Obesity in Adult AML is Not Associated with Inferior Response or Survival Even When Dose Capping Anthracyclines: An ECOG-ACRIN analysis

Abstract

Background:

Obesity [body mass index (BMI) ≥30kg/m2)] is an important epidemiological risk factor for acute myeloid leukemia (AML) development, and we therefore studied its association with clinical and genetic phenotype and its impact on outcome in adult AML.

Methods:

We analyzed BMI in 1,088 adults receiving intensive remission induction and consolidation therapy in 2 prospective, randomized therapeutic clinical trials: E1900 (NCT00049517) (age <60 years) and E3999 (NCT00046930) (age ≥60 years).

Results:

Obesity was prevalent at diagnosis (33%), and in comparison with non-obesity was associated with intermediate cytogenetics risk group (p=0.008), with poorer performance status (p=0.01) and trend with older age (p=0.06). Obesity was not associated with somatic mutations among a selected 18-gene panel tested in a subset of younger patients. Obesity was not associated with clinical outcome (including complete remission, early death or overall survival), and we did not identify any patient subgroup with inferior outcomes based on BMI. Obese patients were significantly more likely to receive less than 90% of the intended daunorubicin dose despite protocol specification, particularly in the E1900 high-dose (90mg/m2) daunorubicin arm (p=0.002); however, this did not correlate with inferior overall survival on multivariate analysis (hazard ratio, 1.39, 95% CI, 0.90–2.13) (p=0.14).

Conclusion:

Obesity is associated with unique clinical and disease-related phenotypic features in AML, and may influence physician treatment decisions regarding daunorubicin dosing. However, our study shows that obesity is not a factor in survival and strict adherence to BSA based dosing is not necessary since dose adjustments do not affect outcomes.

Precis:

Obesity is prevalent in adults with acute myeloid leukemia and these patients are more likely to have intermediate risk cytogenetics and worse performance status, however, obesity does not impact clinical outcomes including complete remission, early death or overall survival. Dose modifications of daunorubicin (dose reductions and dose capping) in obese patients does not affect clinical outcomes and therefore strict adherence to full dose based on BSA is not required for intensive therapy in adult AML.

Introduction

Obesity has been identified as a significant risk factor for acute myeloid leukemia (AML) development, and case-control studies have demonstrated an approximate 1.5- to 2-fold increase in adult AML risk associated with obesity[1–5], particularly in women.[6–8] The degree of obesity, its development over the course of life, and more recently, obesity in adolescents are each associated with increasing risk of adult AML.[2, 9, 10] The observation of an increasing incidence of leukemia by birth cohort and by generation supports the hypothesis that adult leukemia risks are appreciably modulated by environmental and lifestyle exposures,[11] and therefore, the obesity-associated lifestyle represents an important and potentially modifiable factor (or behavior) associated with leukemia risk.

Past studies have demonstrated that abnormal body mass index (BMI), including both underweight and obesity, and in particular morbid obesity (BMI ≥40 kg/m²), is associated with inferior survival in pediatric AML studies,[12, 13] confirmed by a study at St. Jude’s Hospital[14] and a recent meta-analysis.[15] However, the impact in adult AML is less clear. Several recent reports do not show any significant difference in clinical outcome (i.e., complete remission [CR], disease-free survival [DFS], and overall survival [OS]) associated with obesity.[16–21] In contrast, high BMI has been associated with significantly lower CR and OS rates in younger patients (age <65 years)in the German Study Alliance Leukemia prospective trials.[22] This has been attributed, at least in part, to dose capping of chemotherapy to a maximum body surface area of 2 m² and consequent under-treatment, suggesting a potentially important therapy dosing effect.[23] Importantly, in acute promyelocytic leukemia, where obesity is more prevalent,[24, 25] a higher complication and relapse rate has also been observed with obesity,[26–28] confirmed in a retrospective Cancer and Leukemia Group B analysis.[29]

A key epidemiologic question is whether obesity is associated with a unique leukemic clinical or genetic phenotype. Obese patients were found to be more likely to have intermediate risk cytogenetics[20, 22] while a Texas leukemia cohort[7] suggested an association of obesity with recurrent cytogenetics abnormalities.[30] It is not known whether obesity is associated with a specific pattern of somatic mutations or with other genetic lesions, although a possible association with FLT3 mutation was recently suggested in a recent retrospective analysis of The Cancer Genome Atlas cohort.[31]

To better characterize the association of obesity with AML clinical, cytogenetic and genetic phenotype, and with treatment and outcome, we performed an analysis of BMI and its impact in adult AML patients receiving trial-specified therapy on 2 large and mature ECOG-ACRIN Cancer Research Group intensive treatment protocols, including both younger (E1900, NCT00049517)[32] and older (E3999, NCT00046930)[33] adults.

