Human papillomavirus typing and viral gene expression analysis for the triage of women with abnormal results from papanicolaou test smears to colposcopy.
Abstract: * Context.--A cascade of molecular tests for human papillomavirus (HPV), as a follow-up to Papanicolaou test screening, could eliminate unnecessary colposcopy. Tests based on detection of HPV E6 messenger RNA (mRNA) are already being used as screening tools, but there is a good biological rationale for expecting that an increase in the relative amounts of HPV E6 mRNA in cervical samples may better predict cancerous transformation.

Objective.--To compare some of the available diagnostic methods and our novel method of relative quantification (RQ) of HPV gene expression for the effective triage of women with abnormal results from Papanicolaou tests to colposcopy.

Design.--Sensitivities, specificities, and likelihood ratios were calculated for repeat Papanicolaou test smears, HPV DNA polymerase chain reactions, HPV genotyping, HPV16 E6 mRNA detection, and the RQ of HPV-16 E6 mRNA calibrated to cellular RNA and DNA levels and standardized to viral load.

Results.--Human papillomavirus genotype in combination with a repeat Papanicolaou test can be used to categorize most women (96%) with cervical intraepithelial neoplasia of grade 2 or higher for colposcopy while eliminating 44% of women with cervical intraepithelial neoplasia 1 or less. The presence of HPV-16 E6 mRNA (P < .001) and RQ of HPV-16 E6 mRNA (P < .001) displayed significant median differences among the various grades of cervical intraepithelial neoplasia. Further testing of women who are positive for HPV-16 demonstrated that the RQ of E6 mRNA has diagnostic potential when combined with Papanicolaou testing in populations with higher disease prevalence.

Conclusions.--The RQ of HPV E6 mRNA and HPV genotype could be useful in a cascade of diagnostic testing designed to refer women with findings of cervical abnormalities for colposcopy or treatment while reducing triage numbers.

(Arch Pathol Lab Med. 2009;133:1577-1586)
Article Type: Report
Subject: Colposcopy (Usage)
Pap test (Usage)
Gene expression (Research)
Gene expression (Physiological aspects)
Papillomavirus infections (Diagnosis)
Papillomavirus infections (Genetic aspects)
Cervical cancer (Diagnosis)
Cervical cancer (Case studies)
Cervical cancer (Genetic aspects)
Triage (Medicine) (Management)
Authors: Antonishyn, Nick A.
Horsman, Greg B.
Kelln, Rod A.
Severini, Alberto
Pub Date: 10/01/2009
Publication: Name: Archives of Pathology & Laboratory Medicine Publisher: College of American Pathologists Audience: Academic; Professional Format: Magazine/Journal Subject: Health Copyright: COPYRIGHT 2009 College of American Pathologists ISSN: 1543-2165
Issue: Date: Oct, 2009 Source Volume: 133 Source Issue: 10
Topic: Event Code: 200 Management dynamics; 310 Science & research Computer Subject: Company business management
Accession Number: 230247152
Full Text: The family Papillomaviridae comprises a highly diverse collection of more than 180 viruses, and more than 100 types of Human papillomavirus (HPV) have been identified and shown to be associated with a wide variety of benign and malignant epithelial lesions. Genital HPV types have been subdivided into low-risk types, which typically produce benign genital warts, and high-risk types, such as HPV-16, which are more frequently associated with invasive cervical cancer. (1,2) In women, infection with HPV, considering all possible types, achieves a lifetime cumulative incidence of up to 70%, whereas cervical cancer is a relatively rare disease, with a lifetime incidence range of 1.1% to 3.0% worldwide. (3,4) Nevertheless, cervical cancer caused more than 270 000 deaths worldwide in 2002, making it the second most-common cancer among women. (4) Cervical cancer caused by HPV is considered largely preventable and, in developed countries, has one of the best prognoses of all cancers, given its highly successful treatment options. (4,5) However, improvements in early diagnosis are required because cervical cancer screening programs based on Papanicolaou testing have their limitations.

An audit of a UK program found that 47% of stage IB cervical cancer or worse occurred in women younger than 70 years with an adequate screening history. (6) A Papanicolaou test is very specific for determining the presence of neoplastic transformation, but it is not very sensitive. (7) In gynecologic oncology, a review of pathologic reports found reproducibility to be only 84% with 2% of the corrected diagnoses consequentially affecting the proper treatment decisions for the individual patient. (8) There is significant reporting variation in diagnoses of cervical intraepithelial neoplasia (CIN) and in differentiation between findings of normal tissue and borderline abnormalities. (9) Severely abnormal Papanicolaou test results usually suggest colposcopic examination, but it is highly subjective and can result in nonevidenced-based patient management. (10) Although 99.7% of cervical cancers possess HPV DNA, the simple detection of HPV DNA is a poor predictor for the risk of cancerous transformation (7-11) because there is poor correlation between certain low-risk HPV types and cervical cancer. (1) Even HPV infection with a high-risk type does not provide a reasonably useful marker for transformation because most infections are transient, and persistent infection is required for transformation. (12) To date, no biomarker has been found, to our knowledge, that defines the persistence of HPV infection.

High-risk HPV types more effectively achieve the abrogation of cell-cycle checkpoints than low-risk types. (13) In certain conditions, the viral oncoproteins E6 and E7 can become overexpressed by both up-regulation of the major early promoter of the virus and by increased stability of the mRNAs for these proteins. (14,15) During progression to malignancy, infected cells lose their ability to differentiate, and they display the histomorphologic characteristics of high-grade squamous intraepithelial lesions, which are known precursors of cervical cancer. (16) Furthermore, transcription of capsid proteins is restricted to terminally differentiated keratinocytes. (17) As the grade of neoplasia increases, cellular differentiation decreases, and L2 and L1 transcripts become undetectable. (18) Consequently, malignant transformation is expected to be accompanied by an overexpression of E6 and E7 and a reduction of L1 and viral DNA.

It is reasonable to expect that an increase in the relative amounts of E6 RNA or E7 RNA or a decrease in L1 RNA in cervical samples may correlate with a risk of cancerous transformation. The targeted biomarkers E6 and E7 are required for the transformed phenotype in continuous cell lines derived from HPV-infected tissue. (19,20) The amount of HPV E6 and E7 mRNA has been shown to be associated with the presence of CIN. (21-29) However, these studies reveal that the predictive value of HPV E6 RNA remains low. The lack of predictive power could be attributed to one or more problems with the research to date: low sample numbers, inappropriate use of technology, poor quality assurance and insufficient control for variables, or intrinsic variation of E6 and E7 RNA levels in cervical samples. An improvement in triage practice may be achieved by using different tests in combination, particularly when no single test is satisfactory for both specificity and sensitivity. Studies of colposcopy practice have shown that predictors of CIN 3 can be defined, but there is a problem with the performance of available diagnostic tools in clinical practice. One such study has concluded that the most satisfactory management of women with abnormal Papanicolaou test results is likely going to need a proficient combination of tests. (30)

This study set out to resolve the question of whether or not any characteristic of HPV infection has utility for the triage of women to colposcopy for a definitive diagnosis of cervical dysplasia. We took the novel approach of measuring the level of HPV-16 RNA relative to viral DNA, host DNA, and host RNA. This relative quantification (RQ) approach is compared with HPV typing and repeat Papanicolaou testing for its utility in triage of women with HPV-16-positive results to colposcopy and for its ability to predict high-grade lesions. We show that a well-designed cascade of molecular tests as a follow-up to Papanicolaou test screening of the general population could identify women who need further follow-up at colposcopy while potentially increasing the screening period for women who are not at risk.

