To the Editor: The outcome measures appear to have been chosenpost hoc by Mayrand et al. (Oct. 18 issue)1 in their screeningtrial comparing human papillomavirus (HPV) testing with Papanicolaou(Pap) testing.1,2 The "conservative" outcome definition excludesbiopsy-confirmed lesions identified at colposcopy but not verifiedon final excision or on biopsy immediately before ablation.Such lesions are included in the "liberal" definition, alongwith lesions identified by random biopsy and endocervical curettage.With the conservative definition, HPV testing is clearly superior,whereas with the liberal definition, neither strategy is clearlysuperior. Lesions included in the liberal definition may haveno potential to progess, but it is unusual to extend this rationaleto lesions seen at colposcopy. The authors suggest that thesevisible lesions may have been squamous metaplasia histologicallymisinterpreted as high-grade cervical intraepithelial neoplasia(grade 2 or higher), but this hypothesis was never confirmedby pathological re-review of the targeted biopsy specimens.We propose an alternative outcome measure: the presence of histologicallyverified, high-grade cervical intraepithelial neoplasia in eitherthe colposcopically targeted biopsy specimen or the final excisionspecimen.
Alice Lytwyn, M.D. Laurie Elit, M.D. McMaster University Hamilton, ON M9B 5A2, Canada lytwyn{at}hhsc.ca
John W. Sellors, M.D. Program for Appropriate Technology inHealth Seattle, WA 98107
References
Mayrand M-H, Duarte-Franco E, Rodrigues I, et al. Human papillomavirus DNA versus Papanicolaou screening tests for cervical cancer. N Engl J Med 2007;357:1579-1588. [Free Full Text]
Mayrand MH, Duarte-Franco E, Coutlée F, et al. Randomized controlled trial of human papillomavirus testing versus Pap cytology in the primary screening for cervical cancer precursors: design, methods and preliminary accrual results of the Canadian Cervical Cancer Screening Trial (CCCaST). Int J Cancer 2006;119:615-623. [CrossRef][Web of Science][Medline]
To the Editor: The interpretation by Mayrand et al. of cases that were negative on HPV testing, positive on biopsy, and negativeon excision warrants further discussion. The authors consideredthese cases to represent false positive biopsy results. Firstof all, the criteria for grade 2 to 3 cervical intraepithelialneoplasia on biopsy are well established, and pathologists areunlikely to overdiagnose them. Second, was HPV DNA in situ hybridizationperformed on these specimens? Third, were deeper sections ofthe excision specimens examined, to detect dysplasia not seenon initial sections? The data could also be interpreted as falsenegative results of HPV testing. In some cases, HPV types nottested may be present or lesions may be small (sampling error).In other cases, the lesion may have been removed by biopsiesor lost to tissue inflammatory reaction. An abnormal Pap smearor HPV test result should be followed by visualization of thelesion, confirmatory biopsies, and definitive treatment withexcision or ablation. The result of HPV testing should not beused as the sole criterion for rejecting the interpretationof a biopsy finding.
Bryan T. Lin, M.D., Ph.D. Encino Tarzana Regional Medical Center Tarzana, CA 91356 btltarzana{at}hotmail.com
To the Editor: The study by Mayrand and colleagues suffers fromselection bias that precludes generalization to the known screeningpopulation. Exclusion of young women and pregnant women normallyscreened for cervical cancer introduces a bias against Pap testingand for HPV DNA testing, negating the authors' sensitivity andspecificity comparisons for general screening.
Data from the Surveillance, Epidemiology, and End Results (SEER)program1 indicate that 15.5% of cases of cervical cancer occurin women who are 20 to 34 years old, closely matching the 15.1%reported among women who are 45 to 54 years old. Exclusion ofpregnant women is equally questionable. Professional and publichealth organizations recommend the initiation of cervical-cancerscreening no later than at the age of 21 years,2,3,4 and colposcopyis performed in young women with screening-test indications,irrespective of pregnancy status, albeit with some proceduralmodification if biopsy is warranted.
