Fax: (216) 636-0636.
Impact of remission induction chemotherapy on survival in older adults with acute myeloid leukemia
Article first published online: 27 AUG 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 8, pages 1752–1759, 15 October 2007
How to Cite
Baz, R., Rodriguez, C., Fu, A. Z., Jawde, R. A., Kalaycio, M., Advani, A., Sobecks, R. and Sekeres, M. A. (2007), Impact of remission induction chemotherapy on survival in older adults with acute myeloid leukemia. Cancer, 110: 1752–1759. doi: 10.1002/cncr.22976
- Issue published online: 19 SEP 2007
- Article first published online: 27 AUG 2007
- Manuscript Accepted: 10 MAY 2007
- Manuscript Revised: 5 MAY 2007
- Manuscript Received: 17 APR 2007
- acute myeloid leukemia;
- supportive care;
- nonintensive therapy;
- older adults
Significant controversy surrounds the use of remission induction chemotherapy (IC) in older adults with acute myeloid leukemia (AML). Earlier clinical trials have yielded conflicting results and possibly a minor survival benefit, often offset by a longer hospitalization time.
To evaluate the role of IC in patients with AML, a case control study of patients 60 years or older treated at the Cleveland Clinic Taussig Cancer Center between 1997 and 2005 was conducted. Forty-four patients who did not receive IC were matched by a propensity analysis to 138 patients who received an anthracycline-based regimen.
The unadjusted median survival of patients who did not receive IC was 53 days, compared with 197 days (P < .001) for those who did. After further adjusting for age, gender, race, leukocyte count at presentation, AML cytogenetics, history of prior hematologic disorder, and assessing for comorbidities, not receiving IC was still associated with worse survival (hazards ratio of 1.88; 95% confidence interval, 1.15–3.05 [P = .01]). Additional predictors of poor outcomes in older adults with AML included higher leukocyte count at presentation, poor-risk cytogenetics, and African-American race (compared with Caucasians).
The study suggests improved outcomes in older adults with AML who undergo remission induction therapy. Cancer 2007. © 2007 American Cancer Society.
It is estimated that approximately 13,400 patients will be diagnosed with acute myeloid leukemia (AML) in the U.S. in 2007.1 AML is a disorder that afflicts mainly older adults, with a median age at diagnosis of 67 years. Standard therapy involves remission induction with intensive combination chemotherapy, which often consists of an anthracycline or anthracenedione and cytarabine.2–5 For younger adults, this approach results in a long-term disease-free survival of approximately 30% at the cost of a treatment-related mortality of 5% to 10%.5–7 These outcomes are significantly worse for older adults, who have a dismal long-term disease-free survival of less than 10%, and a treatment-related mortality as high as 20% to 30%.4, 5, 8, 9
Whereas remission induction therapy prolongs survival and allows for a significant disease-free interval in younger adults, its benefit in older adults is not as obvious.2 Two studies, performed approximately 2 decades ago, have attempted to address this question. The first randomized 71 patients older than 65 years of age with newly diagnosed AML to remission induction therapy (with daunorubicin, vincristine, and cytarabine) or supportive care (including low-dose chemotherapy for the palliation of symptoms of acute leukemia).10 This trial noted a significant median survival benefit of only 10 weeks for patients receiving remission induction chemotherapy—approximately the same amount of time these patients spent in the hospital, compared with those randomized to receive best supportive care or nonintensive chemotherapy. The second study, by Tilly et al.,11 randomized 87 older patients to low-dose cytarabine (nonintensive chemotherapy) or induction chemotherapy with the combination of an anthracycline and cytarabine, but showed no statistically significant survival difference between the 2 groups (median survival 8.8 months in the nonintensive chemotherapy group vs 12.8 months in the induction chemotherapy group).11 In addition, patients receiving induction chemotherapy required greater transfusion support, had a longer hospital stay, and had a higher treatment-related mortality.11 More recently, Menzin et al.12 surveyed the SEER-Medicare database and compared the outcomes of patients treated with chemotherapy to those who had not received chemotherapy after adjusting for age, comorbidities, and geographical region. Whereas patients who received chemotherapy had a superior median survival than those who had not received chemotherapy (6.1 vs 1.7 months; P < .0001), the cohort treated with chemotherapy included patients who received nonintensive chemotherapy as well as remission induction chemotherapy.12
Because of the high treatment-related mortality, and poor overall outcomes of remission induction therapy, many older patients with AML forego induction chemotherapy in favor of less aggressive therapy with hydroxurea, cytarabine, and more recently hypomethylating agents (such as azacitidine and decitabine),2, 13 often at the recommendation of their physicians. With improvements in supportive care, however, older patients can increasingly be supported through induction chemotherapy with greater success. Similarly, advances in nonintensive chemotherapy have occurred: Low-dose cytarabine was found to result in a survival advantage compared with hydroxyurea, immunomodulatory agents have shown efficacy in patients with myelodysplastic syndromes and AML, and the hypomethylating agent decitabine was superior to induction chemotherapy in patients with high-risk myelodysplastic syndrome (MDS).13–16 Given the controversy surrounding the relative benefit of induction chemotherapy, and advances in nonintensive chemotherapy approaches (including supportive care), it is not clear whether or not the decision to withhold remission induction therapy from older adults is valid.