Patients and Methods

In this database analysis, we examined the impact of obesity at AML diagnosis (defined as BMI ≥30 kg/m², overweight as BMI ≥25 kg/m² and morbid obesity as BMI ≥40 kg/m²measured at study registration) on clinical outcomes (including induction CR, early death, OS, and DFS) and induction daunorubicin dosing in our frontline ECOG-ACRIN newly-diagnosed adult AML randomized phase III intensive therapy studies E1900 and E3999. E1900 (patients aged <60 years) was a 1-to-1 randomized study of standard dose infusional cytarabine with either conventional dose (45 mg/m²) or high-dose daunorubicin (90 mg/m²) during the first induction cycle. From December 2002 until November 2008, 657 patients were enrolled.[32] The E3999 trial (patients aged ≥60 years) was a randomized, placebo-controlled study of standard dose infusional cytarabine and conventional dose daunorubicin (45 mg/m²/day) plus or minus the p-glycoprotein modulator, zosuquidar. E3999 was active from July 2002 until September 2005, with 449 patients enrolled.[33] Consolidation therapy in each protocol comprised 2 planned cycles of age-adapted high-dose cytarabine; the E1900 study then included a second randomization to a single-dose of gemtuzumab (vs no gemtuzumab) prior to subsequent consolidation with autologous hematopoietic cell transplantation.[34] Eligible high-risk patients received consolidation with allogeneic transplantation.

Induction chemotherapy dosing and dose guidelines were mandated and controlled in each protocol. Because of the potential concern for excess cardiotoxicity with high-dose daunorubicin, a dose cap at body surface area 2.5 m² (i.e. maximum total daunorubicin total dose of 225 mg) was employed in the E1900 study consistent with prior studies.

Patient baseline characteristics were compared using Fisher’s exact test for categorical variables and Wilcoxon rank sum tests for continuous variables. Early death was defined as death within 28 days of starting induction therapy. One patient lost to follow-up within 28 days was coded as missing or early death. OS was defined as time from randomization to death from any cause. For patients who achieved CR after induction, DFS was defined as the time from documented CR to relapse or death from any cause. Logistic regression models were used to examine the association between obesity and CR or early death. The univariate models included obesity indicator (BMI <30 kg/m² vs ≥/= 30 kg/m²) and induction treatment arms. Multivariate models included obesity indicator, induction treatment, age, race, sex, cytogenetic risk, ECOG performance status (PS), white blood cell count, platelets, marrow blast percentage, hemoglobin, peripheral blood blast percentage, secondary AML (sAML, including therapy-related or arising from an antecedent hematologic disorder), and whether the patients were initially administered at least 90% of intended daunorubicin dose during cycle 1. Kaplan-Meier method was used to estimate event-time distributions for OS and DFS. Univariate and multivariate Cox proportional hazards models, stratified on induction treatment arms, were performed on OS and DFS. The interactions of treatment by obesity were tested using logistic and Cox proportional hazards models, including the induction treatment arms, obesity indicator, and interaction term as covariates.

In addition, genotyping was performed to evaluate for somatic alterations of interest in a subset of the patients in the E1900 cohort using a selected 18-gene myeloid targeted gene sequence panel (TET2, ASXL1, DNMT3A, CEBPA, PHF6, WT1, TP53, EZH2, RUNX1, PTEN, FLT3, NPM1, HRAS, KRAS, NRAS, KIT, IDH1, and IDH2), as previously published.[35] An exploratory analysis of an association of somatic mutations with BMI the Fisher’s exact test was used. All p-values were based on 2-sided tests, and they were not adjusted by multiple comparisons. No mutation panel was captured in E3999.

The cytogenetic risk group was determined and assigned centrally in all cases according to standard Southwest Oncology Group and ECOG-ACRIN criteria,[36] and classified as favorable, intermediate (including normal karyotype), and adverse risk. Remaining patients were assigned to an indeterminate risk group, including some cases with 2 cytogenetic abnormalities, if they could not be classified into an established risk group.

