Grading the severity of cervical neoplasia based on combined histopathology, cytopathology, and HPV genotype distribution among 1,700 women referred to colposcopy in Oklahoma

Authors


Abstract

Diagnosis and treatment of cervical cancer precursors rely on colposcopic biopsy, which is sometimes hampered by incorrect biopsy placement and the unclear prognostic value of poorly reproducible diagnoses such as cervical intraepithelial neoplasia (CIN) Grade 1 and 2. Searching for discrete disease categories that incorporate the value of cytology and that reflect the causal role of particular HPV types, we analyzed histology, cytology and HPV genotype distributions in the Study to Understand Cervical Cancer Endpoints and Early Determinants (SUCCEED). This cross-sectional study comprises ∼1,700 women referred to colposcopy or treatment for the spectrum of cervical disease, including 439 women with <CIN1, 429 CIN1, 322 CIN2, 297 CIN3 and 107 with cancer. Using hierarchical clustering of histology-cytology groups based on HPV genotype distributions, we could plainly distinguish in this referral population 5 increasingly severe diagnostic groups of HPV-positive women: (i) HPV-positive women with <CIN2 histology and normal cytology, (ii) HPV positive women with <CIN2 histology and ASC or LSIL cytology; (iii) CIN2, including histologic CIN2 and HSIL cytology with any histology <CIN2; (iv) CIN3; and (v) invasive cervical cancer. The grouping of women with HSIL cytology, but without histological abnormalities to women with CIN2 suggests that in these cases the worst lesion was missed during colposcopy-biopsy. We are now using these sharpened diagnostic categories to search for novel biomarkers predicting the risk of progression and invasion. © 2008 Wiley-Liss, Inc.

Persistent infections with carcinogenic genotypes of human papillomaviruses (HPV) cause almost all cervical cancers and their precursors, cervical intraepithelial neoplasia (CIN), recognized cytologically as squamous intraepithelial lesions (SIL). Based on epidemiological and experimental data, a natural history model of HPV-driven cervical carcinogenesis has been proposed: (i) HPV infection of the cervical metaplastic epithelium (typically occult, but producing concurrent, clinically observed cytopathology in 20–30%,1) (ii) persistence rather than clearance of infection, (iii) progression to an immediate precursor of cancer (i.e., CIN3 approximating carcinoma in situ) that slowly grows, and (iv) invasion.2, 3

Over 100 HPV types have been identified to date. They are heterogeneous with respect to their epithelial tropism and ability to induce cancer precursors and, in turn, to cause cancer. Of the 40 types that infect genital mucosa, 14 (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68) are strongly associated with risk of developing invasive cancer.4

HPV infections, even those involving carcinogenic types, are very common in the population: it is estimated that over 70% of all women are infected with HPV during their lifetime.5 Most of these infections clear spontaneously within a year and virtually all clear within 2–3 years; however, in fewer than 10%, infections persist with an increasing risk of diagnosis of carcinomas in situ. These gradually grow to large precancerous lesions that have a 30–50% risk of invasion over the remainder of a woman's life.6 Among the carcinogenic HPV infections, there is substantial and important heterogeneity.7 HPV16 and HPV18 have a much higher risk of causing carcinoma in situ and invasive cancer as compared to the remaining HR types.8 Accordingly, large meta-analyses based on cross-sectional data have demonstrated that the distribution of HPV types varies significantly between different disease stages.9

One of the most important remaining tasks in HPV translational science is to discover biomarkers that can predict the small proportion of HPV infections that will persist and cause carcinoma in situ, the accepted surrogate endpoint for cancer risk (and universally accepted treatment threshold). For example, one hallmark of the transition from acute infection to carcinoma in situ is a change in the viral expression pattern with strong increase of viral oncogene expression.10, 11 The viral oncogenes E6 and E7 interfere with cell cycle control and apoptosis and induce chromosomal instability.12 Several candidate biomarkers for this step have been proposed and are being explored13; however, which biomarkers should be used clinically remains unknown.