We performed a case-control study to compare the outcomes of older adults with AML who received induction chemotherapy and those who did not to address this question.
MATERIALS AND METHODS
A total of 457 patients with AML evaluated at the Cleveland Clinic Taussig Cancer Center between 1997 and 2005 were identified. An AML diagnosis was based on morphologic, cytochemical, and immunophenotypic review, with classification according to the French-American-British system17 or, beginning in 2001, by the World Health Organization classification.18 Two patients were excluded for missing data and 48 patients did not receive remission induction chemotherapy. Of the remaining 407 patients, 280 patients received systemic therapy at Cleveland Clinic, whereas the remainder received AML therapy at the referring institution. After restricting the sample to age ≥60 years, 44 patients did not receive remission induction chemotherapy, whereas 138 patients did (Fig. 1). Patients' records were reviewed for demographic data, laboratory data, AML etiology (de novo or secondary [including those after an antecedent hematologic disorder or treatment-related]), and survival outcomes. A patient was considered to have cardiac comorbidities if he/she had a history of coronary artery disease, congestive heart failure, cardiac arrhythmia requiring therapy, history of symptomatic valvular heart disease, or an echocardiogram demonstrating an ejection fraction ≤50%. Renal dysfunction was defined as a serum creatinine ≥1.5 mg/dL. Hepatic dysfunction was defined as a serum alkaline phosphatase or alanine transaminase greater than 2 times the upper limit of normal (150 U/L and 50 U/L, respectively). AML cytogenetics were recorded as favorable, intermediate, adverse, or missing (no growth or not available) based on the Cancer and Leukemia Group B (CALGB) classification system.19 Briefly, good-risk cytogenetics comprised the following: inversion 16 or translocation (16;16) and translocation (8;21); intermediate-risk cytogenetics included: normal karyotype, deletion Y, deletion of the short arm of chromosome 9 (−9q), or addition of chromosomes 11, 13, or 21; adverse-risk cytogenetics included: complex cytogenetics (3 or more abnormalities), addition of chromosome 8, and deletion of chromosome 7. Performance status was not recorded for many patients and hence was not included in this analysis. This study was approved by the Cleveland Clinic Institutional Review Board.
Patients treated with remission induction chemotherapy received combination therapy with an anthracycline or anthracenedione (consisting of daunorubicin, mitoxantrone, or idarubicin) and cytarabine. Postremission therapy consisted of cytarabine-based regimens. Patients who opted against induction chemotherapy received supportive measures, consisting of antibiotics and transfusion support when indicated. In addition, nonintensive chemotherapy patients received low-dose subcutaneous cytarabine, hydroxyurea, azacitidine, or arsenic trioxide-based therapy at the discretion of the treating physicians usually for the palliation of symptoms of AML and control of leukocytosis.
The outcome of the study was survival time measured from the date of diagnosis. A Cox proportional hazards model was developed as a basic approach to predict overall survival. The exponential of the parameter estimate represented the hazards ratio (HR) associated with an increase of 1 unit in an independent variable, assuming all other covariates remained constant. The Cox proportional hazards model takes maximum advantage of each subject's available data even though subjects were tracked for different lengths of time, and the outcome of interest might not have occurred. Therefore, subjects exited the cohort at the time of death or at the last follow-up, whichever occurred first.
