Acute Myeloid Leukemia in the Elderly: Assessment of Multidrug Resistance (MDR1) and Cytogenetics Distinguishes Biologic Subgroups With Remarkably Distinct Responses to Standard Chemotherapy. A Southwest Oncology Group Study

THE AGE-SPECIFIC INCIDENCE of acute myeloid leukemia (AML) increases exponentially after 50 years, resulting in a median age for AML onset of 63 to 65 years.¹ Thus, the largest proportion of AML cases are those that occur in elderly individuals. In contrast to younger patients with AML, AML in the elderly is frequently highly resistant to chemotherapy and overall outcomes remain extremely poor. AML patients under 50 years of age, treated with curative intent chemotherapy, have complete remission (CR) rates that average 70%; median relapse free survival (RFS) in this group is nearly 2 years with a 5-year RFS of 25% to 40%.2-4 In contrast, in AML in the elderly patient, CR rates average 30% to 50%; median RFS in these patients is only 9 to 12 months, and very few elderly patients survive beyond 2 years.2,5-8 

The poor outcome of AML in the elderly may result from reduced patient tolerance to chemotherapy, differences in the biology of the leukemic blasts in older versus younger patients, or a combination of these factors. Several recent clinical trials have sought to improve patient tolerance to therapy by incorporating hematopoietic growth factors into chemotherapeutic regimens; unfortunately, while faster hematologic recoveries and a reduced incidence of infection were frequently seen, there has been no consistent benefit in CR rate or survival.5,9-11 These results suggest that intrinsic biologic differences in the leukemic cells may play a more important role than host factors in conferring the poor outcomes observed in AML in the elderly.

A variety of evidence suggests that AML in the elderly may be biologically different from AML in younger patients. As the frequency of myelodysplasia (MDS) increases with age, so does the frequency of AML evolving from MDS; such “secondary” AML cases are frequently highly therapy resistant.4,12,13 The frequency of “poor prognosis” cytogenetic abnormalities, such as −5/del(5q) and/or −7/del(7q), also increases with age.14-18 A third potentially important biologic factor that might account for the therapeutic resistance of AML in the elderly is an increase in the incidence of intrinsic drug resistance in leukemic blasts, mediated by expression of the multidrug resistance glycoprotein MDR1 (also known as p-glycoprotein) or other alternative resistance mechanisms. MDR1 encodes a transmembrane efflux pump that actively extrudes chemotherapeutic compounds from leukemic cells, such as the anthracyclines that are commonly employed in AML therapy. MDR1 expression has been extensively studied in AML and has been shown to be associated with poorer outcomes,^19-25 although no prior study has focused on elderly AML patients.

The aim of our study was to investigate the biologic characteristics of AML in the elderly using pretreatment samples from a large number of uniformly treated elderly patients with newly diagnosed AML registered to a single recently completed clinical trial (SWOG 9031) to determine if biologic characteristics might be useful in predicting outcome and in providing insight into the overall poor therapeutic response of AML in the elderly.

All biologic samples were obtained at initial diagnosis before therapy from patients registered to a single Southwest Oncology Group study, SWOG 9031, a randomized, double-blind, placebo-controlled trial of daunomycin (45 mg/m², days 1 through 3) and standard dose cytosine arabinoside (200 mg/m² days 1 through 7) with or without recombinant human granulocyte colony-stimulating factor (rhG-CSF) for previously untreated AML patients over 55 years of age (Godwin et al, in preparation).11 Patients with clinical de novo AML and secondary AML were eligible for this trial; however patients with acute promyelocytic leukemia (French-American-British [FAB] M3/M3v) were excluded. The diagnosis of AML was confirmed by central histopathologic review by the SWOG Leukemia Pathology Committee using standard FAB criteria as modified by SWOG.26,27 

