The objectives of this study were to determine the diagnostic yield of percutaneous biopsy of osseous spinal lesions under CT and fluoroscopy guidance and to analyze lesion-related and technical factors affecting higher diagnostic yield. We retrospectively reviewed 247 consecutive percutaneous spinal biopsies and recorded the following information: size, anatomic location, and bone matrix of lesions; guiding modality; years of attending physicians' experience; number of approaches; pathologic result of initial biopsy; and final diagnosis. The pathologic results of the initial biopsies were classified as diagnostic or nondiagnostic. All variables were compared using Pearson chi-square test or Fisher exact test. Multivariate logistic regression was also conducted. Of the initial 247 biopsies, 197 (80%) biopsies were diagnostic. On multivariate analysis, size, bone matrix, and final diagnosis of lesion were significant factors affecting biopsy yield. Biopsies of large lesions (≥ 20 mm) showed higher diagnostic yield than biopsies of small lesions (p = 0.006). Biopsies of lytic lesions had the highest diagnostic yield (88%), followed by biopsies of mixed (84%), sclerotic (67%), and isodense lesions (61%). Differences were significant for diagnostic yields of biopsies of lytic versus sclerotic lesions (p = 0.004) and lytic versus isodense lesions (p = 0.031). Biopsies of metastases had significantly highest diagnostic yield (97%), followed by biopsies of primary malignancies (84%) and benign lesions (39%) (p < 0.05). For percutaneous image-guided biopsies of spinal tumorous lesions, diagnostic yield was 80%. Size, bone matrix, and final diagnosis of lesions affected diagnostic yield of percutaneous image-guided biopsies.