A case-control study was designed in which cases comprised patients opting against induction chemotherapy and controls represented patients who received induction chemotherapy. As it is possible that a patient's baseline characteristics could have influenced both the type of therapy selected (intensive vs nonintensive chemotherapy) and the outcome of interest, a propensity score method was used to address potential sample selection bias and to better balance the patient characteristics.20 A logistic regression was used first to predict the probabilities (P) of receiving no induction chemotherapy versus induction chemotherapy using individual characteristics: age, gender, race, leukocyte count at presentation, AML cytogenetics, secondary AML, and com- orbid conditions including renal dysfunction, hepatic dysfunction, and cardiac comorbidity. The propensity scores were achieved by taking the logit of the predicted probability (ie, log[(1−p)/p]), to normalize the distribution of the scores. The cases and controls were matched by propensity scores 1:3, with the nearest neighbor being matched with a replacement.21 The matching was restricted by a radius of 0.5 to limit the risk of imprecise matches if the closest neighbor was numerically distant.22
The purpose of propensity score matching is to balance the patient characteristics between the comparison groups. Thus, chi-square tests were used for categorical variables and Student t-tests were used for continuous variables to examine the extent to which the matching procedure resulted in comparable samples. The matched samples were used as the analytic dataset for this approach. For matching with replacement, the weight of the sample was considered where case units were weighted at 1 and the weight for a control is the number of times it is matched to a case unit. In addition, robust standard errors were adjusted.
One hundred and eighty-two patients (44 patients who did not receive induction chemotherapy [cases] and 138 patients who received induction chemotherapy [controls]) were the subjects of this study. The median age of patients who did not receive induction chemotherapy was 75 years (range, 60–92 years), and 19 patients were female (43%). Patients who did not receive induction chemotherapy were generally older than patients who received induction chemotherapy before matching (median age, 75 vs 68 years in the nonintensive chemotherapy and induction chemotherapy groups, respectively; P < .0001). Gender and race were not significantly different between the 2 groups. However, patients who did not receive induction chemotherapy were more likely than patients who received induction chemotherapy to have missing cytogenetics (43% vs 17%; P < .0001), but less likely to have intermediate-risk cytogenetics (20% vs 43%; P < .01). The leukocyte count at presentation was not significantly different between the 2 groups. Baseline comorbidities were similar in the 2 groups with the exception of a higher prevalence of renal dysfunction among patients who received induction chemotherapy (27% vs 12%; P < .05). After matching for propensity scores, the patients who received induction chemotherapy and those who did not receive induction chemotherapy were no different with respect to any of the characteristics above. Patient characteristics are summarized in Table 1 (before matching) and in Table 2 (after propensity score matching).
|NIC, % (N = 44)||IC, % (N = 138)||P|
|Median age (range), y||75 (60–92)||68 (60–80)||<.0001|
|WBC at admission|
|Mean ± SD (×106/μL)||31.7 ± 66.2||26.4 ± 43.9||.62|
|Low dose cytarabine||4||—|
|Arsenic based therapy||4||—|
|NIC, % (N = 34)||IC, % (N = 102)||P|
|Median age (range), y||74 (60–84)||74 (60–80)||.66|
|WBC at admission|
|Mean ± SD (×106/μL)||34.3 ± 73.6||24.2 ± 34.7||.45|
Survival and Outcomes
The median survival for patients who received nonintensive chemotherapy was 53 days, compared with 197 days for patients who received induction chemotherapy (Fig. 2) (log-rank P < .0001). Among patients who received induction chemotherapy, 87 (63%) achieved a complete remission (CR), whereas none of the patients who received nonintensive chemotherapy achieved a CR. The 4-week and 8-week mortality was 12% and 21%, respectively, for patients who received induction chemotherapy, whereas the 4-week and 8-week mortality for patients who did not receive induction chemotherapy was 28% and 52%, respectively. After further adjustment for age, gender, race, leukocyte count at presentation, cytogenetics, and AML etiology (de novo vs secondary), withholding induction chemotherapy remained associated with worse outcomes (HR of 1.88; 95% confidence interval [95% CI], 1.15–3.05 [P = .011]). In addition, high white blood cell count (WBC) at presentation and African-American race were also associated with worse survival (HR of 1.004; 95% CI, 1.001–1.008 [P = .025] for every 106/μL increase in WBC and an HR of 2.20; 95% CI, 1.19–4.07 [P = .012] for African-American race vs Caucasian, respectively). As expected, both favorable and intermediate AML cytogenetics was associated with better survival compared with adverse cytogenetics (HR of 0.29;, 95% CI, 0.12–0.71 [P = .007] and HR of 0.38; 95% CI, 0.25–0.59 [P < .0001], respectively). The results of multivariable analyses are reported in Table 3.
