https://orcid.org/0000-0001-8861-0281
TO THE EDITOR:
Novel strategies are needed to assess frailty in the context of treatment intensity for older adults with acute myeloid leukemia (AML).1,2 Age alone is insufficient for predicting tolerability but is often a primary factor in therapeutic decision-making.1,3 Incorporation of validated strategies to assess fitness or frailty into clinical trials and practice can inform the risks and benefits of treatment. One such strategy is the use of pretreatment geriatric assessments (GAs). GA measures are prognostic of survival and toxicity.4-7 How to best use GA data to characterize frailty in the context of AML therapy remains unknown.
Here, we have assessed pretreatment vulnerability using the deficit accumulation frailty index (DAFI). The DAFI approach can use all GA variables to create a summary score of frailty by quantifying the number of detected age-related deficits in health over the total measured.8 This summary score can stratify populations by frailty status. DAFI has been shown to predict health care utilization and mortality in general older adult populations, including cancer populations.9-12 The purpose of our study was to use DAFI to characterize pretreatment frailty and explore associations with outcomes in older patients with AML.
We retrospectively compiled deidentified data from 2 completed Cancer and Leukemia Group B (CALGB) trials that enrolled patients aged ≥60 years with newly diagnosed AML. CALGB is now part of the Alliance for Clinical Trials in Oncology. CALGB 11001 enrolled patients with FLT3-mutated AML who were determined fit per physician assessment to receive intensive chemotherapy with 7+3 (cytarabine and daunorubicin) plus sorafenib.13 Patients on CALGB 11002 received nonintensive therapy with decitabine ± bortezomib.14 Patients who consented to optional companion studies for these trials (CALGB 361006 and 361101) and completed GA within 7 days before starting treatment4,7 were included in our analysis (data lock 30 September 2021).
DAFI was calculated using 49 of 51 previously published items derived from prechemotherapy demographic, GA, and laboratory data.10 Albumin and Hospital Anxiety and Depression Scale scores were not available for all patients and therefore not included. DAFI score was calculated as the cumulative sum of patient item responses (deficits) divided by the maximum potential deficit of their nonmissing items, in which at least 44 nonmissing items (90%) were available. The DAFI is scored from 0 to 1 (0 is absence of the deficit and 1 is the highest level of deficit), with DAFI categories as defined previously: nonfrail, 0 to <0.2; prefrail, 0.2 to <0.35; and frail, ≥0.35.8
DAFI pretreatment frailty categories were summarized using descriptive statistics, frequencies, and percentages. Kaplan-Meier was used to estimate early mortality probability at 30 and 60 days and overall survival (OS). Cox proportional hazards models assessed age-adjusted relationship of frailty on OS. Toxicity was summarized as the maximum grade nonhematologic adverse event (AE) experienced per patient. Logistic regressions explored this relationship in univariable and age-adjusted multivariable models. Fisher exact test was used to assess the rates of grade 3 to 5 AE across frailty categories. Each study was analyzed independently. Data collection and analyses were conducted by the Alliance Statistics and Data Management Center according to Institutional Review Board guidelines and the tenets of the Declaration of Helsinki.
Baseline clinical and demographic characteristics of evaluable patients are presented in Table 1. The median length of patient follow-up was 89.2 months for intensive study and 88 months for nonintensive study.