Results

BMI data were available for 1,088 patients, after 18 patients were excluded from the analysis due to missing or incomplete BMI data at registration (E1900, n=7; E3999, n=11). Median follow-up on patients alive at the time of this analysis was 6.0 years.

Obesity was prevalent (n=358, 33.0%), and balanced across studies and induction randomization arms. The median BMI for the entire study cohort was 27.1 kg/m² (range, 16.7–93.0); for non-obese patients, the median BMI was 25.0 kg/m² (range, 16.7–30.0), while for the obese patients, the median BMI was 34.0 kg/m² (range, 30.0–93.0). Of the patients with BMI > 30 kg/m², 202 (19%) patients had a BMI between 30–34.9 kg/m², 94 (9%) with BMI 35–39.9 kg/m² and 62 (5.7%) with BMI greater than or equal to 40 kg/m². The median BMI for E1900 was 26.7 and 27.8 in E3999 (p=0.02). 324 patients on E1900 received 90mg/m2, with 93 (28.7%) being obese. Of those, 12 (12.9%) received < 90% intended dose. Among the 326 patients on E1900 that received 45mg/m2, 117 (35.9%) were obese. Of those, 13 (11.1%) received < 90% intended dose. Among the 438 patients on E3999, 148 (33%) were obese, and of those, 6 (4%) received less than 90% of the intended daunorubicin 45mg/m2 dose.

Clinical Characteristics

Baseline clinical characteristics for the obese and non-obese patients are included in Table 1. Obese patients were more likely to have intermediate risk cytogenetics (p=0.01) and have worse ECOG performance status (PS) (p=0.01). Obese patients had a higher median age and greater proportion female sex, although not statistically significant (p=0.06 and p=0.07, respectively). There was no difference in the incidence of secondary AML (sAML) or in presenting white blood cell count or bone marrow blast percentage.

Table 1.

Baseline Characteristics by Non-Obese Vs Obese

Characteristic	BMI <30 kg/m2(n=730)		BMI ≥30 kg/2(n=358)		p-value
	Median and Range		Median and Range
Age, years	56 (17–84)		59 (18–93)		0.06
WBC Count/mm3 (x1000)	8.6 (0.4–366.0)		9.1 (0.5–212.8)		0.79
Hemoglobin (g/dl)	9.2 (4.6–30.0)		9.0 (5.1–15.4)		0.17
Peripheral Blasts, %	22.5 (0.0–99.0)		25.5 (0.0–98.0)		0.65
Platelet Count/mm3, (x1000)	60 (1.4–995.0)		49.0 (0.7–975.0)		0.10
Bone Marrow Blasts, %	60.0 (0.0–100.0)		60 (0.0–100.0)		0.90
Characteristic	No.	%	No	%	p-value
Sex					0.07
Male	404	55.3	176	49.2
Female	326	44.7	182	50.8
Race					0.14
Missing/unknown	15	2.1	9	2.5
White	664	91.0	328	91.6
Black	38	5.2	20	5.6
Asian	11	1.5	0	0.0
Other	2	0.3	1	0.3
ECOG PS					0.01
Missing	5	0.7	1	0.3
0	305	41.8	127	35.5
1	348	47.7	174	48.6
2	56	7.7	38	10.6
3	15	2.1	18	5.0
4	1	0.1	0	0.0
Cytogenetic risk group					0.008
Missing	11	1.5	7	2.0
Favorable	65	8.9	37	10.3
Indeterminate	217	29.7	72	20.1
Intermediate	263	36.0	150	41.9
Unfavorable	174	23.8	92	25.7
Secondary AML					0.86
No	625	85.6	305	85.2
Yes	105	14.4	53	14.8

Abbreviations: AML, acute myeloid leukemia; BMI, body mass index; ECOG PS, Eastern Cooperative Oncology Group performance status; WBC, white blood count;

We performed a multivariate logistic regression to examine the association between obesity and clinical and disease characteristics, including age, race, sex, cytogenetic risk group, ECOG PS, sAML, complete blood cell count parameters, and bone marrow blast percentage. Only older age (p=0.02) and lower platelet counts (p=0.03) remained significantly associated with obesity.

