Page 1
Comparison of Material to determine bias in the chemistry laboratory AAcC Poster 1994. Zoe Brooks, George Massarella, David Plaut COMPARISONOFMATERIALTODETERMINE MATERIAL AND METHODS BIASINTHECHEMISTRYLABORATORY PATIENTPOOLS Zoe Brooks,George Massarella(Cochrane Region Serumsampleswereselectedandpooled togivehighand LaboratoryProgram,Sudbury,Ont.Canada)andDavid lowvalues for eachanalyte.High andlowpoolswere Plaut(BaxterDiagnostics,Miami,Fl) mixed toprovide a mid-rangevalue.Poolswere aliquottedandfrozen;freshaliquotswerethawedon Wecomparcdtheresultsobtainedfromcommercial cach day oftesting. controlmaterial,NSTSerum909,proflciencymaterial andpooledpatientsamplesfor17routinechemistry LIQUIDQC1,2,3 analytestestedin1olaboratoriesandplotted therelative DadeLiquid Monitrolwasusedasrecommendedbythe biasforeachmaterial.Patientresultswere compared to manufacturer. eachlab’sreferenceintervals. PROFICIENCY Objectives: Theaveragevalue from the pastyear’s L.P.T.P.surveys (LaboratoryProficiencyTestingProgram,Ontario, 1.Toobscrvetherelativeabilityofeachmaterial to determinemethod-to-methodbias. Canada)was compared tothereferencevalue. 2.To observe the importance oflab-to-labbiasin NISTSRM-909 interpretation ofpatientresults. Materialwaspreparedandhandled accordingto instructions from NIsT. Methods: 1.Each laboratory testedeach samplein duplicate over Freshaliquotsofeachsampleweretestedin duplicate three consecutive days.Datawasanalyzedandgraphs on three daysusingroutinereagents and calibration. prepared. 2.Pooledpatientsamplesweregraphically compared to LAB ANALYZER REAGENT A KODAK KODAK upperandIowerreferenceintervals. B HITACHI717 B.ML.C. C BECKMAN CX5 BECKMAN Conclusion: D TECHNICONRA-1000 SIGMA 1.Othermaterials donotprovide thesameinformation E TECHNICONRA-1000 SIGMA aspooledpatientsampleswhendetermining F COBAS MIRA SIGMA lab-to-labbias. G COBASMIRA SIGMA 2.Bias and selectionofappropriate intervals are H COBASMIRA SYNERMED important factorsinthereportingofmeaningful COBAS MIRA SIGMA laboratoryresults. Page 1 Page 2
Page 2
Comparison of Material to determine bias in the chemistry laboratory AAcC Poster 1994. Zoe Brooks, George Massarella, David Plaut CHOLESTEROL UREA BIASVARIESWITHMATERIAL BIASVARIESWITHMATERIAL 115% 150% SERUM5mmol/195mg 140% PATIENTPOOLS 110% PERCENTOFTARGET LIQUIDQC1,2,3 PERCENTOFTARGET 130% LIQUIDQC1,2.3 105% 120% PROFICIENCY PROFICIENCY 110% NIST SRM-909 A 100% 100% NISTSRM-909 1 90% 95% TARGETBIAS<3% 80% 70% 90% A C E G B D F H 60% A C 百 G LABORATORIES B D F H LABORATORIES MAXMUM&MINIMUMBIAS-ALLLABS MAXIMUM&MINIMUMBIAS-ALLLABS MATERIAL: SERUM Q.C. P.T. SRM-909 MATERLAL: SERUM Q.C. P.T. SRM-909 “TRUE”: Core Lipid Avgof ReferenceReference Lab,Toronto All Labs Valuc Value “TRUE”: Avgof Avgof Refercnce Reference BIAS-MIN: -1% -6% %- 0% AllLabs AllLabs Value Value BIAS-MAX: +3% +3% 1% +9% BIAS-MIN: -10% -17% -8% -27% BIAS-MAX: +7% +31% +2% +5% Individual testing materials produce variable results in relaton to serum pools. Estimatesofbiasvarywithinindividuallaboratories All laboratoriesshowed acceptablebiaswithnormal dependingonmaterialused. serum;most appeared unaceptable with othermaterial. Page 3 Page 4
Page 3