Demographic and clinical characteristics
| Intensive treatment (n = 36) | Nonintensive treatment (n = 75) | |
|---|---|---|
| Age, median (range), y | 67.5 (60.5-82.1) | 72.4 (60.6-92.3) |
| Age group, n (%) | ||
| 60-64 | 11 (30.6%) | 12 (16.0%) |
| 65-69 | 12 (33.3%) | 15 (20.0%) |
| 70-74 | 7 (19.4%) | 16 (21.3%) |
| 75-79 | 4 (11.1%) | 15 (20.0%) |
| ≥80 | 2 (5.6%) | 17 (22.7%) |
| Female sex, n (%) | 15 (41.7%) | 22 (29.3%) |
| Race, n (%) | ||
| White | 35 (97.2%) | 65 (86.7%) |
| Black/African American | 0 (0.0%) | 3 (4.0%) |
| Native Hawaiian or Pacific Islander | 0 (0.0%) | 1 (1.3%) |
| Unknown | 1 (2.8%) | 6 (8.0%) |
| Hispanic ethnicity, n (%) | 1 (2.8%) | 0 (0.0%) |
| ECOG PS, n (%) | ||
| 0 | 23 (63.9%) | 29 (38.7%) |
| 1 | 8 (22.2%) | 36 (48.0%) |
| 2 | 4 (11.1%) | 9 (12.0%) |
| Missing | 1 (2.8%) | 1 (1.3%) |
| FLT3, n (%) | ||
| FLT3 ITD | 27 (75.0%) | 1 (1.3%) |
| FLT3 TKD | 9 (25.0%) | 1 (1.3%) |
| No FLT3 gene mutation | 0 (0.0%) | 73 (97.3%) |
| Clinical onset of AML, n (%) | ||
| De novo | 29 (80.6%) | 44 (58.7%) |
| Therapy-related myeloid | 3 (8.3%) | 9 (12.0%) |
| MDS related | 4 (11.1%) | 12 (16.0%) |
| Antecedent hemolytic disorder | 0 (0.0%) | 10 (13.3%) |
| ELN category, n (%) | ||
| Adverse | 2 (5.6%) | 27 (36.0%) |
| Normal/intermediate | 29 (80.6%) | 37 (49.3%) |
| No diagnosis cytogenetics | 5 (13.9%) | 11 (14.7%) |
| Marrow blasts, median (range), %∗ | 57.5 (0.0-96.0) | 10.0 (0.0-91.0) |
| Peripheral WBC count, median (range), ×103/μL† | 13.5 (0.8-343.6) | 3.0 (0.0-212.7) |
| LDH, median (range), U/L | 526.5 (103.0-2813.0) | 234.0 (68.0-1467.0) |
| Creatinine, median (range), mg/dL | 0.9 (0.4-1.7) | 0.9 (0.1-7.0) |
| LVEF, median (range), %‡ | 62.0 (42.0-76.0) | 61.0 (29.0-81.0) |
| Intensive treatment (n = 36) | Nonintensive treatment (n = 75) | |
|---|---|---|
| Age, median (range), y | 67.5 (60.5-82.1) | 72.4 (60.6-92.3) |
| Age group, n (%) | ||
| 60-64 | 11 (30.6%) | 12 (16.0%) |
| 65-69 | 12 (33.3%) | 15 (20.0%) |
| 70-74 | 7 (19.4%) | 16 (21.3%) |
| 75-79 | 4 (11.1%) | 15 (20.0%) |
| ≥80 | 2 (5.6%) | 17 (22.7%) |
| Female sex, n (%) | 15 (41.7%) | 22 (29.3%) |
| Race, n (%) | ||
| White | 35 (97.2%) | 65 (86.7%) |
| Black/African American | 0 (0.0%) | 3 (4.0%) |
| Native Hawaiian or Pacific Islander | 0 (0.0%) | 1 (1.3%) |
| Unknown | 1 (2.8%) | 6 (8.0%) |
| Hispanic ethnicity, n (%) | 1 (2.8%) | 0 (0.0%) |
| ECOG PS, n (%) | ||
| 0 | 23 (63.9%) | 29 (38.7%) |
| 1 | 8 (22.2%) | 36 (48.0%) |
| 2 | 4 (11.1%) | 9 (12.0%) |
| Missing | 1 (2.8%) | 1 (1.3%) |
| FLT3, n (%) | ||
| FLT3 ITD | 27 (75.0%) | 1 (1.3%) |
| FLT3 TKD | 9 (25.0%) | 1 (1.3%) |
| No FLT3 gene mutation | 0 (0.0%) | 73 (97.3%) |
| Clinical onset of AML, n (%) | ||
| De novo | 29 (80.6%) | 44 (58.7%) |
| Therapy-related myeloid | 3 (8.3%) | 9 (12.0%) |
| MDS related | 4 (11.1%) | 12 (16.0%) |
| Antecedent hemolytic disorder | 0 (0.0%) | 10 (13.3%) |
| ELN category, n (%) | ||
| Adverse | 2 (5.6%) | 27 (36.0%) |
| Normal/intermediate | 29 (80.6%) | 37 (49.3%) |
| No diagnosis cytogenetics | 5 (13.9%) | 11 (14.7%) |
| Marrow blasts, median (range), %∗ | 57.5 (0.0-96.0) | 10.0 (0.0-91.0) |