At present, many clinicians and researchers continue to rely on traditional, histologic gradations—CIN1, CIN2 and CIN3 (including carcinoma in situ)—to estimate the cervical cancer risk posed by specific HPV-related changes. CIN grading is based on the percentage of the epithelium replaced by abnormal appearing, immature, proliferating cells. Although related to future cancer risk and useful clinically, it is difficult to validate new molecular biomarkers using the CIN scale alone.

CIN grading is limited by subjectivity and poor reproducibility especially in diagnosing CIN1 and CIN2.14 The accuracy of histopathologic diagnosis is limited further by the tendency of colposcopic biopsy to miss small CIN3 lesions almost 50% of the time.15 To compound the problem, noncarcinogenic HPV types cause a substantial fraction of “look alike” CIN1 and even CIN2 lesions which are unrelated to cancer risk.

Discovering critical cervical biomarkers to clarify the risk of progressive steps in the pathogenesis of cervical cancer, particularly those that can eventually be assessed for the entire cervix using an exfoliative specimen, is the major goal of the Study to Understand Cervical Cancer Endpoints and Early Determinants (SUCCEED), a large cross-sectional study analyzing epidemiological, pathologic and molecular biologic alterations in women with the full spectrum of HPV infections, precancer and cancer (Wang submitted).

The first task in SUCCEED was to define what we consider to be the best possible categories that represent disease severity. We hypothesized that combining histology, cytology, and HPV genotyping data would permit the most biologically meaningful disease stratifications.

In the present analysis, we present revised disease stages to set the framework for SUCCEED's future goal of novel biomarker discovery and analysis.

Material and methods

Study population

Enrollment into SUCCEED started in November 2003 and ended for the main phase comprising full population accrual in September 2007. The population includes all women referred to colposcopy at the University of Oklahoma Dysplasia Clinic based at the University of Oklahoma Health Sciences Center (OUHSC), following an abnormal Pap smear diagnosis or a biopsy diagnosis of CIN. Women who were less than 18 years of age, pregnant at the time of their visit, previously treated with chemotherapy or radiation for any cancer, or women scheduled for vaginal colposcopy were excluded from the study. Written informed consent was obtained from all women enrolled into the study and Institutional Review Board approval was provided by OUHSC and the U.S. National Cancer Institute.

At the time of the analysis, 1,899 women had been enrolled; of these, we excluded 16 from the present analyses due to missing HPV results and further excluded 213 women due to unsatisfactory cytology, constituting a study population of 1,670 women (88%) with a median age of 25 years (18–81 years). The baseline characteristics of the study population by histology diagnosis are summarized in Supporting Table I.

Colposcopy and specimen collection

Colposcopic examination was conducted by a gynecologist according to routine practice at OUHSC. Prior to biopsy or loop electrosurgical excision procedure (LEEP), cervical cell samples were first obtained with a Papette™ broom (Wallach Surgical, Orange, CT) and rinsed directly into PreservCyt™ solution (Cytyc Corporation, Boxborough, MA) as described previously.16 The cytology specimen was used for ThinPrep™ (Cytyc Corporation) cytology and for HPV genotyping using the Linear Array (LA) HPV Genotyping System (Roche Diagnostics, Branchburg, NJ). Biopsy specimens were obtained for colposcopically suspicious lesions. Endocervical curettage was performed in cases where the entire transformation zone or extent of a lesion was not visualized adequately. As per standard practice, all histologically confirmed high-grade lesions diagnosed as CIN2 or above (CIN2+) were treated by LEEP of the transformation zone.

Cervical cytology and histology

Liquid-based, ThinPrep cytology slides were prepared from PreservCyt vial specimens according to the manufacturer's standard protocol. Slides were evaluated as previously described by both a cytotechnologist and cytopathologist.17 Cytologic results were recorded on a standardized data collection form based on the Bethesda System.18 Histologic interpretation of biopsy and LEEP specimens was conducted by the study pathologist at OUHSC (REZ) using CIN terminologies. Cytology and histologic diagnoses were masked to each other and to genotyping data.

