Routine Laboratory Values Can Predict Therapy-Related Myeloid Neoplasms in Patients with New Cytopenias after Treatment for Breast Cancer

Background

Therapy-related myelodysplastic syndrome (t-MDS) and therapy-related acute myeloid leukemia (t-AML) occur in 0.6% to 1.8% women treated for breast cancer with chemotherapy and radiation. Cytopenias in breast cancer patients can be secondary to breast cancer marrow infiltration, chemotherapy effect or t-MDS/t-AML, and are usually evaluated with a bone marrow biopsy (BMBx).

Objectives

Identify predictors of t-MDS/t-AML in patients with breast cancer and cytopenias to avoid unnecessary BMBx.

Methods

We used the Memorial Sloan Kettering Cancer Center (MSKCC) institutional database to identify patients with breast cancer and cytopenias who were referred for BMBx between 2002-2018. Characteristics associated with the risk of t-MDS/t-AML in patients with breast cancer and cytopenias were evaluated by multivariate logistic regression and included in a predictive model. The average area under the receiver operating characteristic curve (AUC) of the final predictive model was estimated by 10-fold cross-validation. The same analysis was conducted on breast cancer patients without bone metastasis (mets), and univariate analysis was performed on a smaller subset of women with bone mets.

Results

We identified 240 breast cancer patients who underwent BMBx to evaluate cytopenias. 110 patients were diagnosed with t-MDS/t-AML. By multivariate analysis, 5 variables had a p-value <0.05 (Tab.1). An ANC >1.5 K/mcL, a Hgb level >12.2 g/dL and the presence of known bone mets significantly decreased the likelihood of t-MDS/t-AML (OR: 0.22; 0.23; 0.17). A time from breast cancer diagnosis to BMBx of >15 months was significantly associated with an increased probability of t-MDS/t-AML (OR: 9.71). Race significantly affected the AUC but not the final model. The average AUC was 0.81 with and 0.79 without race.

Then we stratified by presence of bone mets: 185 patients didn’t have bone mets and 100 were diagnosed with t-MDS/t-AML. By multivariate analysis, our 5 variables had similar effects on the risk of having t-MDS/t-AML to those detected in the whole cohort (OR: ANC 0.26; Hgb 0.25; time from diagnosis to BMBx of >15 months 8.75). The average AUC with these 5 predictors was 0.78 with and 0.75 without race. Lastly, 55 patients had bone mets and 10 were diagnosed with t-MDS/t-AML. An ANC >1.5 K/mcL was the only variable associated with a decreased risk of t-MDS/t-AML (OR: 0.08).

Conclusion

In breast cancer women with unexplained cytopenias, an ANC >1.5 K/mcL, a Hgb >12.2 g/dL and the presence of known bone mets anticipated an absence of t-MDS/t-AML on BMBx. Conversely, a time from diagnosis to BMBx of >15 months and both black and white race were predictive of a diagnosis of t-MDS/t-AML on BMBx. These findings can assist clinicians to determine the likely underlying etiology of cytopenias in breast cancer patients and determine the need for and timing of a BMBx.