLT3 inhibition is greater at high-allelic burdens, perhaps because blasts in patients with such burdens are more “addicted” to FLT3 [82]. Because allelic burden is typically higher in relapsed than in untreated patients, the former may benefit more from potent FLT3 specific inhibitors, such as AC220, and the latter from less potent FLT3 inhibitors with however a broader spectrum of kinase inhibition [82].

Management of Intermediate Prognosis Patients (Fig. 1, Intermediate 1 in Table II)

A case for standard therapy is relatively easy to make in most “favorable” patients and similarly for investigational therapy in “unfavorable” patients. However, experience suggests that, given the same expectations with standard therapy, some intermediate patients prefer investigational therapy while others, feeling that the investigational could be worse than standard, opt for standard treatment. If this option is selected, consideration should be given to using escalated doses of daunorubicin and/or ara-C, as described in the section on favorable patients.

For purposes of selecting treatment (standard vs. investigational), it would be desirable to have fewer patients in the intermediate category and more in the favorable/unfavorable categories. In addition to the potential use for this purpose of the “other” markers described in the section on “Molecular Markers,” the future is likely to see increased use of MRD monitoring for this purpose [83–85].

Duration of Postremission Therapy

It is generally held that no more than three to four courses of post remission therapy are needed. Gardin et al. have reported that in patients aged 65 and above six low-intensity post remission courses, given outside a hospital, are associated with superior relapse-free survival and survival, as well fewer transfusions and less time in hospital, than one higher-intensity course given in hospital [86]. MRD monitoring is likely to permit a more sophisticated approach to this issue. In principle quantification of MRD, for example, by multiparameter flow cytometry would permit change of therapy in patients with persistent, or increasing amounts, of MRD, continuation of the same therapy in patients with decreasing amounts of MRD, and discontinuation of therapy in patients with no MRD.

Management of APL

APL is highly curable, although the cure rates of >90% quoted in clinical trials are considerably higher than the rates of 75% reported from large community-hospital databases [87]. The difference to some extent reflects patient selection such as exclusion of patients who die before they can be registered on a trial. Indeed death from CNS and/or pulmonary hemorrhage within the first few days of presentation is the principal obstacle to curing APL and is particularly common in patients with WBC > 10,000. Crucial to prevention of such death is prompt initiation of therapy at earliest suspicion of the diagnosis [88], as discussed in the section on “Diagnosis.” In addition to specific APL therapy management consists of redressing the characteristic thrombopenia and deficiencies of coagulation factors, for example by checking platelet count, INR, and fibrinogen every 8–12 hr and administering platelets and blood products to maintain a platelet count > 30,000, an INR < 1.5, and fibrinogen > 150, while avoiding fluid overload. The most commonly used treatment for APL is ATRA + idarubicin (AIDA) or daunorubicin. However, recent data suggest that addition of ara-C (200 mg/2 daily × 7) to daunorubicin + ATRA improves outcome [89]. A benefit for ara-C might have been more difficult to discern with a daunorubicin dose of 90 mg/2 (rather than 60 mg/m2) daily × 3, but benefit/risk calculations argue for adding ara-C to daunorubicin + ATRA. It is now established that arsenic trioxide (ATO) is a more effective agent in APL than ATRA; in particular single agent ATO can cure APL much more frequently than single agent ATRA [90]. The combination of ATO + ATRA appears at least as effective as AIDA in patients with WBC < 10,000 and is being compared with AIDA in randomized trials in such patients [91]. In patients with WBC > 10,000 benefit/risk considerations suggest the immediate administration of ATO + ATRA, and if the WBC is >20,000–50,000 addition of anthracycline + ara-C. Although routinely unavailable, gemtuzumab ozogamicin is highly effective in APL [92]) and is being combined with ATO + ATRA in the current US Intergroup trial for newly diagnosed patients with WBC > 10,000. A complication of therapy with ATRA, ATO or both is the “APL differentiation syndrome” (APLDS). Potentially fatal, its principal symptoms are fever and dypsnea accompanied by signs of fluid retention such as edema and pleural and/or pericardial effusions. The majority of these patients will have progressive leukocytosis with the WBC being myelocytes, metamyelocytes, bands, and neutrophils. The response rate to steroids is >95%. Some begin steroids prophylactically, and others only when the WBC exceeds 10,000–20,000 or when symptoms develop [9]. The underlying pathology is presumed to be “capillary leak.” The incidence does not appear higher in patients given ATO +ATRA rather than either drug alone.

Once in CR benefit/risk arguments suggest younger patients with initial WBC counts > 10,000 should receive postremission therapy with ara-C in doses of 1–2 g/m2, which has the added benefit of CSF penetration and hence the potential to prevent CSF relapse, which has a cumulative incidence of 5% 3 years after achievement of CR [93]. Older patients, and perhaps younger ones as well, might receive ATO, and ATRA (with intrathecal ara-C if they presented with WBC > 10,000) rather than anthracycline + ara-C; such an approach is being used by several European groups. Finally, MRD monitoring, most sensitively using PCR to detect residual PML-RARα transcripts appears likely to improve outcome, particularly in patients presenting with WBC > 10,000 [94]). The risk of relapse of patients with APL in CR is lower than the risk of early and late complications after HCT, which is thus not justified in patients with APL in CR1.

Management of Relapsed/Refractory AML

Here again the fundamental decision is between a standard (re-induction) regimen such as FLAG or mitoxantrone + etoposide, or an investigational regimen. The principal predictor of response to standard therapy is duration of first CR, with “primary refractory” patients assigned CR duration of 0 [95]. Estey et al. reported that patients about to receive their first re-induction attempt (first salvage) had a CR rates of 60% (15 patients), 40% (30 patients), and 15% (160 patients) if their first CR durations were >2 years, 1–2 years, and <1 year, respectively. Patients who were about to receive > first salvage with a first CR < 1 year had essentially no chance of attaining a CR (58 patients, 96 salvage attempts 0 CR). Focusing more on survival Breems et al. [96] developed a prognostic score based on 667 patients in first relapse (Table V). As seen in Fig. 6, most patients were in the worst group and clearly are candidates for investigational salvage regimens.

Table V. Prognostic Score for Patients with AML in First Relapse [96]
Prognostic factorPoints
RFI, relapse-free interval from first complete remission, months 
CYT, cytogenetics at diagnosis 
 t(16;16) or inv(16)*0
AGE, age at first relapse, years 
SCT, stem-cell transplantation before first relapse 
 No SCT0
 Previous SCT (autologous or allogeneic)2
Figure 6.

Survival after first relapse according to prognostic score in Table V [96].

Future Directions

For many years, the primary goal of therapy was to produce and maintain a CR. The rationale was the demonstration that patients in CR lived longer than other patients. Furthermore, the difference was entirely accounted for by time spent in CR suggesting that the longer life expectancy reflected chemotherapy's ability to produce a CR and not a better underlying prognosis in patients who achieved CR. Recent reports, however, currently include CR and responses the criteria for which are less than stringent than CR; these are known as CRp, CRi, etc. While it appears that prolonged survival likely requires CR, recent data suggest that patients who obtain responses such as CRp, CRi, etc., may live longer than patients who live long enough to have achieved CR, CRp, or CRi but do not do so [60, 97]. This would put less emphasis on CR, and more on survival, as an endpoint of therapy. As a corollary, patients would be advised to stay on a given therapy longer in the absence of CR. This might be important given that multiple courses of newer drugs might be needed to see response. It should be noted however that these patients might be showing improvement short of “response” after a relatively few courses, allowing patients not showing such improvement to be removed from study before the full number of courses had been administered.