We compare the results of 3 dose audits (2013, 2015 and 2017) for bilateral mammography screening exams performed in a full field digital mammography unit GE Senographe DS. We estimated the average entrance surface dose (ESD) for a sample of at least 50 standard breasts. The results were compared with the European DRL - ESD of the 3rd quartile of the dose survey. We estimated average absorbed dose to the glandular tissue within breast (AGD) for a wide range of breast thicknesses, and converted it into effective dose. Data was presented to the staff to induce a process of optimization as team effort.For each dose audit we selected a sample of more than 150 bilateral mammography screening exams. The percentage glandularity was determined using the qualitative method described by Byng (1994) and compared with Dance (2000) empirical formula. AGD was determined from measurements of the output and exposure data using appropriate conversion factors from IAEA (2007), Dance (2000), Dance (2009). Effective doses were determined using ICRP 103 breast tissue weighting factor 0.12.Results (2013): ESD = 5.8 mGy for cranio caudal (CC) and ESD = 6.1 mGy for medio lateral oblique (MLO) projections. These results are considerably below the European DRL’s (10 mGy). E ranged from: E = 0.28 mSv for small breasts (34.5 ± 4.5 mm) to E = 0.69 mSv for very large breasts (84.5 ± 4.5 mm) with larger compressed breast thickness associated with a higher AGD (E); Results (2015): around 30% increase in DRLs: ESD = 8.6 mGy CC and ESD = 8.7 mGy MLO; E ranged from E = 0.38 mSv for small breasts (34.5 ± 4.5 mm) to E = 0.75 mSv for very large breasts (84.5 ± 4.5 mm); Results (2017): within 5% of 2013.The 30% increase in the dose results in 2015 resulted from the increased use of Contrast Enhanced Automatic Optimization of Parameters (AOP CNR) mode. This action was investigated and corrected. Dose audits are important part of a quality assurance program.