Comparison of Material to determine bias in the chemistry laboratory AAcC Poster 1994. Zoe Brooks, George Massarella, David Plaut CREATININE ALBUMINPATIENTPOOLS BIASVARIESWITHMATERIAL CLINICALINTERPRETATIONVARIES 160% 70 150% PATIENTPOOLS 65 HIGH POOL OBSERVEDVALUES-g/L PERCENTOFTARGET 60 140% LIQUIDQC1,2,3 LOWPOOL 55 130% 50 A A PROFICIENCY MIDPOOL 120% 45 NIST SRM-909 40 110% + 35 REFERENCE 100% 30 RANGE 90% 25 LIMITS 80% 20 A C E G A C F H B D F H B E G LABORATORIES LABORATORIES MAXIMUM&MINIMUMBIAS-ALLLABS Althoughtheabsolutevaluesreported on thethree MATERIAL: SERUM Q.C. P.T. SRM-909 patientpoolsdonotvarysignificantlyamong ‘TRUE”: Avgof Avgof Reference Rcference AllLabs AllLabs Value Valuc laboratories,reference intervals and therefore clinical BLAS-MIN: -9% -13% -3% +3% interpretations dovary. BIAS-MAX: +10% +16% +10% +54% Itis obvious thaterroneousreferenceintervalsarebeing Variationsbetween cstimates ofbiasfrom usedinsomelaboratorieswithpotentialforincorrect serumpoolsandcommercial/lyophilized materialsare notconsistentfrom lab tolab. interpretationofclinicalresults. Page 5 Page 6
Page 4
Comparison of Material to determine bias in the chemistry laboratory AAcC Poster 1994. Zoe Brooks, George Massarella, David Plaut GLUCOSEPATIENTPOOLS GLUCOSE-PATIENTCASE#1 VALUES/INTERPRETATIONVARIES PHYSICIANRESPONSEVARIES 14 EACHBAR=1DOCTOR:EACHGROUP=1LABORATORY 13 LOWPOOL 8.0 mmol 12 A 11 MIDPOOL DECISIONLEVELmmol/L OBSERVEDVALUES- 7.5 10 HIGHPOOL 7.0 6.5 REFERENCE RANGE LIMITS 6.0 A B 5.5 E G 6.4 5.8 6.4 6.2 6.4 6.4 LABORATORY LAB’SUPPERR Thevariation inresultsintheselabsissufficientlywideif “A55year-oldmanwithnocomplaintsisseeninyour mirroredat decision levelsto causeproblems when officeasanewpatient.Circlethelowestbloodglucose arbitrarylevelsfor clinicalinterpretation areused,eg: valuewhichwouldpromptyoutorepeatthefasting glucoseandimmediatelyundertakefiurther 1)7.8mmol/L(140mg/dL)fastingglucoselevelontwo investigation forthepossibilityof diabetesmellitus.” separate occasionsissufficientfora diagnosisof (5.8,5.9,6.0,6.1,6.2,6.3,6.4,6.5…>8.0) diabetes mellitus. Twoclinicianswouldtakenoacdonuntilthefasting 2)Alevelof>7.8mmol/Lonehouraftera50goral glucosewas>diagnosticlevelsfordiabetesmellitis. doseofglucoserequiresfurther testingina Manywouldwaitforinappropriatelyhighlevelsbefore pregnantwomantodeterminewhether shehas undertakingfurtherinvestigation. gestatlonal diabetes. The decision to takeactionseems tobearlittlerelation Thus thesamepatientwill/(willnot)be diagnosedas tothelab’supperlevelofthereferenceintervaland there having diabetesmellitussimplybypresentingata isnouniformityofdecisionlevelsamongclinicians differentlaboratory. practisingin thesame institution. Page 7 Page 8
Page 5
Comparison of Material to determine bias in the chemistry laboratory AAcC Poster 1994. Zoe Brooks, George Massarella, David Plaut CONCLUSIONS 1.Othermaterialsdonotprovidethesameinformation aspooledpatientsampleswhen determining lab-to-lab bias. Reproduced March 2014 2.Biasand selection ofappropriateintervalsare important factorsin thereporting ofmeaningfiul as part of laboratoryresults. the Journey to Stats-Free Quality www.awesome-numbers.org COMMENTS In order to provide appropriate patient care,weneed: (a)Accurate laboratoryresults; (b)PreciseLaboratoryresults; (c)Reliable reference ranges;and How much (d)Uniform physician interpretation. has changed (a)Accuracy(absence ofsignificantbias)isusually assumedwithout study. in 20 years? (b)Laboratorieshaveput mosteffortinto the achievementofpreciseresults.Currentprecision oftenexceedsclinicalrequirements. (c)Referencerangesarefrequentlyacceptedwithout verification. (d)Physician interpretation is seldom tested. Page 9