JOURNAL OF CLINICAL MICROBIOLOGY, Nov. 1992, p. 2951-29590095-1137/92/112951-09$02.00/0Copyright © 1992, American Society for Microbiology

Vol. 30, No. 11

Comparison of ViraPap, Southern Hybridization, andPolymerase Chain Reaction Methods for Human

Papillomavirus Identification in an EpidemiologicalInvestigation of Cervical Cancer

ELOISA GUERRERO,1'2 RICHARD W. DANIEL,2 F. XAVIER BOSCH,3 XAVIER CASTELLSAGUE,3NUBIA MUNOZ,3 M. GILI,4 P. VILADIU,5 C. NAVARRO,6 M. L. ZUBIRI,7 N. ASCUNCE,8

L. C. GONZALEZ,9 L. TAFUR,l0 I. IZARZUGAZA,"1 AND KEERTI V. SHAH2*Abbott Cientifica S.A., E-28034 Madrid,1 Catedra de Medicina Preventiva y Social, Facultad de Medicina,

E-41009 Sevilla,4Hospital Sta Caterina, E-1 7001 Girona, Consejeria de Sanidad, Registro de Cancer,E-30008 Murcia Departmento de Sanidad, Diputacion General de Aragon, E-50004 Zaragoza, Planificaciony Ordenacion, Departamento de Salud, Gobiemo de Navarra, E-31001 Pamplona,8 Delegacion Territorial deBienestar Social, E-37013 Salamanca, 9 and Departamento de Sanidad (Edificio Osakidetza), E-01006 Vitoria-Gasteiz,11 Spain; Department ofImmunology and Infectious Diseases, The Johns Hopkins University, School

of Hygiene and Public Health, 615 North Wolfe Street, Baltimore, Maryland 212052; Unit ofField andIntervention Studies, International Agency for Research on Cancer, F-69372 Lyon Cedex 08, France3; and

Universidad del Valle, Cali, ColombialoReceived 15 April 1992/Accepted 4 August 1992

In order to provide a reliable diagnosis for the presence and type of human papillomavirus (HPV) DNA ina case-control study of cervical cancer in Colombia and Spain, 926 cervical scrapes from female subjects wereexamined by ViraPap (VP) and Southern hybridization (SH), and 510 of these (263 cases and 247 controls) werealso tested by polymerase chain reaction (PCR) using the HPV Li consensus primers. HPV DNA prevalencewas much higher in cases than in controls by each of the three tests. There was complete agreement betweenthe results of the three tests for 64.9%o of the 510 specimens; 53.5% were negative and 11.4% were positive(regardless of type) by all tests. An additional 29.0%o of the specimens were positive by PCR: 19.4% by PCRalone, 6.7% by PCR and VP, and 2.9%o by PCR and SH. SH and/or VP gave positive results for 6.0%o of thespecimens for which the PCR finding was negative: 2.7% by SH alone, 2.5% by VP alone, and 0.8% by bothVP and SH. When specimens which were positive byVP alone or only by SH at low-stringency conditions wereexcluded, PCR confirmed all but four specimens which were positive by other tests. The concordance betweentype-specffic diagnosis by SH and PCR was 86% when HPVs were typed in both tests. HPV-16 accounted forover 80%o of the typed HPVs in each test. The presence of blood in case specimens did not appear to inhibit HPVpositivity by VP or by PCR at the dilution tested. Low amounts of cellular DNA of specimens resulted in someunderestimation ofHPV positivity byVP and SH but not by PCR. Compared with that ofPCR, the sensitivitiesfor case specimens were 38% by SH and 50%o by VP; the sensitivity for control specimens, although it couldnot be measured precisely because there were few positive specimens, appeared to be lower than for casespecimens. It was concluded that PCR-based tests are best suited for epidemiological investigation of HPVs.

The evidence linking genital tract human papillomavirus(HPV) infections to cancer of the uterine cervix has beenderived from clinical, epidemiologic, and pathogenetic stud-ies, as well as from laboratory studies of the oncogenicpotential of HPV types (12, 26). Two major difficulties havebeen encountered in the interpretation ofHPV virologic datain epidemiologic investigations. First, because of small sam-ple size and potential bias in the selection of subjects in mostof the studies, it has been difficult to compare results fromdifferent investigations (16). Second, HPV diagnosis, whichrequires nucleic acid hybridization tests, has been made byusing a wide variety of procedures, e.g., Southern hybrid-ization (SH), filter in situ hybridization (25), dot blot hybrid-ization, and polymerase chain reaction (PCR) (14, 19, 24).These tests have not been fully standardized and are subjectto significant interlaboratory variations. For example, pair-wise comparisons of HPV diagnoses of the same specimensby SH in four experienced laboratories showed as much as

* Corresponding author.

34% disagreement (2). In a comparison carried out by twolaboratories, the interlaboratory variation with filter in situhybridization tests was even greater; 11 and 19% of speci-mens were positive with a mixed HPV-16 and HPV-18 probein the two laboratories but only 3% were positive in bothlaboratories (17a, 18). It has been suggested that errors inHPV identification rather than inadequacies of study designare the major source for discrepant results in epidemiologicalinvestigations of the HPV etiology of cervical cancer (6, 21).We had an opportunity to compare three different tests for

HPV diagnosis in the course of a large-scale case-controlstudy of cervical cancer performed by the InternationalAgency for Research on Cancer between 1986 and 1989.Cases and controls of invasive cancer and of cervical intra-epithelial neoplasia grade 3 (CIN3) were studied at twodifferent sites: Cali, Colombia, and nine provinces of Spain(1Sa). The case and control subjects and their male sexualpartners answered detailed questionnaires related to riskfactors for cervical cancer and also provided exfoliated cellsfrom the cervix or the penis for HPV diagnosis and serum

2951

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

2952 GUERRERO ET AL.

specimens for antibody assays. Biopsies from invasive caseswere collected and frozen. This study is based on HPVanalysis of cervical scrapes from cases and controls. Thetests chosen for HPV diagnosis were ViraPap (VP), acommercially available screening test for seven HPV types;SH, which is regarded as the "gold standard" for HPVdiagnosis; and PCR, which is the most sensitive method forHPV detection.

MATERIALS AND METHODS

Specimens. Two cervical specimens were collected fromeach woman, by using an Ayres spatula and cotton swabs orcytobrush. One specimen was used to prepare a Pap smear.The second specimen, as well as the remaining cells of thefirst specimen, was placed in a tube containing phosphate-buffered saline (PBS). Cells were detached from the swab,spatula, and cytobrush by shaking and were centrifuged at500 x g for 10 min. The cell pellet was resuspended in PBSby gentle agitation and, in most instances, was split into twoaliquots with a Pasteur pipette. Both aliquots were stored at-20 or -70°C until they were transported to Lyon, France,on dry ice and stored at -70°C. Aliquot 1 was shipped toBaltimore, Md., for VP and SH tests. Aliquot 2 was storedand processed for PCR in Lyon and the processed specimenswere shipped to Baltimore for PCR amplification and HPVidentification.