MATERIALS AND METHODS

Study Population

Specimens were collected from January 1998 through February 2005 from women who were referred to the colposcopy clinic at the Women's Health Centre, Regina General Hospital, Regina, Saskatchewan, Canada. Archival samples collected from January 1998 through September 2003 were stored frozen at--80[degrees]C and processed at the same time. Women were referred to colposcopy if they had abnormal results, including high-grade squamous intraepithelial lesions, low-grade squamous intraepithelial lesions, and atypical squamous cells of undetermined significance, from 2 conventional Papanicolaou test smears within 6 months or if they had a single Papanicolaou test result indicting high-grade squamous intraepithelial lesions. A diagnosis for each patient was established arising from the procedures and visits as defined by the collection date of the study sample. The diagnosis was classified into 1 of 4 categories, CIN 3, CIN 2, CIN 1, or normal, and was based on histologic grading of the cervical biopsies.

Cervical cells for nucleic acid extraction were taken directly after the Papanicolaou test swab using a Cervex-Brush (Rovers Medical Devices, Amsterdam, Netherlands), which were then immediately frozen in ethanol. All specimens were kept on dry ice until processing, which took place within 4 hours. Study specimens were pulse-vortexed, to remove cells from the brush head, until the suspension appeared homogeneous, typically 1 minute. Three 1-mL aliquots of the cell suspension were centrifuged at 3000g for 5 minutes at 4[degrees]C. The supernatant was removed, and 1 cell pellet was queued for DNA extraction. The other 2 cell pellets were resuspended in TRIzol reagent (Invitrogen, Carlsbad, California) before storage at -70[degrees]C until RNA extraction was performed.

Nucleic Acid Extraction

Cervical cells collected before October 2003 were processed together and resuspended in 90 [micro]L of polymerase chain reaction (PCR) buffer (50 mM potassium chloride, 10 mM Tris at pH 8.3, 2.5 mM magnesium chloride, 0.1 mg/mL gelatin, 0.45% Igepal CA 360, and 0.45% Tween). After addition of proteinase K at the final concentration of 200 [micro]g/mL, the samples were incubated at 55[degrees]C for 1 hour and then boiled for 10 minutes.

Starting in October 2003, DNA was extracted from a pellet of cervical cells using the QIAamp DNA Mini Kit (Qiagen, Mississauga, Ontario, Canada) according to the manufacturer's instructions. RNA was extracted from a different pellet using a modified protocol for the TRIzol reagent. Essentially, cell pellets or control material were disrupted in 1 mL of TRIzol with pulse vortexing and then kept at -70[degrees]C until queued for extraction. In the modified protocol, the final RNA pellet was washed with 75% ethanol and then air-dried before resuspension in a cocktail con taining 2 IU of DNase I (Invitrogen), 1X DNase I buffer, and 20 IU of SUPERase-In RNase Inhibitor (Ambion, Streetsville, Ontario, Canada). The reaction was extracted again using the original TRIzol protocol, and the final RNA pellet was resuspended in 200 [micro]L of diethylpyrocarbonate-treated water.

RNA cleanup was performed if samples were contaminated with DNA by treating the RNA preparation with an additional DNase I digestion and, if required, a final cleanup using the RNeasy Mini Kit (Qiagen). The DNase I treatment was done by adding 11 [micro]L of 10X DNase I buffer (Invitrogen) to the remainder of the RNA preparation along with 40 IU of SUPERase-In RNase inhibitor and 5 IU of DNase I and then incubating at 37[degrees]C for 30 minutes. The reaction was stopped by adding 12 [micro]L of 25mM ethylenediaminetetraacetic acid and incubating the mixture at 65[degrees]C for 10 minutes. The RNeasy cleanup procedure was done according to the manufacturer's protocol with the optional on-column DNase digestion during RNA purification.

Oligonucleotides and Primer Design

Oligonucleotides were synthesized by Sigma-Genosys (Sigma-Aldrich, Oakville, Ontario, Canada), and their sequences and final reaction concentrations are listed in Table 1. Oligonucleotides for HPV-31 E6 gene and HPV-16 L1 gene are original to this study and were designed using Oligo 6 (Molecular Biology Insights, Cascade, Colorado). The sequences of the other oligonucleotides in Table 1 have been published elsewhere. (28,31-33)

Detection of HPV DNA

Preparations of extracted DNA were tested with a hot-start PCR method using consensus primers ([Gp5.sup.+], [Gp6.sup.+]) targeted for a 150-bp semiconserved region within the L1 gene 31. The L1 PCR was used as the primary screening test with negatives queued for testing with a second PCR using consensus primers (E1 350L, E1 547R, E1 847R), which targeted a 180-bp fragment within the E1 ORF 32. The E1 PCR was also used if the sequence data obtained from the L1 PCR did not provide a suitable match to a type strain. A specimen was considered positive for HPV DNA if either PCR generated an amplicon of the correct size with subsequent sequence data matching a known HPV sequence.

The reaction master mixes for both primer sets had the following similarities: Each was adjusted to a final volume of 95 [micro]L using high-pressure liquid chromatography-grade water and contained 2.5 IU of AmpliTaq Gold DNA polymerase (Applied Biosystems, Streetsville, Ontario, Canada), 1X PCR buffer II (Applied Biosystems), 1.5 mM magnesium chloride, 200 [micro] Meachof deoxyadenosine triphosphate, deoxycytidine triphosphate, deoxyguanosine triphosphate, and thymidine monophosphate. The L1 PCR reactions contained 0.5 [micro]M of each primer, [Gp5.sup.+] and [Gp6.sup.+] 31, whereas E1 PCR reactions contained 0.5 [micro]M of each primer, E1 350L and E1 547R, and 0.25 [micro]M of primer E1 847R 32. All reactions contained 5 [micro]L of the extracted DNA preparation (concentration unknown). The E1 PCR also contained 1% NP-40 and Tween-20.

Polymerase chain reaction amplifications were performed with a 9700 thermal cycler (Applied Biosystems). All thermal cycling regimes were preceded by a 10-minute, 95[degrees]C incubation period and followed by an additional extension time of 10 minutes at 72[degrees]C and a final soak at 4[degrees]C. Samples were then removed from the thermal cycler and processed or placed in storage at -20[degrees]C. L1 PCRs were cycled 40 times through the following temperature regimes: 94[degrees]C for 30 seconds, 44[degrees]C for 60 seconds, and 72[degrees]C for 90 seconds. E1 PCRs were cycled using a 2-stage regime with the first-stage cycling of 15 times through: 94[degrees]C for 40 seconds, 45[degrees]C for 40 seconds, and 72[degrees]C for 40 seconds; and the second-stage cycling of 30 times through: 94[degrees]C for 40 seconds, 30[degrees]C for 40 seconds, and 72[degrees]C for 40 seconds.

