comparison of the performance of rapid prescreening, 10random review, and clinical risk criteria as methods of internal quality control in cervical cytopathology.pdf

《comparison of the performance of rapid prescreening, 10random review, and clinical risk criteria as methods of internal quality control in cervical cytopathology.pdf》由会员分享，可在线阅读，更多相关《comparison of the performance of rapid prescreening, 10random review, and clinical risk criteria as methods of internal quality control in cervical cytopathology.pdf（6页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。

1、Comparison of the Performance of RapidPrescreening,10%Random Review,andClinical Risk Criteria as Methods of InternalQuality Control in Cervical CytopathologySuelene B.N.Tavares,MSc1Nadja L.Alves de Sousa,BSc1Edna J.C.Manrique,MSc1Zair B.Pinheiro de Albuquerque,BSc1Luiz C.Zeferino,PhD2Rita G.Amaral,P

2、hD31SchoolofPharmacy,FederalUniversityofGoias,Goiania,Goias,Brazil.2DepartmentofObstetricsandGynecology,School of Medical Sciences,State University ofCampinas-UNICAMP,Campinas,Sa o Paulo,Brazil.3Department of Pathology and Clinical Cytology,School of Pharmacy,Federal University of Goias,Goiania,Goia

3、s,Brazil.BACKGROUND.Rapid prescreening(RPS)is an internal quality-control(IQC)method that is used both to reduce errors in the laboratory and to measure thesensitivity of routine screening(RS).Little direct comparison data are availablecomparing RPS with other more widely used IQC methods.METHODS.Th

4、e authors compared the performance of RPS,10%R-10%),anddirected rescreening of negative smears based on clinical risk criteria(RCRC)over 1 year in a community clinic setting.RESULTS.In total,6135 smears were evaluated.The sensitivity of RS alone was71.3%.RPS detected significantly more(132 cases)fal

5、se-negative(FN)cases thaneither R-10%(7 cases)or RCRC(32 cases).RPS significantly improved the overallsensitivity of the laboratory(71.392.2%;P 5.001);neither R-10%nor RCRCsignificantly changed the sensitivity of RS.RPS was not as specific as the othermethods,although nearly 68%of all abnormalities

6、detected by RPS were verifiedas real.RPS of 100%of smears required the same amount of time as RCRC butrequired twice as much time as R-10%.CONCLUSIONS.The current results demonstrated that RPS is a much more effec-tive IQC method than either R-10%or RCRC.RPS detects significantly moreerrors and can

7、improve the overall sensitivity of a laboratory with either a modestincrease or no increase in overall time spent on IQC.R-10%is an insensitive IQCmethod,and neither R-10%nor RCRC can significantly improve the overall sensi-tivity of a laboratory.Cancer(Cancer Cytopathol)2008;114:16570.?2008American

8、 Cancer Society.KEYWORDS:cervical cytology,false-negative rate,internal quality control,rapidprescreening,sensitivity,review based on clinical risk criteria,10%random review.Cytopathology is used worldwide to screen for cervical cancer,because it is capable of identifying treatable precursor lesions

9、and has been associated with a significant decrease in mortalityfrom this type of cancer over the last 50 years.13However,theimpact on mortality rates in different countries has varied,4and thevalidity of cervical cancer screening programs in some settings hasbeen questioned.1,5,6One possible reason

10、 for this variation in outcome may be therelatively high false-negative(FN)rate of the method.5,79FN casesmay occur because of errors in screening or interpretation.10Factorsthat may be associated with increased FN rates include insufficientclinical information,11poor smear quality,and rare and/or s

11、mallabnormal cells.3,1214How much each of these factors affects the FNSee related Editorial on pages 141?3,this issue.Supported by the National Council for Scientificand Technological Development(Conselho Nacio-nal de Desenvolvimento Cientifico e Tecnologico-CNPq;Grant MCT/CNPq 02/2006-Universal).We