All cervical scrapes from invasive cases and controls wereexamined by VP as well as by SH. All specimens from thesegroups for which a second aliquot was available were alsoexamined by PCR, employing the Li consensus primers forHPV amplification (14). All specimens from CIN3 cases andcontrols were examined by VP, and some of them were alsoexamined by SH.VP. Cells were pelleted from aliquot 1 of each specimen

and digested with the bacterial protease solution provided inthe VP kit for sample preparation. One half of the digestedaliquot was stored at -70°C for future use for SH, and theother half was tested by VP according to the instructions inthe kit. Specimens were placed on filters and hybridized withthe mixed probe containing seven HPV types (HPV-6,HPV-11, HPV-16, HPV-18, HPV-31, HPV-33, and HPV-35).The hybridized filters were autoradiographed for 1 to 10days, according to the age of the probe, as recommended bythe manufacturer. The intensity of the VP signals was scoredas follows: 0, no signal; 1, less than that of the low-positivecontrol; 2, equal to that of the low-positive control; 3,greater than that of the low-positive control but less than thatof the high-positive control; 4, equal to that of the high-positive control; and 5, greater than that of the high-positivecontrol. Specimens were scored as negative for VP when thesignal intensity was 0 or 1 and positive when it was 2 orgreater.SH. Cellular DNA was purified from the stored portion of

aliquot 1 by phenol-chloroform extraction and isopropanol-sodium acetate precipitation. The DNA pellet was resus-pended and digested overnight with PstI at 37°C in a 25-,ulvolume. Five microliters of the digested cellular DNA wasused to estimate the amount of DNA by fluorescence emis-sion with a commercial dye (Hoechst Laboratories) thatspecifically binds DNA. The rest of the digested DNA waselectrophoresed in a 1% agarose gel, transferred by capillar-ity to a nylon membrane (Hybond; Amersham Co.), andfixed by being baked under vacuum at 80°C for at least 30min. The membranes were then hybridized to individual32P-labelled HPV DNA probes under high-stringency condi-

tions (melting temperature [Tm] - 25°C) (23). To prepareprobes, full-length HPV DNAs were purified from plasmidvectors by agarose gel electrophoresis, random primed fol-lowing the manufacturer's instructions (Boehringer Mann-heim Biochemicals), and purified by ethanol precipitation.The filters were tested sequentially with HPV-18, HPV-16,HPV-31, HPV-33, a pool of HPV-6 and HPV-11, and HPV-35. Between hybridizations, filters were boiled to remove theprevious probe. HPVs were specifically identified on thebasis of reactivity with the probes and the sizes of thehybridizing bands, as estimated by a comparison with thoseof an HPV ladder. The sensitivity and specificity of thehigh-stringency hybridizations were estimated from HPVstandards (10, 1, and 0.1 pg of viral DNA of each HPV typein the probe panel), which were electrophoresed into eachgel for 3 min at the end of the electrophoresis period. Thesizes of the DNA bands separated by electrophoresis wereestimated by comparison to those of an HPV ladder in theouter lanes of the gel. The ladder consisted of restrictionfragments of HPV-11, HPV-16, HPV-18, and HPV-31 andwas provided by Attila L6rincz.

Following stringent hybridization with HPV-35, a low-stringency hybridization (Tm - 37°C) (23) was carried outwith a mixture of HPV-16 and HPV-11 probes. Tests underlow-stringency hybridization were not fully satisfactory be-cause of nonspecific hybridization; only approximately halfof the specimens could be interpreted under these condi-tions. Specimens positive only in low-stringency hybridiza-tion were categorized as "HPV-SH, type unknown."PCR amplification and hybridization. In order to minimize

the possibility of contamination, aliquot-2 specimens frominvasive cases and controls were processed for PCR inLyon, France, in a laboratory which had not worked withHPV. Approximately one-half of the aliquot-2 specimenfrom each individual was digested with proteinase K (100,ug/ml) in a 125-,ul reaction at 55°C for 2 h in the presence ofNonidet P-40 (0.5%)-Tween 20 (0.5%). The sample tubeswere then placed in a boiling water bath for 7 min (toinactivate the enzyme) and then centrifuged. The superna-tants were distributed in 25-p,l aliquots and stored at -70°C.Blank tubes (every 20th tube) and simian virus 40-trans-formed rat cells (every 10th tube) were processed with thespecimens to monitor for contamination.The processed specimens were transported to Baltimore,

Md., for PCR amplification and HPV diagnosis. Amplifica-tion was performed taking precautions recommended tominimize the possibility of contamination (14). HPV Liconsensus primers (13), which amplify an approximately450-bp fragment of many HPV types, as well as ,-globinprimers (19) which amplify a 260-bp sequence of the 13-globingene, were included in the same tube. The final 100-pl PCRreaction mixture contained 5 p,l of the template, 50 mM KCl,10 mM Tris (pH 8.3), 6.5 mM MgCl2, 1% Laureth-12 (MazerChemicals, Gurnee, Ill.), 200 puM (each) deoxyribonucleo-tide triphosphates, 2.5 U of AmpliTaq recombinant Taqpolymerase (Cetus Corporation, Norwalk, Conn.), each Liprimer at a concentration of 0.5 mM, and each 3-globinprimer at a concentration of 0.05 mM. Each test included asingle tube with approximately 5 ng of SiHa cell DNA (about500 genome equivalents) or dilutions of SiHa cell DNA, as apositive control for amplification by Li and f-globin prim-ers. Thirty-five amplification cycles were performed in aDNA thermal cycler (Perkin-Elmer Cetus Instruments), byusing thermocycle step parameters of 94°C for 1 min, 55°Cfor 1 min, and 72°C for 1 min. The reaction products (10 p.l)were electrophoresed in a 3% agarose gel (2% Nusieve and

J. CLIN. MICROBIOL.

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

COMPARISON OF METHODS FOR HPV IDENTIFICATION

TABLE 1. Numbers and sources of specimens examined bythree HPV diagnostic methods: VP, SH, and PCR

Results by:

Specimen type Mean age VP SH PCRand source (years) of SH_PCRsubjects No. % Posi- No. % Posi- No. % Posi-

tested tive tested tive tested tivea

Invasive cancerCases 50 368 38.6 368 31.8 302 69.0Controls 49 334 3.3 334 5.4 281 9.1

