Predicting the progression of cervical precursor lesions by human papillomavirus genotyping: A prospective cohort study

Authors


Abstract

Only a subset of cervical precursor lesions progress to cervical cancer and because of the lack of the predictive markers, it cannot be ascertained which lesions will progress or not. To estimate the risk of disease progression associated with human papillomavirus (HPV) genotypes, we followed 570 Japanese women with cytological LSIL (low-grade squamous intraepithelial lesion) and histological CIN (cervical intraepithelial neoplasia) grade 1–2 lesions (479 CIN 1; 91 CIN 2) at 3 to 4 month intervals for a mean follow-up period of 39.1 months. At entry, we detected HPV DNA in cervical samples by polymerase chain reaction-based methodology. Over the period of follow-up period, 46 lesions progressed to CIN 3 while 362 regressed to normal cytology. Women with multiple HPV infections were more likely to have persistent lesions (hazard ratio [HR] for regression, 0.65; 95% confidence interval [CI], 0.42–1.02; p = 0.07); however, multiple infections did not increase the risk of progression (HR for progression, 1.04; 95% CI, 0.37–2.94; p = 0.94). After adjusting for CIN grade and women's age, HRs for progression to CIN 3 (vs. women with low-risk types or negative for HPV DNA) varied markedly by HPV genotype: type 16 (11.1, 95% CI: 1.39–88.3); 18 (14.1, 0.65–306); 31 (24.7, 2.51–243); 33 (20.3, 1.78–231); 35 (13.7, 0.75–251); 52 (11.6, 1.45–93.3); 58 (8.85, 1.01–77.6); other high-risk types (4.04, 0.47–34.7). HPV 45 was not detected in our study subjects. The cumulative probability of CIN 3 within 5 years was 20.5% for HPV 16, 18, 31, 33, 35, 52 and 58; 6.0% for other high-risk types; 1.7% for low-risk types (p = 0.0001). In conclusion, type-specific HPV testing for women with LSIL/CIN 1–2 lesions is useful for identifying populations at increased or decreased risk of disease progression.

Cervical cancer remains the second most common cancer in women worldwide, with nearly 500,000 women developing the disease every year.1 The marked decline in incidence rates of cervical cancer in developed countries is attributed to the development of cytology screening programs that detect women with precursor lesions. However, cancer prevention eventually requires the eradication of such precursor lesions. Vaccination against human papillomavirus (HPV) types 16 and 18 prevents new infection; however, it has no therapeutic effect on women with abnormal Pap results.2

Most low-grade cervical lesions are known to regress spontaneously, whereas only a small fraction progress to cervical cancer.3–7 However, we cannot predict which lesions will regress or progress. Recommendations by the American Society for Colposcopy and Cervical Pathology (ASCCP) suggest that women with cervical intraepithelial neoplasia grade 2 (CIN 2) should be treated.8 However, there is risk of overtreatment, as 40–60% of CIN 2 lesions spontaneously regress to normal.3–7 Grade 1 lesions (CIN 1) are usually followed without treatment;8 however, 10–20% of CIN 1 lesions progress to CIN 3 or more.3–5

Infections by a specific subset of HPVs are a major risk factor in the development of cervical cancer.9, 10 HPV genotypes detected in CIN lesions may serve as predictive markers of disease persistence and progression. The large pooled analyses of International Agency for Research on Cancer (IARC) studies11 and meta-analyses12, 13 have suggested that oncogenic potentials of genital HPVs vary by genotype, although most data were from case–control and cross-sectional studies. Evidence from prospective studies already exists that detection of 13 oncogenic HPVs,14, 15 especially HPV 16 and 18,16 among women with normal cytology raises the risk for subsequent development of precancer. The atypical squamous cells of undetermined significance (ASCUS)/low-grade squamous intraepithelial lesions (LSIL) triage study (ALTS) reported HPV genotype-specific risks of prevalent and incipient CIN 3 diagnosed within the next 2 years among women with ASCUS/LSIL cytology.17 However, prospective data on genotype-specific risk of persistence and progression of biopsy-positive CIN lesions are limited.