DNA Sequencing

Polymerase chain reaction products were processed using Microcon centrifugal filter units (Millipore, Billerica, Massachusetts) according to the manufacturer's protocol. Purified templates were quantified by visually comparing the fluorescence intensity against a Low DNA Mass Ladder (Invitrogen). Fluorescence-based dideoxycycle-sequencing reactions were prepared using the BigDye Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems) according to the manufacturer's protocol. The sequencing primers were the same as those used in the DNA amplification reactions. DNA sequence data for either the L1 or E1 region from each HPV-positive specimen were independently compared with the DNA sequence from known HPV genotypes and confirmed using phylogenetic analysis with a collection of representative mucosal HPV types. An HPV genotype determination was made if the Basic Local Alignment Search Tool search resulted in a similarity score greater than 90% and at least 100 nucleotides of the query providing the score. Human papillomavirus 16-positive specimens were queued for RQ work. Human papillomavirus 16-positive specimens by DNA sequencing were confirmed with real-time PCR. Any specimens that were coinfected with HPV types 16 and 31 were not queued for RQ analysis. Indeterminate genotypes were tested by nested PCR and a multiplex flow-cytometric method based on Luminex technology.

Luminex Technology

Briefly, the method amplifies DNA from mucosal HPV types by a nested-PCR method using PGMY primers 33 followed by [Gp5.sup.+] / [GP6.sup.+] amplification. The products were labeled with the fluorophore phycoerythrin and hybridized to sortable microspheres, carrying specific probes for 45 mucosal HPV types. The probes are sensitive and specific for each type without cross-hybridization. The characteristics of the performance of the assay is comparable to the Linear Array Kit (Roche Molecular Systems, Inc, Branchburg, New Jersey) albeit with a lower sensitivity for samples with more than 4 concurrent HPV types. (34,35)

Risk Classification

The genotype information was used to categorize women based on published risk classification 1. Essentially, there are 15 high-risk HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, and 82), 3 probable high-risk HPV types (26, 53, and 66), and 12 low-risk HPV types (6, 11, 40, 42, 43, 44, 54, 61, 70, 72, 81, and CP6108). For this study, the high-risk categorization included probable high-risk types and unclassified HPV types, whereas the low-risk categorization included known low-risk types and women with negative results for HPV DNA. The inclusion of unclassified HPV in the high-risk categorization enhances the triage of women for the detection of CIN and would be similar in rationale to the current approach for the follow-up of atypical squamous cells of undetermined significance.

Real-Time Nucleic Acid Amplification Reactions

The real-time reverse-transcriptase PCR conditions used in this study were a slight modification to those used by Wang-Johanning et al (2002). (28) All reactions were done in triplicate, with results averaged before use in RQ analysis. The CaSki cell line was used as a source for quality-control material with aliquots of extracted RNA and DNA used as calibration controls for expression analysis and diluted to have reasonable cycle threshold (Ct) values, within the range measured with clinical material. Reaction mixtures contained 12.5 [micro]L of either 2X Quantitect One-Step reverse-transcriptase PCR master mix (Qiagen) with 0.25 [micro]L of enzyme mix reagent or 2X Quantitect PCR master mix reagent (Qiagen). An appropriate volume of oligonucleotide working stock was included in each reaction to achieve the concentrations indicated in Table 1. Each reaction used 5 [micro]L of prepared template (DNA or DNA-free RNA), with high-pressure liquid chromatography-grade water making the final volume of 25 [micro]L as needed. All real-time nucleic acid detection was performed using the ABI 7700 (Applied Biosystems) with thermal cycling consisted of 30 minutes at 48[degrees]C, 15 minutes at 95[degrees]C, and 40 cycles of 15 seconds at 95[degrees]C, and 60 seconds at 60[degrees]C.

Real-Time Data Analysis

The Ct was typically set to 0.2 relative fluorescent units but was determined for each run based on the amplification plots such that Ct values were at the beginning of the linear portion of the logarithmically plotted fluorescent curve versus cycle number. The plots were examined for kinetic outliers and required a measured estimation of the PCR efficiency for each replicate reaction. (36) Outliers, as determined by kinetic outliers or detected manually, were then excluded from the average.

Criteria were established empirically to ensure that only materials of high quality were queued for expression profile analysis. Essentially, RNA extractions were only analyzed if the preparation had measurable levels of ribosomal protein S9 (RPS9), indicating a good total RNA recovery from cervical cells, and no signal in the E6 DNA real-time PCR, thereby indicating no HPV DNA contamination of the RNA preparation. A Ct value in E6 reverse-transcriptase PCR, but not in E6 DNA PCR indicated that the RNA preparation was suitable for HPV mRNA expression analysis. Consequently, Ct value in E6 DNA PCR queued the RNA preparation for DNA cleanup. No amplification in either E6 reverse-transcriptase PCR or RPS9 queued the third cell-pellet aliquot, which was stored in TRIzol at -70[degrees]C, for RNA extraction. This quality check was performed before the sample was queued for analysis and confirmed again during the expression-profile analyses because it was included in the same 96-well plate during the run.

Relative Quantification

Relative quantification was based on the relative expression of a target versus a reference by comparing mathematically transformed Ct values. Human [beta]-actin gene was used as the reference for DNA standardizations, whereas transcripts for human RPS9 were used as the reference for RNA standardizations.

Concentration = Ct = 40 / 3.321 (1)

The absolute concentration of transcript levels were estimated by converting the Ct to a concentration using a theoretic external standard curve as indicated in equation 1. (1,37) Next the ratio of HPV-16 transcript to total human RNA, as measured by the RPS9 levels, was calculated. Finally, the ratio above was divided by a ratio of HPV DNA to total human DNA, as measured by the [beta]-actin levels (equation 2)

[HPV mRNA(E6 or E7 or L1)] / RPS9 / (HPV DNA)/([beta]-actin) (2)

Statistical Analysis

Statistical tests were performed using Prism v4 software (GraphPad, La Jolla, California). All data sets were analyzed using the D'Agostino-Pearson normality test (omnibus K2) to determine whether parametric or nonparametric statistics were appropriate. The statistical significance of mean or median differences was assessed using either a 1-way analysis of variance for Gaussian data or the Kruskal-Wallis test for non-Gaussian data. A value of P < .05 was considered statistically significant. For the Gaussian data, the Bartlett test for equal variances was performed to ensure that a 1-way analysis of variance was appropriate. To calculate the significance between each possible paired category, the Bonferroni or Dunn test was used for Gaussian or non-Gaussian data, respectively. The value of the test statistic (ie, Kruskal-Wallis statistic) was used to determine which normalization or standardization of the transcript demonstrated the greatest correlation with the diagnostic categorization.

Contingency tables were constructed to measure the correlation between research data and diagnosis. The statistical significance was tested using the x2 test. If significance was found (P < .05), then the grouping was carried out to reduce contingency tables to 2 X 2 table size. For example, colposcopy diagnoses were categorized into 2 groups in one of 3 ways: (1) abnormal-high or abnormal-medium with abnormal-low and normal, (2) abnormal-high with abnormal-medium or abnormal-low with normal, and (3) abnormal-high with abnormal-medium and abnormal-low or normal. The best cutoff for RQ data was determined using receiver operator characteristic curves. (38) Statistical significance of the 2 X 2 contingency tables was measured using the Fisher exact test.