| Peripheral WBC count, median (range), ×103/μL† | 13.5 (0.8-343.6) | 3.0 (0.0-212.7) |
| LDH, median (range), U/L | 526.5 (103.0-2813.0) | 234.0 (68.0-1467.0) |
| Creatinine, median (range), mg/dL | 0.9 (0.4-1.7) | 0.9 (0.1-7.0) |
| LVEF, median (range), %‡ | 62.0 (42.0-76.0) | 61.0 (29.0-81.0) |
ECOG, Eastern Cooperative Oncology Group; ELN, European LeukemiaNet (2010); ITD, internal tandem duplication; LDH, lactate dehydrogenase; LVEF, left ventricle ejection fraction; MDS, myelodysplastic syndrome; TKD, tyrosine kinase domain; WBC, white blood cell.
Data available for 62 patients in CALGB 11002.
Data available for 73 patients in CALGB 11002.
Data available for 35 patients in CALGB 11001 and 68 patients in 11002.
Individuals in the CALGB 11001 intensive study (ClinicalTrials.gov identifier: NCT01253070; n = 36) were categorized as nonfrail (75%) or prefrail (25%) based on DAFI score. Nonfrail and prefrail patients had a median age of 67.3 and 69.4 years, respectively. The Spearman correlation between Eastern Cooperative Oncology Group Performance Status (PS) and DAFI score was 0.33 (95% confidence interval [CI], 0.00-0.60). Among patients on this trial, 86% experienced at least 1 grade 3+ nonhematologic AE, and 36% patients had at least 1 grade 4+ nonhematologic AE. Among nonfrail and prefrail individuals, the incidence of nonhematologic grade 3+ AE was 85% vs 89%, respectively (P > .99). However, prefrail adults experienced more grade 4+ nonhematologic toxicity than nonfrail individuals (67% vs 26%; P < .05). For nonfrail patients, both the 30- and 60-day early mortality probabilities were the same at 7.4%; and for prefrail, both were 22.2%, with no difference in OS by frailty category (P = .22; Figure 1). The hazard ratio for death in prefrail compared with nonfrail was 1.83 (95% CI, 0.81-4.12) after controlling for age.
In the nonintensively treated CALGB 11002 study (ClinicalTrials.gov identifier: NCT01420926; n = 75), DAFI score categories were 33% nonfrail, 49% prefrail, and 17% frail. The Spearman correlation between PS and DAFI score was 0.21 (95% CI, –0.02 to 0.42). Nonfrail and prefrail patients had median ages of 71.5 and 72.4 years, respectively. The median age for frail patients was 78.5 years, with nearly 50% aged >80 years. Eighty-nine percent of patients had at least 1 grade 3+ nonhematologic AE; of these, the incidence among nonfrail, prefrail, and frail individuals was 88.0%, 94.6%, and 76.9% (P = .15), respectively. Forty-eight percent of patients experienced at least 1 grade 4+ nonhematologic AE, with 36.0% in nonfrail, 54.1% in prefrail, and 53.8% frail categories (P = .36). The 30- and 60-day early mortality probabilities were 0% and 12% for nonfrail patients, 2.7% and 16.2% for prefrail patients, and 15.4% and 30.8% for frail patients (P = .17). There were 70 deaths observed during follow-up, with no statistical difference in OS between groups (P = .17; Figure 1). Age-adjusted hazard ratios compared with nonfrail were 1.6 (95% CI, 0.74-3.31) for frail and 1.5 (95% CI, 0.84-2.54) for prefrail.