This cohort study (Japan HPV And Cervical Cancer [JHACC] Study) was designed to identify determinants of persistence and progression of cervical precursor lesions that were detected by cytology screening programs. Follow up of >500 women with precursor lesions enabled us to characterize the risks of CIN persistence and progression in relation to HPV genotypes.

Material and Methods

Study design

Between April 1998 and August 2004, 905 women with low-grade cervical abnormalities were enrolled in a prospective non-intervention cohort study (Fig. 1). They were recruited from nine hospitals that performed conventional Pap smears, colposcopy and cervical biopsies. Cervical smears were classified according to the Bethesda System. Only at the entry, two small specimens were taken by colposcopy-directed punch biopsies and histological diagnosis was made using HE (hematoxylin and eosin)-stained sections according to the World Health Organization (WHO) classification. The inclusion criteria of this analysis were evident LSIL cytology, biopsy-positive CIN 1 or CIN 2 lesion, age 18 to 54 years and first detection of cervical abnormality. Women with ASCUS smears were excluded from the analysis. All subjects entered the study only after voluntarily giving signed informed consent. Two cytopathologists (Y.H. and M.T.) and two pathologists (R.F. and T.K.) reviewed all cytological and histological specimens at entry. The review was performed blindly by two cytopathologist and two pathologists. When disagreement was observed in their review results, the final diagnoses were based on their discussion. After the central cytology and pathology review, 225 women were excluded from the study; 34 women with HSIL cytology or CIN 3 histology, 90 women with LSIL cytology alone, 60 women with histological CIN 1–2 findings alone and 41 women without LSIL or CIN 1–2 findings. We excluded women with cytology-positive and biopsy-negative lesions from the analysis because they were more likely to have regression than women with both positive lesions. Also, we excluded women with cytology-negative and biopsy-positive lesions from the analysis because we could not define cytological regression for them. At enrollment, the subjects were tested for cervical HPV DNA. A total of 680 eligible subjects were followed at 3–4 month intervals and received cytology and colposcopic examinations at each visit. During follow up, a cervical biopsy was performed only when women had HSIL smears and major colposcopic changes (dense aceto-white change, iodine negativity, and coarse punctation / wide irregular mosaics of differing size) that were suggestive of progression to CIN 3 or worse. In our study, an increase in lesion size was not regarded as progression. For women that were regarded as progressing based on cytology and histology in the participating hospitals, two cytopathologists and two pathologists reviewed all cytological and histological specimens collected for diagnosis of disease progression. The primary endpoint of our study was progression to CIN 3, defined as histological CIN 3 lesions or worse diagnosed upon rigorous pathology review. Occasionally a few difficult cases were adjudicated by joint review with consideration of cytology as well as histology. The second endpoint was regression to normal cytology. We defined regression as normal colposcopy and at least two consecutive negative smears. In this analysis, women were regarded as having persistent lesions when they did not have either regression or progression over the period of follow up.

Figure 1.

Study design and natural courses of LSIL/CIN1-2 lesions. A total of 680 eligible subjects with cytological LSIL and histological CIN 1-2 lesions were followed with cytology and colposcopy at 3–4 month intervals. The analysis cohort consisted of 570 women with a sufficient number (≥2) of follow-up visits.

As 110 women had an insufficient number (<2) of follow-up visits to be included in the analysis, the analysis cohort consisted of 570 women. Unfortunately, cervical HPV data were not available in 25 women because of insufficient samples. These 25 women were included in the analysis of CIN grade and women's age, but excluded from the HPV analysis.

Institutional ethical and research review boards of the participating institutions approved the study protocol.

