AIM：Toestablishaprognosticformulathatdistinguishesnon-hypervascularhepaticnodules（NHNs）withhigheraggressivenessfromlesshazardousone.METHODS：Seventy-threeNHNsweredetectedingadoliniumethoxybenzyldiethylene-triamine-pentaaceticacidmagneticresonanceimaging（Gd-EOB-DTPA-MRI）studyandconfirmedtochange2mmormoreinsizeand/ortogainhypervascularity.Allimageswereinterpretedindependentlybyanexperienced,board-certifiedabdominalradiologistandhepatologist;bothknewthatthepatientswereatriskforhepatocellularcarcinomadevelopmentbutwereblindedtotheclinicalinformation.AformulapredictingNHNdestinywasdevelopedusingageneralizedestimatingequationmodelwiththirteenexplanatoryvariables：age,gender,backgroundliverdiseases,Child-Pughclass,NHNdiameter,T1-weightedimaging/T2-weightedimagingdetectability,fatdeposition,lowersignalintensityinarterialphase,lowersignalintensityinequilibriumphase,α-fetoprotein,des-γ-carboxyprothrombin,α-fetoprotein-L3,andcoexistenceofclassicalhepatocellularcarcinoma.Theaccuracyoftheformulawasvalidatedinbootstrapsamplesthatwerecreatedbyresamplingof1000iterations.RESULTS：Duringamedianfollow-upperiodof504d,73NHNswithamediandiameterof9mm（interquartilerange：8-12mm）greworshrankby68.5%（fiftynodules）or20.5%（fifteennodules）,respectively,whereashypervascularitydevelopedin38.4%（twentyeightnodules）.Inthefifteenshranknodules,twelvenodulesdisappeared,while11.0%（eightnodules）werestableinsizebutacquiredvascularity.AgeneralizedestimatingequationanalysisselectedfiveexplanatoriesfromthethirteenvariablesassignificantfactorstopredictNHNprogression.Theestimatedregressioncoefficientswere0.36forage,6.51forlowersignalintensityinarterialphase,8.70or6.03forpositivityofhepatitisBvirusorhepatitisCvirus,9.37fordes-γ-carboxyprothrombin,and-4.05forfatdeposition.Aformulaincorpora
