An external surrogate-based respiratory gating technique is a useful method to reduce target margins for the treatment of a moving lung tumor. The success of this technique relies on a good correlation between the motion of the external markers and the internal tumor as well as the repeatability of the respiratory motion. In gated lung stereotactic body radiation therapy (SBRT), the treatment time for each fraction could exceed 30 min due to large fractional dose. Tumor motion may experience pattern changes such as baseline shift during such extended treatment time. The purpose of this study is to analyze tumor motion traces in actual treatment situations and to evaluate the effect of the target baseline shift in gated lung SBRT treatment. Real-time motion data for both the external markers and tumors from 51 lung SBRT treatments with Cyberknife Synchrony technology were analyzed in this study. The treatment time is typically greater than 30 min. The baseline shift was calculated with a rolling average window equivalent to ∼20 s and subtracted from that at the beginning. The magnitude of the baseline shift and its relationship with treatment time were investigated. Phase gating simulation was retrospectively performed on 12 carefully selected treatments with respiratory amplitude larger than 5 mm and regular phases. A customized gating window was defined for each individual treatment. It was found that the baseline shifts are specific to each patient and each fraction. Statistical analysis revealed that more than 69% treatments exhibited increased baseline shifts with the lapse of treatment time. The magnitude of the baseline shift could reach 5.3 mm during a 30 min treatment. Gating simulation showed that tumor excursion was caused mainly by the uncertainties in phase gating simulation and baseline shift, the latter being the primary factor. With a 5 mm gating window, 2 out of 12 treatments in the study group showed significant tumor excursion. Baseline shifts alone could cause up to 20% of tumor excursion outside the gating window. It is concluded that baseline shifts may increase with the treatment time and are more likely to act as a time-dependent systematic error. For phase-based gated lung SBRT, a baseline shift may be one of the major sources of targeting error during treatment.