Clinical Outcome

There was no interaction of the effects on the clinical outcome between obesity and the 4 induction treatment arms in the E1900 and E3999 trials. Importantly, obesity was not associated with lower CR or with early death or induction mortality. Similarly, there was no significant association of obesity with DFS or OS. The median OS for non-obese patients was 1.19 years, 95% (1.046, 1.31) compared to obese patients 1.05 years, 95% (0.914, 1.27). Median DFS for non-obese patients was 1.09 years, 95% (0.997, 1.27) versus obese patients 1.05 years, 95% (0.849, 1.28). (Table 2, Figures 1A and 1B). As expected, factors significantly associated with OS in a multivariate Cox proportional hazards model included ECOG PS 2 or 3, cytogenetic risk group, age, female sex, and peripheral blast percentage and platelet counts (Table 3).

Table 2.

Logistic and Cox Model Fitting Results for BMI (Obese vs Non-Obese)

Outcome	Univariate Model		Multivariate Model
	p-value		p-value
CR	0.74		0.74
Early death	0.60		0.45
	HR (95% CI)	p-value	HR (95% CI)	p-value
OS	1.13 (0.98, 1.30)	0.09	0.99 (0.85,1.16)	0.90
DFS (for CR patients only)	1.19 (0.98, 1.44)	0.08	1.07 (0.86, 1.32)	0.55

Abbreviations: BMI, body mass index; CR, complete remission; DFS, disease-free survival; HR, hazard ratio; OS, overall survival.

Figure 1A. Kaplan–Meier estimates of overall survival.

BMI indicates body mass index;

Figure 1B. Kaplan–Meier estimates of disease-free survival.

BMI indicates body mass index; CNSR: censored

Table 3.

Multivariate Cox Model Results for Overall Survival

Parameter	HR	95% CI	p-value
BMI
BMI ≥30 kg/m2 vs BMI <30 kg/m2	0.990	0.85–1.16	0.90
ECOG PS
1 vs 0	1.13	0.96–1.33	0.15
2 vs 0	1.70	1.29–2.22	<0.001
3 vs 0	2.63	1.71–4.05	<0.001
Cytogenetics
Favorable vs unfavorable	0.26	0.19–0.36	<0.001
Indeterminate vs unfavorable	0.50	0.41–0.61	<0.001
Intermediate vs unfavorable	0.42	0.34–0.50	<0.001
Secondary AML
Yes vs no	1.24	1.01–1.53	0.05
Sex
Female vs male	0.85	0.73–0.99	0.04
Race
Black vs white	1.25	0.92–1.68	0.15
Asian vs white	0.86	0.40–1.83	0.69
Other vs white	1.09	0.15–7.84	0.94
Agea	1.02	1.01–1.03	<0.001
Peripheral WBCa	1.00	0.999–1.003	0.32
Hemoglobina	1.00	0.95–1.04	0.89
Peripheral blastsa	1.01	1.00–1.01	0.002
Peripheral platelet counta	1.00	0.998–1.00	0.04
Marrow blasta	1.00	0.99–1.001	0.15
Initially administrated ≥ 90% of intended daunorubicin dose during cycle 1
No. vs. Yes	1.39	0.90–2.13	0.14

Abbreviations: AML, acute myeloid leukemia; BMI, body mass index; ECOG PS, Eastern Cooperative Oncology Group performance status; WBC, white blood cell.

^aIncluded into the model as a continuous variable

We further looked at the impact on outcome of obesity in subgroups of interest (specifically age, sex, FLT3-ITD status, and cytogenetic risk group). In the univariate analysis (adjusted by study and treatment effect), obesity was not statistically significant in any subgroup, except in female patients. In univariate model, obese female patients tend to have worse survival compared to non-obese (hazard ratio (HR), 1.26 [95% Confidence Intervals (CI) 1.02–1.55], p=0.03). However, after adjusting for other risk factors, the differences were no longer-significant (HR 1.09 [95% CI 0.86–1.39], p=0.48).