12、 are grateful to the team at the Romulo RochaClinical Analysis Center,School of Pharmacy andto Dr.Silvia Helena Rabelo dos Santos andGislaine A.Fonsechi-Carvasan for their contribu-tions to the study.Address for reprints:Rita G.Amaral,PhD,Schoolof Pharmacy,Federal University of Goias,AvenidaBeloHori

13、zonte,Qd39,Lt04,SetorJa?o,CEP 74673-020,Goi?ania,Goi?as,Brazil;Fax:(011)55 62 3209-6037;E-mail:ritaprppg.ufg.brReceived December 9,2007;revision receivedMarch 6,2008;accepted March 11,2008.2008 American Cancer SocietyDOI 10.1002/cncr.23509Published online 2 May 2008 in Wiley InterScience().165in dif

14、ferent settings is not known,because the FNrarely is measured in routine clinical practice.Several different internal quality-control(IQC)strategies have been introduced to reduce the FNrate.Random review of 10%of negative smears(R-10%)is the most common IQC method in the UnitedStates and is recomme

15、nded by the InternationalAcademy of Cytology and the Clinical LaboratoryImprovement Amendments of 1988.15However,in avariety of settings,it has been demonstrated thatthis technique is ineffective in reducing the FNrate,1619and a true measure of the FN rate cannotbe made.Indeed,in a large prospective

16、 clinicalstudy,the R-10%method had a sensitivity of 30%for atypical squamous cells of undetermined signi-ficance or greater and a sensitivity of 0%for low-gradesquamous intraepithelial lesion(LSIL)or greater.20Another IQC approach has been rescreening withclinical risk criteria(RCRC)to select a grou

17、p of nega-tive cases for rescreening that are more likely to haveerrors than a randomly selected group of negativecases.Some studies have suggested that this methodresults in a higher rate of error identification.21,22Like the R-10%method,however,the FN rate cannotbe measured with the RCRC method.A

18、third IQC approach,which has gained tractionin the United Kingdom,has been to rapidly reviewallcases(30120secondseach)afteraroutinereview.Manrique et al.23reported that this methodwas more effective for detecting FN results thanRCRC.However,like the IQC approaches describedabove,with this method,the

19、 FN rate cannot bemeasured.It also has been proposed that rapid prescreening(RPS)of cases may be the best method for evaluatingthe performance of routine gynecologic screening.With this method,cases are screened rapidly beforeroutine screening(RS)in a short time(typically 30120 seconds).A diagnosis

20、of abnormal or normal ismade,and the case is returned for RS without reveal-ing any markings or information to the routinescreener.Because both negative and abnormal casesare included in the preevaluation,this method makesthe task much more interesting for the examiner andalso allows for calculating

21、 the sensitivity of RPS andRS.2428Performance data from Canada are availableand suggest that the RPS method works in the real-life setting.27Nevertheless,to our knowledge,direct compari-sons of the performance of these different methodsin the same practice setting have not been widelyreported previo

22、usly.To address this,we comparedthe relative rate of error detection of R-10%,RCRC,and RPS in a real-life practice setting.MATERIALS AND METHODSThis study was carried out at the Romulo RochaClinical Analysis Center at the School of Pharmacy,Federal University of Goias,Goiania,Goias,Braziland was app

23、roved by the Internal Review Board ofthis institution.In total,6135 conventional cervicalsmears were collected at the basic healthcare centersof the city of Goiania between March 2006 andMarch 2007.These cases formed the basis of the cur-rent study.Two cytopathologists with 2 years and 10 yearsof ex

24、perience participated in the study and wereresponsible for performing RS.Another 2 cytopatho-logists with 6 years and 13 years of experience wereresponsible for RPS,R-10%,and RCRC.One cyto-pathologist was responsible for RPS,and another wasresponsible for both R-10%and RCRC,and the taskswere alterna

25、ted each month.Another 3 cytopatholog-ists with doctoral degrees,all with 15 years of experi-ence,were responsible for a detailed review of allabnormal smears identified by RS or any other of theIQC methods and for a review of discordant cases todefine the final diagnosis(FD).The cytopathologists wh