RESULTS

The study population consisted of women who had 2 previously abnormal Papanicolaou test results within 6 months or had a single Papanicolaou test result of high-grade squamous intraepithelial lesions. In total, 887 samples were collected during the study period. All samples were tested for HPV DNA by conventional L1 and E1 PCR and were genotyped, if positive, by direct sequencing. The results from these molecular tests in relation to the Papanicolaou test result and disease category, as determined by histology, are tabulated in Tables 2 and 3. The HPV DNA results obtained from the 2 extraction methods were similar. The period of specimen collection for January 1998 through September 2003 used the in-house DNA extraction method and yielded a 61% (350 of 573) positive rate. The commercial DNA extraction method for specimens collected between October 2003 through February 2005 yielded a 60% (187 of 314) positive rate.

A total of 121 samples (14%) that were HPV-16 positive were queued for RNA analysis and the relative quantification of E6 transcripts, which produced the most statistically significant difference in measured expression levels among cervical diseases, are shown in Figure 1. Receiver operator characteristic curve 38 for HPV-16 E6 RQ was examined (Figure 2, A through C). A cutoff of .5 was empirically chosen based on receiver operator characteristic curves for the best trade-off between sensitivity (.66) and specificity (.78) with CIN 3 at colposcopy as the endpoint (Figure 2, A). Tables 4 and 5 show the test and reference standard results on the 121 positive HPV-16 study samples for HPV-16 E6 RNA detection and HPV-16 E6 RQ, respectively. Normality testing confirmed that the RQ of the HPV transcripts produced data that were non-Gaussian. The presence of HPV-16 E6 mRNA, ([chi square], P < .001) and the level of its transcription relative to the viral load of host nucleic acid, henceforth called HPV-16 E6 RQ (P < .001, Kolmogorov-Smirnov test = 22.77), emerge as the most statistically significant finding with respect to the correlation with cervical abnormalities. Dunn multiple comparison tests show that the median level of the HPV16 E6 RQ is significantly different for the following pairs of cytologic diagnoses: CIN 3 versus CIN 2 (P < .01), CIN 3 versus CIN 1 (P < .05), and CIN 3 versus normal (P < .001). Table 6 shows the sensitivity, specificity, and likelihood ratios (LRs) for 3 diagnostic tests and combinations of the tests: for repeat Papanicolaou testing, for HPV DNA detection, and for genotyping. The test performance measurements of HPV gene expression analysis are summarized in Table 7. The effect of combination testing on the actual number of women excluded from colposcopy triage is indicated in Table 8. The prevalence of any type of CIN in patients with HPV-16 was .88, and the prevalence of CIN 2 or greater was .65. In this subset of women, 34.7% of the women will still be diagnosed with greater than CIN 1 or normal histology (Table 8).

COMMENT

Effective triage for cervical cancer involves referring women affected with disease to colposcopy while implementing stringent monitoring for women with low-grade disease. Ideally, this occurs without extensive testing or medical treatment of women with disease-free cervixes. Logically, a combination of tests using an either-positive rule will improve sensitivity and capture the most number of women with disease. However, the evaluation here was to measure the impact that combination testing using molecular and viral markers would have on both sensitivity and specificity and the implications of these markers in a cascade testing strategy for triage. The objective of this study was to determine whether additional testing could reduce the number of women who would normally be referred to colposcopy while maintaining sensitivity for capturing disease.

Diagnostic Performance Evaluation

The samples for repeat Papanicolaou testing were taken first, followed by a brush for HPV nucleic acid extractions, and finally a biopsy for histology. Consequently, the performance of the repeat Papanicolaou test or any of the molecular tests could be considered as they might perform when the patient visits a specialist's office and before the scheduled colposcopy. Neither Papanicolaou testing or any single molecular marker demonstrates superiority in both sensitivity and specificity. Consequently, combination testing is warranted. Nevertheless, Papanicolaou testing is more specific for the detection of any abnormality (ie, [greater than or equal to] CIN 1), whereas genotyping had the best specificity for disease of CIN 2 or greater.

Predictive values provide the probability of disease given the test result and, therefore, are often used as a clinically relevant basis to compare test performances in a given population. However, predictive values are influenced by disease prevalence when test dependence between diseased and nondiseased groups is asymmetric. (39) Asymmetry can arise if, for example, there is a negative correlation between test results in the diseased population but a positive correlation in the nondiseased population. Such is the case with HPV infection with its high incidence of cytologically normal cervical smears in women who clear the virus without treatment. Consequently, LRs are useful for performance evaluation because they do not depend mathematically on disease prevalence and encompass the trade-offs between sensitivity and specificity. A combined test would have better test performance when [LR.sup.-.sub.combined] < [LR.sup.-.sub.single] and [LR.sup.+.sub.combined] > [LR.sup.+.sub.single]. The single test is a better choice when [LR.sup.-.sub.combined] > [LR.sup.-.sub.single] and [LR.sup.+.sub.combined] < [LR.sup.+.sub.single], (40) that is, the test or combination of tests that has the highest number for LR+ and the lowest number for [LR.sup.-] is the best choice rather than the others in the comparison. In this study, there was no test or combination of tests with the best pair of LRs (Table 6). Consequently, the choice between basing clinical decisions on a repeat Papanicolaou test or on combining the initial Papanicolaou test with a molecular test is not obvious. In populations such as ours, there will be a trade-off between retaining women with true cervical disease and eliminating false-positives.

[FIGURE 1 OMITTED]

Given the LR information in Table 6, Papanicolaou testing offers an advantage combined with HPV DNA using the either-positive rule for the detection of any grade of abnormality. However, HPV genotype information offers the best advantage when combined with Papanicolaou test for the detection at least CIN 2. Consequently, a repeat Papanicolaou test and HPV genotyping on cells collected at the same time could be considered an auxiliary testing tier before referral for biopsy, colposcopy, or treatment. When using this combination for detection of at least CIN 2, of the 603 women with double negative Papanicolaou tests and HPV results, 267 women (44%), who would later reveal a biopsy with CIN 1 or less, could have been relieved from further testing or colposcopy (Table 8). Therefore, a repeat Papanicolaou test and HPV genotyping (with the either-positive rule) would have retained 96% (852 of 887) of the women who needed further medical workup while deferring 30.1% (267 of 887) from immediate colposcopy. Given that only 11% of women with CIN 1 progress to CIN 3, (41) this auxiliary testing strategy would ultimately save up to 26.8% (n = 238) of the women in our triage population (n = 887) from colposcopy and potential problems associated with biopsy.

