The current study aims to assess the suitability of setup errors during the first three treatment fractions to determine cone-beam computed tomography (CBCT) frequency in adjuvant breast radiotherapy. For this, 45 breast cancer patients receiving non-hypofractionated radiotherapy after lumpectomy, including a simultaneous integrated boost (SIB) to the tumor bed and daily CBCT imaging, were retrospectively selected. In a first step, mean and maximum setup errors on treatment days 1-3 were correlated with the mean setup errors during subsequent treatment days. In a second step, dose distribution was estimated using a dose accumulation workflow based on deformable image registration, and setup errors on treatment days 1-3 were correlated with dose deviations in the clinical target volumes (CTV) and organs at risk (OAR). No significant correlation was found between mean and maximum setup errors on treatment days 1-3 and mean setup errors during subsequent treatment days. In addition, mean and maximum setup errors on treatment days 1-3 correlated poorly with dose coverage of the CTVs and dose to the OARs. Thus, CBCT frequency in adjuvant breast radiotherapy should not be determined solely based on the magnitude of setup errors during the first three treatment fractions.