PurposeTo analyze results from three years of in vivo transit EPID dosimetry of abdominal and pelvic stereotactic radiotherapy and to establish tolerance levels for routine clinical use. Material80 stereotactic VMAT treatments (152 fractions) targeting the abdomen or pelvis were analyzed. In vivo 3D doses were reconstructed with an EPID commercial algorithm. Gamma Agreement Index (GAI) and DVH differences in Planning Target Volume (PTV) and Clinical Target Volume (CTV) were evaluated. Initial tolerance level was set to GAI > 85% in PTV. Fractions Over Tolerance Level (OTL) were deemed to be due to set-up errors, incorrect use of immobilization devices, 4D errors, transit EPID algorithm errors and unknown/unidentified errors. Statistical Process Control (SPC) was applied to determine local tolerance levels. ResultsAverage GAI were (82.7 ± 20.9) % in PTV and (72.9 ± 29.7) % in CTV. 37.8% of fractions resulted OTL and were classified as: set-up errors (3.3%), incorrect use of immobilization devices (2.1%), 4D errors (2.1%), EPID transit algorithm errors (17.1%). OTL causes for the remaining 13.2% of fractions were not identified. The differences between PTV and CTV measured in vivo and calculated mean dose (average difference ± standard deviation) were (−3.3% ± 3.2%) and (−2.3% ± 3.0%). When tolerance levels based on SPC to PTV mean dose differences were applied, the percentage of OTL decreased to 7% and no EPID algorithm error occurred. One error was not identified. ConclusionsThe application of local tolerance levels to EPID in vivo dosimetry proved to be useful for detecting extra-lung SBRT treatment errors.