HPV genotyping

Details of DNA isolation and HPV genotyping employed in SUCCEED have been previously described.19 Briefly, DNA was isolated from 1 mL aliquots of PreservCyt-fixed cells using the QIAamp DNA Blood Mini Kit (Qiagen Sciences, Germantown, MD) following a brief rinse in Hanks' Balanced Salt Solution (HBSS). Isolated DNA was stored at −70°C until PCR amplification using the Linear Array® HPV Genotyping System (Roche Diagnostics). The LA assay is capable of detecting 37 HPV genotypes (6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 58, 59, 61, 62, 64, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82, 83, 84, IS39, CP6108). Two different concentrations of β-globin probes are present on each strip as internal positive controls for assuring adequate amplification of each specimen. Up to 80 patient specimens and up to 4 HPV-positive and -negative control specimens were amplified at one time using the LA assay, following the manufacturer's instructions. Control specimens were processed through DNA isolation, amplification and detection similar to patient specimens. Hybridization of PCR products to linear arrays and subsequent signal detection were performed using the Auto-LiPA automated staining system (Innogenetics N.V., Belgium). Hybridization to both β-globin probes was required to report genotyping results. A positive hybridization signal was called when an unambiguous, continuous band was observed on the array. A single evaluator subjectively graded the intensity of each hybridization band as strong (s), moderate (m), weak (w), very weak (vw) or extremely weak (ew) as previously described in Ref. 20. HPV genotyping was performed masked to all other patient data.

Statistical analyses

As a starting place for our exploration of optimal disease categories, the histology categories used for the initial analysis were <CIN1, CIN1, CIN2, CIN3 and Cancer. The cytology categories used were as follows: negative for intraepithelial lesion or malignancy (NILM), atypical squamous cells (ASC, including ASC-US and ASC-H), low grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion and greater (HSIL+).18 ASC mainly represented ASC-US; we included ASC-H as part of ASC in this epidemiologic classification analysis because it is an uncommon cytologic interpretation (n = 113 out of 1,670 total samples = 6.8%; 52 out of 749 <CIN2 samples = 6.9%). However, as properly noted in recent guidelines,21 ASC-H is associated with a higher prevalence of HPV16 and other carcinogenic HPV types, as well as a higher risk of more severe disease categories. Excluding ASC-H did not change the conclusions.

On the basis of the published literature, we divided HPV infection status into categories of decreasing risk: HPV16, other carcinogenic (18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68), noncarcinogenic (6, 11, 26, 40, 42, 53, 54, 55, 61, 62, 64, 67, 69, 70, 71, 72, 73, 81, 82, 83, 84, IS39, CP6108), and negative.

Independent of this a priori risk approach to use HPV categories, we applied hierarchical clustering based on the frequencies of all individual HPV types to analyze the “distances” between possible disease groups defined by histology and cytology. This permitted 3-variable (histology, cytology, HPV) prediction of which disease categories might best be combined or kept separate. HPV frequencies were used to create a self-organizing map followed by hierarchical clustering. For cluster analysis, we used the Cluster 3.0 program,22 the results were visualized using heat maps and dendrograms generated by the TreeView program.23

Rare histology-cytology combinations with less than 20 cases were excluded from the clustering (CIN2-NILM, CIN3-NILM, CA-NILM, CA-ASC, CA-LSIL). Because the HPV distributions were very similar, we combined CIN3-ASC and CIN3-LSIL into one category (CIN3-A/L). To explore potential confusion in clustering based on multiple infections, we reanalyzed frequencies of HPV types for single infections separately.