CIN3Cases 37 343 31.8 159 _bControls 38 489 8.2 65

Total 43 1,534 926 583a For percentage of positivity by PCR, 55 specimens which tested negative

for both ,B-globin and HPV were excluded.b_, Percentage of positivity was not reported because these specimens

composed a selected sample based on VP results.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

A aB I P 44

FIG. 1. Autoradiograph of PCR products hybridized withHPV-16 oligomer probe. The hybridizing fragment was approxi-mately 450 bp in size. In row A, all three dilutions of SiHa cell DNArepresenting 10, 100, and 1,000 genome equivalents of SiHa cells(lanes 1, 2, and 3, respectively) were positive. Lanes 4 to 10 were

clinical specimens containing known HPV types (HPV-35, lane 4;HPV-33, lane 5; HPV-31, lane 6; HPV-18, lane 7; HPV-16, lane 8;HPV-6, lane 9; and HPV-11, lane 10); only the specimen containingHPV-16 (lane 8) was positive. In row B, negative controls (lanes 7and 15) were negative; all other lanes contained clinical specimens,of which five (lanes 1, 2, 11, 13, and 17) were positive for HPV-16.

1% ME agarose; FMC Corporation, Rockland, Maine),visualized by ethidium bromide staining, and transferred bycapillarity to nylon membranes. The filters were hybridizedsequentially to end-labelled type-specific oligomer probes(19 to 22 bases) of HPV-6, HPV-11, HPV-16, HPV-18,HPV-31, HPV-33, and HPV-35 and ,-globin (1) under high-stringency conditions (Tm - 15'C) for 1 h. Subsequent to thestringent hybridization with specific probes, the filters werescreened with an HPV generic probe, which was a mixtureof amplimers of HPV-16 and HPV-18. The amplimers for thegeneric probe were synthesized, as described previously (1),by using nested primer pairs, MY74 and MY75 for HPV-16and MY76 and MY77 for HPV-18. This generic probe hasbeen shown to be capable of detecting all the HPV typesincluded in the VP, as well as many additional types, but itdetects a narrower spectrum of HPVs than did the four-amplimer generic probe used by Bauer et al. (1, 13a). Theprimers for the generic probe were kindly supplied byMichele Manos. Specimens hybridizing with the genericprobe, but not with the specific probes, were characterizedas "HPV-PCR, type unknown."

Statistical analysis. For descriptive analyses, the SASprogram was used (20). Comparisons between proportionsas well as tests for linear trends were done by using theSTATCALC module of the Epi Info program (4). Fisher'sexact test was performed when at least one expected cellvalue was less than 5, otherwise the uncorrected chi-squaretest was employed. For all analyses, an alpha value of 0.05was used to establish statistical significance and, whenappropriate, 95% confidence intervals were computed forproportions.

RESULTS

The numbers and sources of the specimens and the meanages of the case and control subjects are listed in Table 1. Allspecimens (n = 1,534) were tested by VP. All specimensfrom invasive cases and controls (n = 702) were alsoexamined by SH. PCR was performed for all invasive casesand controls for whom a second aliquot was available (n =583). In addition, SH was performed with 224 specimensfrom CIN3 cases and controls; these comprised 199 speci-mens (148 cases and 51 controls) which gave any signal in VP(score of 1 to 5) and 25 specimens (11 cases and 14 controls)

which gave no signal in VP (score of 0). All tests wereperformed in a blinded manner with respect to source ofspecimen and results of the other assays. For all three tests,HPV prevalence rates were much higher for cases than forcontrols (Table 1).

Test performance. In SH tests, the HPV standard of 10 pgwas always detected, that of 1 pg was detected in mostinstances, and that of 0.1 pg was detected rarely. The probeshybridized only with the homologous HPV standards. InPCR tests (Fig. 1), HPV-negative controls were consistentlynegative for HPV and ,-globin amplification. Positive con-trols (SiHa cells) were consistently positive for both HPVand ,3-globin amplification. Of 583 specimens tested by PCR,,B-globin was amplified in 510 (88%). Eighteen of the 733-globin-negative specimens (17 of 45 case specimens andone of 28 control specimens) were HPV positive; these wereincluded in the analysis. The fifty-five 3-globin-negative,HPV-negative specimens were excluded when test resultswere compared.SH and PCR results by VP scores. The correlations be-

tween VP score and SH data (based on 926 specimens) andbetween VP score and PCR data (based on 528 specimens)are presented in Table 2. For case specimens, HPV positiv-ity by SH and by PCR increased with increasing signalintensity of VP. HPV positivity by SH increased from 10%for specimens which gave no signal in VP (score of 0) to 74to 79% for specimens with high VP signals (scores of 4 and5) (P for trend, <0.0001) but did not reach 100% for any ofthe VP scores. Although scores of both 0 and 1 are inter-preted as negative by VP, the positivity by SH for the scoreof 0 (10%) was clearly different from that for the score of 1(36%) (P, <0.0001). This indicates that in case specimens aVP signal that was less intense than that of the low-positivecontrol was often due to the presence of HPV. In a compar-ison of VP and PCR, a majority of case specimens that werenegative by VP (score of 0 or 1) and nearly 100% ofspecimens with VP scores of 4 or 5 were positive by PCR (Pfor trend, <0.0001).A relatively small proportion of the control specimens was

positive in any of the three tests. Positivity by SH increasedwith an increasing VP score (P, <0.0001) and was signifi-cantly higher in control specimens that were positive by VP(29%) than in control specimens that were negative by VP

VOL. 30, 1992 2953

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

2954 GUERRERO ET AL.

TABLE 2. HPV positivity as measured by SH and PCR by VPscores among cases and controls

SH result for a: PCR result for b:

VP Cases Controls Cases Controlsscore %%No. Posi- No. Posi- No. Pos NO. Posi-tested tiv tested tested ts tested t

tive tive tive tive

0 202 10.4 319 5.6 152 54.6 231 9.51 77 36.4 29 3.4 22 63.6 14 7.12 43 51.2 22 31.8 19 84.2 3 0.03 101 73.3 16 12.5 26 84.6 4 0.04 38 78.9 9 44.4 18 100.0 1 0.05 66 74.2 4 50.0 37 97.3 1 0.0

Total 527 _c 399 - 274 69.0 254 9.1

P for trend <0.0001 <0.0001 <0.0001

a Percent positive results include those specimens which were positive bylow-stringency and/or high-stringency conditions.

b Percent positive results include those specimens which were positive bytype-specific probes and/or generic probe.c-, total percent positivity is not reported because these specimens

composed a selected sample based on VP results.