Analysis of MDR1 expression and functional efflux.Blasts from pretreatment bone marrow/peripheral blood samples were enriched by density gradient separation and assays were performed either on fresh cells or after cryopreservation and thawing; our previous studies have reported successful assessment of MDR and functional dye/drug efflux on appropriately cryopreserved samples.28 MDR1 expression by leukemic blasts was measured using the MDR1-specific antibody MRK16 (Kamiya, Thousand Oaks, CA) in three-color flow cytometric assays where blasts were costained with MRK16, the hematopoietic stem/progenitor cell antigen CD34, and the pan-myeloid antigen CD33, as previously described.28 This approach allows accurate analysis of MRK16 staining in a phenotypically gated myeloid blast population and correlation of MDR1 protein, CD34, and CD33 expression.28 Appropriately matched isotype controls were used in all assays. To assess functional drug efflux and correlate efflux with MDR1 expression, the ability of leukemic blasts to efflux a fluorescent dye, DiOC₂ , was measured in single-color flow cytometric assays, as described.28 The fluorescent dye, DiOC₂ , is an MDR1 substrate, but unlike other MDR1 substrates such as doxorubicin and Rhodamine 123, it does not appear to be transported by the multidrug resistance protein (MRP), one of the more recently identified drug transporters, and thus may be more specific than these other drugs/dyes for MDR1-mediated transport.29,30 Briefly, leukemic blasts were incubated in media containing DiOC₂ to allow uptake for 30 minutes; the blasts were then washed, baseline dye uptake measured, and resuspended in fresh dye-free media with or without the MDR1-modulator cyclosporine A (CsA; 2500 ng/mL; Sandoz Pharmaceuticals, Basel, Switzerland) and incubated for 90 minutes at 37°C to allow efflux. Cells were then resuspended in fresh 4°C media for immediate flow cytometric analysis. The MDR1(+) DOX cell lines and MDR1(−) 8226/S parental line (kindly provided by W.S. Dalton, University of Arizona, Tucson) were used as controls in all experiments.31 

Analysis of MDR1 expression and efflux data.Analyses were performed on a FACScan flow cytometer using Lysis II software (Becton Dickinson, Thousand Oaks, CA). MRK16 staining of gated leukemic blasts compared with control cells was measured using the Kolmogorov-Smirnov (KS) statistic, denoted D, which measures the difference between two distribution functions and generates a value ranging from −1.0 to 1.0.32 This method accurately identifies small differences in fluorescence and is useful in detection of low level MDR1 expression, which frequently occurs in primary patient samples.28,33 MRK16 staining intensity was categorized for descriptive purposes as follows: bright (D ≥ 0.25), moderate (0.15 ≤ D > 0.25), dim (0.10 ≤ D < 0.15), and negative (D < 0.10); however, correlations with clinical outcome were largely performed using the D value as a continuous variable. DiOC₂ efflux was assessed by analyzing cellular fluorescence of gated leukemic blasts after efflux in the presence/absence of CsA; differences in fluorescence were analyzed with KS statistics and a D value of ≥ 0.25 was used to define a case as efflux (+).

Cytogenetic analysis.Cytogenetic studies on pretreatment bone marrow or unstimulated blood samples were performed using standard G-banding with trypsin-Giemsa or trypsin-Wright's staining in SWOG-approved cytogenetics laboratories. Karyotypes were interpreted using International System for Cytogenetic Nomenclature (ISCN) criteria (1995).34 Karyotypes were considered normal diploid if no clonal abnormalities were detected in a minimum of 20 metaphases examined and if two growth/harvesting methods were used. Each karyotype was independently reviewed by at least three members of the SWOG Cytogenetics Committee.

Demographic and clinical data for patients in this study were collected with quality control review according to standard procedures of the SWOG. MDR1 expression and efflux were represented as either quantitative variables using the KS statistic D, or were dichotomized as positive versus negative. Unweighted least squares (LS) and logistic regression (LR) analyses were performed to identify variables predictive of MDR1 expression or functional efflux.35,36 The two methods gave similar results, so only LS results are reported here. Standard criteria were used to define CR and relapse.37 Overall survival (OS) was measured from randomization until death from any cause, with observation censored for patients last known alive. RFS was measured from establishment of CR until relapse or death from any cause, with observation censored for patients last known alive without report of relapse. Distributions of OS and RFS were estimated by the method of Kaplan and Meier.38 Analyses of prognostic factors for treatment outcomes were based on LR models for CR and proportional hazards (PH) regression models for OS and RFS.36,39 Prognostic factors considered in the analysis included the clinical parameters listed in Tables 1 and 2, MDR1 and CD34 expression, functional efflux, and cytogenetics. Statistical significance is represented by two-tailed P values. Analyses were based on clinical and biologic data available September 23, 1996.