26、o performed RPS had noprior experience with the method.Thus,they weretrained in the technique,which consisted of practicein rapid screening pf 40 slides per day(1 minuteeach)over the course of 3 months with feedbackregarding the FD of the case.Thus,during the studyperiod,RPS was limited to 40 slides

27、,and each slidewas examined for a mean of 1 minute at the start ofthe working day.The workflow during the study period was as fol-lows:All cases were submitted for RPS.The resultswere classified as abnormal,negative,or unsatisfac-tory;and the results were recorded on a spreadsheet.The prescreening e

28、xaminer had no access to anyclinical information,and the rapid screening waslimited to a mean of 1 minute per slide.No marks ofany kind were made on the slide,and the interpreta-tion was not made available to the cytopathologistwho performed RS.After RPS,all cases were submitted to RS.No timelimit w

29、as imposed for this,and the time used for thisprocess ranged from 6 minutes to 10 minutes.AfterRS,all smears that were classified as abnormal orunsatisfactory were submitted for detailed review byboth cytopathologists.In addition,10%of all negativesmears were submitted to R-10%,and the slides clas-s

30、ified as negative that had some clinical risk criteriawere submitted to RCRC.All cases with an abnormalor unsatisfactory diagnosis from RPS,R-10%,or RCRCalso were submitted to 2 cytopathologists for FD.When the diagnoses were concordant,an FD was166CANCER(CANCERCYTOPATHOLOGY)June 25,2008/Volume 114/

31、Number 3issued.For cases in which the 2 cytopathologists dis-agreed,the cases were reviewed with a third cyto-pathologist at a consensus meeting,and an FD basedon this consensus was issued.Criteria for RCRC were postmenopausal genitalbleeding,contact ectocervical bleeding,evidence ofsexuallytransmit

32、teddiseaseatgynecologicexamination(including human immunodeficiency virus),significantmacroscopic abnormalities at specular examination orcolposcopy,previous radiotherapy and/or chemo-therapy,or previous abnormal cytology.The resultswere classified in accordance with the 2001 BethesdaSystem.29Statis

33、tical analysis was performed using either a2-tailed Fisher exact test or a chi-square test,asappropriate.A threshold of P.05 was used todetermine significance.The sensitivity and specificitywere calculated with their respective 95%confidenceintervals(95%CIs).RESULTSThe initial classification of the

34、cases by the differentreview methods are shown in Table 1.RS detected377 of 529(71%)of all abnormal cases determinedby the FD.The numbers of additional cases detectedby each IQC method are summarized in Table 2.Onthe basis of the FD in Table 2,RPS detected signifi-cantly more FN cases(n 5 132)than e

35、ither R-10%(n 5 7;P.001)or RCRC(n 5 32;P.001).Determining the sensitivity of each techniquewas difficult,because none of the review methodshad high sensitivity;thus,the FD listed in Table 1and 2 does not include the total number of abnormalTABLE 1Corrected Results for Routine Screening and Multiple

36、Internal Quality-Control MethodsCategoryRSRPSR-10%RCRCFDNo.%No.%No.%No.%No.%Unsat721.17751.220030.26841.37Negative544788.79533386.9354698.73113897.01552290.01Abnormal3776.153946.4271.27322.735298.62ASCUS1121.831201,9530.55151.271873.05ASCH550.90520.850030.26811.32LSIL1221.991352.2040.72110.941612.62

37、HSIL721.17761.240030.26801.30AGC160.26110.180000200.33Total5896580255311736135RS indicates routine screening;RPS,rapid prescreening;R-10%,10%random review of negative smears;RCRC,review of smears based on clinical risk criteria;FD,final diagnosis;Unsat,unsatisfactory;ASCUS,atypical squamous cells of

