OBJECTIVES: The postoperative recurrence rate of sinonasal inverted papilloma (IP) is high, and the residual tumor at the attachment site (the pedicle of the tumor) is considered the main cause of recurrence. Therefore, a surgical approach tailored to the tumor attachment is crucial. Localized hyperostosis is an imaging characteristic of the attachment observed on computed tomography (CT). This study aimed to determine the tendency of hyperostosis at the IP attachment location according to the detailed anatomical site and to improve the prediction accuracy of the preoperative attachment site. MATERIALS AND METHODS: This single-center retrospective cohort study was conducted at the Jikei University Hospital from April 2018 to March 2024, targeting patients diagnosed with IP. The attachment distribution and details of the hyperostosis at the attachment on CT were investigated. RESULTS: Among the 127 included patients, the attachments were identified in the nasal septum (n=3), ethmoid sinus (n=51), maxillary sinus (n=59), frontal sinus (n=8), sphenoid sinus (n=5), and multiple or broad attachments (n=1). Hyperostosis was observed in all cases of the skull base, uncinate process, infraorbital wall, posterior wall, floor of the maxillary sinus, and floor of the sphenoid sinus. In contrast, non-hyperostosis was observed in all cases of the nasal septum, supraorbital cell, anterior wall of the maxillary sinus, and posterior wall of the sphenoid sinus. CONCLUSIONS: This study revealed significant hyperostosis in the posterior wall of the maxillary sinus, infraorbital wall, skull base, uncinate process, and floor of the sphenoid sinus, whereas the infraorbital wall, posterior wall, floor of the maxillary sinus, skull base, uncinate process, and floor of the sphenoid sinus exhibited significant bone hyperostosis. These findings can lead to improved accuracy of preoperative attachment prediction using CT, appropriate surgical approach selection, and better explanations for patients, especially in the maxillary sinus and skull base IP.