No discernible difference in therapy-related toxicity was observed in obese versus non-obese patients on review of reported grade 3 to 5 hematologic and non-hematologic toxicities, including cardiac toxicities, reported for patients from the 2 studies. (supplement table 1 and 2)

Obesity and Intended Daunorubicin Dose

Daunorubicin dosing was mandated per protocol,[32, 33] and included a dose cap of 2.5 m² in the E1900 study, and the actual dose administered was collected in the protocol case report forms. Only 3.9% (25 patients) were dose capped because of BSA > 2.5 m^2. We evaluated the proportion of patients who received at least 90% of the intended daunorubicin dose to determine the effect of obesity on actual dosing. We found that, despite pre-specified protocol dosing guidelines, obese patients were significantly more likely to receive less than 90% of the planned relative dose intensity for daunorubicin in cycle 1; in E1900, 25 obese patients (11.9%) received less than 90% of the intended daunorubicin dose (11.1% obese vs 0% non-obese in the 45 mg/m² arm [p<0.001] and 12.9% obese vs 2.6% non-obese in the 90 mg/m² arm [p=0.001]), while in E3999, 6 obese patients (4.1%) received less than 90% of the intended daunorubicin dose (vs 1.0% for non-obese patients; p=0.07). Of the patients who received < 90% of the intended dose, the lowest dose reduction was 46%. No differences in OS were observed between patients who received at least 90% of intended daunorubicin dose vs. <90% intended dose (Figure 2A). When adjusted for other risk factors in the multivariate Cox proportional hazards model, administration of less than 90% of intended daunorubicin dose was not associated with significantly inferior OS (HR 1.39 [95% CI, 0.90–2.13], p=0.14) (Table 3). Restricting OS analysis to obese patients alone, no difference was observed for those receiving at least 90% of intended daunorubicin dose versus those who did not (univariate model HR 1.09 [95% CI 0.70–1.69], p=0.71; multivariate model HR 0.86 [95% CI 0.50–1.48], p=0.59); similar results were observed among non-obese patients (univariate model HR 0.73 [95% CI 0.34–1.54], p=0.41; multivariate model HR 0.55 [95% CI 0.25–1.19], p=0.13) (Figure 2B).

Figure 2A. Overall Survival According to Administered Daunorubicin Dose.

Figure 2B. Overall Survival According to Daunorubicin Dose in Obese and Non-Obese Patients.

DNC: Daunorubicin; CNSR: censored

Obesity and Genetic Alterations

Apart from a borderline lower incidence of MLL gene partial tandem duplication (MLL-PTD) (1.7% obese vs 3.6% non-obese; p=0.06), we did not find an association of somatic mutations with obesity, or with mutations involved in specific pathways of interest, which included aberrant signaling (FLT3, NRAS, or KIT; p=.92), DNA methylation (IDH1, IDH2, TET2, and DNMT3A; p=0.18), or histone methylation (ASXL1; p=0.10) (Table 4). When the analysis was confined to patients with intermediate risk cytogenetics, there was similarly no association of obesity with individual somatic mutations tested, apart from a non-significant increase in the frequency of ASXL1 mutations (7.4% obese vs 2.1% non-obese; p=0.08).

Table 4.

Somatic Mutations and Obesity (E1900 Subset Only)