A total of 234 patients entered SWOG study 9031 between January 1992 and February 1994. In 23 cases (10%) the initial diagnosis of AML was not confirmed by central pathology review; these patients were excluded from the present analysis. The 211 remaining patients included 89 women and 122 men with a median age at study entry of 68 years (range, 56 to 88). AML cases were most frequently classified as FAB M1 or M2 (Table 1). Fifty patients (24%) were considered to have secondary AML based on a history of antecedent MDS or exposure to potentially leukemogenic drug/radiation therapy, while 161 patients (76%) were considered clinically as de novo AML. Sufficient pretreatment biologic samples were available to examine one or more of the following in these 211 patients: MDR1 protein expression on leukemic blasts, functional efflux, cytogenetics and CD34 expression. All four parameters were examined in 130 cases. Data was not available for all four parameters in 81 cases either because the specimens received contained insufficient blasts for all analyses or because the analyses performed were deemed unsatisfactory after review (for flow cytometric assays: low viability; for cytogenetic analysis: no metaphases detected or < 20 normal metaphases detected or normal metaphases detected when only a single growth/harvesting time was used).

Among 164 cases classified cytogenetically, abnormalities were found in 90 cases (55%) including 54 (33%) with two or more abnormalities. The most common cytogenetic abnormalities in these elderly AML cases were those associated with MDS, as well as AML, including −7/7q− (24 cases [15%]), −5/5q− (21 cases [13%]), +8 (25 cases [15%]). In contrast, abnormalities associated with de novo AML were uncommon: t(8; 21) was found in only three cases (2%), while inv(16)/t(16; 16) was found in only seven cases (4%). As the number of patients with any single abnormality was small, karyotypic abnormalities were grouped into favorable, intermediate, or unfavorable categories based on published literature for correlation with clinical outcome (Table 3).14-18,40-42 Using this scheme, the cytogenetics in 52 cases (32%) were classified as unfavorable, while only nine patients (5%) had a favorable karyotype. The remaining 103 cases (63%) had cytogenetic abnormalities considered intermediate in prognosis. Because of the small number of favorable cases, this category was combined with the intermediate category for most analyses with clinical outcome. Although the unfavorable cytogenetic category included abnormalities such as −7/7q−, which are associated with MDS-related secondary AML, the correlation between secondary AML and unfavorable cytogenetics was only marginal (P = .059) in multivariate analysis after accounting for a significant association between unfavorable cytogenetics and both increasing CD34 expression (P = .0032) and decreasing platelet count (P = .0001).

MDR1 protein expression on selected leukemic blasts was examined using multiparameter flow cytometry in 189 cases and detected in 135 (71%) including 69 (37%) staining brightly with the MDR1-specific antibody, MRK16. Moderate or dim positive staining was seen in 33 (17%) each. Functional dye/drug efflux, inhibited by cyclosporine (an MDR1 efflux inhibitor), was detected in 101 (58%) of 175 samples studied (Table 4). As expected, functional efflux was strongly correlated with MDR1 expression: 82 (67%) of 122 MDR1(+) cases were efflux(+), while 32 (67%) of 48 MDR1(−) cases were efflux(−) (P < 0.0001). However, as we and others have previously described,28,43,44 discrepant cases were identified including 16 MDR1(−)/efflux(+) cases and 40 MDR1(+)/efflux(−) cases. In multiple LS regression analyses performed to identify factors predictive of MDR1 expression or efflux, MDR1 expression was found to be significantly and independently associated with efflux (P < .0001) and FAB subtype (P = .0037). Associations of MDR1 and FAB subtype were largely due to lower MDR1 expression in the M4 and M5 categories; 14 of 36 (39%) M4 or M5 cases were MDR1(+) in contrast to 121 of 153 (79%) non-M4/M5 cases. In addition to its significant association with MDR1 expression, functional efflux was also independently associated with CD34 (P < .0001), but not with any other factor measured. Surprisingly, although CD34 expression was associated with MDR1 expression in univariate analysis (P = .0001), after accounting for all other factors in multiple LS regression analysis including the significant effects of functional efflux and FAB type on MDR1 expression, there was no longer an association between these variables (P = 0.60). Similarly, multiple LS regression analysis failed to find an independent association of MDR1 expression with secondary disease (P = .76) or unfavorable cytogenetics (P = .28).

