Risk Stratification of Elderly Patients with Newly Diagnosed Acute Myeloid Leukemia Unfit for Intensive Chemotherapy

Abstract

Background. Acute myeloid leukemia (AML) incidence increases with age, but elderly patients are often too frail to receive intensive chemotherapy (IC). Instead, treatment with hypomethylating agents (HMAs) is usually proposed, even if they have not demonstrated any real improvement over best supportive care. However, clinical practice has shown a significant difference in overall survival (OS) between responding and non-responding patients to HMAs. Thus, we aimed to assess predictive criteria of HMA response in elderly and frail AML patients. We reviewed the literature for any existing scores: none was found for elderly unfit patients. On the other hand, 3 models/scores existed for fit elderly patients undergoing IC : in the ALFA 9803 trail published by Malfuson and al, the HOCSG model in the MRC AML11 and LRF AML trails by Wheatley and al and finally the MDACC model by Kantarijan and al. Methods. We reviewed all patients aged ≥ 60 years old (y.o) diagnosed with AML, unfit for IC and treated with Azacytidine (AZA) alone in first line from July 2015 to December 2019 in Poitiers' University Hospital. The ineligibility to IC was defined by an age > 75 y.o and/or the appreciation by the physician based on performance status (ECOG) and the evaluation of comorbidities with Charlson Comorbidity Index (CCI). The type of AML (de novo versus secondary) and the 2017 ELN prognosis score were also assessed. Results. Among 63 patients recruited, 86% were older than 70 y.o and 29% older than 80. Frailty criteria were found with a ECOG ≥ 2 in 11 (17%) and a CCI ≥ 3 in 29 (45%) patients. Secondary AML was found in 41 (63%) of patients. Adverse karyotype was detected in 21 (32%) patients and 2017 ELN score was quoted as adverse in 25 (38%) patients. After a median follow up of 10.75 months (IQR 3.98 - 17.94) for the whole cohort, median OS was 10.75 months (95%CI 6.37476 - 14.984). Among 54 (86%) evaluable patients, 50 reached some response: at least hematological improvement according to 2003 International Working Group criteria in 20 of them (37%), and stable disease in 30 (55.6%) patients. Patients reaching at least hematological improvement were able to complete a median of 11.5 cycles (IQR 7.5 - 17) vs 6.5 cycles (IQR 3 - 12). Sadly 23 (35%) patients died before achieving 6 cycles. In our cohort the 2017 ELN score failed to predict risk assessment in elderly frail patients. The ALFA model and HOCSG score were also not able to identify treatable patients neither to predict mortality in our cohort. Interestingly the MDACC high-risk category was able to identify the patients who would not benefit from AZA treatment with worst survival rates (HR 3.01; 95%CI 1.04 - 8.79). Conclusion. To date, stratification scores were developed for young fit patients, undergoing IC. Those scores predict poorly response to HMA and OS in unfit patients. In the era of new treatments and combinations with AZA, there is an urgent and growing need for dedicated prognosis model for unfit elderly AML patients. Disclosures Leleu: Bristol-Myers Squibb: Honoraria; Carsgen Therapeutics Ltd: Honoraria; Gilead Sciences: Honoraria; Celgene: Honoraria; Janssen-Cilag: Honoraria; Karyopharm Therapeutics: Honoraria; Merck: Honoraria; Mundipharma: Honoraria; Novartis: Honoraria; Amgen: Honoraria; AbbVie: Honoraria; Oncopeptides: Honoraria; Pierre Fabre: Honoraria; Roche: Honoraria; Sanofi: Honoraria; Takeda: Honoraria, Other: Non-financial support.

The transcriptomic landscape of elderly acute myeloid leukemia identifies B7H3 and BANP as a favorable signature in high-risk patients

“…In this context core binding factor (CBF) leukemias, NPM1 and biallelic CEPBA mutations are considered of good prognosis, while complex or monosomal karyotype, TP53 , FLT3-ITD , RUNX1 and ASXL1 mutations, other recurrent translocations and KMT2A rearrangements confer a poor prognosis. However, this classification is not validated for elderly patients treated with a non-intensive approach, and its performance seems to be suboptimal in this context ( 7 ).…”