(5.5%) (P, <0.0001). Surprisingly, this was not true forpositivity by PCR, which was 0 for nine controls that werepositive by VP but 9.4% (23 of 245) for controls that werenegative by VP. This difference was not statistically signifi-cant (P = 0.42).Specimens positive by VP (score of 2 to 5) were confirmed

more often when they were derived from cases than whenthey were derived from controls. SH confirmed 71% of thecases that were positive by VP and 29% of the controls thatwere positive by VP (P, <0.0001), and PCR confirmed 92%of the cases that were positive by VP but none of the ninecontrols that were positive by VP.HPV diagnosis by SH and PCR. In two of the three tests

employed (SH and PCR), HPV-positive specimens wereclassified as having one of six specific types (16, 18, 31, 33,35, or 6 and/or 11) or as having HPV, type unknown. Thesediagnoses for invasive carcinoma case and control speci-mens (702 specimens by SH and 528 specimens by PCR) areshown in Table 3. HPV-16 was the predominant type iden-tified in cases by both tests. It was found in 20.9% of thecases by SH and in 48.2% of the cases by PCR; it constituted

TABLE 3. SH and PCR diagnoses for invasive carcinoma casesand controls

Positive result by:

SH PCRHPV type

Cases Controls Cases Controls

No. % No. % No. % No. %

16 77 20.9 1 0.3 132 48.2 14 5.518 5 1.4 2 0.6 11 4.0 2 0.831 6 1.6 0 0.0 10 3.6 1 0.433 4 1.1 0 0.0 6 2.2 1 0.435 3 0.8 0 0.0 5 1.8 1 0.46/11 0 0.0 0 0.0 0 0.0 1 0.4Unknown 22 6.0 15 4.5 25 9.1 3 1.2Negative 251 68.2 316 94.6 85 31.0 231 90.9

Total 368 100.0 334 100.0 274 100.0 254 100.0

TABLE 4. Distribution of invasive carcinoma cases and controlsby their HPV status by VP, SH, and PCR

Specimen with indicated result for:Result by:

Cases Controls

PCRa SHb VP No. % No. % Totalno.

+ + + 58 22.0 0 0.0 58- - - 66 25.1 207 83.8 273

+ - - 81 30.8 18 7.3 99+ + - 12 4.6 3 1.2 15+ - + 34 12.9 0 0.0 34

- + + 3 1.1 1 0.4 4- + - 4 1.5 10 4.0 14- - + 5 1.9 8 3.2 13

Total 263 100.0 247 100.0 510a PCR positive includes those specimens which were positive by type

specific probes and/or generic probe.b SH positive includes those specimens which were positive under low

stringency and/or high stringency conditions.

83.7% of the types identified by SH and 80.5% of the typesidentified by PCR. HPV-18 was encountered relatively rare-ly; it constituted 5.4% of the types identified by SH and 6.7%of the types identified by PCR. Among controls, only alimited number of specimens were shown as having a spe-cific HPV type: three specimens (0.9%) by SH and 20specimens (7.9%) by PCR. HPV-SH, type unknown classi-fications were detected frequently in both case and controlspecimens, but HPV-PCR, type unknown classificationswere found mostly in case specimens.

Analysis of specimens examined by all three tests. Thedistribution of the 510 specimens from invasive carcinomacases and controls by their HPV diagnoses in the three tests(HPV present or absent, regardless of type) is shown inTable 4. These specimens tested satisfactorily by all threemethods. Complete agreement between results of the threetests was found for 47.1% of the case specimens (22%positive and 25.1% negative) and for 83.8% of the controlspecimens (all negative). An additional 48.3% of the casespecimens were positive by PCR: 30.8% by PCR alone and17.5% by PCR and either VP or SH. VP and/or SH resultswere positive for 4.5% of the case specimens, for which thePCR result was negative: 1.9% by VP alone, 1.5% by SHalone, and 1.1% by both VP and SH. Thus, 74.8% of thecases were positive by at least one test and 69.3% werepositive by PCR. Eight and one-half percent of the controlspecimens were positive by PCR: 7.3% by PCR alone and1.2% by PCR and SH. VP and/or SH results were positivefor 7.6% of the control specimens, for which the PCR findingwas negative: 3.2% by VP alone, 4% by SH alone, and 0.4%by VP and SH. Thus, 16.2% of the controls were positive byat least one test and 8.5% were positive by PCR. The PCRgave positive results for 58 of 61 (95%) cases which werepositive by VP and SH, 34 of 39 (89%) cases which werepositive by VP but negative by SH, 12 of 16 (75%) caseswhich were negative by VP but positive by SH, and 81 of 147(55%) cases which were negative by both VP and SH. Withcontrol specimens, PCR was positive for 3 of 22 (14%)specimens which were positive either by VP alone (8 spec-imens) or by SH alone (13 specimens) or by both (1 speci-

J. CLIN. MICROBIOL.

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

COMPARISON OF METHODS FOR HPV IDENTIFICATION 2955

Complete agreement

Positive by PCR alone

Positive by PCR & VP

Positive by PCR & SH

Positive by SH alone

Positive by VP alone

Positive by VP & SH

64.9

19.4

2.9

27

2.5

0.8

0 10 20 30 40 50 60 70

Percent of specimens

FIG. 2. Distribution of the invasive carcinoma cases and controls by results of the three HPV diagnostic methods.

men) and for 18 of 225 (8%) specimens which were negativeby VP and SH.The contribution of the different tests to the results for the

510 specimens (cases and controls combined) is shown inFig. 2. If specimens positive by VP alone or by SH alone areconsidered equivocal and excluded from analysis, then thePCR confirmed all but 4 (0.8%) of the 510 specimen results.The degree of agreement between type-specific diagnoses

by SH and PCR is shown in Table 5. Among cases, type-specific diagnosis was made by both SH and PCR in 58instances. For 50 of these cases (86%), the diagnoses wereidentical, and for an additional 5 cases, the types identifiedby the two tests, although different, were in the samesubgroup of viruses (HPV-16, HPV-31, HPV-33, and HPV-35) known to be closely related (17). In the remaining threeinstances, the specimens were identified as having HPV-16by one test and HPV-18 by the other. There was littleagreement between the classifications HPV-SH, unknowntype and HPV-PCR, unknown type. Among specimensdiagnosed as HPV-SH, unknown type, PCR was positive for

TABLE 5. Concordance between HPV diagnoses by PCR andSH for invasive carcinoma cases and controls

Specimen sources HPV trpe by PCRand HPV type Total

by SH 16 18 31 33 35 6/11 Unknown Negative

Cases16 46 1 1 1 1 2 5218 2 2 431 1 2 333 0 1 135 1 1 0 26/11 0 0Unknown 7 4 4 15Negative 72 8 7 4 5 19 71 186

