In older patients with acute myeloid leukemia, the more frequent presence of biologically inherent therapy-resistant disease and increased comorbidities translate to poor overall survival and ...therapeutic challenges. Optimal front-line therapies for older patients with acute myeloid leukemia remain controversial. We retrospectively evaluated survival outcomes in 980 elderly (≥70 years) acute myeloid leukemia patients from a single institution between 1995 and 2016. Four treatment categories were compared: high-intensity (daunorubicin/cytarabine or equivalent), hypomethylating agent, low-intensity (low-dose cytarabine or similar without hypomethylating agents), and supportive care therapy (including hydroxyurea). At a median follow up of 20.5 months, the median overall survival for the entire cohort was 7.1 months. Multivariate analysis identified secondary acute myeloid leukemia, poor-risk cytogenetics, performance status, front-line therapy, age, white blood cell count, platelet count, and hemoglobin level at diagnosis as having an impact on survival. High-intensity therapy was used in 360 patients (36.7%), hypomethylating agent in 255 (26.0%), low-intensity therapy in 91 (9.3%), and supportive care in 274 (28.0%). Pairwise comparisons between hypomethylating agent therapy and the three other treatment groups demonstrated statistically significant superior median overall survival with hypomethylating agent 14.4 months)
high-intensity therapy 10.8 months, hazard ratio 1.35, 95% confidence interval (CI): 1.10-1.65;
=0.004, low-intensity therapy (5.9 months, hazard ratio 2.01, 95%CI: 1.53-2.62;
<0.0001), and supportive care (2.1 months, hazard ratio 2.94, 95%CI: 2.39-3.61;
<0.0001). Our results indicate a significant survival benefit with hypomethylating agents compared to high-intensity, low-intensity, or supportive care. Additionally, high-intensity chemotherapy resulted in superior overall outcomes compared to low-intensity therapy and supportive care. Results from this study highlight the need for novel therapeutic approaches besides utilization of intensive chemotherapy in this specific aged population.
Introduction
Patients older than 70 years with acute myeloid leukemia (AML) are generally considered to have poor prognosis. As a result, many patients are routinely not offered active treatment ...and/or are referred to palliative hospice care based on the assumption that their expected survival will be well below 6 months. However, a substantial number of patients live beyond 6 months indicating that management decisions ought to be individualized taking into considerations patients' preferences about benefits and harms of treatments and estimated survival prognostication.
Methods
Using large Moffitt AML database we identified all consecutive patients (n=305) with AML older than 70 who received high or low intensity chemotherapy to develop a multiple logistic regression model to assess the probability of survival at 12 month since diagnosis of AML. Patients who were censored prior to 12 months were considered not eligible (n=300). The final model was determined by the backward elimination method. We assessed discrimination of the model by performing ROC (receiver operating characteristic) analysis and calibration by using Hosmer-Lemeshow (H-L) goodness-of-fit test. We also performed regret-based decision curve analysis (DCA) to compare three decision strategies over all possible patient's preferences: "Do Not Treat/Refer to Hospice" vs. "Treat All" with chemotherapy vs. "Use Model" to guide decision about treatment (to treat or not to treat depending on the survival estimates in a relationship to the patient's preferences). In DCA, the preferences are captured by determining the threshold probability (T) of disease (AML) outcome at which a patient is indifferent between benefits (B) and harms (H) of treatment according to: T=1/1+B/H. The T can be elicited by asking a simple question concerning regret of omission (failure to benefit) vs. regret of commission (causing unnecessary harm): "how many more times would you regret not receiving a health intervention that could improve disease outcome (survival) compared with unnecessary and potentially harmful administration of treatments?" (http://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/1472-6947-10-51) Based on our previous study, we assumed that if treatment is administered, it is associated with hazard ratio (HR) of death reduction by 0.35 in the baseline analysis. The best strategy is the one associated with the least amount of regret.
Results
The prognostic model consisted of the following variables; cytogenetic status, ECOG PS, type of AML (De Novo vs. Secondary), and WBC level. A total of 112 patients (37%) survived at least 12 months. The model has good discrimination (area under curve=0.80) and excellent calibration HL (chi2)=4.3; p=0.83 (Fig 1). DCA analysis showed that strategy "Do Not Treat/Refer to Hospice" was always inferior to the strategies "Treat" vs. "Use Model" (Fig 2). The decision strategy "Treat All" patients with AML older than 70 was best strategy for the threshold probability ranging from 1 to 46%. That is, as long as the patient would regret of not receiving benefit of treatment between 99 to 1.17 more than unnecessary receiving potentially harmful chemotherapy treatment, "Treat All" represents the best decision strategy for the management of elderly patients with AML. If the harms of treatments are more important to the patient (B/H<1.17), then "Use Model" becomes the best management choice. If we assumed that HR of treatment effect is equal to 0.1 and 0.65, then the threshold varied from 1% to 39% and 1% to 75%, respectively. That is, under these circumstances "Treat All" becomes best strategy if the patient values benefit of treatment 99 to 1.56 (or, 99 to 0.33 when HR of treatment=0.65) times more than avoiding harms of treatment.
