Radionuclide bone scanning (RNB) is considered to be the most practical screening technique for assessing the entire skeleton for skeletal metastases. However, RNB has been shown to be of lower sensitivity than MRI and CT in detecting osteolytic metastases. A prospective study was designed to evaluate the accuracy of planar RNB versus tomographic bone imaging with ¹⁸F-labeled NaF and PET (¹⁸F PET) in detecting osteolytic and osteoblastic metastases and its dependency on their anatomic localization. Methods: Forty-four patients with known prostate, lung or thyroid carcinoma were examined with both planar RNB and ¹⁸F PET. A panel of reference methods including MRI of the spine, ¹³¹I scintigraphy, conventional radiography and spiral CT was used as the gold standard. RNB and ¹⁸F PET were compared by a lesion-by-lesion analysis using a five-point score for receiver operating characteristic (ROC) curve analysis. Results: ¹⁸F PET showed 96 metastases (67 of prostate carcinoma and 29 of lung or thyroid cancer), whereas RNB revealed 46 metastases (33 of prostate carcinoma and 13 of lung or thyroid cancer). All lesions found with RNB were also detected with ¹⁸F PET. Compared with ¹⁸F PET and the reference methods, RNB had a sensitivity of 82.8% in detecting malignant and benign osseous lesions in the skull, thorax and extremities and a sensitivity of 40% in the spine and pelvis. The area under the ROC curve was 0.99 for ¹⁸F PET and 0.64 for RNB. Conclusion: ¹⁸F PET is more sensitive than RNB in detecting osseous lesions. With RNB, sensitivity in detecting osseous metastases is highly dependent on anatomic localization of these lesions, whereas detection rates of osteoblastic and osteolytic metastases are similar. Higher detection rates and more accurate differentiation between benign and malignant lesions with ¹⁸F PET suggest the use of ¹⁸F PET when possible.