Total (cases) 129 11 10 6 5 0 24 78 263

ControlsUnknown 2 1 11 14Negative 12 2 1 1 1 1 215 233

Total 14 2 1 1 1 1 1 226 247(controls)

11 of 15 (73%) case specimens but for only 3 of 14 (21.4%)control specimens. Overall, complete agreement betweenthe two tests was found for 125 (71 negative and 54 positive)of the 263 case specimens (47.5%) and for 215 (all negative)of the 247 control specimens (87.0%).The characteristics of the 31 specimens (12 cases and 19

controls) positive by VP or SH but not confirmed by PCRwere as follows. Of the 12 case specimens, 5 were positiveonly by VP, 4 were positive only by SH at low stringencyand 3 were positive by SH at high-stringency conditions aswell as by VP. Of the 19 control specimens, 8 were positiveby VP alone, 10 were positive only by SH at low-stringencyconditions, and 1 was positive by VP as well as by SH atlow-stringency conditions. Most of the specimens that werepositive by SH but that were not confirmed by PCR werepositive by SH only at low-stringency conditions.

Multiple infections. In the analysis described above, spec-imens with multiple infections by SH (13 specimens) and byPCR (5 specimens) were counted as having only one HPVtype; the specimen was considered to be positive for the typewhich was more prevalent in the study. All but four multipleinfections were found in case specimens. The distribution ofspecimens positive for multiple types of HPV infections bySH was as follows: four for types 16 and 31; three for types16 and 18; two for types 16 and 33; one for types 16 and 35;one for types 6 and/or 11 and 35; one for types 16, 31, and 33;and one for types 16, 18, and 35. By PCR, specimens werepositive as follows: three for types 16 and 33, one for types16 and 31, and one for types 16 and 6 and/or 11. None of thespecimens which tested positive for more than one type bySH tested positive for multiple types by PCR. Seventeenspecimens in which HPV-16 was detected as one of the typeswere classified as being type 16 positive. One specimen thattested positive for types 6 and/or 11 and 35 was classified asbeing type 35 positive.

Effect of blood in case specimens on HPV identification.Many of the specimens contained blood on visual examina-tion. The insert in the VP kit warns that the presence ofblood in the specimen may produce false-negative resultsand blood is also known to inhibit PCR amplification (10).We therefore examined whether the presence of bloodaffected the test results. The spots on VP filters where the

VOL. 30,1992

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

2956 GUERRERO ET AL.

TABLE 6. Correlation of VP and SH results for presence ofblood in case specimens of invasive carcinoma

Specimen with blood Specimen withoutblood

VP PfrXscore No. No. (%) No. No. (%) Pfor x2

tested positive tested positivebySH bySH

0 69 3 (4.3) 122 13 (10.7) 0.1311 16 6 (37.5) 19 4 (21.1) 0.2832 13 5 (38.5) 16 7 (43.7) 0.7743 20 14 (70.0) 22 13 (59.1) 0.4614 11 8 (72.7) 12 8 (66.7) 0.7525 21 16 (76.2) 27 20 (74.1) 0.867

Total 150 52 (34.7) 218 65 (29.8) 0.326

bloody specimens were placed appeared green, rather thanthe conventional blue. The percentages of specimens thatcontained blood, as judged by green spots on VP filters, were41, 13, and 4%, respectively, for invasive cases, CIN3 cases,and controls. The effect of blood on VP results was exam-ined by comparing positivity by SH for each VP score forcase specimens with and without the presence of blood. Thepresence of blood in the specimen is not expected toinfluence the results of SH because the specimen DNA ispurified for the SH procedure. There was no evidence thatthe presence of blood affected positivity by VP for casespecimens; positivity rates by SH for specimens with andwithout blood were comparable for each VP score (Table 6).A similar analysis was not conducted for control specimensbecause few of them were positive by VP or containedblood. Also, in our study, in which the original specimenswere diluted 1:400 for PCR, the presence of blood in theoriginal specimen seemed to have no demonstrable effect onP-globin amplification in case or control specimens. Amongcase specimens, 118 of 140 specimens (84.3%) with bloodwere positive for ,-globin compared with 139 of 162 speci-mens (85.8%) without blood. Among control specimens, 11of 12 specimens (91.7%) with blood were positive forP-globin compared with 242 of 269 (90.0%) without blood.

Relationship between 13-globin amplification and HPV pos-itivity. The HPV positivity rate was higher for P-globin-positive case specimens (69% of 257 specimens) comparedwith 3-globin-negative case specimens (38% of 45 speci-mens) (P, <0.001). Similarly, for control specimens, HPVpositivity in 3-globin-positive specimens (8.7% of 253 spec-

imens) was greater than in P-globin-negative specimens(3.6% of 28 specimens), but this difference was not statisti-cally significant (P = 0.3).Amount of cellular DNA and HPV positivity. The amounts

of cellular DNA were measured by spectrofluorometry forspecimens which were tested by SH. The median amountand interquartile range of cellular DNA for 527 case speci-mens (1,850 ng and 560 to 5,760 ng, respectively) was

significantly greater than those for 399 control specimens(1,100 ng and 320 to 2950 ng, respectively) (P, <0.001). Also,28% of the case specimens, but only 14% of the controlspecimens, had greater than 5,000 ng of DNA (P, <0.0001).Therefore, we analyzed whether HPV positivity of case andcontrol specimens was affected by the amount of cellularDNA (Table 7). The amounts of cellular DNA used for SHand PCR were, respectively, 80 and 1% of those used for VP.Among case specimens, the proportions of specimens posi-tive by VP and SH with the lowest amount (<100 ng) ofcellular DNA were 19 and 9%, respectively; these valueswere significantly lower than those for specimens withhigher amounts of DNA (P, <0.005 for VP and <0.001 forSH). For specimens with more than 100 ng of DNA, theHPV positivity rates for case specimens by VP did notincrease significantly with increasing amounts of cellularDNA (P = 0.23), but they did for SH (P = 0.02). In contrastto these findings by VP and SH, the amount of cellular DNAdid not influence the rate of HPV positivity by PCR. Theseresults suggest that HPV positivity by VP and SH may havebeen underestimated for case specimens with low amountsof cellular DNA. There were few HPV-positive controlspecimens, but by both VP and SH, there was a significanttrend of increasing HPV positivity with increasing amountsof specimen DNA. As with case specimens, the HPVpositivity in control specimens by PCR was not affected byamounts of cellular DNA.

