All authors reviewed and approved the final article. Dr. Harb reviewed all cases and wrote the article. Ms. Tan and Dr. Wilding performed the statistical analyses. Ms. Ford constructed the database. Drs. Sait and Block reviewed all karyotype analyses. Dr. Barcos reviewed the pathology specimens. Dr. Wallace reviewed the flow cytometry data. Dr. Wang contributed to the care of the patients. Dr. Wetzler oversaw the conduct of the study and contributed to article preparation.
Treating the octogenarian and nonagenarian patients who have acute myeloid leukemia (AML) with intensive chemotherapy is controversial. Several models to predict outcome were proposed, including the use of a comorbidity index. However, it is unclear whether the Charlson comorbidity index (CCI) or the hematopoietic cell transplant comorbidity index (HCTCI) is more sensitive.
The authors analyzed their experience with 92 patients aged ≥80 years who had AML. Patients' pretreatment characteristics and their treatment outcomes were recorded.
All patients were offered intensive treatment; 59 patients (64%) were treated intensively with a variety of regimens, whereas 33 patients (36%) elected to receive supportive care. The CCI and the HCTCI had similar predictive ability for outcome in both groups. A multivariate analyses of prognostic factors identified near-normal albumin (48% of patients; 1-year survival rate, >27%) as a favorable factor for the whole cohort, age <83 years (47% of patients; 1-year survival rate, >25%) and nonmonocytic morphology (75% of patients; 1-year survival rate, >26%) as favorable factors for the intensively treated cohort, and bone marrow blasts <46% (50% of patients; 1-year survival rate, >19%) as a favorable factor for patients who received supportive care.
The treatment of acute myeloid leukemia (AML) in patients aged ≥60 years is controversial, and the majority of these patients probably do not undergo intensive treatment.1-4 A recent systematic review of the literature of large population-based investigations analyzed clinical trials that included >40 patients.3 In a total of 36 AML trials that involved 12,370 patients with a median age of 70 years, the median overall survival (OS) approached 30 weeks for intensively treated patients, whereas the median survival for patients who received best supportive care alone or best supportive care and nonintensive therapy was 7.5 weeks and 12 weeks, respectively. The complete remission (CR) rate after induction therapy was 44%; and, for the patients who achieved a CR, age no longer influenced prognosis.3 We were interested specifically in a subset of elderly patients with AML, those aged ≥80 years, because we have a significant number of these patients; and, to the best of our knowledge, very few articles have concentrated specifically on this age group.
Four different groups have addressed treatment outcomes of patients aged ≥80 years with AML.5-8 In those reports, the CR rate after induction ranged between 20% and 37%; however, only a few patients survived beyond 1 year. Each of the 4 groups attempted to define the precise risk factors that could be used as prognostic tools to identify the subgroup of patients who would benefit from induction regimens and those who should be offered supportive care. Whereas Mori and colleagues7 emphasized the importance of karyotype as a risk factor, Wahlin and colleagues6 assigned risk groups based on karyotype, the presence of an antecedent hematologic disorder, and leukocytosis. DeLima and colleagues5 observed that poor performance status and low albumin were significant risk factors in very elderly patients with AML. However, a more recent analysis from the same group8 in a larger cohort of patients revealed that karyotype, performance status, antecedent hematologic disorder, and the use of laminar air-flow rooms all were important prognostic factors. Those studies included between 24 and 82 patients with AML aged ≥80 years, and none of assessed the role of comorbidities on the outcome of these elderly patients.
Comorbid illness was defined in the 1960s by Feinstein9 as a distinct additional clinical entity that has existed or may occur during the clinical course of a patient with a primary disease. Specifically in cancer, comorbid illnesses have led to minimal enrollment of cancer patients onto clinical trials and often have led to substandard treatment. The Charlson comorbidity index (CCI) was developed in 198710 and was validated in many types of cancers. It takes into consideration 19 medical conditions and assigns a different weight to each. The total score is between 0 and 4: The higher the score, the worse the outcome.11 It recently was demonstrated that the CCI is an independent predictor of CR in patients with AML aged ≥70 years.12 The hematopoietic cell transplantation comorbidity index (HCTCI) was developed by Sorror and colleagues13 to predict the outcome of patients who undergo blood and bone marrow transplantation; it is an adapted and more developed version of the CCI with a score from 0 to 11. The HCTCI recently was used to assess patients with AML aged ≥60 years.14, 15 Whereas 1 study14 indicated that the HCTCI was predictive of early death and event-free survival (EFS), the other study15 did not. It is noteworthy that no comparison was done between the 2 scoring systems (CCI and HCTCI) in patients with AML.
