A Consensus Approach toDiagnosing Coronary Artery Disease Basedon Clinical andExercise TestData

Objective: Todemonstrate that a consensusapproach forcombining prediction equations based on clinical andexercise testvariables derived fromdifferent populations canstratify patients referred for possible coronary artery disease (CAD) intolow-, intermediate-, andhigh-risk groups. Design: Retrospective analysis ofconsecutive patients withcomplete datafromexercise testing and coronary angiography referred forevaluation ofpossible CAD.After derivation ofalogistic equation inour own training setofpatients, thisequation, along withtwootherequations developed independently byotherinvestigators, was validated ina testset.Thevalidation strategy forthe consensusapproach included thefollowing: (1) calculation ofprobability scoresforeachpatient using eachlogistic equation independently; (2) determination ofprobability thresholds inthetraining setto divide thepatients intothreegroups.low risk(prevalence CAD 70%prevalence ofCAD); (3) using agreement amongatleast twoofthree ofthe prediction equations togenerate "consensus" foreachpatient; and(4) application oftheconsensus approach thresholds tothetestsetofpatients. Settings: Twouniversity-affil iated Veteran's Affairs medical centers. Patients: We studied 718consecutive men between 1985and1995whohadcoronary angiography within 3months ofan exercise treadmill testforsuspected CAD.Thepopulation was randomly divided into a training setof429patients anda testsetof289patients. Patients withprevious myocardial infarction or coronary artery bypass surgery, valvular heart disease, left bundle branch block, or anyQwaves present on their resting ECG were excluded fromthestudy. Measurements: Recording ofclinical andexercise testdataalong with visual interpretation oftheECG recordings on standardized forms andabstraction ofvisually interpreted angiographic datafrom clinical catheterization reports. Results: We demonstrated thatbyusing simple clinical andexercise testvariables, we couldimprove on thestandard useofECGcriteria during exercise testing fordiagnosing CAD.Using the consensus approach divided thetestsetintopopulations withlow, intermediate, andhigh risk forCAD.Since thepatients intheintermediate group wouldbesentforfurther testing andwouldeventually be correctly classified, thesensitivity oftheconsensus approach is94%andthespecificity is92%.The consensus approach controls forvarying disease prevalence, missing data, inconsistency invariable definition, andvarying angiographic criterion forstenosis severity. Thepercent ofcorrectdiagnoses increased fromthe67%forstandard exercise ECG analysis andfromthe80%formultivariable