Dose–volume parameters of MRI-based active bone marrow predict hematologic toxicity of chemoradiotherapy for rectal cancer

PURPOSE: Magnetic resonance imaging (MRI) is routinely used for locoregional staging of rectal cancer and offers promise for the prediction of hematologic toxicity. The present study compares the clinical utility of MRI-based active bone marrow (BMact) delineation with that of CT-based bone marrow total (BMtot) delineation for predicting hematologic toxicity. METHODS: A prospective cohort study was performed. Eligible patients had stage II/III rectal cancer and qualified for preoperative chemoradiotherapy. The BMact areas on T1-weighted MRI were contoured. The impact of the dose–volume parameters of BMact/BMtot and clinical data on hematologic toxicity were assessed. Basic endpoints were the occurrence of grade 3/4 hematologic toxicity and peripheral blood parameters reaching a nadir. Linear regression models were generated for the nadirs and receiver operating characteristic (ROC) curves for the occurrence of grade 3/4 hematologic toxicity. RESULTS: Thirty-five patients were enrolled. Women presented higher dose–volume parameters of BMact, BMtot, and lymphocyte nadir (ALCnadir%) than men. Models for the prediction of ALCnadir% (V5-V20BMtot, V5-V30BMact) and platelet nadir (PLTnadir%; V5-V10BMtot, V5-V20BMact) were statistically significant. In the ROC curves, a baseline lymphocyte level of 1.81 × 10(3)/ml was adopted as the cutoff for predicting grade 3/4 lymphopenia, with specificity of 77.8% and sensitivity of 73.1%. The multivariate linear regression model for ALCnadir% had R(2) = 0.53, p = 0.038. In the tenth step of selection, V5BMact (p = 0.002) and gender (p = 0.019) remained. The multivariate linear regression model for PLTnadir% had R(2) = 0.20, p = 0.34. In the sixth step of selection, V15BMact remained (p = 0.026). CONCLUSION: The dose–volume parameters of BMact serve as better predictors of ALCnadir% and PLTnadir% than BMtot.