Psychiatric and substance use disorders predict poor outcomes in older veterans with acute myeloid leukemia

TO THE EDITOR:

The combination of venetoclax with either a DNA methyltransferase inhibitor or low-dose cytarabine has become the standard frontline treatment for older or unfit adults with acute myeloid leukemia (AML).^1-3 However, real-world studies consistently demonstrate inferior outcomes compared with those reported in the landmark phase 3 trials.^4,5 The prevalence and impact of psychiatric and substance use disorders (SUDs) and the role they play in driving worse outcomes in real-world relative to trial populations are understudied.

Veterans with AML represent a frail group of older adults with unique wartime exposures who would likely be excluded from many clinical trials.^6,7 In veterans treated for multiple myeloma, psychiatric disorders and SUDs emerged as a lethal pattern of comorbidities coexisting together.⁸ Thus, we aimed to measure the prevalence of psychiatric disorders and SUDs in veterans with AML, determining the impact of these conditions on treatment response and outcomes.

We conducted a retrospective cohort study using the Veterans Affairs (VA) Corporate Data Warehouse, which captures electronic health record (EHR) information from VA facilities throughout the United States.⁹ We selected veterans with newly diagnosed AML who received one of the Federal Drug Association–approved venetoclax-based therapies in the frontline setting between 2018 and 1 April 2022. To limit our population to veterans who were consistently using the VA, at least 1 non-AML diagnosis or procedure code in each of the 3 years preceding treatment initiation was required. Using the algorithm of the Centers for Medicare and Medicaid’s Chronic Conditions Corporate Data Warehouse,¹⁰ we extracted claims representing preexisting psychiatric disorders and/or SUDs within a 3-year period leading up to AML treatment initiation. Additional details on cohort selection can be found in the supplemental Methods.

We identified 452 veterans (Table 1); 98.5% were male with a median age of 74 years at AML diagnosis (interquartile range, 71.2-78.5), and 72.8% were deemed frail by the VA frailty index.^7,11,12 By the European LeukemiaNet 2017 risk stratification, 61% presented with adverse-risk disease. Secondary AML accounted for 41.8% of all new cases, with 35.8% evolving from a previous myeloid disorder. In our cohort, 46% had at least 1 psychiatric diagnosis, and SUDs were reported in 19% (supplemental Results, Table 1). Psychiatric disorders were significantly more prevalent in younger veterans (68% in <65 years vs 50% in 65-74 years vs 39% in ≥75 years; P < .003); the same was true for SUDs (38% in <65 years vs 26% in 65-74 years vs 9% in ≥75 years; P < .001).

In the entire cohort, 57% achieved either complete remission (CR) or CR with incomplete hematologic recovery (CRi), whereas 14% had progressive disease. Only 38% of veterans with SUDs achieved CR/CRi. Within 60 days of treatment initiation, 19% of veterans (n = 87) died. Early mortality rate was the lowest in those with neither psychiatric disorders or SUDs and the highest in veterans with SUDs (13% vs 35%; P = .006). When comparing veterans with vs without comorbid psychiatric diagnoses, those with psychiatric disorders had ∼40% lower odds of achieving remission (odds ratio [OR], 0.6; 95% confidence interval [CI], 0.36-0.98), had a 64% higher likelihood of dying within 60 days of initiating AML treatment (OR, 1.64; 95% CI, 0.95-2.88), and had a 22% higher hazard of death (hazard ratio [HR], 1.22; 95% CI, 0.95-1.58). Veterans with SUDs had nearly 50% lower likelihood of achieving remission after venetoclax (OR, 0.53; 95% CI, 0.27-1.0) and a 40% higher hazard of overall mortality (HR, 1.4; 95% CI, 1.0-1.99) than those without a preexisting SUD. Although there were no significant differences in intensive care unit admission rates after induction between veterans with and without psychiatric disorders (2.8 vs 2.5 per 5 person-years; risk ratio, 1.1; 95% CI, 0.8-1.52), the admission rates were significantly higher among veterans with vs without SUDs (5.0 vs 2.2 per 5 person-years; risk ratio, 2.26; 95% CI, 1.56-3.2). At the time of study analysis, 69% of patients (n = 310) had died, and the median overall survival (OS) for the entire cohort was 216 days. In our cohort of veterans, the number at risk of death after 1 year was low. Kaplan-Meier analyses (Figure 1) demonstrated inferior 1-year OS in veterans with psychiatric disorders and SUDs compared with those without these comorbidities (log-rank P = .018).

