Epidemiologicstudiesuseoutcome-dependentsampling(ODS)schemeswhere,inadditiontoasimplerandomsample,therearealsoanumberofsupplementsamplesthatarecollectedbasedonoutcomevariable.ODSschemeisacost-effectivewaytoimprovestudyefficiency.Wedevelopamaximumsemiparametricempiricallikelihoodestimation(MSELE)fordatafromatwo-stageODSschemeundertheassumptionthatgivencovariate,theoutcomefollowsagenerallinearmodel.Theinformationofbothvalidationsamplesandnonvalidationsamplesareused.Whatismore,weprovetheasymptoticpropertiesoftheproposedMSELE.