Validation of a Frailty Score Predicting Survival of Elderly, Non-Fit AML Patients Receiving Hypomethylating Therapy: Results of the Decider Trial

Introduction

In older patients (pts), host factors such as functional deficits, comorbidities and other age-related factors are increasingly recognized as predictors for outcome of leukemia treatment. Thus, prospective clinical trials increasingly implement functional studies (often termed geriatric assessment) in their pretreatment diagnostic workup. In older, fit AML pts receiving standard chemotherapy, Klepin et al. demonstrated a predictive value of this approach for overall survival (OS) in a single-center study (Blood 2013). We also developed a Frailty Score, assessed at 3 study centers and predicting OS of elderly AML/MDS pts receiving either hypomethylating agents (n=66) or sole best supportive care (n=35; Deschler et al., Haematologica 2013). To validate this score, which is composed of performance status (PS), activities of daily living (ADL) and fatigue, we prospectively assessed these parameters in the DECIDER trial (AMLSG 14-09, NCT00867672).

Methods

In the DECIDER trial, 200 non-fit AML pts aged >60 years (yr) were randomized between four treatment arms with either decitabine (DAC) alone, or DAC plus all-trans retinoid acid (ATRA) or DAC plus valproic acid (VPA) or DAC plus ATRA and VPA. We assessed PS via ECOG, ADL via Barthel index and fatigue via the EORTC QLQ-C30 questionnaire, which were available for 200, 175 and 156 pts, respectively. Pts with missing ADL and/or fatigue assessments tended to have a lower PS than those with complete data, which may explain at least in part why the assessments could not be performed. The Frailty Score was calculated using ECOG PS as a substitute for the Karnofsky Index (KI, as applied when establishing this score), with ECOG 0-1 taken as corresponding to KI ≥80, and ECOG 2-3 to KI <80%. The influence of the single factors (ECOG, ADL, fatigue) and of the Frailty Score on OS was investigated with Cox regression models adjusted for treatment, and in analyses adjusted additionally for established disease-related prognostic factors.

Results

For 141 pts with complete data in all 3 parameters building the Frailty Score, median follow-up time was 28 months, and 122 pts had died. The median age of pts at AML diagnosis was 76 yr (range 61-90) and 96 pts (68.1%) were males. Median WBC was 3.100/µl (interquartile range [IQR] 1.400 to 11.500/µl), median platelet count 49.000/µl (IQR 30.000 to 98.000/µl), median hemoglobin (Hb) 9.2g/dl (IQR 8.3 to 10.1g/dl), with median peripheral blasts of 48% (IQR 25 to 76%) and median LDH 287 U/l (IQR 203 to 459 U/l). 32.6% of the pts had adverse genetic risk according to ELN 2010 criteria. 56.7% had relevant comorbidities (HCT-CI ≥3). The majority of pts had a reduced PS (21.3% ECOG 0, 62.4% ECOG 1 and 16.3% ECOG 2-3). When assessing ADL, 30.5% had limitations (ADL <100). Fatigue ≥50 was detected in 58.2%. This resulted in 45 pts (31.9%) with score 0, 61 pts (43.3%) with score 1, 18 pts (12.8%) with score 2 and 17 pts (12.1%) with score 3. In analyses adjusted for treatment, ECOG PS >1, ADL <100 and fatigue ≥50 had significant adverse effects on OS, with estimated hazard ratios (HRs) of 1.78, 1.91, and 1.62, respectively (Table 1). When the analyses were additionally adjusted for the established disease-related factors HCT-CI, LDH, hemoglobin, and genetics, effects were reduced with estimated HRs of 1.38, 1.37, and 1.40, respectively. Our Frailty Score, applied to this cohort of AML pts, showed that with the defined cut-off values of PS via ECOG, ADL and fatigue, pts can be grouped into three risk categories (combining pts with score 2 and 3 into one group) with significantly different OS (adjusted for treatment, trend test p<0.0001, Table 1, Figure 1). When the analysis was adjusted for the stated disease-related prognostic factors, this effect was also reduced, but still present with an estimated HR of 1.99 comparing 1 vs. 0 points, and of 1.70 comparing 3 vs. 0 points, trend test p=0.033.

Conclusions

This previously described pt fitness assessment to calculate the Frailty Score could now be implemented in a multicenter academic trial setting. In our independent, large AML cohort, the score was able to separate pts into groups with different OS, underlining the emerging roles of ADL and fatigue as predictors of outcome, in addition to the already well-established role of PS. A second prospective validation study, in AML pts aged ≥ 60 yr receiving standard "7+3" induction therapy or 10-day DAC (EORTC AML21 "inDACtion vs. induction" trial, NCT02172872), is ongoing.

View largeDownload slide

View largeDownload slide

Close modal