ONCOLOGY Multicenter external validation of two malignancy risk prediction models in patients undergoing 18F-FDG-PET for solitary pulmonary nodule evaluation Simone Perandini 1 & G. A. Soardi 1 & A. R. Larici 2 & A. del Ciello 2 & G. Rizzardi 3 & A. Solazzo 4 & L. Mancino 5 & F. Zeraj 5 & M. Bernhart 6 & M. Signorini 1 & M. Motton 1 & S. Montemezzi 1 Received: 28 December 2015 /Revised: 27 June 2016 /Accepted: 26 August 2016 # European Society of Radiology 2016 Abstract Objectives To achieve multicentre external validation of the Herder and Bayesian Inference Malignancy Calculator (BIMC) models. Methods Two hundred and fifty-nine solitary pulmonary nod- ules (SPNs) collected from four major hospitals which underwent 18-FDG-PET characterization were included in this multicentre retrospective study. The Herder model was tested on all available lesions (group A). A subgroup of 180 SPNs (group B) was used to provide unbiased comparison between the Herder and BIMC models. Receiver operating characteristic (ROC) area under the curve (AUC) analysis was performed to assess diagnostic accuracy. Decision analy- sis was performed by adopting the risk threshold stated in British Thoracic Society (BTS) guidelines. Results Unbiased comparison performed In Group B showed a ROC AUC for the Herder model of 0.807 (95 % CI 0.742–0.862) and for the BIMC model of 0.822 (95 % CI 0.758–0.875). Conclusions Both the Herder and the BIMC models were proven to accurately predict the risk of malignancy when test- ed on a large multicentre external case series. The BIMC model seems advantageous on the basis of a more favourable decision analysis. Key Points • The Herder model showed a ROC AUC of 0.807 on 180 SPNs. • The BIMC model showed a ROC AUC of 0.822 on 180 SPNs. • Decision analysis is more favourable to the BIMC model. Keywords 18 F-fluorodeoxyglucose positron emission tomography . Lung cancer . Solid pulmonary nodule . Decision analysis . Computed tomography Introduction Lung nodules are being detected more and more often owing to the widespread use of computed tomography (CT) of the chest [1]. While multiple nodules are usually easier to charac- terize, a solitary pulmonary nodule (SPN) still constitutes a challenging task for the physician. A SPN is defined, as stated in the Fleischner glossary, as B… a rounded or irregular opac- ity, well or poorly defined, measuring up to 3 cm in diameter^ [2]. Although some of these nodules can be easily characterized because of supplemental findings such as peculiar calcifica- tion patterns, perifissural location or accessory adenopathy, most SPNs are difficult to characterize, posing a serious man- agement issue. There is a growing body of literature that recognizes the importance of obtaining a pre-test probability of malignancy * Simone Perandini mail@simoneperandini.com 1 UOC Radiologia, Ospedale Maggiore di Borgo Trento, Azienda Ospedaliera Universitaria Integrata di Verona, Piazzale Stefani 1, Verona 37124, Italy 2 Dipartimento di Scienze Radiologiche, Università Cattolica del Sacro Cuore, Roma, Italy 3 UO Chirurgia Toracica, Ospedale Humanitas Gavazzeni, Bergamo, Italy 4 UO Radiologia, Ospedale Humanitas Gavazzeni, Bergamo, Italy 5 UO Pneumologia, Ospedale dell’Angelo di Mestre, Venezia, Italy 6 UO Radiologia, Ospedale dell’Angelo di Mestre, Venezia, Italy Eur Radiol DOI 10.1007/s00330-016-4580-3