Lapatinib and capecitabine (L-CAP) is effective in HER-2 positive patients with metastatic breast cancer (MBC). However, moderate to severe diarrhea and rash (≥ grade 2) are problematic dose limiting toxicities. Since risk may vary over the course of therapy, we developed repeated measures models to predict the risk of ≥ grade 2 diarrhea and rash prior to each cycle of L-CAP. Data from 197 patients who received the L-CAP as part of a clinical trial were reviewed (Cameron, Breast Cancer Res Treat 112:533-543, 2008). Generalized estimating equations were used to develop the risk models using a backward elimination process. Risk scoring algorithms were then derived from the final model coefficients. Finally, a receiver operating characteristic curve (ROC) analysis was undertaken to measure the predictive accuracy of the scoring algorithms. Patient age, presence of skin metastases at baseline, treatment being initiated in the spring, earlier cycles, and grade I diarrhea in the prior cycle were identified as being significant predictors for ≥ grade 2 diarrhea. The ROC analysis indicated good predictive accuracy for the diarrhea algorithm with an area under the curve of 0.78 (95 %CI: 0.72-0.82). Prior to each cycle of therapy, patients with risk scores > 125 units would be considered at high risk for developing ≥ grade 2 diarrhea. A similar prediction index was also derived in the case of ≥ grade 2 rash. Our models provide patient-specific risk information that could be helpful in assessing the risks and benefits of L-CAP in the MBC patients.