Somatic Mutations	BMI <30 kg/m2 (n=441)		BMI ≥30 kg/m2 (n=210)		Total (N=651)		p-value
	N	%	N	%	N	%
FLT3							0.84
Missing	112	25.4	38	18.1	150	23.0
MUT	109	24.7	59	28.1	168	25.8
WT	220	49.9	113	53.8	333	51.2
MLL-PTD							0.06
Missing	182	41.3	72	34.3	254	39.0
MUT	18	4.1	3	1.4	21	3.2
WT	241	54.6	135	64.3	376	57.8
NPM1							0.24
Missing	112	25.4	38	18.1	150	23.0
MUT	79	17.9	50	23.8	129	19.8
WT	250	56.7	122	58.1	372	57.1
DNMT3A							0.23
Missing	123	27.9	39	18.6	162	24.9
MUT	83	18.8	36	17.1	119	18.3
WT	235	53.3	135	64.3	370	56.8
NRAS							0.40
Missing	183	41.5	72	34.3	255	39.2
MUT	30	6.8	12	5.7	42	6.5
WT	228	51.7	126	60.0	354	54.4
CEBPa							0.58
Missing	183	41.5	75	35.7	258	39.6
MUT	22	5.0	14	6.7	36	5.5
WT	236	53.5	121	57.6	357	54.8
TET2							0.87
Missing	116	26.3	39	18.6	155	23.8
MUT	27	6.1	15	7.1	42	6.5
WT	298	67.6	156	74.3	454	69.7
WT1							0.70
Missing	187	42.4	73	34.8	260	39.9
MUT	22	5.0	10	4.8	32	4.9
WT	232	52.6	127	60.5	359	55.1
IDH2							0.12
Missing	112	25.4	39	18.6	151	23.2
MUT	38	8.6	12	5.7	50	7.7
WT	291	66.0	159	75.7	450	69.1
IDH1							0.72
Missing	112	25.4	39	18.6	151	23.2
MUT	25	5.7	11	5.2	36	5.5
WT	304	68.9	160	76.2	464	71.3
KIT							>0.99
Missing	182	41.3	72	34.3	254	39.0
MUT	16	3.6	9	4.3	25	3.8
WT	243	55.1	129	61.4	372	57.1
RUNX1							0.82
Missing	193	43.8	76	36.2	269	41.3
MUT	15	3.4	7	3.3	22	3.4
WT	233	52.8	127	60.5	360	55.3
PHF6							0.11
Missing	183	41.5	74	35.2	257	39.5
MUT	10	2.3	1	0.5	11	1.7
WT	248	56.2	135	64.3	383	58.8
KRAS							0.75
Missing	184	41.7	72	34.3	256	39.3
MUT	8	1.8	3	1.4	11	1.7
WT	249	56.5	135	64.3	384	59.0
PTEN							0.72
Missing	182	41.3	73	34.8	255	39.2
MUT	6	1.4	2	1.0	8	1.2
WT	253	57.4	135	64.3	388	59.6
TP53							>0.99
Missing	187	42.4	75	35.7	262	40.2
MUT	7	1.6	3	1.4	10	1.5
WT	247	56.0	132	62.9	379	58.2
Signaling							0.92
Missing	168	38.1	67	31.9	235	36.1
All FLT3 and RAS and KIT WT	124	28.1	66	31.4	190	29.2
Any FLT3 or RAS or KIT mutated	149	33.8	77	36.7	226	34.7
DNA methylation							0.18
Missing	120	27.2	41	19.5	161	24.7
All IDH1/2 and TET2 and DNMT3A WT	182	41.3	107	51.0	289	44.4
Any IDH1/2 or TET2 or DNMT3A mutated	139	31.5	62	29.5	201	30.9
Histone methylation							0.10
Missing	114	25.9	38	18.1	152	23.3
ASXL1 MUT	10	2.3	11	5.2	21	3.2
ASXL1 WT	317	71.9	161	76.7	478	73.4
Any mutation							0.78
Missing	143	32.4	56	26.7	199	30.6
All genes listed above WT	43	9.8	24	11.4	67	10.3
Any mutation in genes listed above	255	57.8	130	61.9	385	59.1

Abbreviations: MUT, mutation; PTD, partial tandem duplications; WT, wild type.

Discussion

Obesity has reached pandemic levels, now affecting 12% of the global adult population 42% of US population,[21] and is a major contributor to mortality, including increased cancer deaths.[37] Obesity represents a common risk factor for leukemia development, and in this analysis, was both prevalent (33%) in AML patients and associated with some unique clinical and leukemia phenotypic features (i.e., older, female, worse ECOG PS score, and differences in cytogenetics risk group). In this study, obesity did not impact DFS and OS after intensive therapy. However, some clinical features associated with obesity (specifically age and PS) were independently associated with worse survival, while female sex (also associated with obesity) was conversely associated with better survival on multivariate analysis[38].

Interestingly, obesity did influence the administered daunorubicin dosing, despite its use in a controlled clinical trial setting. Obese patients were more likely to receive less than 90% of the intended daunorubicin dose with the first cycle of induction therapy. Despite protocol dosing guidelines, particularly in E1900 (which explored high dose daunorubicin),[32]obese patients were still under dosed due to lack of physician and/or patient comfort with administration of full dose. At the time this study was done, case report forms did not capture reasons for dose reduction, therefore reasons for dose reduction are based on speculation. However, under-dosing did not negatively affect OS in our study, therefore serves as a foundation for education regarding medical decision making based on weight and BMI. As the prevalence of obesity rises in the population, this will become a more important topic, and consensus regarding dosing is necessary to optimize patient outcomes. While some prior studies have suggested an adverse impact of dose capping on outcome, especially in more favorable risk AML,[22, 23] the degree of variation in daunorubicin dosing in our study was not clinically significant in this population including older adults. Importantly, we did not observe differences in treatment toxicity in obese versus non-obese patients. While the updated 2021 American Society of Clinical Oncology Guidelines on dosing for obese patients with cancer recommends full, weight based cytotoxic chemotherapy dosing, review of the AML specific literature includes five studies showing no outcome differences in patients that were dose reduced or dose capped, and only one study demonstrating worse OS in obese patients with dose reductions.[39, 40] [41]Therefore, our study is one of the largest with over 1000 patients and shows no impact on outcome when patients are dosed with less than 90% daunorubicin compared to those who received more than 90%. Compared to other cancer patients, obese AML patients are unique and should be treated with leukemia specific guidelines instead of general cancer guidelines.