Sensitivity and specificity. In view of the effect of cellularDNA amount on HPV positivity by SH and VP, estimates ofsensitivity and specificity were made after stratification byDNA amount. Only those specimens which were examinedby all three tests were considered; VP was compared withSH and both VP and SH were compared with PCR. Esti-mates were made separately for cases and controls (Table 8).For cases, VP had an identical overall value of 79% for bothsensitivity and specificity against SH as the reference test;against PCR, both VP and SH showed high levels of speci-ficity (87 to 100%), but at each cellular DNA amountcategory, VP had a higher sensitivity than SH, with anoverall value of 50% for VP versus 38% for SH. Both VP and

TABLE 7. HPV positivity by VP, SH, and PCR for cases and control by amount of cellular DNA

Case result by Control result by:

Cellular DNA VP SH PCR VP SH PCRamt" (ng)

No. % Posi- No. % Posi- No. % Posi- No. % Posi- No. % Posi- No. % Posi-tested tive tested tive tested tive tested tive tested tive tested tive

'100 32 18.7 32 9.4 19 63.2 41 2.4 41 0.0 25 4.0101-500 84 42.9 84 33.3 46 76.1 90 6.7 90 1.1 60 6.7501-1,000 68 44.1 68 36.8 33 63.6 57 8.8 57 3.5 36 11.1

1,001-5,000 193 52.8 193 51.3 90 65.6 155 16.1 155 11.6 98 7.1>5,000 150 49.3 150 46.0 75 77.3 56 25.0 56 23.2 28 17.9

Total 527 47.1 527 42.5 263 70.3 399 12.8 399 8.5 247 8.5P for trend <0.005 <0.001 0.50 <0.001 <0.001 0.17

a Amount estimated for specimens for SH.

J. CLIN. MICROBIOL.

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

COMPARISON OF METHODS FOR HPV IDENTIFICATION 2957

TABLE 8. Sensitivity and specificity of VP with respect to SH and PCR and of SH with respect to PCR by cellular DNA amount

Specimen source and cellular No. of VP vs SH VP vs PCR SH vs PCRDNA amt (ng) specimnenstested Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity

Cases5100 19 89 (76-100)a 17 (0-38) 100 0 100

101-1,000 79 86 (71-100) 79 (69-90) 48 (35-61) 87 (73-100) 36 (23-48) 96 (87-100)1,001-5,000 90 81 (68-93) 80 (69-90 61 (49-73) 87 (75-99) 54 (42-67) 87 (75-99)

>5,000 75 70 (50-90) 75 (63-86) 47 (34-59) 94 (83-100) 31 (19-43) 88 (73-100)All cases 263 79 (70-88) 79 (73-85) 50 (43-57) 90 (83-96) 38 (31-45) 91 (85-97)

Controls5100 25 100 0 100 0 100

101-1,000 96 0 96 (92-100) 0 95 (91-100) 0 99 (97-100)1,001-5,000 98 13 (0-35) 96 (91-100) 0 95 (90-99) 14 (0-40) 92 (87-98)

>5,000 28 0 100 0 100 40 (0-83) 87 (73-100)All controls 247 7 (0-21) 97 (94-99) 0 96 (93-99) 14 (0-29) 95 (92-98)

All cases and controls 510 68 (59-78) 89 (86-92) 45 (38-51) 94 (92-97) 35 (29-42) 94 (91-97)

a 95% confidence interval range is given in parentheses.

SH had low sensitivity for cellular DNA amounts of <100ng. For controls, the estimated specificity was high (87 to100%) in all comparisons. The estimate of sensitivity forcontrols was imprecise because few specimens were posi-tive. Nevertheless, in all comparisons, sensitivity for con-trols was lower than that for cases, raising the possibility ofdifferential misclassification.

DISCUSSION

The major goal of these studies was to provide reliableHPV diagnoses in the epidemiological investigation of cer-vical cancer in Colombia and Spain. In view of the knownintra- and intertest variability of the diagnostic procedures, alarge number of specimens from invasive carcinoma casesand controls were examined by each of three procedures:VP, which is commercially available; SH, which is consid-ered the gold standard; and PCR, which is capable ofdetecting very small amounts of virus. The goal of providinga reliable HPV diagnosis was largely met. The differences inHPV prevalence rates between cases and controls werehighly significant when judged by the results of any of thethree tests individually or by any combination of the resultsof the three tests. The low HPV prevalence rate in controlswas confirmed by all three tests. The predominance ofHPV-16 and the relative paucity of HPV-18 were confirmedby both PCR and SH, the two tests capable of providingtype-specific diagnosis. Thus, virologic examination of thespecimens by three tests allowed a confident interpretationof the data in the case-control study (1Sa).As expected, HPV detection rate was higher among

0-globin-positive specimens than among P-globin-negativespecimens. In the test system, the HPV primer concentra-tions are higher than ,-globin primer concentrations so as tofavor HPV amplification over 3-globin amplification (13a). Itwas therefore not surprising that HPV but not 3-globin wasamplified from some specimens. Eighteen 3-globin-negative,HPV-positive specimens (17 cases and 1 control) wereincluded in the analysis in order to minimize misclassifica-tion of true positives. If these cases were excluded, the HPVprevalence rate would have decreased from 69 to 67% amongcases but would have remained unaffected among controls.HPV diagnosis by PCR provided the highest prevalence

rates in both cases and controls. PCR rarely failed to confirm

the presence of HPV in specimens which were positive bySH under high-stringency conditions or specimens whichwere positive by both VP and SH. On the other hand, it oftendid not confirm the presence of HPVs in specimens whichwere positive only by VP or only by SH under low-strin-gency conditions; these cases may represent false-positivediagnoses by VP and SH. The failure of PCR to confirmspecimens positive by SH and VP was more frequent forcontrols than for cases. This raised the possibility of differ-ential misclassification, but we did not have a large enoughnumber of positives among control specimens to furtherevaluate this question.The inhibitory action of blood on PCR amplification is

clearly established (10). However, in our study, the bloodcontaminating the cervical scrapes appeared not to inhibitamplification probably because the specimens were diluted1:400 prior to being tested. Since a significant proportion ofthe case specimens were bloody, the PCR component of thestudy would have been compromised if bloody specimenshad to be excluded from analysis. Two additional commonlyexpressed concerns about PCR were not encountered in thisstudy. First, in the numerous controls interspersed through-out the tests to monitor contamination, not a single instanceof contamination was discovered. Second, the suspicion thatHPV infections are so persistent and widespread that ahighly sensitive technique like PCR will detect clinicallyirrelevant HPVs in a large number of controls was not borneout. The low HPV prevalence seen in our controls (meanage, 49 years) is comparable to that found in other PCR-based investigations of populations in similar age groups(23a). While PCR provided the most valuable data for thestudy, several inadequacies of the test were also evident.First, P-globin could not be amplified from 12% of thespecimens. Second, there was a 14% disagreement betweenPCR and SH when specific types were identified by bothtests, and PCR and SH results did not identify the samespecimens as having multiple infections. Disagreements ofsimilar magnitude have been noted in other studies of PCRand SH comparisons (21). In three instances, samples scoredHPV positive and typed by SH were scored negative byPCR. These disagreements may result from misdiagnosis ineither test, presence of multiple infections, or technicalerrors such as mislabeling of specimens. Third, the 69%HPV prevalence rate in cases by PCR is probably an