Patients aged ≥80 years are a precarious group of patients, because they are beyond the life expectancy in the United States (77.8 years).16 There is a clear need to help clinicians and other healthcare providers decide when to offer intensive approaches and when to refer patients to supportive care and/or hospice.17 Therefore, for the current study, we analyzed our institutional experience with these patients to define their prognosis based on pretreatment factors.
MATERIALS AND METHODS
Adults who had a diagnosis of AML at aged ≥80 years were evaluated at the Roswell Park Cancer Institute between 1991 and 2007. Patients with acute promyelocytic leukemia were excluded from this analysis. We recorded patients' age, sex, performance status, albumin, lactic dehydrogenase level, white blood cell count, morphology (monocytic [including myelomonocytic, monoblastic, and monocytic] vs nonmonocytic), including the presence of extramedullary disease, bone marrow and peripheral blood blast percentages, CD34 and CD56 positivity of leukemic blasts, karyotype (either favorable and intermediate18 were grouped together, or favorable and normal karyotype vs all others), de novo versus secondary AML, CCI and HCTCI, and treatment outcome. Induction therapy varied based on protocol availability (Table 1). Supportive care consisted of blood product support with or without hydroxyurea. This analysis was approved by the Roswell Park Cancer Institute Scientific Review Committee and the Institutional Review Board.
Table 1. Induction Treatment Regimens
No. of Patients (%)
Cytarabine 100 mg/m2 as a continuous infusion over 7 days, daunorubicin 60 mg/m2 daily for 3 days, and etoposide 100 mg/m2 daily for 3 days.
Cytarabine 1.5 g/m2 every 12 hours for a total of 12 doses and idarubicin 12 mg/m2 daily for 3 consecutive days.
Cytarabine 100 mg/m2 continuous infusion over 7 days and daunorubicin 60 mg/m2 daily for 3 days.
Cytarabine 100 mg/m2 continuous infusion over 7 days, daunorubicin 60 mg/m2 daily for 3 days, and oblimersen 7 mg/kg daily as a continuous infusion over 10 days.
Hematologic CR was defined according to previously established criteria.19 OS was defined as the interval between the day of diagnosis for supportively treated patients or the first day of treatment for intensively treated patients and death. Patients were censored at the date they were last seen alive. Disease recurrence after CR was defined by the appearance of peripheral blood blasts, >5% leukemic cells in bone marrow aspirates, or the development of extramedullary leukemia.
Descriptive statistics, such as frequencies and relative frequencies, were computed for all categorical variables. Numeric variables were summarized using simple descriptive statistics, such as the mean, median, standard deviation, range, etc. The medians were used as cutoffs for analyses. A variety of graphic techniques also was used to display data. To compare the CCI and the HCTCI, the area under the receiver operating characteristic (ROC) curve20 was computed for a variety of survival-based outcomes (eg, 3-month survival). The ROC curve is constructed through calculation of the sensitivity and specificity associated with all possible cutoffs of the individual test scores. The area under the ROC curve (AUC) is a measure of the predictive power of a given variable and, thus, is a measure of the diagnostic accuracy of each test in terms of predictability of outcome. The 2 AUCs were derived from the same set of patient data; therefore, the measures themselves are correlated, so that testing was based on the use of bootstrap methodology, which incorporated this information. The bootstrapping algorithm involved resampling of the sample pairs. A nominal significance level of .05 was used. The estimated OS distributions were obtained using the Kaplan-Meier method for binary variables. By using this distributional estimate, summary descriptive statistics, such as the median survival, were obtained. Statistical assessment of observed differences in the survival distributions of different groups of interest was done using the log-rank test. Cox proportional hazards models were used to assess the effect of study variables on OS in both univariate and multivariate analyses.
