Descriptive Evidence That Risk Profiles for Cervical Intraepithelial Neoplasia 1, 2, and 3 Are Unique

Infection of the cervix with human papillomavirus (HPV) does not uniformly lead to histologic changes. HPV viral replication within cells causes a range of cellular changes seen microscopically in cervical biopsy specimens. Cervical lesions as ascertained by histology, termed cervical intraepithelial neoplasia (CIN), are classified in three grades based on increasing degrees of cellular change and disorganization. Although CIN was initially described as a continuum of histologic changes (all of which were considered true precursors of invasive disease if untreated) (1), more recent data on the natural history of cervical neoplasia have since been revised to acknowledge that CIN does not always inevitably progress through each higher grade of CIN to cancer (1, 2). Progression from, for example, CIN 1 to 2, is the exception rather than the rule. Consequently, the classification of cervical cancer precursors may need revision to more accurately reflect the natural history of disease progression and to inform clinical decision making (1). Currently, a two-tiered histologic grouping has been adopted by many (CIN 1 and 2/3; ref. 3) to reflect putative differences in the natural history between CIN 1 (lower probability of progressing) and CIN 2/3 (a dysplastic process with a greater likelihood of eventual progression to cancer; ref. 3).

Recent research suggests CIN 2 may not be a discrete histologic grade that can be reliably classified but, instead, an overlap between CIN 1 (a nonneoplastic HPV infection) and CIN 3 (a neoplastic cervical cancer precursor; refs. 4, 5). It has been hypothesized that a proportion of CIN 2 lesions are biologically similar to CIN 1 and have greater probability of regressing over time, whereas other CIN 2 lesions are more characteristic of CIN 3 lesions (5). If the current construct of CIN 2, or an identifiable subset, is confirmed to be more similar to CIN 1 than 3, research categorization and clinical classifications may need to change. Future treatment guidelines for women, especially young women with CIN 2 diagnoses, would then need to be reevaluated. Under consideration would be differential rescreening and treatment options, as well as incorporation of more sophisticated molecular markers of risk for progression.

To describe whether CIN 2 lesions are epidemiologically more similar to CIN 1 or 3 lesions in the reproductive risk factors among women with these diagnoses, we used data from a cross-sectional study in China: the Shanxi Province Cervical Cancer Screening Study (SPOCCS) II (6). In this population-based study, women were screened for cervical cancer, and those with cytologic abnormalities and/or who tested positive for cervical HPV DNA were triaged to colposcopy for universal histologic confirmation of cervical disease status. Our primary aim was to estimate if histologically confirmed CIN 2 lesions have risk factor profiles that more closely resemble CIN 1 or 3 cases. We present here results examining differences in the risk factor profiles among HPV DNA–positive women with histologic diagnoses of CIN 1, 2, and 3.

Between May 2001 and June 2002, SPOCCS II was conducted in Yangcheng and Xiangyuan counties, Central China, as previously described (6). In brief, women of ages 35 to 50 years were invited to participate in a cross-sectional, cervical cancer collaborative research project between the Cancer Institute of the Chinese Academy of Medical Sciences in Beijing, China, and the Cleveland Clinic Foundation in Cleveland, Ohio. Eligibility criteria included being a resident of 1 of 15 county communes, nonpregnant, having an intact uterus, no history of pelvic irradiation or cervical cancer, and no history of cervical cancer screening within the last 5 years. HIV status was not identified. Of 9,034 women attending the Xiangyuan Women's and Children's Clinic and the Yangchen Cancer Hospital, 8,798 (97%) women met study eligibility criteria.

All participating women performed a self-collected HPV test in their village. Approximately 10 months later (range, 3-16 months), women underwent a pelvic examination at the study clinic by a study staff physician to collect cervical cell samples for cytologic and HPV testing. A standardized questionnaire was administered in a confidential setting by trained health workers to elicit information on risk factors potentially associated with cervical cancer and its precursors, including sexual, diet, and reproductive history. Women with abnormal cytologic results and/or HPV-positive test results were examined by a colposcope, and biopsies were collected with a 2-mm bronchoscopy biopsy instrument (6). The cervix was examined by quadrant, and all colposcopically detected cervical abnormalities were biopsied.