Utility of HPV E6 RNA as an Adjunct Test

The quality and quantity of all RNA preparations were measured before being queued for expression analysis, which is a distinctive element of this study relative to most previously published attempts to evaluate the diagnostic potential of HPV-16 E6 mRNA RQ. Taking the cascade testing to the next level (eg, Tier 2), the potential utility of HPV-16 E6 RNA, or the more refined analysis of HPV-16 E6 RQ, was examined with HPV-16 as the prototype. That is, could HPV gene expression information be useful as a complement test for triage? Again, the either-positive rule was chosen to improve test performance in the group with disease and was compared with repeat cytology. Tables 4 and 5 show the test and reference-standard results on 121 HPV-16 positive study samples for HPV-16 E6 RNA detection and HPV-16 E6 RQ, respectively. In this subset of women, there are still 34.7% of women (42 of 121) with at least CIN 1 (Table 8). Additional testing of clinical material already collected could be a useful reflex testing strategy to further eliminate unnecessary biopsy and colposcopy.

[FIGURE 2 OMITTED]

The test performance measurements of HPV gene-expression analysis are summarized in Table 7. Again, no single test demonstrated the best combination of sensitivity and specificity. The simple detection of HPV-16 E6 RNA (ie, positive/negative scoring for E6 transcript presence) offered no advantage as a single axillary test or in combination with Papanicolaou testing. However, the RQ of the E6 transcript (ie, HPV-16 E6 RQ) did offer some further advantage for triage, when standardized to HPV viral load, the number of cells collected, and the yield of nucleic extraction--as done here for the first time, to our knowledge. Examination of the LR for the combination of Papanicolaou testing with HPV-16 E6 RQ shows that, for higher disease prevalence, the combined test would find more true positives than simply an additional Papanico laou test (Table 7). The addition of HPV-16 E6 RQ could be used to further identify 31% (13 of 42) of women with positive HPV-16 and at least CIN 1 while retaining 92.4% (73 of 79) of women with CIN 2 or greater for triage to colposcopy (Table 8).

Interestingly, E6 expression levels can be used to increase the specificity beyond simply the presence/absence scoring of the transcript, which has been the focus of a new molecular test, the PreTect HPV Proofer (NorChip AS, Klokkarstua, Hurum, Norway). (42) It is reasonable to expect that HPV E6 RNAs would make good markers because they appear early in the staging toward carcinogenesis. These transcripts and their presence, or even their relative levels, correlate with the cause of disease, and therefore, they could be a diagnostic marker for cervical cancer. An application of the E6 RQ could be to confirm an abnormal diagnosis (positive predictive value, 95%). In this circumstance, the LR supports the RQ of E6 transcript as better than simply positive or negative E6 RNA for the detection of cervical neoplasia of any grade. The utility of E6 RQ for other non-HPV-16 types will need to be determined. Improvement to E6 RQ analysis would be made if the actual number of infected, versus uninfected cells, could be included in the standardization.

The utility of E6 RQ as triage tool is shown here, but its potential for prognosis is, perhaps, most intriguing. Further study with patient follow-up may find that RNA molecular tests are better for prognosis than diagnosis. In this study, 3 of 3 patients (100%) whose results were less than the 0.5 cutoff for RQ of E6 and who had biopsies of CIN 1 had disease that remained at CIN 1 after 4 to 5 months, when another biopsy was taken. On the other hand, 4 of 8 patients with CIN 1 (50%) were positive for E6 RQ 0.5, which progressed to at least CIN 2 in 3 to 10 months. A recent survey of archival tissue from patients with stages IB and IIA cervical cancer found that high E6/ E7 mRNA expression was associated with a poor prognosis. (43) Even so, the true diagnostic and prognostic value of HPV-16 E6 RQ will require further evaluation with a prospective cohort study.

In conclusion, the use of HPV genotyping, in combination with a Papanicolaou test screening program, demonstrates excellent diagnostic potential. Human papillomavirus DNA testing has already been demonstrated to be useful for reflex testing of Papanicolaou test smears with atypical squamous cells of undetermined significance. (44) The addition of HPV genotyping, regardless of Papanicolaou test smear result, to any triage protocol is shown here. We have shown that there is a diagnostic potential in HPV gene expression analysis and that it benefits from standardization for variables, such as the amount of HPV DNA and total cellular nucleic acid. The analysis of E6 transcript levels, combined with Papanicolaou testing, has a potential diagnostic value in a triage protocol. The added gain in diagnostic performance is attributable to the high specificity of E6 RQ analysis.

We thank L. A. Brydon, MD; R. Bhargava, MD; J. Bastian, MD; D. Shepherd, MD; A. Akinbiyi, MD; R. Cardoso, MD; C. Jabs, MD; and J. Guillemin, PhD; who were invaluable in the acquisition of clinical material. We thank B. Cargill, PhD; and J. Kirsch, PhD; for pulling together the Papanicolaou test and histology results. This research was partially supported by the Natural Sciences and Engineering Research Council of Canada, Ottawa, Ontario (Dr Kelln).

Accepted for publication January 6, 2009.

References

(1.) Munoz N, Bosch F, De Sanjose S, et al. Epidemiologic classification of human papillomavirus types associated with cervical cancer. N Engl J Med. 2003; 348(6):51 8-527.

(2.) Severson J, EvansTY, Lee P, Chan T, Arany I, Tyring S. Human papillomavirus infections: epidemiology, pathogenesis, and therapy. J Cutan Med Surg. 2001; 5(1):43-60.

(3.) Koutsky L. Epidemiology of genital human papillomavirus infection. Am J Med. 1997;102(5A):3-8.

(4.) Parkin D, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin. 2005;55(2):74-108.

(5.) Rydstrom C, Tornberg S. Cervical cancer incidence and mortality in the best and worst of worlds. Scand J Public Health. 2006;34(3):295-303.

(6.) Sasieni P, Cuzick J, Lynch-Farmery E. Estimating the efficacy of screening by auditing smear histories of women with and without cervical cancer: the national co-ordinating network for cervical screening working group. Br J Cancer. 1996;73(8):1001-1005.

(7.) Clavel C, Masure M, Bory JP, et al. Human papillomavirus testing in primary screening for the detection of high-grade cervical lesions: a study of 7932 women. Br J Cancer. 2001;84(12):1616-1623.

(8.) Santoso J, Coleman R, Voet RL, Bernstein S, Lifshitz S, Miller D. Pathology slide review in gynecologic oncology. Obstet Gynecol. 1998;91(5, pt 1):730734.

(9.) Creagh T, Bridger J, Kupek E, Fish DE, Martin-Bates E, Wilkins M. Pathologist variation in reporting cervical borderline epithelial abnormalities and cervical intraepithelial neoplasia. J Clin Pathol. 1995;48(1):59-60.

(10.) Nelson G, Duggan M, Nation J. Controversy in colposcopic management: a Canadian survey. J Obstet Gynaecol Can. 2006;28(1):36-40.

(11.) Walboomers J, Jacobs M, Manos M, et al. Human papillomavirus is a necessary cause of invasive cervical cancer worldwide. J Pathol. 1999;189(1):1219.

(12.) Cuschieri K, Cubie HA, Whitley M, et al. Persistent high risk HPV infection associated with development of cervical neoplasia in a prospective population study. J Clin Pathol. 2005;58(9):946-950.

(13.) Syrjanen S, Syrjanen K. New concepts on the role of human papillomavirus in cell cycle regulation. Ann Med. 1999;31(3):175-187.

(14.) zur Hausen H. Papillomavirus infections--a major cause of human cancers. Biochim Biophys Acta. 1996;1288(2):F55-F78.