Overall 95 (45%) of the 211 elderly patients with AML achieved a complete remission (CR). The CR rate after the first induction attempt was 86 of 211 (41%) patients. Of the remaining 125 patients, 48 received the second protocol induction attempt and 9 (19%) achieved CR. In univariate analysis, disease onset (de novo v secondary AML) was the only clinical parameter strongly predictive for achievement of CR (P = .0005; Table 4); no other clinical parameter including age, white blood count, blast count, or platelet count at presentation, or treatment arm (G-CSF v placebo) was correlated with CR rate. Among laboratory parameters, there was a marginally significant association between CR rate and FAB subtype (P = .048) with CR rates ranging from 34% (24 of 70) for M2 to 65% (26 of 40) for M4 and M5 combined. However, as shown in Table 4, in univariate analysis the CR rate was highly significantly associated with CD34 and MDR1 expression and with functional efflux. CR rate significantly decreased with increasing expression of CD34 (P = .0027) or MDR1 expression (P = .0019) and with increasing strength of efflux (P = .0039). The CR rate was also significantly worse in patients with unfavorable cytogenetics (P < .0001).

In multiple LR analysis, CR rate was highly significantly and independently associated with each of three factors: secondary AML (P = .0035), MDR1 expression (P = .0041), and unfavorable cytogenetics (P = .0031). After accounting for these three factors, there was no significant association of CR rate with CD34 expression (P = .80), functional efflux (P = .75), or FAB subtype (P = .081), or any of the other factors in Table 1. Only 2 (12%) of 17 MDR1(+) patients with secondary AML and unfavorable karyotype achieved CR. In contrast, 22 of 27 (81%) patients with de novo MDR1(−) AML and favorable/intermediate cytogenetics achieved CR (Table 5).

Of the 116 patients who failed to achieve CR, 73 (63%) had documented resistant disease, including 41 patients who received one and 32 patients who received two induction courses. The remaining 43 patients included 19 who died during aplasia, 19 who died before marrow examination was performed, and five who were unevaluable for various reasons. Resistant disease was significantly and independently associated with unfavorable cytogenetics (P = .017) and MDR1 expression (P = .0007), but not with secondary AML status (P = .11).

Of the 211 patients, 180 have died. The remaining 31 were alive between 18 and 53 months (median, 33 months) after registration. Multivariate proportional hazards regression analysis of OS identified three significant independent prognostic factors: OS was significantly poorer for patients with unfavorable cytogenetics (P < .0001) and decreased significantly with increasing age (P = .014) and increasing white blood cell count (WBC) (P = .029). After accounting for these factors, none of the other variables considered had independent prognostic significance, including secondary AML (P = .29), MDR1 expression (P = .93), efflux (P = .10), or CD34 (P = .45).

Of the 95 patients who achieved CR, 78 relapsed and 4 others died without report of relapse (all 4 from consolidation toxicities). In multivariate analysis, only one marginally significant prognostic factor was identified: RFS was poorer for patients with unfavorable cytogenetic status (P = .028).

None of the other variables considered had independent prognostic significance for RFS, including secondary AML (P = .44), efflux (P = .73), CD34 (P = .27), or MDR1 expression (P = .19).

Our studies indicate that the consistently poor outcomes achieved with conventional chemotherapeutic regimens in elderly AML patients may result predominantly from the distinct biologic features of the leukemic blasts in older patients. In our studies, we observed a notably high frequency of: (1) secondary AML (24% of our cases), (2) cytogenetic features traditionally associated with an unfavorable outcome (32%), and (3) MDR1 protein expression by leukemic blasts (71%). Each of these factors was independently and significantly linked to lower CR rates in elderly patients treated with the conventional regimen employed in SWOG trial 9031. Additionally, we demonstrate that assessment of these biologic parameters at diagnosis can be used to identify groups of patients with quite different responses to therapy. Using these factors, we could identify elderly AML patients with de novo MDR1(−) AML and favorable or intermediate cytogenetics who had a very high likelihood of achieving CR, over 80% in our study. This unusually high CR rate achieved in elderly patients who lack the typical elderly AML biologic profile is comparable to that seen in younger AML patients with “good risk” features.6,17 At the other extreme, we could identify patients with de novo or secondary MDR1(+) AML and unfavorable karyotypes who had very poor CR rates. Our studies suggest that these patients rarely benefit from current conventional chemotherapeutic regimens.