|HR (95% CI)|
|Basic approach (N = 182)|
|NIC/ IC||1.88 (1.15–3.05)*|
|Per year increase||1.00 (0.98–1.03)|
|Race missing/white||1.32 (0.31–5.60)|
|Secondary/de novo||1.24 (0.87–1.77)|
|WBC at presentation|
|Per unit 106/μL increase||1.005 (1.001–1.008)*|
Additional Testing to Validate the Model
Because knowledge of the patient's cytogenetics may impact the decision to deliver or withhold induction chemotherapy, additional analyses were performed. The interaction between AML cytogenetics and treatment decision was added to the proportional hazards model, but it did not achieve a level of statistical significance (data not shown). In addition, subgroup analyses in which adverse risk or missing cytogenetics were respectively excluded were performed. When patients with adverse risk cytogenetics were excluded from the model, and after controlling for all other covariates (age, gender, race, WBC at presentation, other cytogenetics, and AML etiology), not receiving induction chemotherapy remained associated with worse outcomes; because of the smaller sample size, this did not achieve the level of statistical significance (HR of 1.62; 95% CI, 0.85–3.10 [P = .14]). However, when patients with missing cytogenetics were excluded, not receiving induction chemotherapy remained significantly associated with worse outcomes (HR of 2.35; 95% CI, 1.26–4.40 [P = .007]).
The propensity score matching yielded more balanced groups in terms of the measured covariates. All of the variables with statistically significant differences before matching had no significant differences after matching (Tables 1 and 2). The Cox proportional hazards model with 1:3 matching sample demonstrated that nonintensive chemotherapy remains associated with shorter survival (HR of 2.14; 95% CI, 1.23–3.75 [P = .007]) (Table 4).
|HR (95% CI)|
|Propensity score matching (N = 136)|
|Per year increase||1.07 (1.00–1.14)†|
|Race missing/white||1.91 (0.37–9.95)|
|Secondary/de novo||1.34 (0.68–2.61)|
|WBC at presentation|
|Per unit 106/μL increase||1.00 (0.99–1.01)|
To account for the fact that patients with poor performance status might not survive greater than 1 week regardless of whether they receive chemotherapy, patients who had died within 1 week of diagnosis were excluded (n = 6) and nonintensive chemotherapy remained associated with shorter survival compared with intensive chemotherapy (HR of 1.76; 95% CI, 1.06–2.90 [P = .028]).
The management of older adults with AML is far from straightforward. For patients not thought to be good candidates for intensive chemotherapy, the use of low-dose cytarabine is currently considered a standard treatment strategy.14 For older adults who are eligible for induction chemotherapy, considerable controversy persists. Whereas approximately 40% to 50% of patients will achieve a CR with this approach, a significant number will not survive induction chemotherapy. Less aggressive chemotherapy (such as low-dose cytarabine) will achieve a lower CR rate, but to our knowledge has not been compared with remission induction chemotherapy.14 Kantarjian et al.15 compared the outcomes of a cohort of patient with high-risk MDS treated with decitabine with a historic cohort of patients who received intensive chemotherapy. After matching for age, International Prognostic Scoring System, and cytogenetics, patients who received decitabine had a similar CR rate and survival benefit of approximately 10 months compared with patients treated with intensive chemotherapy.15 It is unclear whether these findings can be applied to older patients with AML.
Because prior studies had suggested an overall survival benefit of induction chemotherapy, which is often completely or partially offset by prolonged hospitalization and decreased quality of life, and because of the very small number of patients who derive long-term disease-free survival from induction chemotherapy, this approach may not be offered to, or accepted by, patients.2, 3, 23 Our study suggests a more substantial survival benefit to induction chemotherapy in a more contemporary series.
In this study, age, leukocyte count at presentation, race, AML cytogenetics, and treatment (induction chemotherapy vs withholding induction chemotherapy) were found to influence outcomes. These predictive factors have been validated among patients who received remission induction chemotherapy by other investigators.5, 24–28 African-American race was identified as a poor prognostic factor in this study. This is in agreement with data from the CALGB that noted a worse complete remission rate and survival of African-American patients, particularly African-American men, with AML.24 In that study, because all patients were enrolled on CALGB leukemia trials, access to care and treatment variability were eliminated as possible causative factors for the difference in outcome, leaving the possibility of true biologic heterogeneity. It is difficult to know in the current study what factors play into the worse outcome experienced by African-Americans.