In total, 92 patients were analyzed. All patients were offered intensive treatment without waiting for karyotype analysis, and treatment decisions were made within 24 to 48 hours from diagnosis. Fifty-nine patients (64%) were treated intensively on a variety of regimens (Table 1), and 33 patients (36%) elected to receive supportive care. The patients' characteristics are described in Table 2; except for age and karyotype, the characteristics were similar between the 2 groups. Extramedullary disease was identified in 5 patients (2 patients had central nervous system involvement of extramedullary disease, and 3 patients had extramedullary disease at other sites); only 2 of these patients had monocytic morphology. Their response to treatment and OS are shown in Table 3. Because patients were evaluated after 1991, we compared the outcome by time (both dichotomous and continuous). There was no statistical difference in outcome according to the time of patient accrual among the whole cohort, those who received intensive treatment, and those who received supportive treatment (P = .2140, P = .2623, and P = .8369, respectively). There also was no significant difference between intensively and supportively treated patients with regard to OS (Fig. 1A), although patients who achieved CR had a longer survival (median survival, 39.5 weeks) than patients who did not (median survival, 9.3 weeks) (P = .0014) (Fig. 1B).
ROC curves were constructed for both the CCI and the HCTCI to examine the sensitivity and specificity of the comorbidity index scores in the prediction of 3-month survival (Fig. 2). The AUCs (0.564 for the CCI and 0.501 for the HCTCI) did not differ statistically (P = .35). Both comorbidity indexes had poor diagnostic accuracy. Results at 1 month, 6 months, 9 months, and 12 months were similar.
A univariate analysis of pretreatment characteristics associated with OS is shown in Table 4. Multivariate analysis, including all pretreatment characteristics (Table 5), identified near-normal albumin as an independent favorable prognostic factor for the whole cohort of patients (48% of patients; 1-year survival rate, >27%) (Fig. 3A). Similarly, age <83 years (47% of patients; 1-year survival rate, >25%) (Fig. 3B) and nonmonocytic morphology (75% of patients; 1-year survival rate, >26%) (Fig. 3C) were identified as independent favorable prognostic factors for the intensively treated cohort; and bone marrow blasts <46% (50% of patients; 1-year survival rate, >19%) (Fig. 3D) was identified as an independent favorable prognostic factor for those who received supportive care. Combining the prognostic factors did not result in better outcome prediction.
Table 4. Univariate Analysis of Prognostic Factors Associated With Overall Survival
This retrospective analysis of octogenarian and nonagenarian patients with AML represents our attempt to define the pretreatment characteristics that will predict who will benefit from treatment. It is noteworthy that, in our analyses, we detected 4 different prognostic variables: albumin for the whole cohort, age and nonmonocytic morphology in the intensively treated patients, and bone marrow blast count in the supportively treated patients.
This study could be criticized for the relatively small number of patients; however, to the best of our knowledge, this is the largest single-institution cohort of patients with AML aged ≥80 years. The importance of a single-institution approach should be underscored here, because all patients were offered a uniform treatment approach by a relatively small group of physicians and were cared for by the same mid-level practitioners, nurses, social worker, case manager, and psychologist.
The current study also may be censured for the long period of accrual (16 years). However, because AML treatment has not changed significantly in the last 2 decades,21 and because we did not detect any difference in outcome by time-to-accrual to this retrospective analysis, we believe that the time period does not impact this analysis.
The age difference between the intensively treated and supportively treated patients is intriguing. It may represent patient and/or physician choice. It is possible that the “younger” octogenarians preferred intensive treatment. Similarly, attitude disparities among us, the physicians, and our support staff unknowingly may have affected decision making when we discussed treatment options with the “older” octogenarians and nonagenarians.22 Such quibbles also were raised at the other end of the age spectrum with regard to treating young adults with acute lymphoblastic leukemia.23 Prospectively analyzing patient and physician attitudes may help answer questions about the role of the patient versus the physician in decision making about treatment assignment.
The difference in karyotype between the intensively and supportively treated patients was unexpected, especially because we did not wait for cytogenetic results before offering treatment to our patients. We propose to evaluate this question prospectively.
Our results about similar outcomes between intensively and supportively treated patients differ from the results reported by other groups. Specifically, in the study by Lowenberg et al,24 randomizing patients between intensive induction therapy and supportive care demonstrated a significantly longer survival duration for the intensively treated patients. However, the median age in that study was 72 years, and the cohort included only a dearth of patients ≥80 years old. Similarly, in 3 retrospective studies,25-27 longer survival was reported for intensively treated patients compared with supportively treated patients, although all of those studies included only a few patients aged ≥75 years. Conversely, our results are in line with 3 other retrospective studies,28-30 which demonstrated only a marginal, if any, advantage in OS for intensively treated patients. Finally, our study included the largest number octogenarian and nonagenarian patients.
