A Nomogram Built on Clinical Factors and CT Attenuation Scores for Predicting Treatment Response of Acute Myeloid Leukemia Patients.

Abstract

Background: Acute myeloid leukemia (AML) is an aggressive cancer with variable treatment responses. While clinical factors such as age and genetic mutations contribute to prognosis, recent studies suggest that CT attenuation scores may also predict treatment outcomes. This study aims to develop a nomogram combining clinical and CT-based factors to predict treatment response and guide personalized therapy for AML patients. Methods: This retrospective study included 74 newly diagnosed AML patients who underwent unenhanced abdominal CT scans within one week before receiving their first induction chemotherapy. Clinical biomarkers of tumor burden were also collected. Patients were classified into two groups based on treatment response: complete remission (CR; n = 24) and non-complete remission (NCR; n = 50). Multivariable logistic regression was used to identify independent predictors of treatment response. Predictive performance was evaluated using receiver operating characteristic (ROC) curves, and model consistency was assessed through calibration and decision curve analysis (DCA). Results: Significant differences in hemoglobin (Hb), platelets (Plt), and CT attenuation scores were observed between the CR and NCR groups (all p < 0.05). Multivariable logistic regression identified Hb, Plt, and CT attenuation scores as independent predictors of treatment response. A nomogram incorporating these factors demonstrated excellent predictive performance, with an area under the curve (AUC) of 0.912 (95% CI: 0.842-0.983), accuracy of 0.865 (95% CI: 0.765-0.933), sensitivity of 0.880 (95% CI: 0.790-0.970), and specificity of 0.833 (95% CI: 0.684-0.982). The CR nomogram displayed significant clinical value and excellent goodness of fit. Conclusions: The nomogram, which incorporates Hb, Plt, and CT attenuation scores, provides valuable insights into predicting treatment response in AML patients. This model offers strong discriminatory ability and could enhance personalized treatment planning and prognosis prediction for AML.