38、 undetermined significance;ASCH,atypical squamous cells,cannot exclude high-grade squamous intraepithelial lesion;LSIL,low-grade squamous intraepithelial lesion;HSIL,high-grade squamous intraepithelial lesion;AGC,atypical glandular cells.TABLE 2Frequency of False-negative Results Identified by the M

39、ethods of Rapid Prescreening,10%Random Review of Negative Smears,and Review ofSmears Based on Clinical Risk Criteria According to Final DiagnosisCytopathologydiagnosesRPSR-10%RCRCDetected onlyby RPS,No.Detected onlyby R-10%,No.Detected onlyby RCRC,No.FDNo.%No.%No.%No.%ASCUS641.0420.03150.2450121873.

40、05ASCH230.370030.052000811.32LSIL340.5650.08110.1820021612.62HSIL70.110030.05400801.30AGC40.070000300200.33Total1322.1570.11320.5297145298.62RPS indicates rapid prescreening;R-10%,10%random review of negative smears;RCRC,review of smears based on clinical risk criteria;FD,final diagnosis;ASCUS,atypi

41、cal squamous cells of undeterminedsignificance;ASCH,atypical squamous cells,cannot exclude high-grade squamous intraepithelial lesion;LSIL,low-grade squamous intraepithelial lesion;HSIL,high-grade squamous intraepithelial lesion;AGC,atypical glandular cells.Rapid Prescreening of Cervical Smears/Tava

42、res et al.167cases in this dataset.By using the precedent setbefore for analyzing the results of RPS,30in thisdataset,RPS detected 262 of the 377 abnormalcases that were detected by RS,which means thatRPS had a sensitivity of 69.5%.The 132 additionalabnormal cases that were detected by using RPSrepr

43、esented 86.8%of all abnormal cases that weremissed by RS,and there was a total of 152 FNcases.Thus,the true FD should have included 529abnormalcases(377 1 152cases 5 529cases).Seven of those abnormal cases were detected withR-10%,and 32 were detected by RCRC.By usingthis analysis,the sensitivities o

44、f the different meth-ods are listed in Table 3.The sensitivity of RS plusRPS was significantly higher than that of RS alone(P 5.001).However the sensitivity of RS combinedwith either R-10%(P 5.81)or RCRC(P 5.32)wasnot significantly higher than the sensitivity of RSalone.The specificity of each techn

45、ique based on theFD in Table 1 is shown in Table 4.The specificityof RPS in this study(96.7%)was lower than thespecificity of R-10%(99.3%)and RCRC(99.6%).Inthis study,approximately 75%of all cases that wereidentified as abnormal on RPS were confirmed onthe FD.The number of hours necessary for each t

46、ech-nique is summarized in Table 5.Both RPS andRCRC took approximately twice as long to completeas R-10%.DISCUSSIONThe results we demonstrate in this study are in linewithpreviousstudies.20,22,27,30,31RPSisapproxi-mately twice as sensitive as R-10%.However,becauseRPS reviews so many more cases,RPS i

47、dentifies 19times as many errors and can improve the overallsensitivity of the laboratory significantly.Twenty-seven percent of these errors are either LSIL or high-grade squamous intraepithelial lesion.Although R-10%detected a few missed cases,this did not resultin any significant increase in the o

48、verall sensitivity ofthe laboratory.It is noteworthy that the results fromthe current study indicated that RCRC is a muchmoresensitivemethodthanR-10%.However,because it also reviews so few cases,RCRC also wasnot able to increase the sensitivity of the laboratorysignificantly.RPS,as reported previous

49、ly,30was theonly method that allowed for determining the sensi-tivity of the laboratory or any other IQC method.Overall,the results strongly reinforce previous studiessuggesting that RPS can improve the performance ofa laboratory significantly and that the R-10%methodmakes nosignificant difference i

50、n overall performance.RPS does have some limitations.The specificityof the technique was lower than any other IQC.Nevertheless,even with this difference,68%of allabnormal cases that were detected with RPS wereidentified as abnormal on FD.Because each case isreviewed for only 1 minute,we believe that