In quadrants that seemed normal and those that seemed abnormal, biopsies were still obtained at the squamocolumnar junction at 2, 4, 8, and 10 o'clock depending on the quadrant. Endocervical curettage was also done. Therefore, all participants colposcoped had a minimum of five biopsies, including the endocervical curettage. Biopsies, obtained using the four-quadrant biopsy protocol, were interpreted by a senior pathologist at the Cancer Institute Hospital of Beijing without knowledge of the screening test results. Two previous studies, the SPOCCS I pilot study and SPOCCS I (7), had previously established the quality of the slide preparation and diagnoses obtained in this laboratory in China. Based on these validated results, all histologic diagnoses in the SPOCCS II study were based on histologic review by two senior pathologists in China. The histologic diagnoses used in the analyses contained within this section are those determined in China for the SPOCCS II study. Final diagnosis in SPOCCS II was based on the highest-grade biopsy obtained, and high-grade lesions were treated according to standard protocols.

Study protocol and questionnaires were approved by human subject review boards of the Cancer Institute/Hospital of the Chinese Academy of Medical Sciences, and the Cleveland Clinic Foundation.

Both self-collected cervical-vaginal specimens and physician-collected samples from the cervix were evaluated for 13 high-risk oncogenic HPV DNA types (HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68) using a Digene second-generation hybrid capture assay (8) blinded to all clinical results. Per manufacturers' instructions, a value of ≥1.0 pg/ml HPV DNA was used as the cutoff for positivity. Oncogenic HPV DNA will be referred henceforth to as HPV.

Estimates of median values and proportions of women with suspected risk factors for CIN were calculated, stratified by histologic status. Prevalence odds ratios and corresponding 95% confidence intervals (95% CI) for grades of CIN were calculated for each grade of CIN by polytomous and multivariate polytomous logistic regression, with a four-level nominal response variable (controls or normal histology, CIN 1, 2, or 3), and the logit estimator always compared grade of CIN with control subjects. Using χ² tests generated by polytomous logistic regression, we compared odds ratios for CIN 3 to 2 and 1 to 2. HPV DNA–positive women with normal histology were used as the reference group for all analyses. The variables considered in these analyses included 3-year age groups (35-39, 40-44, 45-50 years), education, age at first intercourse, lifetime number of sexual partners, number of pregnancies, number of live births, intercourse within 4 months of childbirth, number of abortions, and current form of contraception. Models were restricted to HPV-positive women to reduce residual confounding by HPV infection (9). Models were reduced using backward elimination (10, 11) until only those variables producing a change of ≥10% remained. Pregnancy and live births were highly correlated and thus not included in the same model. The final multivariate model included terms for age, lifetime sexual partners, number of pregnancies, intercourse within 4 months of childbirth, and number of induced abortions. Current form of contraception was not included in the final model due to the small number of current users.

Parity is defined as a woman's number of live births, and gravidity is the number of pregnancies. Because few women reported having never been pregnant (72 women, 0.82%), and few were nulliparous (e.g., no previous live births, 1.2%), the parity variable was collapsed into a binary variable (0-2 and ≥3).

A sensitivity analysis was conducted to compare results of models restricted to HPV-positive women to those that adjusted for HPV status. Analyses were done using both Stata 9.0 analytic software (Stata Corp LP) and PROC CATMOD of the SAS statistical package (SAS Institute, Inc.).

