BACKGROUND: Relapses frequently occur in giant cell arteritis (GCA), and long-term glucocorticoid therapy is required. The identification of associated factors with poor treatment outcomes is important to decide the treatment algorithm of GCA. METHODS: We enrolled 139 newly diagnosed GCA patients treated with glucocorticoids between 2007 and 2014 in a retrospective, multi-center registry. Patients were diagnosed with temporal artery biopsy, 1990 American College of Rheumatology classification criteria, or large vessel lesions (LVLs) detected by imaging based on the modified classification criteria. Poor treatment outcomes (non-achievement of clinical remission by week 24 or relapse during 52 weeks) were evaluated. Clinical remission was defined as the absence of clinical signs and symptoms in cranial and large vessel areas, polymyalgia rheumatica (PMR), and elevation of C-reactive protein (CRP) levels. A patient was determined to have a relapse if he/she had either one of the signs and symptoms that newly appeared or worsened after achieving clinical remission. Re-elevation of CRP without clinical manifestations was considered as a relapse if other causes such as infection were excluded and the treatment was intensified. Associated factors with poor treatment outcomes were analyzed by using the Cox proportional hazard model. RESULTS: Cranial lesions, PMR, and LVLs were detected in 77.7%, 41.7%, and 52.5% of the enrolled patients, respectively. Treatment outcomes were evaluated in 119 newly diagnosed patients who were observed for 24 weeks or longer. The mean initial dose of prednisolone was 0.76 mg/kg/day, and 29.4% received any concomitant immunosuppressive drugs at baseline. Overall, 41 (34.5%) of the 119 patients had poor treatment outcomes; 13 did not achieve clinical remission by week 24, and 28 had a relapse after achieving clinical remission. Cumulative rates of the events of poor treatment outcomes in patients with and without LVLs were 47.5% and 17.7%, respectively. A multivariable model showed the presence of LVLs at baseline was significantly associated with poor treatment outcomes (adjusted hazard ratio [HR] 3.54, 95% CI 1.52-8.24, p = 0.003). Cranial lesions and PMR did not increase the risk of poor treatment outcomes. CONCLUSION: The initial treatment intensity in the treatment algorithm of GCA could be determined based upon the presence or absence of LVLs detected by imaging at baseline.