(15.) Graham SV. Late events in the life cycle of human papillomaviruses. In: Campo MS, ed. Papillomavirus Research: From Natural History to Vaccines and Beyond. Norfolk, England: Caister Academic Press; 2006:193-211.

(16.) McCance D, Kopan R, Fuchs E, Laimins L. Human papillomavirus type 16 alters human epithelial cell differentiation in vitro. Proc Natl Acad Sci USA. 1988; 85(19):7169-7173.

(17.) Stoler M, Wolinsky S, Whitbeck A, Broker T, Chow L. Differentiation-linked human papillomavirus types 6 and 11 transcription in genital condylomata revealed by in situ hybridization with message-specific RNA probes. Virology. 1989;172(1):331-340.

(18.) Stoler M, Rhodes C, Whitbeck A, Wolinsky S, Chow LT, Broker T. Human papillomavirus type 16 and 18 gene expression in cervical neoplasias. Hum Pathol. 1992;23(2):1 17-128.

(19.) Demers G, Halbert C, Galloway D. Elevated wild-type p53 protein levels in human epithelial cell lines immortalized by the human papillomavirus type 16 E7 gene. Virology. 1994;198(1):169-174.

(20.) Nauenburg S, Zwerschke W, Jansen-Durr P. Induction of apoptosis in cervical carcinoma cells by peptide aptamers that bind to the HPV-16 E7 oncoprotein. FASEB J. 2001;15(3):592-594.

(21.) Park J, Hwang E, Park S. et al. Physical status and expression of HPV genes in cervical cancers. Gynecol Oncol. 1997;65(1):121-129.

(22.) Hsu E, McNicol P, Guijon F, Paraskevas M. Quantification of HPV-16 E6E7 transcription in cervical intraepithelial neoplasia by reverse transcriptase polymerase chain reaction. Int J Cancer. 1993;55(3):397-401.

(23.) Czegledy J, Evander M, Hernadi Z, Gergely L, Wadell G. Human papillomavirus type 18 E6* mRNA in primary tumors and pelvic lymph nodes of Hungarian patients with squamous cervical cancer. Int JCancer. 1994;56(2):182 186.

(24.) Daniel B, Mukherjee G, Seshadri L, Vallikad E, Krishna S. Changes in the physical state and expression of human papillomavirus type 16 in the progression of cervical intraepithelial neoplasia lesions analysed by PCR. J Gen Virol. 1995; 76(pt 10):2589-2593.

(25.) Selinka H, Sotlar K, Klingel K, Sauter M, Kandolf R, Bultmann B. Detection of human papillomavirus 16 transcriptional activity in cervical intraepithelial neoplasia grade III lesions and cervical carcinomas by nested reverse transcriptionpolymerase chain reaction and in situ hybridization. Lab Invest. 1998;78(1):9 18.

(26.) Sotlar K, Selinka H, Menton M, Kandolf R, Bultmann B. Detection of human papillomavirus type 16 E6/E7 oncogene transcripts in dysplastic and nondysplastic cervical scrapes by nested RT-PCR. Gynecol Oncol. 1998;69(2):114 121.

(27.) Ke L, Adler-Storthz K, Mitchell M, Clayman G, Chen Z. Expression of human papillomavirus E7 mRNA in human oral and cervical neoplasia and cell lines. Oral Oncol. 1999;35(4):415-420.

(28.) Wang-Johanning F, Lu D, Wang Y, Johnson M, Johanning G. Quantitation of human papillomavirus 16 E6 and E7 DNA and RNA in residual material from thinprep Papanicolaou tests using real-time polymerase chain reaction analysis. Cancer. 2002;94(8):21 99-2210.

(29.) Sotlar K, Stubner A, Diemer D, et al. Detection of high-risk human papillomavirus E6 and E7 oncogene transcripts in cervical scrapes by nested RTpolymerase chain reaction. J MedVirol. 2004;74(1):107-116.

(30.) Ciotti M, Sesti F, Paba P, etal. Human papillomavirus (HPV) testingin the management of women with abnormal Pap smears: experience of a colposcopy referral clinic. Eur J Gynaecol Oncol. 2004;25(5):577-584.

(31.) De Roda Husman AM, Walboomers J, Van Den Brule AJ, Meijer C, Snijders P. The use of general primers GP5 and GP6 elongated at their 3' ends with adjacent highly conserved sequences improves human papillomavirus detection by PCR. JGenVirol. 1995;76(pt 4):1057-1062.

(32.) Ylitalo N, Bergstrom T, Gyllensten U. Detection of genital human papil lomavirus by single-tube nested PCR and type-specific oligonucleotide hybridization. J Clin Microbiol. 1995;33(7):1822-1828.

(33.) Gravitt P, Peyton C, Alessi T, et al. Improved amplification of genital human papillomaviruses. JClin Microbiol. 2000;38(1):357-361.

(34.) Goleski V, Dawood M, Smart G, et al. A microsphere based multiplexed assay for genotyping 46 mucosal human Papiloomavirus types. In: Canadian Association of Clinical Microbiologists and Infectious Disease Annual Conference. Hamilton, ON: Canadian Association of Clinical Microbiologists and Infectious Disease; 2007;6. Abstract SP28.

(35.) Goleski V, Dawood M, Ratnam S, Severini A. Luminex[R]-based assay for multiplex genotyping of 45 mucosal human papillomavirus types: FC 3a-5. Paper presented at: Eurogin: The 8th International Multidisciplinary Congress--Joining Forces for Cervical Cancer Prevention; November 13, 2008; Nice, France.

(36.) Bar T, Stahlberg A, Muszta A, Kubista M. Kinetic outlier detection (KOD) in real-time PCR. NucleicAcids Res. 2003;31(17):e105.

(37.) Rutledge R, Cote C. Mathematics of quantitative kinetic PCR and the application of standard curves. Nucleic Acids Res. 2003;31(16):e93.

(38.) Nettleman MD. Receiver operator characteristic (ROC) curves. Infect Control Hosp Epidemiol. 1988;9(8):374-377.

(39.) Gunnarsson R, Lanke J. The predictive value of microbiologic diagnostic tests if asymptomatic carriers are present. Stat Med. 2002;21(12):1773-1785.

(40.) Macaskill P, Walter S, Irwig L, Franco E. Assessing the gain in diagnostic performance when combining two diagnostic tests. Stat Med. 2002;21(17):2527 2546.

(41.) Ostor A. Natural history of cervical intraepithelial neoplasia: a critical review. Int JGynecol Pathol. 1993;12(2):186-192.

(42.) Andersson S, Hansson B, Norman I, etal. Expression ofE6/E7 mRNAfrom 'high risk' human papillomavirus in relation to CIN grade, viral load and p16INK4a. Int JOncol. 2006;29(3):705-711.

(43.) De Boer MA, Jordanova ES, Kenter GG, et al. High human papillomavirus oncogene mRNA expression and not viral DNA load is associated with poor prognosis in cervical cancer patients. Clin Cancer Res. 2007;13(1):132-138.

