A non-invasive myelodysplastic syndromes (MDS) diagnostic model would allow for care while avoiding invasive bone marrow examinations (BME). BME-established MDS patients were compared to non-MDS (BME-excluded) patients. Variables (gender, age, hemoglobin (Hb), mean red blood cell corpuscular volume (MCV), platelet (PLT), and white blood cell (WBC)) were combined with multivariate logistic regression; a probability score (Y) was calculated. MDS (n = 48) and non-MDS (n = 63) patients were used to establish the model. The ROC was drawn, giving an AUC of 0.748 (95% CI: 0.656-0.84). Two cutoff values were used for Y. Y ≥ 0.633: high likelihood (positive predictive value (PPV) = 85%); Y ≤ 0.288: low likelihood (negative predictive value (NPV) = 81%) of MDS. The first group is defined as probable MDS (pMDS); the second, probably not MDS (pnMDS). The model was validated with 40 additional patients (20 with and 20 without MDS). Using clinical and lab data, we could diagnose or exclude MDS in about half of the patients, avoiding BME. Future work will use larger cohorts of patients to improve and further validate the model.
Keywords: Myelodysplastic syndromes; diagnosis; logistic regression; model; noninvasive.