We used the χ2 test for trend to analyze the percentage of strong signals across disease categories. All statistical tests were 2-sided and considered to be significant at p < 0.05. Statistical analyses were performed using SAS, version 9.1 (SAS Institute, Cary, NC).

Results

HPV prevalence in histology-cytology combinations

As a first step toward revising diagnostic categories that integrate cytology and HPV results, we analyzed the HPV type distributions for all combinations of cytology and histology. Table I shows the case numbers for each cytology–histology category together with frequencies of HPV types, classified in a hierarchical manner based on their a priori assumed relation to cancer risk: HPV16 (including multiple infections with other carcinogenic or noncarcinogenic types), other carcinogenic types without HPV16 (including multiple infections with noncarcinogenic types), and noncarcinogenic types alone.

Table I. HPV Prevalence by Histology and Cytology Results
CytologyStatistics<CIN1CIN1CIN2CIN3CANATotal
  1. HR, carcinogenic HPV types; LR, noncarcinogenic HPV types; NA, not available.

NILMn156701910112268
HPV1627 (17.3%)6 (8.6%)8 (42.1%)6 (60%)1 (100%)5 (41.7%)53 (19.8%)
HR(-HPV16)69 (44.2%)34 (48.6%)9 (47.4%)4 (40%)0 (0%)6 (50%)122 (45.5%)
LR19 (12.2%)14 (20%)1 (5.3%)0 (0%)0 (0%)1 (8.3%)35 (13.1%)
Total HPV115 (73.7%)54 (77.1%)18 (94.7%)10 (100%)1 (100%)12 (100%)210 (78.4%)
ASCn1631347134617425
HPV1637 (22.7%)31 (23.1%)31 (43.7%)25 (73.5%)3 (50%)5 (29.4%)132 (31.1%)
HR(-HPV16)81 (49.7%)77 (57.5%)33 (46.5%)9 (26.5%)1 (16.7%)6 (35.3%)207 (48.7%)
LR18 (11%)16 (11.9%)3 (4.2%)0 (0%)0 (0%)5 (29.4%)42 (9.9%)
Total HPV136 (83.4%)124 (92.5%)67 (94.4%)34 (100%)4 (66.7%)16 (94.1%)381 (89.6%)
LSILn57169591836312
HPV1614 (24.6%)32 (18.9%)31 (52.5%)15 (83.3%)0 (0%)1 (16.7%)93 (29.8%)
HR(-HPV16)34 (59.6%)111 (65.7%)26 (44.1%)3 (16.7%)1 (33.3%)4 (66.7%)179 (57.4%)
LR9 (15.8%)24 (14.2%)2 (3.4%)0 (0%)0 (0%)1 (16.7%)36 (11.5%)
Total HPV57 (100%)167 (98.8%)59 (100%)18 (100%)1 (33.3%)6 (100%)308 (98.7%)
HSIL+n49571722429635651
HPV1625 (51%)28 (49.1%)77 (44.8%)174 (71.9%)54 (56.3%)18 (51.4%)376 (57.8%)
HR(-HPV16)18 (36.7%)29 (50.9%)90 (52.3%)65 (26.9%)32 (33.3%)17 (48.6%)251 (38.6%)
LR3 (6.1%)0 (0%)5 (2.9%)2 (0.8%)4 (4.2%)0 (0%)14 (2.2%)
Total HPV46 (93.9%)57 (100%)172 (100%)241 (99.6%)90 (93.8%)35 (100%)641 (98.5%)
NAn81011314
HPV163 (37.5%)0 (0%)0 (0%)1 (100%)1 (100%)1 (33.3%)6 (42.9%)
HR(-HPV16)2 (25%)0 (0%)0 (0%)0 (0%)0 (0%)0 (0%)2 (14.3%)
LR1 (12.5%)1 (100%)0 (0%)0 (0%)0 (0%)0 (0%)2 (14.3%)
Total HPV6 (75%)1 (100%)0 (0%)1 (100%)1 (100%)1 (33.3%)10 (71.4%)
Totaln433431321305107731,670
HPV16106 (24.5%)97 (22.5%)147 (45.8%)221 (72.5%)59 (55.1%)30 (41.1%)660 (39.5%)
HR(-HPV16)204 (47.1%)251 (58.2%)158 (49.2%)81 (26.6%)34 (31.8%)33 (45.2%)761 (45.6%)
LR50 (11.5%)55 (12.8%)11 (3.4%)2 (0.7%)4 (3.7%)7 (9.6%)129 (7.7%)
Total HPV360 (83.1%)403 (93.5%)316 (98.4%)304 (99.7%)97 (90.7%)70 (95.9%)1,550 (92.8%)

