Background and purpose: Evaluation of set-up accuracy and analysis of target reproducibility in the stereotactic body frame (SBF), designed by Blomgren and Lax from Karolinska Hospital, Stockholm. Different types of targets were analyzed for the risk of target deviation. The correlation of target deviation to bony structures was analyzed to evaluate the value of bones as reference structures for isocenter verification. Materials and methods: Thirty patients with 32 targets were treated in the SBF for primary or metastatic peripheral lung cancer, liver metastases, abdominal and pelvic tumor recurrences or bone metastases. Set-up accuracy and target mobility were evaluated by CT-simulation and port films. The contours of the target at isocenter level, bony structures and body outline were compared by matching the CT-slices for treatment planning and simulation using the stereotactic coordinates of the SBF as external reference system. The matching procedure was performed by using a 3D treatment planning program. Results: Set-up accuracy represented by bony structures revealed standard deviations (SD) of 3.5 mm in longitudinal, 2.2 mm in anterior–posterior and 3.9 mm in lateral directions. Target reproducibility showed a SD of 4.4 mm in longitudinal, 3.4 mm ap and 3.3 mm in lateral direction prior to correction. Correlation of target deviation to bones ranged from 33% (soft tissue targets) to 100% (bones). Conclusion: A security margin of 5 mm for PTV definition is sufficient, if CT simulation is performed prior to each treatment to correct larger target deviations or set-up errors. Isocenter verification relative to bony structures is only safe for bony targets but not for soft tissue targets.