Overall, 8,798 women participated, of whom 4% had histologically confirmed CIN 1, 2% had CIN 2, and 2.1% had CIN 3. HPV prevalence was 85.2% among women with CIN 1, 96.5% among women with CIN 2, and 97.4% for those with CIN 3. Of the 8,062 women with normal histology, 1,405 (17.4%) were HPV positive. Infection with HPV was associated with an increased risk for CIN 2 versus 1 [odds ratio = 5.0 (2.1, 11.9)]. No difference in HPV prevalence was found between CIN 3 and 2. Among 2,055 women who tested positive for HPV, the prevalence of histologically confirmed CIN was 14.4% for CIN 1, 8% for CIN 2, and 8.9% for CIN 3 (Table 1). Most women with HPV infection (67.7%) were histologically normal (negative for CIN). Participants' median age ranged from 40 years for women who were HPV-positive <CIN 2 to 43 years for HPV-positive CIN 3.

More than 96% of women were married; most reported having a single lifetime sexual partner. Independent of the grade of cervical disease, women had a median number of two live births and three pregnancies, with a median age at first intercourse of 20 years. Most women reported using female sterilization as their contraceptive method, and this also did not vary by grade of CIN. The proportion of women reporting intercourse within 4 months of childbirth ranged from 28% to 44% and differed by histologic grade.

For the oldest age group, there was a statistically significant 1.55-fold higher risk for CIN 2 and a 2.2-fold higher risk for CIN 3 compared with younger women, but age was not associated with risk for CIN 1 (Table 2). Conversely, both an earlier age at first intercourse (19-20 versus ≥21) and having two lifetime sexual partners versus one were associated with a higher risk for CIN 1, whereas neither risk factor was associated with CIN 2 or 3. Compared with women who had zero to two pregnancies, women who had at least three pregnancies were at higher risk for CIN 3 [odds ratio = 1.47 (1.04, 2.07)], but pregnancy history was not associated with CIN 1 or 2. In addition, compared with women who delayed sexual intercourse, those who had sexual intercourse within 4 months of childbirth had a higher risk for CIN 3 [odds ratio = 1.74, (1.27, 2.38)] and CIN 1 [odds ratio = 1.34 (1.04, 1.75)]; no association was observed for CIN 2. There was a lower risk for CIN 1 associated with one abortion versus zero, but abortion history was not associated with CIN 2 or 3. Current use of an IUD versus female sterilization also showed a 50% lower risk for CIN 3. No associations were found for educational attainment or reported number of live births. Results from HPV-restricted models and HPV-adjusted models were nearly identical (data not shown).

Most multivariate-adjusted associations were of similar magnitude (Table 3) to those from age-adjusted models for most risk factors except for lifetime sexual partners. In multivariate polytomous regression analysis, there was nearly a 1.5-fold higher risk for CIN 3 for women with three or more partners compared with those with only one partner. Similar associations were observed for CIN 1. Other than age, there were no associations of any risk factors with CIN 2. In multivariate polytomous logistic regression, we directly compared odds ratios across histologic grade. There were statistically significant differences in the associations of age groups with CIN 1 and 2 such that there were positive associations with CIN 2 but no associations for CIN 1 (P for difference = 0.03 for each age group). Conversely, the magnitudes of associations with age were similar between CIN 3 and 2 (P for difference > 0.4). Odds ratios for reported number of pregnancies were significantly higher for CIN 3 compared with 2 (P for difference = 0.02), but there was no difference in associations with number of pregnancies for CIN 2 and 1. Having intercourse within 4 months of childbirth was positively associated with both CIN 1 and 3, but not with CIN 2; the differences in associations between CIN 1 and 2 and between CIN 3 and 2 were highly significant (P < 0.001, respectively). There were no differences in associations of number of sexual partners and reported number of abortions between CIN 1 and 2 or between 3 and 2.