Patients who did not receive induction chemotherapy had an overall survival of 53 days (approximately 8 weeks) in this study. This is in contrast to the study by Lowenberg et al.,10 which reported a median survival of 11 weeks for patients who did not receive induction chemotherapy. However, the former study included patients eligible to receive induction chemotherapy, who were likely healthier than in the present study. It is interesting to note that patients who received induction chemotherapy in the present study had a comparable overall survival and CR rate to that reported by Lowenberg et al. in 1990 (overall survival, 28 weeks vs 21 weeks; CR rate of 63% vs 58%). The group that did not receive induction chemotherapy had a shorter survival than patients who received nonintensive chemotherapy in the study by Tilly et al.11, although patients in that study received low-dose cytarabine, which has been shown to produce a CR rate of 18% and to have a survival advantage over treatment with hydroxyurea.14 Survivals noted among induction chemotherapy patients in the present study, as well as risk factors for outcome, are similar to other large studies in older AML patients, and to the study by Tilly et al.8, 11, 29, 30 Improved CR rates and survival duration in the present study may also reflect improvement in supportive care in older adults in the current era, including the routine use of better antifungal agents, compared with older studies.
This study has several potential limitations. The nonrandomized, retrospective nature may have resulted in the introduction of bias. Specifically, treating physicians may have advocated to patients an approach avoiding intensive chemotherapy in the presence of indicators suggesting poor outcomes with this approach. Important factors when considering intensive chemotherapy in this group of patients include patient age and comorbidities, performance status, secondary AML, and cytogenetics. With the exception of patient performance status, these factors were accounted for in this analysis and the benefit of induction chemotherapy persisted. Whereas exploratory analysis excluding patients experiencing an early death were performed to nullify the impact poor performance status patients may have on the results, and comorbidities were taken into account in comparing the groups, these analyses likely do not account fully for the impact of the performance status on survival, which cannot be inferred from this study.
In addition, although a retrospective analysis of induction chemotherapy versus no induction chemotherapy in older AML patients is not ideal, it is likely the only source of evidence for this question, as prospective randomized trials are unlikely to be performed. One attempt to randomize patients to nonintensive chemotherapy and induction chemotherapy was terminated early because of poor patient accrual.31 In particular, case-control studies are often criticized for the selection of controls because this group can make cases look either better or worse if not well matched. Because nearly identical results were obtained from both the basic regression approach and propensity score approach, this confirms the validity of the study results.
In conclusion, remission induction chemotherapy in older adults with AML is associated with a significant survival benefit (of approximately 28 weeks) compared with nonintensive chemotherapy. Older adults with AML should not be denied induction chemotherapy because of age alone, and can be informed that induction chemotherapy does appear to provide a survival advantage, albeit based on retrospective data. Because few older AML patients achieve long-term disease-free survival, however, clinical trial participation should be encouraged in this group of patients.
- 9A placebo-controlled study of recombinant human granulocyte-macrophage colony-stimulating factor administered during and after induction treatment for de novo acute myelogenous leukemia in elderly patients. Groupe Ouest Est Leucemies Aigues Myeloblastiques (GOELAM). Blood. 1998; 91: 2722–2730., , , et al.
- 10On the value of intensive remission-induction chemotherapy in elderly patients of 65+ years with acute myeloid leukemia: a randomized phase III study of the European Organization for Research and Treatment of Cancer Leukemia Group. J Clin Oncol. 1989; 7: 1268–1274., , , et al.
- 12Effects of initial treatment on survival among elderly AML patients: findings from the SEER-Medicare database. Blood. 2006; 108. Abstract 1973., , , .
- 14A comparison oflow-dose cytarabine and hydroxyurea with or without all-trans retinoic acid for acute myeloid leukemia and high-risk myelodysplastic syndrome in patients not considered fit for intensive treatment. Cancer. 2007; 109: 1114–1124., , , et al.
- 19Pretreatment cytogenetic abnormalities are predictive of induction success, cumulative incidence of relapse, and overall survival in adult patients with de novo acute myeloid leukemia: results from Cancer and Leukemia Group B (CALGB 8461). Blood. 2002; 100: 4325–4336., , , et al.
- 21Evaluating the econometric evaluations of training programs. Am Econ Rev. 1986; 76: 604–620..
- 31Treatment of AML in the elderly: who is “fit” for intensive chemotherapy? Blood. 2006; 108. Abstract 1974., , .