(44.) Srodon M, ParryDilworth H, RonnettBM. Atypical squamouscells,cannot exclude high-grade squamous intraepithelial lesion: diagnostic performance, human papillomavirus testing, and follow-up results. Cancer. 2006;108(1):32-38.

Nick A. Antonishyn, PhD; Greg B. Horsman, MD; Rod A. Kelln, PhD; Alberto Severini, MD

From the Department of Chemistry and Biochemistry, University of Regina, Regina, Saskatchewan, Canada (Drs Antonishyn and Kelln); the Saskatchewan Disease Control Laboratory, Regina (Dr Horsman); and the Public Health Agency of Canada, National Microbiology Laboratory, Winnipeg, Manitoba, Canada (Dr Severini).

The authors have no relevant financial interest in the products or companies described in this article.

Reprints: Nick A. Antonishyn, PhD, Molecular Diagnostics, Saskatchewan Disease Control Laboratory, 3211 Albert St, Regina, SKS4S5W6, Canada (e-mail: nantonis@health.gov.sk.ca).
Table 1. Oligonucleotides

                                                             Final
Name                     Sequence (5' to 3')             Concentration

Gp5 +                    TTT GTT ACT GTG GTA GAT              0.5
                           ACT AC
Gp6 +                    GAA AAA TAA ACT GTA AAT              0.5
                           CAT ATT C
E1 350L                  TRY RKG YYY TAA AAC GAA AGT          0.5
E1 547R                  TTC CAC TTC AGW AYW GCC ATA          0.5
E1 847R                  CAA ATC DSW ACA BST KSW TTT          0.25
                           ATY RCT YTK AAA
HPV-16 L1 forward (a)    GCT GGT TTG GGC CTG TGT AG           0.3
HPV-16 L1 reverse (a)    GGC CAC TAA TGC CCA CAC C            0.3
HPV-16 L1 probe (a,b)    ATG GCT GAC CAC GAC CTA CCT          0.2
                           CAA CA
HPV-16 E6 forward        CTG CAA TGT TTC AGG ACC CA           0.1
HPV-16 E6 reverse        TCA TGT ATA GTT GTT TGC AGC          0.1
                           TCT GT
HPV-16 E6 probe (b)      AGG AGC GAC CCG GAA AGT TAC          0.15
                           CAC AGT T
HPV-16 E7 forward        AAG TGT GAC TCT ACG CTT CGG          0.1
                           TT
HPV-16 E7 reverse        GCC CAT TAA CAG GTC TTC CAA A        0.1
HPV-16 E7 probe (b)      TGC GTA CAA AGC ACA CAC GTA          0.15
                           GAC ATT CGT A
HPV-31 E6 forward (a)    AAC CTA CAG ACG CCA TGT              0.1
HPV-31 E6 reverse (a)    AAT GCC GAG CTT AGT TCA              0.1
HPV-31 E6 probe (a,b)    AAT CCT GCA GAA AGA CCT CGG A        0.15
RPS9 forward             ATC CGC CAG CGC CAT ATC              0.2
RPS9 reverse             TCG ATG TGC TTC TGG GAA TCC          0.2
RPS9 probe (b)           AGC AGG TGG TGA ACA TCC CGT          0.1
                           CCT T

Abbreviations: HPV, human papillomavirus; RPS9, ribosomal
protein S9.

(a) Original design.

(b) All probes were labeled with FAM (6-carboxy-fluorescein;
Applied Biosystems, Streetsville, Ontario, Canada) as the
reporter dye and with Black Hole Quencher (BHQ-1; Biosearch
Technologies Inc, Novato, California) as the quencher.

Table 2. Cross-Classified Repeat Papanicolaou lest and Human
Papillomavirus (HPV) DNA With Categories of Disease (n = 887)

                        Disease Category, No. (a)

Papanicolaou     Normal HPV DNA        CIN 1 HPV DNA
Test Result
               Positive   Negative   Positive   Negative

HSIL               5          2         13         2
LSIL              14          8         84         24
ASCUS             20         19         30         21
Normal            59        147         58         97
Total             98        176        185        144

               Disease Category, No. (a)

Papanicolaou     CIN 2/3 HPV DNA
Test Result
               Positive   Negative

HSIL             156          7
LSIL              51          9
ASCUS             17          3
Normal            30         11
Total            254         30

Abbreviations: ASCUS, atypical squamous cells of undetermined
significance; CIN, cervical intraepithelial neoplasia; HSIL,
high-grade squamous intraepithelial lesion; LSIL, low-grade
squamous intraepithelial lesion.

(a) Derived from histologic grading of biopsies obtained during
colposcopic examinations with CIN grades 1, 2, and 3
corresponding to  abnormal-low, abnormal-medium, and
abnormal-high, respectively.

Table 3. Cross-Classified Repeat Papanicolaou Test and Human
Papillomavirus (HPV) Genotype With  Categories of Disease
(n = 887)

                     Disease Category, No. (a)

                 Normal       CIN 1        CIN 2/3

                Genotype     Genotype     Genotype
                Risk (b)       Risk         Risk
Papanicolaou
Test Result     High   Low   High   Low   High   Low

HSIL              5      2    13     2    147    16
LSIL             14      8    68    40     47    13
ASCUS            17     22    21    30     14     6
Normal           50    156    44    111    30    11
Total            86    188   146    183   238    46

Abbreviations: ASCUS, atypical squamous cells of undetermined
significance; CIN, cervical intraepithelial neoplasia; HSIL,
high-grade squamous intraepithelial lesion; LSIL, low-grade
squamous intraepithelial lesion.

(a) Derived from histologic grading of biopsies obtained during
colposcopic examinations with CIN grades 1, 2, and CIN 3,
corresponding to abnormal-low, abnormal-medium, and
abnormal-high, respectively.

(b) Genotype information was used to categorize women based on
published risk classification. High-risk category included
probable high-risk types, HPV-23, HPV-53, and HPV-66, and
unclassified HPV types. Low-risk included women who had negative
findings for HPV DNA+.

Table 4. Cross-Classified Repeat Papanicolaou Test and Human
Papillomavirus (HPV) 16 E6 RNA With Categories of Disease for
Women With Positive HPV-16 Results (n = 121)

                       Disease Category, No. (a)

                      Normal                CIN 1

                  HPV-16 E6 RNA         HPV-16 E6 RNA
Papanicolaou
Test Result    Positive   Negative   Positive   Negative

HSIL              0          1           1          1
LSIL              2          1          10          4
ASCUS             0          1           1          1
Normal            4          5           4          6
Total             6          8          16         12

               Disease Category, No. (a)

                        CIN 2/3

                     HPV-16 E6 RNA
Papanicolaou
Test Result       Positive   Negative

HSIL                 50          4
LSIL                 14          1
ASCUS                 2          0
Normal                2          6
Total                68         11

Abbreviations: ASCUS, atypical squamous cells of undetermined
significance; CIN, cervical intraepithelial neoplasia; HSIL,
high-grade squamous intraepithelial lesion; LSIL, low-grade
squamous intraepithelial lesion.

(a) Derived from histologic grading of biopsies obtained during
colposcopic examinations with CIN grades 1, 2, and CIN 3,
corresponding to abnormal-low, abnormal-medium, and
abnormal-high, respectively.