HPV detection and genotyping

We detected HPV DNA in cervical samples by polymerase chain reaction (PCR)-based methodology described previously.18 In brief, exfoliated cells from the ectocervix and endocervix were collected in a tube containing 1 ml of PBS (phosphate buffered saline) and stored at −30°C until DNA extraction. Total cellular DNA was extracted from cervical samples by a standard sodium dodecyl sulfate (SDS)-proteinase K procedure. HPV DNA was amplified by a PCR method, using the consensus L1 primers for HPV L1 region, L1C1 (5′-CGT AAA CGT TTT CCC TAT TTT TTT-3′) (1 μM), L1C2 (5′-TAC CCT AAA TAC TCT GTA TTG-3′) (0.5 μM) and L1C2M (5′-TAC CCT AAA TAC CCT ATA TTG- 3′) (0.5 μM). Forty amplication cycles were run in Perkin- Elmer 2400 Thermal Cycler (Perkin-Elmer, Norwalk, CT) with a 95°C denaturation step (1.5 min), a 48°C annealing step (1.5 min) and a 72°C extension step (2 min), including an initial denaturation step of 9 min and a final extension step of 5 min. The length of the resulting amplicon was ˜250 bp (e.g., 244bp for HPV 6 and HPV 11; 253 bp for HPV 16 and HPV 18; 256 bp for HPV 31 and HPV 33; 260 bp for HPV 52 and HPV 58). A reaction mixture without template DNA was included in every set of PCR runs as a negative control. Also, primers for a fragment of the β-actin gene were used as a control to rule out false-negative results for samples in which HPV DNA was not detected. Husnjak et al. reported that the sensitivity and specificity of HPV detection were similar between PCR methods using the L1C1/L1C2+L1C2M and MY09/MY11 primer sets.19 The sensitivity of our PCR method was 83% and that of the MY09/MY11 PCR method was 89%, compared to the reference PCR method using pI-1/pI-2 primer sets. The specificity was 51% for our PCR and 55% for MY09/MY11 PCR. Another study reported that the sensitivity of HPV detection was higher in our PCR method (91%) than in PCR methods using MY09/MY11 (69%) and GP17/GP18 primers (62%).20 HPV types were identified by restriction fragment length polymorphism (RFLP), which has been shown to identify at least 26 types of genital HPVs.21 HPV genotyping was performed blinded to the clinical data collected from the study subjects.

Based on HPV type prevalence data from the IARC-pooled analyses11, 22, and meta-analyses,12, 23 13 high-risk (HR) HPV types detected by Hybrid Capture 2 test (Digene, Gaithersburg, MD) and Amplicor HPV test (Roche Molecular Systems, Branchburg, NJ) were separated into two groups in the present study. Group 1 high-risk (G1HR) types included HPV 16, 18, 31, 33, 35, 45, 52 and 58, the eight most common oncogenic types associated with invasive cervical cancer worldwide. Group 2 high-risk (G2HR) types consisted of HPV 39, 51, 56, 59 and 68. All other HPV types were classified as low-risk types (non-oncogenic).

Statistical analysis

All time-to-event analyses were based on the actual date of the visits because analyses estimating date of event occurrence as the midpoint between two visits did not change the findings. For regression or progression, time to event was measured from the date of the index visit (i.e., the first instance of an abnormal cytology result) to the date of the visit at which cytological transition to normal or CIN 3 was first detected. Women whose lesions persisted or who dropped out of the study were censored at their last recorded return visit dates. Subjects who had only one negative colposcopy / cytology result before loss to follow-up were censored at the last date of positive Pap tests. Subjects who were biopsied were censored at the time of their biopsy, regardless of the biopsy results, to reduce the potential for interference by the biopsy procedure on estimates of time of regression. Cumulative probability of LSIL regression or progression was estimated using the Kaplan–Meier method and compared to a log-rank test, and the Cox regression model was used for statistical adjustments. CIN grade, women's age and HPV risk category were included in the multivariate model for adjustments. Since the results did not differ among the nine hospitals, the study sites were not included in the multivariate models. All analyses were carried out using the STATA 9 (StataCorp LP, College Station, TX) statistics packages. Two-sided p values were calculated throughout and considered to be significant at less than 0.05.