The main carcinogenic type, HPV16, was observed in 9–25% of the least severe abnormalities (<CIN1 or CIN1 histology combined with NILM, ASC or LSIL cytology). Interestingly, the frequencies of HPV16 and of all other carcinogenic HPV types combined were not higher in CIN1 histology than in <CIN1 histology. In fact, the lowest percentage of HPV16 was found in CIN1 + NILM (9%). Thus, we combined the histologic categories of <CIN1 and CIN1 as <CIN2 because of their virologic similarity, the known intractable problems in making reliable histologic or cytologic distinctions24 and the lack of large differences in prospective risk of subsequent CIN3+.25

Hierarchical clustering of disease groups by HPV genotype profiles

We used hierarchical clustering by HPV genotype distributions to explore without a priori risk assumptions distinctions between the different histology-cytology disease groups. Figure 1a shows the dendrogram of the disease categories based on the HPV genotype frequencies including all infections displayed in the heat map. The prominence of HPV16 is evident across the disease spectrum. HPV18 was observed to be the second most frequent type among cancers but relatively lacking in CIN3 lesions. HPV types within phylogenetic species tended to be similarly distributed with regard to disease category.

Figure 1.

Clustering of disease groups by HPV genotypes. (a) Hierarchical clustering of disease groups by HPV type frequencies, including multiple infections. (b) Hierarchical clustering of disease groups by HPV type frequencies, single-type infections only LC2: < = CIN2; N: NILM; A: ASCUS; L: LSIL; H: HSIL. Types are grouped by species.

The clustering revealed the greatest distance between combinations of <CIN2 histology with <HSIL cytology and cancers or CIN3 with HSIL+. Most interestingly, the disease group with <CIN2 histology and HSIL cytology was grouped with CIN2 and CIN3. Thus, in the <CIN2 group, cytology was important (NILM, vs. ASC/LSIL vs. HSIL) while CIN1 was not important in determining the virologic patterns. But when histology was CIN2+, the distribution of HPV types in the histological categories did not differ by cytology results.

Figure 1b shows the same analysis, restricted to cases with single HPV genotype infections. Because of the low number of cases with single genotype HPV infections, additional disease categories (CIN2-LSIL, CIN2-ASC, CIN3-ASC/LSIL) had to be excluded. Still the grouping of disease categories was highly similar compared to the analysis including multiple genotype infections, most importantly, <CIN2-HSIL was again grouped with CIN2 and CIN3. As expected, in high-grade lesions, the types that appear as single infections are also predominant among the multiple infections.

HPV signal strength (viral load)

We used visual signal intensity data as an approximation for viral load, as previously described for a similar PCR-based HPV genotyping method.26 We analyzed signal intensity by type and by disease category for 2,346 carcinogenic infections (Table II). On the basis of the previous cluster analysis, we established the following disease groups: <CIN2-NILM, <CIN2-ASC/LSIL, CIN2 + HSIL (if histology was <CIN2), CIN3, and cancer. All carcinogenic types showed the lowest percentage of strong signals in the <CIN2-NILM group (40%) and the highest percentage in cancers (71–88%). There was a significant increase of strong signals from <CIN2-NILM to <CIN2-ASC/LSIL for all carcinogenic types (p < 0.0001). For HPV16, increasing signal strength in a steady, stepwise fashion through all disease categories (p < 0.0001) was observed. For the remaining carcinogenic types, no significant increase from <CIN2-ASC/LSIL to CIN2 and to CIN3 was observed (p = 0.18) although, as noted, signals tended to be strongest in cancer.