Table 5. Cross-Classified Repeat Papanicolaou Test
and Human Papillomavirus 16 (HPV-16) E6 Relative
Quantification (RQ) 0.5 With Categories of Disease
for Women With Positive HPV-16 Results (n = 121)

                   Disease Category (a,b)

                Normal      CIN 1       CIN 2/3
Papanicolaou
Test Result    RQ+   RQ-   RQ+   RQ-   RQ+   RQ-

HSIL            0     1     1     1    43    11
LSIL            2     1     9     5     9     6
ASCUS           0     1     1     1     2     0
Normal          2     7     4     6     2     6
Total           4    10    15    13    56    23

Abbreviations: ASCUS, atypical squamous cells of undetermined
significance; CIN, cervical intraepithelial neoplasia; HSIL,
high-grade squamous intraepithelial lesion; LSIL, low-grade
squamous intraepithelial lesion.

(a) Derived from histologic grading of biopsies obtained during
colposcopic examinations with CIN grades 1, 2, and CIN 3,
corresponding to abnormal-low, abnormal-medium, and abnormal-high,
respectively.

(b) Relative quantification of HPV-16 E6 messenger RNA using a
cutoff of 0.5 for positive results.

Table 6. Comparison of Dianostic Test Performance for Women With
Initial Abnormal Papanicolaou Test Results During Cervical Cancer
Screening (n = 887)

Histology and Method                  Sensitivity   Specificity

Any abnormality
  Repeat Papanicolaou test (c)           0.68          0.75
  HPV DNA                                0.72          0.64
  HPV genotype                           0.63          0.69
  Papanicolaou test or HPV DNA (d)       0.82          0.54
  Papanicolaou test or genotype (d)      0.80          0.57
  Papanicolaou test and HPV DNA (e)      0.57          0.78

[greater than or equal to] CIN 2
  Repeat Papanicolaou testc              0.86          0.60
  HPV DNA                                0.89          0.53
  HPV genotype                           0.84          0.62
  Papanicolaou test or HPV DNA (d)       0.96          0.40
  Papanicolaou test or genotype (d)      0.96          0.44
  Papanicolaou test and HPV DNA (e)      0.79          0.72

Histology and Method                  LR+ (a)   LR- (b)

Any abnormality
  Repeat Papanicolaou test (c)         2.74      0.43
  HPV DNA                              2.00      0.44
  HPV genotype                         2.00      0.54
  Papanicolaou test or HPV DNA (d)     1.78      0.33
  Papanicolaou test or genotype (d)    1.86      0.35
  Papanicolaou test and HPV DNA (e)    2.59      0.55

[greater than or equal to] CIN 2
  Repeat Papanicolaou testc            2.13      0.24
  HPV DNA                              1.91      0.20
  HPV genotype                         2.18      0.26
  Papanicolaou test or HPV DNA (d)     1.61      0.10
  Papanicolaou test or genotype (d)    1.73      0.09
  Papanicolaou test and HPV DNA (e)    2.82      0.29

Abbreviations: CIN, cervical intraepithelial neoplasia; HPV, human
papillomavirus; LR, likelihood ratio.

(a) The positive LR is given by LR+ = sensitivity/(1--specificity).

(b) The negative LR is given by LR- = (1--sensitivity)/specificity.

(c) Repeat Papanicolaou test includes high-grade squamous
intraepithelial lesions, low-grade squamous intraepithelial lesions,
and atypical squamous cells of undetermined significance.

(d) Combined methods used the either-positive rule.

(e) Combined methods used the both-positive rule.

Table 7. Comparison of Diagnostic Test Performance for Patients
Screened Who Had Abnormal Papanicolaou Test Results and Findings for
Human Papillomavirus 16 (HPV-16) Were Positive (n = 121)

      Histology and Method (a)         Sensitivity   Specificity

Any abnormality
  Repeat Papanicolaou test (d)            0.83          0.64
  HPV-16 E6 RNA                           0.79          0.57
  HPV-16 E6 RQ                            0.66          0.71
  Papanicolaou test or HPV-16 E6 RNA      0.89          0.36
  Papanicolaou test or HVP-16 E6 RQ       0.89          0.50

[greater than or equal to] CIN 2
  Repeat Papanicolaou test (d)            0.90          0.45
  HPV-16 E6 RNA                           0.86          0.48
  HPV-16 E6 RQ                            0.71          0.55
  Papanicolaou test or HPV-16 E6 RNA      0.92          0.26
  Papanicolaou test or HPV-16 E6 RQ       0.92          0.31

      Histology and Method (a)         LR+ (b)   LR- (c)

Any abnormality
  Repeat Papanicolaou test (d)          2.33      0.26
  HPV-16 E6 RNA                         1.83      0.38
  HPV-16 E6 RQ                          2.32      0.47
  Papanicolaou test or HPV-16 E6 RNA    1.38      0.31
  Papanicolaou test or HVP-16 E6 RQ     1.78      0.22

[greater than or equal to] CIN 2
  Repeat Papanicolaou test (d)          1.64      0.22
  HPV-16 E6 RNA                         1.64      0.29
  HPV-16 E6 RQ                          1.57      0.53
  Papanicolaou test or HPV-16 E6 RNA    1.25      0.29
  Papanicolaou test or HPV-16 E6 RQ     1.34      0.25

Abbreviations: CIN, cervical intraepithelial neoplasia; LR,
likelihood ratio; RQ, relative quantification.

(a) Combined methods used the either-positive rule.

(b) The positive LR is given by LR+ = sensitivity/(1--specificity).

(c) The negative LR is given by LR- = (1--sensitivity)/specificity.

(d) Repeat Papanicolaou test includes high-grade squamous
intraepithelial lesions, low-grade squamous intraepithelial lesions,
and atypical squamous cells of undetermined significance.

Table 8. Effect of Cascade Testing on Triage Numbers

                                  Normal, CIN 1      CIN 2/3
Cascade Options                   (No. Excluded)   (No. Missed)

Tier 1
  Initial Screen (n = 887)          603 (NA)         284 (NA)
  Repeat Papanicolaou test (a)      242 (361)        243 (41)
  HPV DNA                           283 (320)        254 (30)
  Papanicolaou test or
    HPV DNA (b)                     359 (244)        273 (11)
  Papanicolaou test or
    genotype (b)                    336 (267)        273 (11)
  Papanicolaou test and HPV (a)     166 (437)        224 (60)
Tier 2 (HPV-16 positive)
  Initial screen (n = 121)           42 (NA)          79 (NA)
  Repeat Papanicolaou test           23 (19)          71 (8)
  HPV-16 E6 RNA                      22 (20)          68 (11)
  HPV-16 E6 RQ                       19 (23)          56 (23)
  Papanicolaou test or
    HPV-16 E6 RNA (b)                31 (11)          73 (6)
  Papanicolaou test or
    HPV-16 E6 RQ (b)                 29 (13)          73 (6)

Abbreviations: HPV, human papillomavirus; LR, likelihood ratio; RQ,
relative quantification.

(a) Combined method used the both-positive rule.

(b) Combined method used the either-positive rule.
Gale Copyright: Copyright 2009 Gale, Cengage Learning. All rights reserved.