Construction and external validation of a nomogram that predicts lymph node metastasis in early gastric cancer patients using preoperative parameters

Objective:Tocreateanomogramtopredicttheincidenceoflymphnodemetastasis(LNM)inearlygastriccancer(EGC)patientsandtoexternallyvalidatethenomogram.Methods:Toconstructthenomogram,weretrospectivelyanalyzedaprimarycohortof272EGCpatients.Univariateanalysisandabinarylogisticregressionwereperformed.AnomogrampredictingtheincidenceofLNMinEGCpatientswascreated.Thediscriminationabilityofthenomogramwasmeasuredusingtheconcordanceindex(c-index),andthenomogramwasalsocalibrated.Then,anotherprospectivecohortof81patientswasanalyzedtovalidatethenomogram.Results:Intheprimarycohort,LNMwaspathologicallyconfirmedin37(13.6%)patients.Inmultivariateanalysis,thepresenceofanulcer,themaximumlesiondiameterobservedviagastroscopy,thethicknessofthelesionobservedviaendoscopicultrasonography,andthepresenceofenlargedlymphnodesoncomputedtomography(CT)wereindependentriskfactorsforLNM.Anomogramwasthencreatedbasedontheregressionmodelwiththec-indexof0.905,andthecalibrationcurveofthenomogramfellapproximatelyontheideal45-degreeline.Thecut-offscoreofthenomogramwas110,andthesensitivity,specificity,positivepredictiveandnegativepredictivevaluesofthenomogramintheprimarycohortwere81.1%,86.0%,47.6%and96.7%,respectively,andintheprospectivevalidationcohortwere75.0%,91.0%,60.0%and95.5%,respectively.Thecalibrationcurveoftheexternalvalidationcohortwasalmostonthe45-degreeline.Conclusions:WedevelopedaneffectivenomogrampredictingtheincidenceofLNMforEGCpatients.