Table II. Signal Strength Distribution by Disease Category for Carcinogenic Types
 Carcinogenic types without HPV16HPV16 only
Disease groupnStrongMediumWeakV. weaknStrongMediumWeakV. weak
  1. All carcinogenic infections are presented (n = 2,346). Women with multiple carcinogenic infections can be represented more than once. Visual categories of Linear Array signal strength are shown. The visual categories very weak and extremely weak were combined to v. weak. <CIN2 is separated into <CIN2 without cytological abnormalities (NILM) and with cytological findings (ASC/LSIL). Row percentages for signal strength distribution by disease category are given. HR, carcinogenic.

<CIN2 and NILM18240.1%31.9%13.2%14.8%3339.4%33.3%12.1%15.2%
<CIN2 and ASC-LSIL65958.4%20.3%7.4%13.8%11464.0%21.1%7.9%7.0%
CIN2 or HSIL+53457.9%20.8%8.4%13.0%20077.0%12.5%3.0%7.5%
CIN328952.6%25.6%5.5%16.3%22190.5%7.7%1.4%0.5%
Cancer5570.9%7.3%1.8%20.0%5988.1%8.5%1.7%1.7%
p trend 0.18    <0.0001   

Revision of disease scale based on histology, cytology, and HPV genotyping

On the basis of the hierarchical clustering and the signal strength distribution, we developed the following disease groups as distinct strata for disease progression in the SUCCEED referral population (Fig. 2): (i) a small group of women referred but found to have normal histology, normal cytology and HPV negativity, (ii) HPV positive women with <CIN2-NILM that had only low-level viral activity, (iii) HPV positive women with productive (high-level viral particle production) viral infections (including HPV-positive <CIN2-ASCUS, <CIN2-LSIL), (iv) CIN2 (including CIN2 histology regardless of cytology and <CIN2-HSIL), (v) CIN3 (regardless of cytology) and (vi) cancer (regardless of cytology).

Figure 2.

Distinct disease stages identified in the SUCCEED population based on histology, cytology, and HPV genotyping. (a) Dendrogram of histology-cytology combination clustered by HPV genotype frequencies (derived from Fig. 1). (b) Viral load as determined by Linear Array signal strength in the 6 disease stages (percentage of strong signals): (i) women with normal histology, normal cytology and HPV negativity, (ii) HPV positive women with <CIN2-NILM and low-level viral activity, (iii) HPV positive women with productive viral infections (including HPV-positive <CIN2-ASCUS, <CIN2-LSIL), (iv) CIN2 (including CIN2 histology regardless of cytology and <CIN2-HSIL), (v) CIN3 (regardless of cytology) and (vi)cancer (regardless of cytology). Although there is a gradient of HPV16 viral load with increasing disease stages, this is not observed for the remaining HR-HPV genotypes. (c) Functional viral-host interaction during cervical carcinogenesis correlating to the findings in the SUCCEED population.

Discussion

We present a comprehensive analysis of HPV genotypes in SUCCEED, a cross-sectional study of women referred to colposcopy due to recently abnormal cytology or histology. The study includes 1,670 women spanning all stages of cervical squamous neoplasia, with a high proportion of cervical cancer and precursor lesions. The focus of our analysis was to combine histology, cytology and HPV genotype data in order to create optimal disease severity categories for use in our search for biomarkers of risk of progression and invasion. By adding cytology and extensive HPV genotyping using a standardized assay detecting 37 HPV genotypes, we sought to extend the previous cross-sectional reports on HPV typing and histology.27, 28