Our analysis used a novel DAFI clinical data aggregation approach to distinguish between groups of patients with varying vulnerabilities. The DAFI score identified heterogeneity in accumulated vulnerabilities among patients enrolled on both intensive and less intensive trials. There were no frail patients enrolled in the intensive therapy trial, suggesting that clinical screening may have excluded patients with the highest cumulative deficits. However, the DAFI was able to identify a subgroup of prefrail patients in this trial who may be at higher risk for toxicity. In the nonintensive trial, the DAFI categorized 33% of patients as fit, despite their older age, but did not associate with toxicity or mortality. It is possible that lack of associations between DAFI and outcomes in the nonintensive setting may be related to limited power to detect differences or that a DAFI summary score has greater utility in more intensive treatment settings. It is also possible that DAFI may associate with other treatment tolerability outcomes that were not available in this data set. Additionally, there was a minimal correlation between DAFI score and PS, suggesting DAFI could provide additional insights into health status
Our analysis was designed to generate hypotheses and suggests that further investigation is needed to clarify the use of DAFI as a tool to identify subsets of older adults in clinical trial populations with varying risks of toxicity or mortality. The predictive utility of DAFI should be further evaluated in larger samples, ideally using data from a standardized GA, to assess its ability to predict patient outcomes over time and evaluate its utility in different clinical settings. Future comparisons between frailty measures and GA domains will be important.
The limitations of this study are that this retrospective, post hoc secondary analysis was performed using available data from 2 completed clinical trials and constrained to data previously collected. Similarly, the sample size was limited to patients enrolled in the trials and not adequately powered to detect differences between deficit accumulation groups. Real-world data would provide an important setting to test frailty assessment strategies. Lastly, this study uses data from a period before venetoclax.
DAFI score provides a potential strategy for categorization of frailty in AML. In our assessment, frailty by DAFI score was associated with toxicity during intensive induction. If supported by larger studies, a DAFI score could be a tool to facilitate comparative effectiveness research to understand intervention impact across frailty states.
Acknowledgments: Research reported in this publication was supported by National Cancer Institute of the National Institutes of Health award numbers UG1CA189823 (Alliance for Clinical Trials in Oncology National Cancer Insitute Community Oncology Research Program grant), UG1CA189824, and UG1CA23975), National Institute on Aging (NIA) of the National Institutes of Health grants R01AG068193 and R56AG068086 (J.S.M.), National Cancer Institute of the National Institutes of Health grants R01CA129769, R35CA197289, R35CA283926, and R01CA127617 (J.S.M.), R50CA275927 (G.L.U.), and NIA grant R33AG059206-03 (H.D.K.). This study was also supported, in part, by Bayer Healthcare/Berlex (CALGB 11001).
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Contribution: C.L., J.S.M., and H.D.K. contributed to concept and design, manuscript writing, data interpretation, and approval of the final article; M.L. and O.B. contributed to data analysis, manuscript writing, data interpretation, and approval of the final article; J.L.-R. contributed to concept and design, data analysis, manuscript writing, data interpretation, and approval of the final article; and S.M., G.J.R., and G.L.U. contributed to manuscript writing, data interpretation, and approval of the final article.
Conflict-of-interest disclosure: G.L.U. reports consulting fees from Jazz Pharmaceuticals. H.D.K. reports royalties from UpToDate as a contributor. The remaining authors declare no competing financial interests.
Correspondence: Heidi D. Klepin, Department of Internal Medicine, Wake Forest School of Medicine, Medical Center Blvd, Winston Salem, NC 27157; email: hklepin@wakehealth.edu.
