Statistical Controversies in Clinical Research. Statistical significance - too much of a good thing

《Statistical Controversies in Clinical Research. Statistical significance - too much of a good thing》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、AnnalsofOncologyAdvanceAccesspublishedFebruary9,20161StatisticalControversiesinClinicalResearch:Statisticalsignificance-toomuchofagoodthing…M.Buyse1,S.A.Hurvitz2,F.Andre3,Z.Jiang4,H.A.Burris5,M.Toi6,W.Eiermann7,M-A.Lindsay8,andD.Slamon21DepartmentofBiostatistics,InternationalDrugDevelopmentIn

2、stitute(IDDI),LouvainLaNeuve,Belgium2DivisionofHematology-Oncology,UniversityofCalifornia,LosAngeles(UCLA),LosAngeles,California,USA3DepartmentofMedicalOncology,InstitutGustaveRoussy,UniversitéParisSud,Villejuif,France4DepartmentofBreastOncology,Beijing307HospitalofPLA,Beijing,China5DrugDevel

3、opmentDepartment,SarahCannonResearchInstitute,Nashville,Tennessee,USA6DepartmentofBreastSurgery,GraduateSchoolofMedicine,KyotoUniversity,Kyoto,Japan7BreastSurgeryClinic,IsarklinikumMünchen,Munich,Germany8ScientificDevelopmentDepartment,TranslationalResearchinOncology(TRIO),Edmonton,Canada©The

4、Author2016.PublishedbyOxfordUniversityPressonbehalfoftheEuropeanSocietyforMedicalOncology.Allrightsreserved.Forpermissions,pleaseemail:journals.permissions@oup.com.2Addressforcorrespondence:DrMarcBuyse,IDDI,AvenueProvinciale,30,1340,Louvain-la-Neuve,Belgium.Tel:+3210614444,Fax:+3210618888,ema

5、il:marc.buyse@iddi.comAbstractTheuseandinterpretationofP-valuesisamatterofdebateinappliedresearch.WearguethatP-valuesareusefulasapragmaticguidetointerprettheresultsofaclinicaltrial,notasastrictbinaryboundarythatseparatesrealtreatmenteffectsfromlackthereof.WeillustrateourpointusingtheresultofB

6、OLERO-1,arandomized,double-blindtrialevaluatingtheefficacyandsafetyofaddingeverolimustotrastuzumabandpaclitaxelasfirst-linetherapyforHER2+advancedbreastcancer.Inthistrial,thebenefitofeverolimuswasseenonlyinthepre-definedsubsetofpatientswithhormonereceptor-negativebreastcanceratbaseline(progre

7、ssion-freesurvivalhazardratio=0.66,P=0.0049).Astrictinterpretationofthisfinding,basedoncomplex“alphasplitting”rulestoassessstatisticalsignificance,ledtotheconclusionthatthebenefitofeverolimuswasnotstatisticallysignificanteitheroverallorinthes