We analyzed HPV types in all relevant cytology/histology combinations and found that the distinction of CIN1 and <CIN1 is not relevant in women with low grade cytology. This is consistent with our analysis of HPV-cofactors that also show no difference between <CIN1 and CIN1 (unpublished data). The term, “histologically confirmed CIN1” is still widely used and we offer the suggestion that the clinical value of this term may be questioned. For example, a postcolposcopic definition of CIN1 is actually a lower-risk diagnosis than precolposcopic LSIL, where obvious, prevalent CIN2 or worse has not yet been ruled out.25

One hundred and six of the 1,670 women (6.3%) in SUCCEED had a HSIL+ cytology with a <CIN2 histology. We clearly demonstrated that the HPV type distribution in <CIN2 histology resembles CIN2 when the concurrent cytologic interpretation is HSIL. This corroborates the widely held view that HSIL is an important cytologic interpretation with high predictive value for cancer risk. Our data suggest that in such cases, a small high-grade lesion may have been missed during colposcopy and biopsy, whereas cells derived from that lesion (and the responsible HPV genotypes) were present in the cytology sample. Several studies recently demonstrated that colposcopy guided biopsy frequently misses the worst lesion present on the cervical surface.15 Although it was shown that visual patterns are not very helpful in guiding biopsy placement, obtaining more biopsies increased the chance of finding the worst lesion.29 Currently, it is not clear how many biopsies are most efficient for reliable disease detection.

Of course, cytology is also prone to error: 45% of the CIN2 lesions and 21% of the CIN3 lesions had a cytology result <HSIL. We believe that both histology and cytology add important information for disease classification. We therefore created the combined disease categories of <CIN2 (when cytology was NILM/ASC/LSIL), CIN2 including HSIL even when histology did not show CIN2, CIN3 and cancer.

The category of <CIN2 is heterogeneous and includes 2 functionally different disease stages: (i) HPV-positive women with <CIN2 histology and no cytological abnormalities (NILM), characterized by lower viral load as compared to the other group, as well as (ii) HPV-positive <CIN2 with ASC or LSIL cytology. The first group is predominant in a screening population, but uncommon in a referral population such as SUCCEED. The 2 groups share a low-risk profile (<HSIL cytology, <CIN2 histology, similar HPV genotype patterns). In a screening setting, cytologic normalcy and very low viral load are correlated; the combination predicts a lower risk for subsequent disease as compared to viral loads high enough to be associated with ASC or LSIL cytology.30 When focusing on biomarker identification for risk of more serious disease in a referral population, we therefore believe that it is reasonable to combine these 2 subcategories into <CIN2, recognizing that the number of women found on referral to have NILM cytology will be low.

We again observed that the category of CIN2, which includes HSIL by our analysis, is a heterogeneous group consisting of transient infections that will resolve and early precancers (CIN3, including carcinoma in situ).14, 31 To understand which CIN2 lesions are truly precancers, more disease specific markers will be necessary; although HPV genotypes can provide useful risk stratification (e.g., HPV16 CIN2 in biopsy is frequently upgraded to CIN3 in LEEP31), HPV genotyping is not perfectly predictive by any means. Many investigators are researching HPV oncogene mRNA expression, proliferation markers, p16 or novel markers that SUCCEED or similar projects might uncover. The molecular epidemiologic clarification of CIN2 is an important, remaining interdisciplinary task.

As we continue our molecular studies in SUCCEED, we may further refine the disease stages, e.g., by analyzing host and HPV gene expression patterns. This approach, guided by the disease categories we propose here, may allow to more precisely address the problem of heterogeneity, especially within CIN2. Importantly, these proposed diagnostic categories of disease progression will be critical for our search for novel biomarkers that predict the risk of progression and invasion to cervical cancer.

Acknowledgements

We would like to thank the laboratory personnel of the Surgical Pathology and Cytopathology Laboratories of OU Medical Center for their conscientious attention to specimen processing and Pap test interpretation.

Ancillary