As mentioned previously, recent data suggest that CIN 2 may not be a discrete histologic grade but an overlap of CIN 1 and 3. It also has been hypothesized that CIN 2 may be more biologically similar to CIN 1 than to 3. Therefore, using data from a large, cross-sectional Chinese study of 2,055 HPV-positive women among whom at least five cervical biopsies were obtained, and culturally competent private interviews were conducted to collect risk factors information, we compared risk factor profiles among women with CIN 1, 2, and 3. If CIN 2 is biologically similar to CIN 1 or 3, we would expect the strength of associated factors to be comparable. However, the results of our study showed no common patterns of associations between CIN 1 and 2 or between 3 and 2. In our study, older age was associated with the risk for CIN 2 and 3 but not with CIN 1. These findings are consistent with the fact that older age has long been identified as a risk factor for cervical precancer and cancer, both of which are more common in older women. Although CIN 2 was not associated with any other risk factor, in multivariate analyses, a higher number of pregnancies were associated with CIN 1 and 3, and only the difference in association between CIN 3 and 2 was statistically significant. Similarly, women engaging in intercourse within 4 months of childbirth also were at higher risk for CIN 1 and 3. Lifetime number of sexual partners and reported number of abortions were not associated with any grade of CIN in these data.

Our study is one of the few that directly compares CIN 2 to 1 and 3, and builds on previous findings that CIN 2 lesions may not be biologically similar to CIN 3, at least as reflected by risk factor profiles (4, 12, 13). For example, among 2,366 women with oncogenic HPV enrolled in the U.S. ASCUS LSIL Triage Study (ALTS) trial, smoking was strongly associated with a diagnosis of ≥CIN 3 but was not associated with a diagnosis of CIN 2, except among women with HPV-16 (4). In our study from China, too few women smoked, limiting the power to examine smoking as a risk factor. Unlike the results from our study, data from the ALTS trial showed no significant differences in reproductive factors between women diagnosed with CIN 2 or 3. However, few ALTS participants reported three or more pregnancies or live births. Further evidence that the characteristics and behaviors of CIN 2 may be distinct from CIN 3 include data from a large population-based prospective study among women with HPV infection, which found that time from last normal cytologic smear was a risk factor for prevalent CIN 3 but not for =CIN 2 (14).

Our observed association between CIN 3 and 1 with intercourse within 4 months of birth is unexpected. A similar variable was tested among 759 women with invasive cancer and 1,430 controls in a case-control study in Latin America, where no association was found (14). To gain understanding of this association, we conducted analyses to identify predictors of intercourse within 4 months of birth. Older age was strongly associated with having intercourse within 4 months of childbirth, and having more than one lifetime sexual partner was protective. However, the association of earlier return to intercourse remained even after adjustment for age and number of sexual partners. Although the possible associations between intercourse within 4 months of delivery and CIN 3 and 1 are intriguing, the meaning of these associations is unclear. Additional research is needed to clarify both the actual timing of intercourse and the potential biological mechanisms. Intercourse within 6 to 8 weeks of delivery would be concurrent with a significant inflammatory response from birth trauma and tissue remodeling. Alternatively, exposure to semen or additional microtrauma from intercourse could exacerbate processes that facilitate cervical lesion progression (15). Conversely, assuming that most women in Shanxi province breast-feed, hormonal changes related to lactation may affect immune response. In this study, ∼30% of women had resumed intercourse within 4 months of childbirth. This is substantially lower than was reported for women in an urban population in Zibo, China, where 93% of women had resumed sexual intercourse at 4 months postpartum (16).

Similarly, it is unclear to what extent the association between number of pregnancies and CIN 3 is due to a factor related to the pregnancy (such as effects of pregnancy hormones on a woman's immune function or physical properties of the cervix affecting progression risk) or an independent host factor associated with both risk for HPV progression and the ability to achieve pregnancy. If it is the former, we would have expected both the number of pregnancies and number of live births to be associated with CIN 3 risk, which was not the case.