The authors are straightforward in explaining why young womenwere excluded from their study: HPV DNA testing for cervical-cancerscreening was expected to perform poorly in young women. However,since the purpose of the study was to compare the two testsfor cervical-cancer screening, exclusion of women on the basisof expected test performance seems irreparably confounding.
Allan P. Frank, M.D. 5648 Cedar Crest Dublin, CA 94568 afrankmd{at}msn.com
References
National Cancer Institute. Surveillance, Epidemiology, and End Results (SEER) home page. (Accessed January 17, 2008, at http://www.seer.cancer.gov.)
Saslow D, Runowicz CD, Solomon D, et al. American Cancer Society guideline for the early detection of cervical neoplasia and cancer. CA Cancer J Clin 2002;52:342-62. (Also available at http://caonline.amcancersoc.org/cgi/content/full/52/6/342.)
To the Editor: Mayrand and colleagues present 95% confidenceintervals for the sensitivities of cytologic examination andHPV testing. For the HPV test, the lower bound of the intervalis 84.2%. We think that this lower bound gives an overly optimisticimpression of the test's sensitivity. The 95% confidence intervalswere computed with the use of normal reference distributions.Such confidence intervals may have poor coverage.1 Suppose thatMayrand and colleagues found 1 woman instead of none with grade2+ cervical intraepithelial neoplasia in the group of 652 womenwith two negative screening tests and histologic verification.Then the sensitivity of the HPV test would drop below 80%, indicatingthat the confidence interval may be too narrow. We think thatin order to make a reliable statement about the test sensitivity,the sample of women with two negative tests and histologic verificationneeds to be substantially larger than 652. We understand thatsuch data are difficult to obtain because of costs and patientdiscomfort. We nonetheless think that the coverage of the confidenceinterval should have been discussed, since this informationis important for health care decision makers.
Johannes Berkhof, Ph.D. Chris J. Meijer, M.D. VU University Medical Center 1007MB Amsterdam, the Netherlands h.berkhof{at}vumc.nl
Dr. Meijer reports receiving consulting fees from Digene. Noother potential conflict of interest relevant to this letterwas reported.
References
Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion. Stat Sci 2001;16:101-117. [Web of Science]
To the Editor: Runowicz, in the editorial accompanying the articleby Mayrand et al., wonders whether widespread HPV DNA testingwould result in more colposcopies, "thereby considerably increasingthe use of health care resources."1 I believe that with propertriage algorithms, rates of detection of cervical cancer wouldsignificantly improve, and systemwide expenses would go down.
Currently, many physicians recommend an annual Pap smear forall women. Yet if we now know that more than 99% of women inwhom cervical cancer ultimately develops are HPV-positive,2and if we determine the HPV status of all women, why continueto perform an annual Pap smear in HPV-negative women who arenot at risk for this disease?
Robert P. Fields, M.D. 18109 Prince Philip Dr. Olney, MD 20832
References
Runowicz CD. Molecular screening for cervical cancer -- time to give up Pap tests? N Engl J Med 2007;357:1650-1653. [Free Full Text]
Walboomers JM, Jacobs MV, Manos MM, et al. Human papillomavirus is a necessary cause of invasive cervical cancer worldwide. J Pathol 1999;189:12-19. [CrossRef][Web of Science][Medline]
The authors reply: Lytwyn et al. and Lin underscore the importanceof outcome definition in screening trials. Despite Lin's assertion,pathologists often overinterpret grade in cervical-biopsy specimens.The literature on cervical pathology leaves no doubt about theissue of interobserver variability.1 Likewise, there is substantialvariation in colposcopic diagnoses,2 a point with which Lytwynet al. would probably agree. Mindful that variations in diseaseascertainment could differ in the test groups and thus biasour results, we adopted a standardized colposcopy protocol requiringendocervical curettage and biopsies regardless of the reasonfor referral (blinded), whether because of a test result orrandom assignment. This approach to ascertaining disease ismore thorough than community colposcopy practiced throughoutNorth America. Therefore, the disease prevalence in our trialincluded many low-volume, incipient lesions, which requiredseparate consideration. Although it is unlikely because of ourprotocol's heightened scrutiny, pathologists may have occasionallymissed dysplastic tissue in excisional or ablative specimens.However, a lesion that spontaneously resolved or was "cured"by a simple biopsy does not have the same influence on the riskof subsequent cancer as a lesion whose size and grade causedit to remain detectable in the confirmatory specimen. We reportedthe results of the first screening round. Once the trial's secondround and follow-up have been completed,3 pathological adjudicationwill permit an evaluation of different outcome definitions,as proposed by Lytwyn et al. Also planned is typing of all HPV-positivespecimens, which will shed light on Lin's point.
