BackgroundThe aim of this study was to develop a prediction model of progressive disease (PD) in breast cancer patients without measurable disease in first-line chemotherapy. MethodsWe developed a model to predict PD using carcinoembryonic antigen (CEA) and carbohydrate antigen (CA) 15-3 in metastatic breast cancer patients who were enrolled in a phase III trial. The model was determined using the area under the receiver operating characteristic curve (AUC) calculated by the bootstrap method as internal validation. We verified the model for those who received first-line chemotherapy in a clinical setting as external validation. We categorized patients without measurable disease into PD and non-PD groups and compared the time to progression (TTP). ResultsThe model consisted of percent changes in CEA and CA 15-3 levels from second to third chemotherapy course and baseline abnormality of them. The AUC after external validation was 0.90. Patients without measurable disease were categorized into PD (N = 10) and non-PD groups (N = 53) by the model. The difference in TTP between the two groups was statistically significant (hazard ratio, 0.437; P = 0.021). ConclusionThe model may be useful to determine PD in metastatic breast cancer patients without measurable disease.