Limitation: Our model requires external validation before it can be exported for the use in a routine practice.
Conclusion
Our analysis indicates that not offering treatment to patients older than 70 with AML is never acceptable. The optimum decision is driven by the patients' preferences and estimated survival. If the patient regrets not receiving the potential benefits of treatments more than avoiding harms, treatment should be offered. The treatment should be avoided if the patient places more weight on harms than on benefits of treatment. Predictive model can help guide this decision.
This research is supported by NIH grant 1-R01-CA168677-01A1
Extermann:GTx: Research Funding.
Introduction
Treatment of elderly patients with acute myeloid leukemia (AML) is a therapeutic challenge. Elderly patients frequently have more biologically inherent resistant disease, along with ...comorbidities that together result in poor overall survival (OS). Optimal frontline therapy for elderly patients with AML remains controversial, and choice of regimen varies among clinicians. In this large, single-institution retrospective cohort study of AML patients over age 70, we present survival analysis and comparison amongst a variety of commonly used initial regimens.
Methods
602 AML patients aged 70 or greater who received treatment between 1995 and 2014 in a single-large institution were retrospectively analyzed. Patients were categorized into 4 different treatment groups: High Intensity Therapy (defined as daunorubicin/cytarabine or equivalent), Hypomethylating Agent (HMA) Therapy, Low intensity Therapy (defined as low-dose cytarabine or similar without HMAs), and Supportive Care (including hydroxyurea if indicated). Age, type of AML, history of previous hematological disease, cytogenetics, ECOG performance status, comorbidities (Charlson Index), complete blood counts and blast percentage at time of diagnosis were obtained for each treatment category. Pairwise comparison of survival between different treatment groups was performed, using the stratified log-rank test and propensity score matching to adjust for potential treatment indication bias between groups. Within pairwise comparison groups, the stratified Cox proportional hazards regression model was used to assess correlation of the clinical variables with overall survival.
Results
Median age was 77 years (range 70 - 95) with a male predominance (M:F=68%:32%). ECOG Performance Status was 0 to 1 in 80% and 2 to 4 in 20% of patients. Per NCCN criteria, cytogenetics risk category was intermediate or favorable in 67% and unfavorable in 33%. Baseline median WBC, hemoglobin and platelet counts were 3.31 k/uL, 9.40 g/dL and 43 k/uL respectively. Median baseline bone marrow blast percentage was 35% (range 2% - 95%), and the large majority (445 of 550 patients - 81%) had peripheral blood blasts at the time of diagnosis. The majority of patients had secondary AML (61%) compared with de novo AML (39%). Of those with secondary AML, myelodysplastic syndrome (MDS) was the most common antecedent hematologic disease (97%), for which 36% had received prior HMAs.
For frontline therapy, 238 (40%) patients received High Intensity Therapy; 110 (18%) received HMA Therapy, 67 (11%) received Low Intensity Therapy, and 187 (31%) received Supportive Care. Pairwise comparison between HMA Therapy and the 3 other treatment groups individually demonstrated statistically significant superior OS with HMA Therapy (median 13.3 mo; 95% CI 10.6 - 16.8 mo) compared to High Intensity Therapy (median 9.5 mo; 95% CI 7.4 - 10.9 mo), Low Intensity Therapy (median 5.9 mo; 95% CI 4.2 - 7.8 mo) and Supportive Care (median 2.5 mo; 95% CI 2.1 - 3.0 mo.). In addition, pairwise comparisons demonstrated superior OS with high vs low-intensity (p=0.0007), but no significant difference between Low-Intensity and Supportive Care (p=0.10). A pairwise comparison between HMA and High Intensity Therapy in the small subset of patients who had received prior HMA for MDS revealed extremely poor outcomes in both arms, with < 6 month median OS.
Conclusion
In this analysis of a very large data set of patients over age 70 with AML, using pairwise comparison with propensity score matching, our results indicate a survival benefit with high intensity therapy or HMAs compared to supportive care or low-intensity (non-HMA) therapy. Interestingly, treatment with HMAs also resulted in better OS than traditional high intensity therapy. These data are contributing to an ongoing effort to design a comprehensive decision analysis model comparing treatment effectiveness according to baseline characteristics in AML patients 70 and older.
Display omitted
Lancet:Kalo-Bios: Consultancy; Boehringer-Ingelheim: Consultancy; Celgene: Consultancy, Research Funding; Amgen: Consultancy; Seattle Genetics: Consultancy; Pfizer: Consultancy.