We are, to our knowledge, the first to report the prevalence of psychiatric comorbidities in veterans with AML and the first to describe their co-occurrence with SUDs and how these affect outcomes, specifically in those treated with venetoclax combinations. Our identification of psychiatric disorders and SUDs as important predictors of poor outcomes even after adjusting for age, AML-specific risk factors, and frailty reinforces previous examination of patients with AML that found an association of anxiety and depression with worse OS after intensive chemotherapy induction.¹³ Although cigarette smoking has been associated with inferior survival in patients with AML,^14,15 we identified novel associations between negative outcomes and other substances independent of tobacco use.

A limitation of our study is that administrative claim data may underestimate the prevalence of certain health conditions,¹⁶ but the nationally integrated VA EHR is constantly updated, making measurement error less of a concern than in private health systems with less connected EHRs. Our cohort was largely male, and additional studies with more females are needed to explore how gender may affect the outcomes of patients with AML with psychiatric disorders or SUDs. Finally, residual confounding is difficult to exclude despite rigorous multivariable adjustment for potential covariates, a known limitation of EHR- and claims-based observational studies. However, we bring our attention to the outcomes of a group of patients who are often excluded or underrepresented in randomized clinical trials to bridge existing gaps between expectations and reality.

We hypothesize several potential mechanisms that may explain the inferior outcomes of patients with AML with psychiatric disorders or SUDs. First, patients may experience re-emergence or acute worsening of their preexisting psychological comorbidities with a new cancer diagnosis.¹⁷ Patients with psychiatric disorders or SUDs may engage in detrimental lifestyle behaviors that increase the risk of nonadherence with oral chemotherapy or non-AML mortality.¹⁸ Second, individuals with psychiatric disorders and SUDs may be at greater risk of not receiving guideline-concordant care. In our study, it is worth noting that despite 57% achieving remission, just 3% received allogeneic hematopoietic stem-cell transplantation (HSCT), which is the only potentially curative therapy for AML in the setting of venetoclax-based therapy. None of the veterans with preexisting SUDs received transplantation. Although the determination of eligibility for HSCT involves a very comprehensive evaluation, the low rate of HSCT among veterans with AML is worth investigating in the future. Third, treatment response and mortality may be attributable to biological mechanisms underlying the complex interplay between mental illnesses and AML. Nontobacco substance use has been implicated in the genesis and acceleration of several cancers potentially through inflammation¹⁹ or premature biological aging through telomere length reduction.²⁰ Interestingly, patients with SUDs were younger at diagnosis, and despite lower chronologic age, they presented with disease features such as adverse cytogenetics and TP53, which are often associated with older age (supplemental Results, Table 2). Examination of the mechanisms that mediate substance use, inflammation, and cellular aging in the context of AML are necessary.

Overall, our data suggest that psychiatric disorders and SUDs (conditions underrepresented in clinical trials²¹ and thus underappreciated in the safety of novel AML treatments) represent vulnerabilities that confer worse outcomes in real-world populations. Prospective studies are needed to explore the means by which psychiatric disorders and SUDs mediate the higher risk observed in this study.

Acknowledgments: The authors thank the Department of Veterans Affairs and the Massachusetts Veterans Epidemiology Research and Information Center that made it possible for this work to be conducted.

This work was supported by the VA Office of Research and Development, Cooperative Studies Program; American Society of Hematology (Scholar Award number 246246), American Heart Association (grant/award number 857078), and VA Career Development Award (grant/award number IK2CX002218).

Contribution: M.H.L., J.L., and C.V.E. conceptualized and designed the study; M.H.L., J.L., N.V.D., N.R.F., and C.D. provided administrative support; M.H.L., J.L., N.V.D., M.T.B., N.R.F., and C.D. collected and assembled the data; M.H.L., J.L., G.S.H., and C.D. performed data analysis and interpretation; M.H.L., G.S.H., and C.D. wrote the manuscript; and all authors completed final editing and approved the manuscript.

Conflict-of-interest disclosure: M.H.L. served on the advisory board of MorphoSys and reports consulting fees from PharmaEssentia. J.L. reports research funding from Merck. G.S.H. received grant support from Incyte; served on the advisory board of or received consulting fees from PharmaEssentia, AbbVie, GlaxoSmithKline, Pfizer, Novartis, MorphoSys, Cogent, Sobi, Merck, and Bristol Myers Squibb; and has stock/stock options in Regeneron Pharmaceuticals. N.R.F. reports research funding from Merck. The remaining authors declare no competing financial interests.

Correspondence: Michelle H. Lee, Department of Medical Oncology, Mass General Cancer Center, Massachusetts General Hospital, 55 Fruit St, Boston, MA 02114; email: mlee37@mgb.org.