Results

The clinical outcomes of 570 women with LSIL cytology were monitored by cytologic and colposcopic testing at intervals of 3 to 4 months (Fig. 1). The mean age of the study subjects was 36.0 years (range, 18–54). They completed 4,830 visits and the mean follow-up time was 39.1 months (median, 40.9 months; range, 6.8–86.4).

Over the follow-up period, 46 lesions progressed to CIN 3 while 362 regressed to normal cytology. For all study subjects, the cumulative probability of regression in the 2-year follow-up was 62.3% (95% confidence interval [CI]: 58.0–66.7%). The median time to regression was 14.2 months (95% CI: 11.7–16.6 months). The vast majority (89%) of regression occurred within the first 24 months. The estimated probability of progression in the 5-year follow-up was 12.1% (95% CI: 9.0–16.3%). For women whose lesions progressed to CIN 3, the median time to progression was 17.9 months (95% CI: 13.3–23.9 months) and 72% of progression was detected after 12 months. In our study, no progression to invasive cervical cancer was observed.

Histopathological CIN grade

We confirmed that CIN 2 lesions were less likely to regress and more likely to progress than CIN 1 lesions, even after adjustment for age at diagnosis and HPV risk category. The cumulative probability of regression in the 2-year follow-up was 64.0% (95% CI: 59.2–68.7%) for CIN 1 and 53.7% (95% CI: 42.7–65.5%) for CIN 2 (Table 1). Women with CIN 2 were at significantly higher risk of lesion persistence compared to women with CIN 1 (hazard ratio [HR] for regression, 0.71; 95% CI, 0.50–0.99; p = 0.04). The median time to regression was 12.4 months (95% CI: 10.3–15.3 months) for CIN 1 and 21.2 months (95% CI: 15.8–41.5 months) for CIN 2. The estimated probability of progression in the 5-year follow-up was 9.7% (95% CI: 6.7–14.0%) for CIN 1 and 24.2% (95% CI: 14.5–38.8%) for CIN 2 (Table 2). Significant risk for progression to CIN 3 was found among women with CIN 2 compared to women with CIN 1 (HR, 2.15; 95% CI, 1.14–4.06; p = 0.02). Among women whose lesions progressed to CIN 3, the median time to progression was 24.6 months (95% CI: 11.2–39.6 months) for CIN 1 and 17.7 months (95% CI: 13.3–23.9 months) for CIN 2.

Table 1. Cumulative probabilities of LSIL regression within 2 years in relation to biopsy results, age and HPV genotypes
inline image
Table 2. Cumulative probabilities of LSIL progression within 5 years in relation to biopsy results, age and HPV genotypes
inline image

Women's age

In younger women aged 18–29 years, the 2-year probability of spontaneous regression (74.8%; 95% CI: 65.3–83.4%) was significantly higher than in older women aged 30–39 (56.9%; 95% CI: 50.4–63.5%) or 40–54 (62.6%; 95% CI: 55.2–70.0%), even after adjustment for CIN grade and HPV risk category (Table 1). Increased likelihood of LSIL persistence was found among women aged 30–39 (HR for regression, 0.56; 95% CI, 0.42–0.74; p < 0.001) and among women aged 40 to 54 (HR for regression, 0.52; 95% CI, 0.39–0.71; p < 0.001). The 5-year progression probability was much higher in women aged 30–39 years (17.2%; 95% CI: 11.6–25.1%), compared to women aged 18–29 (7.4%; 95% CI: 2.8–18.7%) and 40–54 (8.5%; 95% CI: 4.9–14.4%), although the difference was not statistically significant after adjusting for CIN grade and HPV risk category (Table 2).

