运用多准则决策分析得出
全身性红斑狼疮（SLE）的全新分类标准
背景EULAR和ACR大会都在支持SLE基于加权准则和持续性概率量表来制定不同阶段的分类标准。此前的步骤包括标准制定、名义群体和特尔非技术进行标准修改、针对敏感性/候选特殊性标准写出文献综述，将候选标准划分为7个临床、3个免疫领域。
目标修改候选标准定义、运用多准则决策分析确定相应权数并确定SLE分类的最低标准。
方法SLE专家小组（由9名北美人、8名欧洲人组成）提交个不同SLE概率的病例。专家运用候选标准为20个典型病例打分并按分数进行排列。在为期两天的会议中，专家会评估评分人打分是否可信、修改标准定义并运用Minds软件进行多准则决策分析（MCDA）。专家会对两个病例做出多个决策，每种决策都根据口腔溃疡、急性浆膜炎程度以及秃发情况和胸腔积液两大标准进行打分，并匿名投票哪种更贴近SLE。之后专家会对投票进行讨论并达成一致。Minds软件记录专家共识并计算出权重标准并将分数应用于其余个病例中，对它们进行排序。专家对每个病例是否是SLE进行投票。对标准的权重与专家分析不一致的病例重复MCDA，直到达成一致意见。然后，Minds软件重新计算标准权重并对病例重新排序，专家达成一致意见的最后一个病例的分数即为最低分数。
结果评分人可信度打分这一点很好，人为数据操作失误或者未遵循指示操作或是对标准的不同解读都可能导致差异。关节炎和心包炎的定义会根据小组讨论进行修改。MCDA包括74组决策。由于对SLE的定义意义不大，颅神经病变和六型狼疮性肾炎已经被排除。对关节炎和皮肤领域重复MCDA是因为初始权重与专家意见不符。在重新计算标准权重和分数以后，专家对分数大于83分的SLE分类标准达成了一致意见。
结论在反复操作的过程中，专家小组修正了定义、候选标准权重并确立了最低分数大于83分作为SLE分类标准，将会进行进一步证实。
原文MULTICRITERIADECISIONANALYSISFORDEVELOPINGNEWCLASSIFICATIONCRITERIAFORSYSTEMICLUPUSERYTHEMATOSUS
S.Tedeschi1,S.Johnson2,D.Boumpas3,D.Daikh4,B.Diamond5,T.Dorner6,S.Jacobsen7,D.Kamen8,W.McCune9,M.Mosca10,R.Ramsey-Goldman11,G.Ruiz-Irastorza12,M.Schneider13,J.Smolen14,M.Urowitz2,D.Wofsy4,M.Aringer15,R.Naden16,K.Costenbader1.1HarvardMedicalSchool,Boston,UnitedStates;2Univ.Toronto,Ontario,Canada;3Univ.Athens,Athens,Greece;4Univ.California,SanFrancisco;5FeinsteinInstitute,Manhasset,UnitedStates;6ChariteUniv.Hospitals,Berlin,Germany;7CopenhagenUniv.Hospital,Copenhagen,Denmark;8MedicalUniv.S.Carolina,Charleston;9Univ.Michigan,AnnArbor,UnitedStates;10Univ.Pisa,Pisa,Italy;11NorthwesternUniv.,Chicago,UnitedStates;12BiocrucesInstitute,Barakaldo,Spain;13Univ.Dusseldorf,Dusseldorf,Germany;14MedicalUniv.Vienna,Vienna,Austria;15Univ.MedicalCenterCarlGustavCarus,Dresden,Germany;16McMasterUniv.,Hamilton,Canada
Background:
EULARandACRaresupportingmulti-phasedevelopmentofSLEclassificationcriteriabasedonweightedcriteriaandacontinuousprobabilityscale.Priorstepsincludedcriteriageneration,criteriareductionthroughDelphiandNominalGroupTechniqueexercises,literaturereviewforsensitivity/specificityofcandidatecriteria,andorganizationofcandidatecriteriaintosevenclinicalandthreeimmunologicdomains.
Objectives:
Torefinedefinitionsofcandidatecriteria,determinerelativeweightsusingmulticriteriadecisionanalysis,anddetermineathresholdscoreforSLEclassification.
Methods:
AnSLEExpertPanel(9NorthAmerican,8European)submitteduniquecaseswitharangeofSLEprobability.Expertsscored20representativecasesusingthecandidatecriteriaandrank-orderedthem.Ina2-daymeeting,expertsreviewedinter-raterreliabilityofscoring,refinedcriteriadefinitions,andparticipatedinamulticriteriadecisionanalysis(MCDA)exerciseusingMindsTMsoftware.Expertswerepresentedaseriesofdecisionsbetweentwocases,eachwithdifferentcriteriafromtwodomains(e.g.oralulcers[cutaneous]andacutepericarditis[serositis]vs.alopecia[cutaneous]andpleuraleffusion[serositis])andanonymouslyvotedforthecasemorelikelytobeclassifiedasSLE.Voteswerediscusseduntilconsensuswasreachedforeachdecision.Usingtheconsensusdecisions,Minds?calculatedcriteriaweights,assignedatotalscoretoeachofremainingcasesandrank-orderedthecases.ExpertsvotedonwhethereachcaseshouldbeclassifiedasSLE.MCDAwasrepeatedforcriteriawhosecalculatedweightswereinconsistentwithexpertopinionuntilgroupconsensuswasachieved.MindsTMthenre-calculatedcriteriaweightsandre-rankedcasesonce.Thescoreofthelastcaseforwhichexpertconsensuswasachievedwasthethresholdscore.
Results:
Inter-raterreliabilitywasgood;humandataentryerror,notfollowinginstructions,anddifferinginterpretationsofcriteriadefinitionsaccountedfordiscrepancies.Arthritisandpericarditisdefinitionsweremodifiedthroughgroupdiscussion.TheMCDAinvolved74pairwisedecisions.CranialneuropathyandClassVIlupusnephritiswereremovedastheyaddedlittletoSLEclassification.MCDAwasrepeatedforthearthritisandcutaneousdomainsasinitialweightsdidnotmatchexpertopinion.Aftercriteriaweightsandscoreswerere-calculatedonce,expertsreachedconsensusforSLEclassificationforcasescore83.
Conclusions:
Usinganiterativeprocess,theexpertpanelrefineddefinitions,weightedcandidatecriteriaanddeterminedathresholdscoreof83forSLEclassification,whichwillundergovalidation.
Acknowledgements:JointsupportfromEULARandACR
DisclosureofInterest:Nonedeclared
DOI:10./annrheumdis--eular.
▼
更多专业资讯