Although population-based, our findings from two rural counties in Shanxi province cannot be considered to be representative of the entire Chinese population. Reliance on self-reported data may have led to underreporting or overreporting of reproductive variables. Self-reported data are always difficult to ascertain, although because of the one-child policy in China, there has been strong incentive for women in China to reliably report reproductive information and pregnancies, abortions, live, and stillbirths to the government (17). Although there was a large range in the numbers of pregnancies (0-10) and live births (0-10), these numbers for this rural province decrease within reported ranges.

We realize there may be concern with our histologic diagnoses because of inherent difficulties in obtaining standardized pathologic diagnoses. Many studies report on the poor reproducibility of CIN 2 versus 3 and CIN 1 versus HPV changes. Interobserver variation for diagnosing CIN 2 is greater than for CIN 3 (12). A potentially large proportion of CIN 2 cases could be misclassified and actually be true CIN 1 cases. To describe potential changes in estimates related to misclassification, estimates were recalculated using data from a review of study biopsies. Two hundred seventy-one biopsies from SPOCCS I (a study preceding this SPOCCS II study) were recently reviewed by two expert pathologists, Dr. Bin Yang, a gynecologic pathologist from the Cleveland Clinic Foundation, and Xun Zhang, a pathologist from the Cancer Institute Hospital of Beijing, China. Of the 271, 59 were initially read as CIN 3, 83 as CIN 2, 62 as CIN 1, and 67 as normal pathology. On pathologic review, 118 of 129 slides initially interpreted as normal or CIN 1 were again interpreted as normal or CIN 1, 10 were reinterpreted as CIN 2, and 1 was interpreted as CIN 3. Of 142 slides, 135 initially interpreted as CIN 2 or 3 were again interpreted as CIN 2 or 3, 4 were reinterpreted as normal, and 1 was interpreted as CIN 1. We applied these proportions to the pathology in SPOCCS II and estimated the rates of misclassification in the pathology that we used for these analyses; 92% of CIN 2 remained CIN 2, 2.4% changed to CIN 1, 4.8% changed to normal histologic finding, and 1.2% was changed to CIN 3. When we recalculated the odds ratios and 95% CI for higher gravidity (≥3 pregnancies), no change in the magnitude of the effect was observed.

Given results of this quality control study, including double reading of pathologic slides, the change in the magnitude of the effect for number of pregnancies was small, suggesting that misclassification of CIN was negligible in this study. The fact that 92% of CIN 2 remained classified as CIN 2 would suggest it may be sorted into a histologic classification distinguishable from CIN 1 and 3, at least using the four-quadrant biopsy protocol implemented here (6).

Our study results are consistent with others that suggest that CIN 2 may represent a more heterogeneous group of lesions than originally believed. If confirmed by future prospective studies, different treatment options may need to be considered for a proportion of women with CIN 2 diagnoses to avoid potential overtreatment. Current recommendations are to treat women with histologically confirmed CIN 2 or 3 with either ablative (e.g., cryotherapy) or excisional treatment (e.g., loop electrosurgical excision procedure, conization; ref. 18). Whereas clinical trials have generally failed to show differences in treatment modalities, excisional treatment allows for pathologic diagnosis of the excised tissue. Excision is often the treatment of choice and is routinely recommended for women with unsatisfactory colposcopy or recurrent disease (18). Treatment of CIN is not without complications, especially for women who are still in their childbearing years (19). Although uncommon, significant pregnancy complications have been associated with larger loop electrosurgical excision procedure (20-22).

In conclusion, our data show unique profiles of characteristics associated with CIN 1, 2, and 3. These data suggest that grouping CIN 2 with 3 might misclassify CIN 2 and attenuate the associations of risk factors associated with CIN 3. Data from prospective studies are needed that follow women with untreated CIN 2 to document its natural history to determine if it behaves more like CIN 3 or 1 or is its own distinct histologic grade. We encourage those with prospective data to further examine these potential differences.

J.S. Smith: MERCK, GlaxoSmithKline, commercial research grant, Genprobe, commercial research support, GlaxoSmithKline, speakers bureau/honoraria. The other authors disclosed no potential conflicts of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.