Concerning the issue raised by Frank, exclusion of young andpregnant women does not bias the results (interval validity)but does preclude extension of the findings to these groups(external validity). We excluded pregnant women on ethical groundsbecause of our augmented colposcopy referral criteria. Our aimwas to compare tests in women who were 30 years of age or older,a restriction that maximizes both Pap and HPV performance butdoes not confound the results. By analogy with breast screening,mammography does not perform as well in premenopausal womenas it does in postmenopausal women, but this does not precludeits broad recommendation. Optimal cervical-cancer screeningin women younger than 30 years of age is still an area in needof much research.
Berkhof and Meijer allude to the issue of precision stemmingfrom our estimates corrected for verification bias. Althoughwe agree that cases among subjects with negative tests are veryinfluential with respect to the estimates, the need for ascertainingdisease unconditionally on the basis of test status to correctfor this bias is a tradeoff between ethical concerns and externalvalidity. Ultimately, we compared the tests on the basis ofstatistical significance, which provides the evidence for healthtechnology assessment.
Eduardo L. Franco, Dr.P.H. Marie-Hélène Mayrand, M.D. McGill University Montreal, QC H2W 1S6, Canada eduardo.franco{at}mcgill.ca
Sam Ratnam, Ph.D. Newfoundland Public Health Laboratory St. John's, NF A1B 3T2, Canada
for the Canadian Cervical Cancer Screening Trial Study Group
References
Castle PE, Stoler MH, Solomon D, Schiffman M. The relationship of community biopsy-diagnosed cervical intraepithelial neoplasia grade 2 to the quality control pathology-reviewed diagnoses: an ALTS report. Am J Clin Pathol 2007;127:805-815. [Free Full Text]
Jeronimo J, Massad LS, Castle PE, Wacholder S, Schiffman M, National Institutes of Health (NIH)-American Society for Colposcopy and Cervical Pathology (ASCCP) Research Group. Interobserver agreement in the evaluation of digitized cervical images. Obstet Gynecol 2007;110:833-840. [CrossRef][Web of Science][Medline]
Mayrand MH, Duarte-Franco E, Coutlée F, et al. Randomized controlled trial of human papillomavirus testing versus Pap cytology in the primary screening for cervical cancer precursors: design, methods and preliminary accrual results of the Canadian Cervical Cancer Screening Trial (CCCaST). Int J Cancer 2006;119:615-623. [CrossRef][Web of Science][Medline]
The editorialist replies: Although we may one day use HPV testingfor primary screening, we are not ready to do so at this time.Before we can accept HPV testing for primary screening, we willneed to develop a rapid, simple, accurate, and affordable HPVDNA test. New algorithms, including a triage for HPV DNA tests,will need to be developed and tested. The duration of protectionafforded by a negative HPV DNA test will require further long-termfollow-up of studies like the one reported by Mayrand and colleagues.
As noted in my editorial, the ultimate goal of cervical screeninghas to be to reduce the incidence of and mortality from invasivecervical cancer worldwide with the use of a cost-effective andreadily available test. The optimal approach will depend onthe prevalence of disease, access to screening, and availableresources. We are not there yet.
Carolyn D. Runowicz, M.D. University of Connecticut Health Center Farmington, CT 06030
Previous