VOL. 30, 1992

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

2958 GUERRERO ET AL.

underestimate. In a study in progress, HPVs have beenrecovered from approximately 50% of the biopsy tissues ofinvasive cases of this study whose cervical scrapes wereHPV negative, by using identical amplification and detectionsystems (22a). Fourth, the generic probe available to usdetected a narrower spectrum of HPV types than the probeused by Bauer et al. (1). Therefore, some HPV types mayhave escaped detection. Underdetection of HPVs makes itdifficult to assess the role of other risk factors of cervicalcancer (la).

In previous investigations, HPV detection rates in casespecimens by VP and SH have varied over a wide range. Ourdetection rate of 38.6% for VP and 31.8% for SH are in thelower part of that range. The type-specific SH tests weresatisfactory in that the HPV standard of 1 pg of viral DNA(equivalent to 105 copies of viral genome) was almost alwaysdetectable. As described in Materials and Methods, SHunder low-stringency conditions was not always satisfac-tory. A factor that contributed partially to the lower detec-tion rates by VP and SH was the low amount of cellularDNA for some specimens. An analysis of HPV DNA posi-tivity by amount of cellular DNA showed that HPV preva-lence by both SH and VP was underestimated for the lowestcellular DNA category of < 100 ng (which accounted for lessthan 10% of the specimens). At other categories of cellularDNA amounts, there was some underestimation of HPVpositivity by SH but not by VP. However, in our study, theHPV detection rate by VP and SH did not approach that ofPCR even when 5,000 ng of cellular DNA per specimen wastested. In this respect, our results differ from those ofSchiffman (21), who found that for case specimens SH wasnearly as efficient as PCR for HPV diagnosis. In comparisonwith PCR, VP had a greater sensitivity than SH for allcategories of cellular DNA amounts of case specimens. Ourfinding of a much higher detection rate by PCR comparedwith that of VP is similar to those of other studies in whichthese two tests were compared (1, 3, 7; for a review, seereference 9).

It has been reported that cases have higher amounts ofHPV DNA than controls and that the amount of HPV DNAin the genital tract specimen is a risk factor for preinvasiveand invasive disease (15, 18). The amount of HPV DNA wasestimated in VP by the signal intensity and in SH by theestimation of copy number of genome per cell. These datafor the Colombia-Spain study will be reported elsewhere. Itwas also anticipated that SH data may allow the classifica-tion of some hybridization patterns (e.g., with fragments ofunexpected size) as suggestive of integration. Such patternswere seen in a few instances in HPV-16-positive samplesfrom invasive cases. However, in most cases, the determi-nation of the physical state of the viral DNA is not straight-forward (5) and could not be accomplished on the basis ofthe digestion pattern with a single enzyme.

In a comparison of VP and SH, the sensitivity andspecificity estimates for VP against SH for cases and con-trols combined were 68 and 89%, respectively, for the 510specimens tested by all three methods. This specificityestimate is probably lower than the true value becausespecimens that were positive by VP but negative by SH arenot all false positives. As shown in Table 3 for specimensexamined by all three tests, the PCR positivity for cases thatwere positive by VP but negative by SH is higher than thatfor cases that were negative by both VP and SH (34 of 39versus 81 of 147, P, <0.001), an indication that some of thespecimens in this group may have been true positives missedby SH. Kiviat et al. (11) have reported a higher sensitivity

and specificity (both greater than 90%) of VP, as measuredagainst the gold standard of SH, than those estimated in ourstudy. One reason for their higher values may be that theyperformed the two tests using identical reagents (includingRNA probes), whereas we performed SH using DNA probesand reagents not identical with those in the VP kit.

In conclusion, the results of our study strongly imply thatPCR-based HPV diagnosis is the method of choice for futureepidemiological investigation. However, there is a strongneed to validate and standardize reagents for PCR tests andto have one or more reference centers which can assist inquality control of the test. The results of PCR tests areheavily dependent on the choice of primers and probes (22).It is necessary that different PCR protocols now in use (1, 8,24) be compared with each other in tests of clinical speci-mens. Questions that need attention include the best way toprocess a cervical scrape sample for PCR, the comparabilityof amplification by consensus primers with that by type-specific primers, whether employment of more than oneprimer pair in the same reaction tube decreases the sensitiv-ity of HPV detection, the differences in detection of HPVtypes by the different generic probes, validity of the type-specific diagnosis of PCR products with the oligomer probes,and development of reagents to identify HPVs which arenow classified as unknown types. In addition to the PCR-based test, it would also be useful in future studies to test atleast a subset of the specimens with a second test which isnot amplification based and which can provide type-specificdiagnosis as well as an estimate of the quantity of HPV in thesample. In this way, it will be possible to monitor thePCR-based diagnosis when both tests are positive and toevaluate the importance of virus load in distinguishing be-tween cases and controls.

ACKNOWLEDGMENTS

We are deeply grateful to Attila L6rincz and Lutz Gissmann forvaluable advice during the course of the study and to A. Lorincz forproviding reagents and technical advice for Southern hybridizationassays. We are also grateful to Michele Manos for assistance insolving problems in the PCR test.

This study was supported in part by U.S. Public Health Service(grants P01 AI16959 and R03 CA52543), by the European Commu-nity (grant CI 1-0371-F [CD]) and the Fondo de InvestigacionesSanitarias (FIS) of the Spanish government (grants 86/753, 87/1513,88/2049, and 90/0901), and by the IARC.

REFERENCES1. Bauer, H., Y. Ting, C. E. Greer, J. C. Chambers, C. J. Tashiro,

J. Chimera, A. Reingold, and M. M. Manos. 1991. Genitalhuman papillomavirus infection in female university students asdetermined by a PCR-based method. J. Am. Med. Assoc.265:472-477.

la.Bosch, F. X., et al. Risk factors for cervical cancer in Colombiaand Spain. Int. J. Cancer, in press.

2. Brandsma, J., R. D. Burk, W. D. Lancaster, H. Pfister, andM. H. Schiffman. 1989. Inter-laboratory variation as an expla-nation for varying prevalence estimates of human papillomavi-rus infection. Int. J. Cancer 43:260-262.