Comparison of the biologic features that we have described herein in elderly AML patients with those historically associated with younger patients strongly supports the hypothesis that AML in the elderly is a biologically distinct disease. In our trial, 24% of the elderly patients had documented secondary AML; however, we found that most of the elderly patients who presented clinically with de novo disease shared similar biologic features with these secondary AML patients. In particular, our studies indicate that many cases of de novo AML occurring in the elderly are strikingly similar to secondary AML cases occurring after alkylating agent therapy or myelodysplasia.14,16,45-49 There was a similarly high frequency of unfavorable cytogenetic abnormalities such as −7/7q− or −5/5q− (which are frequently detected in MDS and secondary AML) in both our de novo, as well as secondary AML patients, while in contrast, the frequency of cytogenetic abnormalities more traditionally associated with de novo AML in younger patients [such as inv(16) or t(8; 21)] was strikingly low.14-18,45,46 The frequency of MDR1 expression was also extremely high in our study (71%) and moreover occurred at similar frequencies in patients with de novo or secondary disease. This high frequency of MDR1 expression detected in elderly AML patients stands in striking contrast to the frequency of 30% that we have found in a recent study of 400 younger AML patients registered to SWOG study 8600 (median age, 45 years) using the identical sensitive methodology reported herein (C. Leith et al, in preparation). Thus, our studies show that most cases of de novo AML in elderly patients show a similar constellation of biologic features more traditionally associated with secondary AML; these features are quite distinct from those associated with true de novo AML in younger patients. We thus speculate that the majority of AML cases arising in elderly patients that present clinically with de novo disease actually arise from a setting of prior bone marrow injury or clinically undetected antecedent MDS.

The identification of biologic and clinical features associated with a poor prognosis is an essential first step for developing a rational approach to overcome the poor therapeutic outcome of AML in elderly patients. Our identification of MDR1 expression as an important independent predictor of response to therapy is particularly interesting, as it suggests that therapies that incorporate MDR1 modulators may be potentially beneficial to many elderly patients with AML. The use of MDR1 reversing agents might have a large impact on those elderly patients who have MDR1(+) leukemic cells, but otherwise favorable prognostic indicators for CR (including a de novo disease presentation and intermediate/favorable cytogenetics); such patients constituted 38% of the elderly AML patients entering our trial.

In our study, although MDR1 expression was associated with a lower CR rate and resistant disease, it did not predict for OS or RFS. These results differ from previous studies in predominantly younger AML patients, in which MDR1 expression was associated with both CR rate and OS.^19,21,23 This lack of correlation between MDR1 expression and OS and RFS may, in part, be because the small number of survivors in this study precluded identification of all biologic parameters associated with outcome in statistical analysis. In addition, other as yet unidentified biologic factors may contribute to disease resistance in AML in elderly patients. Patterns of drug resistance in elderly AML are particularly complex: the MDR1(−)/efflux(+) cases identified by us and others suggest that alternative efflux pumps (such as the more recently recognized multidrug resistance associated protein (MRP) or the lung resistance protein (LRP), may be important in conferring resistance. ^{22,30,43,44,50,51} The relative expression of these resistance proteins in older versus younger AML patients is currently unknown. The poor OS even among patients who achieved CR points out the need for better postremission therapies in these elderly patients with AML.

In conclusion, we have identified biologic disease characteristics that can be used to help identify those elderly AML patients with a high likelihood of response, as well as those with a poor response to current conventional chemotherapeutic regimens. Recognition of these disease characteristics leads the way to the development of risk adapted therapies designed to circumvent these biologic disease factors. The incorporation of MDR1 modulators into therapeutic regimens represents a first step in this direction.49,52