A particular strength of this analysis is that it capitalized on robust ECOG-ACRIN data from large and uniformly-treated prospective studies, including both younger and older patients, where therapy and outcomes were strictly defined and controlled. This makes the observations of the significant impact of obesity on daunorubicin dosing all the more remarkable.

We utilized BMI as a marker of obesity, consistent with WHO criteria, but acknowledge that this may not reflect true body habitus, as BMI includes both body fat and muscle. An assessment of visceral fat and ectopic fat - to more accurately identify obesity (rather than increased BMI)[42] - and of muscle mass was not possible in this cohort, as patients did not undergo standardized testing or imaging to assess the degree of adiposity.[43, 44]

It is not surprising that obesity is associated with older age, female sex, and inferior ECOG PS. While we did observe differences in cytogenetic risk group, this difference disappeared in the multivariate analysis, and we did not identify a distinct cytogenetic lesion or group associated with obesity. This is in apparent contrast to the retrospective Mayo Clinic AML epidemiology study, where obesity was associated with intermediate risk abnormal (non-diploid) cytogenetics,[20] and the Texas case-control study where it was associated with recurrent cytogenetic abnormalities.[7] The latter included acute promyelocytic leukemia patients with t(15;17), among whom obesity is highly prevalent,[24, 25, 29] but those patients were excluded from the E1900 and E3999 studies, and therefore, are not included in this analysis. There are notable differences in the patient population and treatment strategies between these reports, nevertheless, the current analysis does not suggest that obesity is clearly associated with a distinct pattern of cytogenetic lesions.

We did not find a clear-cut association of obesity with specific somatic mutations, and instead observed a similar frequency of common myeloid mutations among obese patients,.[45] Similarly, we did not confirm an association of obesity with FLT3-ITD mutation in AML (observed esp. in APL), although in contrast to that report - and as noted already - APL was excluded from the E1999 and E1900 studies and therefore not included in this analysis.[31] Additionally, this exploratory analysis was based on a somewhat limited 18-gene panel that did not include some common lesions, such as spliceosome mutations,[35] and was confined to younger E1900 patients. It is not known therefore whether this is true also for older patients, who appear to have a different disease biology.[46] Increased BMI in adults of European origin is also associated with aberrant methylation, with perturbation of hypoxia inducible transcription factor pathways in blood cells, which may also be relevant.[47] AML has been associated with unique aberrant DNA methylation clusters,[48] and obesity was associated with a unique methylation cluster on a multivariate analysis in a Mayo Clinic AML epidemiology study.[49] However, this effect was not comprehensively studied in this cohort, and was meant to be hypothesis generating. Such an analysis requires a prospective epidemiologic study, which is embedded in the recently completed ECOG-ACRIN E2906 Intergroup AML trial, for older adults (≥60 years), from which these results are awaited. Other potential limitations include combining two distinct AML populations, both in terms of age and treatment received, the retrospective nature of the study, and the lack of mutation data for the E3999 patients. These could potentially affect the results and merit further discussion

Obesity is an expanding pandemic and global medical concern, contributes to leukemia development, and is associated with specific AML clinical characteristics. While obesity did not play a role in DFS and OS, the association of obesity with administration of lower than intended daunorubicin dose - and thereby with protocol adherence - highlights the need for a better understanding of patient-related factors and provider perceptions that influence treatment decisions. Leukemia specific guidelines regarding dose adjustments are needed as our study further supports previous literature showing no association between dose adjustments giving less than intended doses and clinical outcomes. This is especially important for dosing in clinical trials. Future studies must concentrate on a better understanding of epidemiologic exposures and genetic and genomic lesions associated with obesity contributing to leukemogenesis.