Performance status was a predictor of outcome in most studies of patients with AML aged ≥60 years,1, 8, 31 but not in all such studies.24 Because most studies concentrated on sexagenarians and older patients, the role of performance status in the patients with AML aged ≥80 years will need to be evaluated prospectively.
It was somewhat disappointing to realize that neither the CCI nor the HCTCI had any predictive value in this patient population despite the finding that 24% of patients had a CCI of 0 (8% had a HCTCI of 0). Although 1 group of investigators reported that the CCI was predictive of CR,12 the HCTCI was described as predictive of early death and EFS by 1 group,14 but not by another group.15 None of those studies concentrated on the patients aged ≥80 years. The scoring variability between the CCI and the HCTCI is inherent within the different systems. Both systems were developed for other purposes; thus, the lack of predictability suggests that a new system for analyzing the role of comorbidities in this age group needs to be devised.
Several groups of investigators have demonstrated that karyotype has prognostic significance in patients with AML aged ≥80 years.6-8 One possibility for the discrepancy between our results and others may be the system used to define intermediate versus unfavorable karyotype cohorts. We used the Cancer and Leukemia Group B system,18 and we also separated the patients into 2 groups: favorable and normal karyotype versus all others (data not shown). Neither method resulted in a significant value to predict outcome. Thus, the finding that karyotype lacked predictability power here may reflect the relatively smaller size of the intensively treated cohort compared with the other groups.
Three of the outcome predictors, albumin, age, and bone marrow blast count, were identified previously as statistically significant in elderly patients with AML. Albumin had prognostic value in elderly patients with AML according to at least 1 additional group.5 Its role in predicting outcome may be related to its relation to the patients' nutritional status and their hepatic function. It is noteworthy that albumin was an outcome predictor in several other malignancies.32, 33 Age is a well known prognostic factor in all patients with AML and specifically in the elderly.8 Similarly, at least 1 group of investigators observed that bone marrow blasts had prognostic significance in elderly patients with AML.34 The percentage of bone marrow blasts may reflect disease activity; patients with higher blast counts may have more aggressive disease. Finally, our data demonstrate that patients with nonmonocytic morphology had a better outcome among the intensively treated patients. It also worth noting that, in contrast to previous published data,35, 36 extramedullary disease was less common in the patients who had monocytic morphology in this age group. If the current results can be reproduced in wider patient cohorts in a prospective manner, then these 4 variables may be helpful in assigning treatments to patients aged ≥80 years with AML.
Recently, it was demonstrated that low-dose cytarabine was beneficial for supportively treated patients.37 None of our patients received this treatment modality. Furthermore, the new term, “unfit for intensive treatment,” was not coined until recently.8 This term describes those patients with an 8-week mortality that exceeds 50%. It includes not only induction death/treatment toxicity but also lack of efficacy of a particular approach. Our current model, if confirmed prospectively, could identify octogenarian and nonagenarian patients with normal albumin levels and nonmonocytic morphology as those who may be “fit for chemotherapy” despite their advanced age.
The question remains which treatment to offer patients with AML aged ≥80 years who are fit for chemotherapy. Several new drugs are currently in development, eg, clofarabine, cloretazine, decitabine, and arsenic trioxide. Their role will be determined in the next few years, when the clinical trials will be completed. In the interim, we recommend enrolling all octogenarian and nonagenarian patients with AML onto clinical trials.
In summary, this retrospective study defined prognostic groups among intensively treated and supportively treated octogenarian and nonagenarian patients with AML. If the results are confirmed in prospective studies, then physicians, allied care support staff, and patients should be able to consider treatments in a more informed manner.
Conflict of Interest Disclosures
Supported in part by grant CA16056 from the National Cancer Institute (W.T., G.E.W., S.N.J.S., A.W.B. M.B., P.K.W., E.S.W., and M.W.); by the Szefel Foundation and the Roswell Park Cancer Institute (E.S.W.); and by the Heidi Leukemia Research Fund, Buffalo, New York (M.W.).