Using a Comorbidity-Based Composite Model to Predict Mortality in Patients With AML

Researchers developed a new composite model incorporating patient- and disease-specific features to predict mortality in patients with acute myeloid leukemia (AML), according to a report published in JAMA Oncology. The novel composite model (termed AML-CM) outperformed conventional prognostic factors such as age and Karnofsky performance score in predicting early and late mortality, and could prove more useful than these factors for designing eligibility criteria for clinical trials of AML, noted the authors, led by Mohamed L. Sorror, MD, from the clinical research division at Fred Hutchinson Cancer Research Center in Seattle, Washington.

“Decisions about the choice of therapy have largely been based on age (e.g., ≥65 years or <65 years) [and] performance status often influences decisions and is frequently used to define the vague notion of ‘unfit for intensive therapy,’” the researchers explained. “Formal evaluation of comorbidities has played a small role in decisions about initial therapy.”

In this retrospective cohort study, the investigators reviewed the electronic medical records of 1,100 patients (median age = 60 years; age range = 20-89 years) treated at five academic centers between January 2008 and December 2012. All patients were newly diagnosed with AML and received initial treatment with either low-, moderate-, or high-intensity regimens during the study period; patients who received only palliative care were excluded from the analysis.

Death within one year of initial therapy was the primary endpoint, given that it included “both early deaths due to regimen-related toxic effects or to lack of response and/or relapse of AML, as well as later deaths following treatment of relapsed or refractory disease.” Death within eight weeks was a secondary endpoint.

During the study period, 679 of the total cohort (62%) died, with 379 of the deaths (65%) occurring in the first year after initial therapy. In multivariate analyses in the training set, nine comorbidities met the predetermined cutoff for association with one-year mortality (hazard ratios [HRs] > 1.2), and constituted “the new AML-Comorbidity Index (AML-CI)”:

  • cardiac dysfunction
  • hepatic dysfunction
  • infection
  • peptic ulcer
  • heart valve disease
  • albumin value <3.5 g/dL
  • thrombocytopenia
  • lactate dehydrogenase (LDH) >200-1,000 U/L
  • LDH >1,000 U/L

Additionally, age and cytogenetic/molecular risk groups were associated with one-year mortality. To incorporate these factors into the AML-CM, the researchers converted adjusted HRs for older age and cytogenetic/molecular risk groups (per the European Leukemia Net classification) to weighted scores. See TABLE 1 for a description of the components in each risk model.

“The strong impact of cytogenetic/molecular risks on mortality is not a surprise,” the authors commented, but “why increasing age continues to have a significantly independent impact on mortality after accounting for comorbidities is unclear.” They suggested that the acquisition of additional adverse molecular AML markers with aging could explain this association.

In the validation set, the authors compared the prognostic ability of the novel AML-CM with the AML-CI, the original hematopoietic cell transplantation comorbidity index (HCT-CI), and an augmented HCT-CI (constructed from the 17 comorbidities in the original HCT-CI plus hypoalbuminemia, thrombocytopenia, and high LDH values – predictive factors that were identified in multivariate analyses).

The prognostic models’ performances were compared using C statistics for continuous outcomes and area under the curve (AUC) for binary outcomes.

As seen in TABLE 2, the augmented HCT-CI performed better than either AML-CI or the original HCT-CI in predicting early and late mortality, “[validating] that comorbidities have a significant impact on early and one-year mortality [in this patient population],” the authors noted. And, when comorbidities, age, and cytogenetic/molecular risks were incorporated into the AML-CM, it performed better than any of the individual risk components alone (C statistic = 0.72 and AUC = 0.76 for 1-year mortality; AUC=0.78 for 8-week mortality).

“Just as using the HCT-CI before allogeneic HCT has had a major impact on the decision to proceed to HCT, we believe use of the AML-CM could inform decisions as to whether patients with newly diagnosed AML should receive more intensive or less intensive therapies for their disease,” the researchers concluded. “This model could prove useful to the U.S. Food and Drug Administration when monitoring clinical trials to ensure adequate representation of high-risk patients in these trials and, hence, generalizability of trial results to the whole AML population.”

The model did not incorporate treatment intensity (though it was shown to predict mortality), because patients received therapies of differing intensities and because “this is the decision that we plan to improve based on the AML-CM scores,” they added. The study also is limited by the retrospective nature of data collection.

The authors report no conflicts.


Sorror ML, Storer BE, Fathi AT, et al. Development and validation of a novel acute myeloid leukemia–composite model to estimate risks of mortality. JAMA Oncol. 2017 September 7. [Epub ahead of print]

TABLE 1. Components of the AML-CM and Their Corresponding Scores
Comorbidity Score
Arrhythmia 1
Cardiac dysfunction (coronary artery
disease, congestive heart failure, myocardial infarction, or EF ≤50%)
Inflammatory bowel disease 1
Diabetes 1
Cerebrovascular disease (transient ischemic attack or cerebrovascular accident) 1
Psychiatric disturbance 1
Mild hepatic dysfunction 1
Obesity 1
Infection 1
Rheumatologic comorbidity 2
Peptic ulcer 2
Moderate/severe renal dysfunction 2
Moderate pulmonary comorbidity 2
Prior malignancy 3
Heart valve disease 3
Severe pulmonary comorbidity 3
Moderate/severe hepatic dysfunction 3
The Augmented HCT-CI (all of the above, plus the following)
Hypoalbuminemia <3.5 g/dL 1
Thrombocytopenia 1
LDH >200-1,000 U/L 1
LDH >1,000 U/L 2
The AML-CM (all of the above, plus the following)
Age 50-59 years 1
Age ≥60 years 2
ELN intermediate cytogenetic/molecular risk 1
ELN adverse cytogenetic/molecular risk 2
HCT = hematopoietic cell transplantation; CI = comorbidity index; EF = ejection fraction; AML = acute myeloid leukemia; LDH = lactate dehydrogenase; ELN = European Leukemia Net


TABLE 2. Comparisons of the Performance of Risk Factors and Indices in the Validation Set
Risk factor C-statistic for 1-year mortality True AUC for 1-year mortality True AUC for 8-week mortality
n (SD) n (SD) n (SD)
AML-CI 314 0.596
297 0.606 (0.039) 305 0.659 (0.043)
Original HCT-CI 352 0.649 (0.025) 326 0.674 (0.028) 339 0.684 (0.042)
Augmented HCT-CI 305 0.664 (0.023) 289 0.687 (0.035) 296 0.721
Age (groups) 367 0.640 (0.020) 340 0.682 (0.029) 354 0.640 (0.040)
Cytogenetic/molecular risks (groups) 350 0.614
324 0.654 (0.023) 337 0.597 (0.042)
AML-CM 292 0.719
277 0.758 (0.030) 283 0.776
KPS (groups) 291 0.619
266 0.646 (0.035) 279 0.676 (0.048)
AUC = area under the curve; SD = standard deviation; AML = acute myeloid leukemia; CI = comorbidity index; CM = composite model; HCT-CI = hematopoietic cell transplantation comorbidity index; KPS = Karnofsky performance status