3. Burmer, G. C., J. D. Parker, J. Bates, K. East, and B. G.Kulander. 1990. Comparative analysis of human papillomavirusdetection by polymerase chain reaction and ViraPap/ViraTypekits. Am. J. Clin. Pathol. 94:554-560.

4. Centers for Disease Control. 1991. Epi info. USD, Inc., Centersfor Disease Control, Atlanta.

5. Cullen, A. P., R. Reid, M. Campion, and A. T. Lbrincz. 1991.Analysis of the physical state of different human papillomavirusDNAs in intraepithelial and invasive cervical neoplasm. J.Virol. 65:606-612.

J. CLIN. MICROBIOL.

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

COMPARISON OF METHODS FOR HPV IDENTIFICATION 2959

6. Franco, E. L. 1991. The sexually transmitted disease model forcervical cancer: incoherent epidemiologic findings and the roleof misclassification of human papillomavirus infection. Epi-demiology 2:98-106.

7. Gravitt, P., A. Hakenewerth, and J. Stoerker. 1991. A directcomparison of methods proposed for use in widespread screen-ing of human papillomavirus infections. Mol. Cell. Probes5:65-72.

8. Gregoire, L., M. Arella, J. Campione-Piccardo, and W. D.Lancaster. 1989. Amplification of human papillomavirus DNAsequences by using conserved primers. J. Clin. Microbiol.27:2660-2665.

9. Guerrero, E., and K. V. Shah. 1991. Polymerase chain reactionin HPV diagnosis. Papillomavirus Rep. 2:115-118.

10. Higuchi, R. 1989. Simple and rapid preparation of samples forPCR, p. 31-38. In H. A. Erlich (ed.), PCR technology. M.Stockton Press, New York.

11. Kiviat, N. B., L. A. Koutsky, C. W. Critchlow, D. A. Galloway,D. A. Vernon, M. L. Peterson, P. E. McElhose, S. J. Pendras,C. E. Stevens, and K. K. Holmes. 1990. Comparison of Southerntransfer hybridization and dot filter hybridization for detectionof cervical human papillomavirus infection with types 6, 11, 16,18, 31, 33, and 35. Am. J. Clin. Pathol. 94:561-565.

12. Koutsky, L. A., D. A. Galloway, and K. K. Holmes. 1989.Epidemiology of genital human papillomavirus infection. Epi-demiol. Rev. 10:122-162.

13. Kwok, S., and R. Higuchi. 1989. Avoiding false positives withPCR. Nature (London) 339:237-238.

13a.Manos, M. Personal communication.14. Manos, M. M., Y. Ting, D. K. Wright, A. J. Lewis, T. R. Broker,

and S. M. Wolinsky. 1989. The use of polymerase chain reactionamplification for the detection of genital human papillomavi-ruses. Cancer Cells 7:209-214.

15. Morrison, E. A. B., G. Y. F. Ho, S. H. Vermund, G. L.Goldberg, A. S. Kadish, K. F. Kelley, and D. Burk. 1991. Humanpapillomavirus infection and other risk factors for cervicalneoplasia: a case-control study. Int. J. Cancer 49:6-13.

15a.Munfoz, N., et al. The causal link between human papillomavirusand invasive cervical cancer: a population-based case-controlstudy in Colombia and Spain. Int. J. Cancer, in press.

16. Mufioz, N., X. Bosch, and J. M. Kaldor. 1988. Does humanpapillomavirus cause cervical cancer? The state of the epidemi-ological evidence. Br. J. Cancer 57:1-5.

17. Pfister, H. 1990. General introduction to papillomaviruses, p.1-9. In H. Pfister (ed.), Papillomaviruses and human cancer.

CRC Press, Boca Raton, Fla.17a.Rawls, W. Personal communication.18. Reeves, W. C., L. A. Brinton, M. Garcia, M. M. Brenes, R.

Herrero, E. Gaitan, F. Tenorio, R. C. deBritton, and W. E.Rawls. 1989. Human papillomavirus (HPV) infection and cervi-cal cancer in Latin America. N. Engl. J. Med. 320:1437-1441.

19. Resnick, R. M., M. T. E. Cornelisssen, D. K. Wright, G. H.Eichinger, H. S. Fox, J. ter Schegget, and M. M. Manos. 1990.Detection and typing of human papillomavirus in archival cer-vical cancer specimens by DNA amplification with consensusprimers. J. Natl. Cancer Inst. 82:1477-1484.

20. SAS. 1986. SAS user's guide: statistics, 5th ed. SAS InstituteInc., Cary, N.C.

21. Schiffman, M. H. 1992. Validation of hybridization assays:correlation of filter in situ, dot blot, and PCR with Southern blot.In N. Mufioz, F. X. Bosch, K. V. Shah, and A. Meheus (ed.),The epidemiology of human papillomavirus and cervical cancer.International Agency for Research on Cancer publication no.119. International Agency for Research on Cancer, Lyon,France.

22. Schiffman, M. H., H. M. Bauer, A. T. LMrincz, M. M. Manos,J. C. Byrne, A. G. Glass, D. M. Cadell, and P. M. Howley. 1991.Comparison of Southern blot hybridization and polymerasechain reaction methods for the detection human papillomavirusDNA. J. Clin. Microbiol. 29:573-577.

22a.Shah, K. Unpublished data.23. Shah, K. V., and L. Gissmann. 1989. Papovaviruses, p. 1067-

1102. In N. J. Schmidt and R. W. Emmons (ed.), Diagnosticprocedures for viral, rickettsial and chlamydial infections, 6thed. American Public Health Association, Washington, D.C.

23a.van den Brule, A. J. C. 1991. Ph.D. thesis. Free UniversityHospital, Amsterdam.

24. van den Brule, A. J. C., P. J. F. SnUders, R. L. J. Gordijn, 0. P.Bleker, C. J. L. M. MeUer, and J. M. M. Walboomers. 1990.General primer-mediated polymerase chain reaction permits thedetection of sequenced and still unsequenced human papilloma-virus genotypes in cervical scrapes and carcinomas. Int. J.Cancer 45:644-649.

25. Wagner, D., H. Ikenberg, N. Boehm, and L. Gissmann. 1984.Identification of human papillomavirus in cervical swabs bydeoxyribonucleic acid in situ hybridization. Obstet. Gynecol.64:767-772.

26. zur Hausen, H. 1991. Human papillomaviruses in the pathogen-esis of anogenital cancer. Virology 184:9-13.

VOL. 30, 1992

on August 15, 2020 by guest

http://jcm.asm

.org/D

ownloaded from