HPV genotypes

At enrollment, HPV DNA was detected in 482 women (88.4% of all subjects), with HPV52 (16%), 16 (14%), 51 (12%) and 58 (12%) being the most prevalent HPV strains. In this analysis, 25 women who did not have an HPV diagnosis due to insufficient samples were excluded. Of 13 oncogenic HPVs, HPV45 was not detected in women who participated in our study.

CIN lesions with multiple infections were more likely to persist than those with single infections (2-year regression probability, 49.2% vs. 63.4%; HR for regression, 0.65; 95% CI, 0.42–1.02; p = 0.07), whereas multiple infections did not correlate with an increased risk of progression (5-year progression probability, 11.7% vs. 12.4%; HR for progression, 1.04; 95% CI, 0.37–2.94; p = 0.94).

We found that the risk of LSIL persistence and progression varied considerably among high-risk genotypes (Tables 1 and 2). Multiple infections were excluded from the analysis of risks associated with individual genotypes. After adjusting for CIN grade and women's age, hazard ratios for progression to CIN 3 (vs. women with low-risk types or negative for HPV DNA) were obviously higher in women with HPV 16 (11.1; 95% CI: 1.39–88.3), 18 (14.1; 0.65–306.1), 31 (24.7; 2.51–242.6), 33 (20.3; 1.78–230.6), 35 (13.7; 0.75–250.8), 52 (11.6; 1.45–93.3) or 58 (8.85; 1.01–77.6) than in women with HPV 39 (3.08; 0.19–49.3), 51 (5.74; 0.59–55.8), 56 (1.95; 0.12–31.2), 59 (no progression) or 68 (no progression) (Table 2).

Among women with cytological LSIL and histological CIN 1-2 lesions, the cumulative probability of CIN 3 within the subsequent 5 years varied greatly according to HPV risk category: 20.5 % (95% CI: 14.6–28.4%) in women with G1HR types; 6.0% (95% CI: 2.3–15.3%) in those with G2HR types; 1.7% (95% CI: 0.2–11.4%) in those with LR types or negative for HPV DNA (p = 0.0001; Fig. 2b). In addition, probability of regression to normal cytology within 2 years was 51.6% (95% CI: 44.7–58.8%) in women with G1HR types; 68.7% (95% CI: 60.0–77.0%) in those with G2HR types; 77.4% (95% CI: 67.4–86.1%) in those with LR types or negative for HPV DNA (Fig. 2b). These differences were statistically significant (p = 0.0001). Similarly, median time to regression also varied markedly by HPV risk category: 22.7 months (95% CI: 18.7–38.7 months) in women with G1HR types; 9.4 months (95% CI: 6.0–13.2 months) in women with G2HR types; 9.0 months (95% CI: 7.0–11.3 months) in those with LR types or negative for HPV DNA.

Figure 2.

Which high-risk HPV types should we distinguish for women with LSIL/CIN1-2 lesions? A Kaplan–Meier plot was used to estimate cumulative risks of LSIL persistence and progression in relation to HPV genotypes. Women with undetermined types or multiple infections were excluded from these analyses. In the management of women with LSIL/CIN 1-2 lesions, distinguishing HPV 16, 18, 31, 33, 35, 45, 52 and 58 may be more useful than distinguishing only HPV 16 and 18.

Among women with CIN 2 histology, the absolute risk of CIN 3 within the subsequent 5 years was 40.5% (95% CI: 23.7–63.1%) when they were positive for G1HR types but only 8.3% (95% CI: 2.0–33.5%) when negative for G1HR types (Table 2). Similarly, among women with CIN 1 lesions, the risk of CIN 3 within 5 years was 16.6% (95% CI: 10.8–25.1%) when they were positive for G1HR types but only 3.3% (95% CI: 1.6–7.0%) when negative for G1HR types.

In young women (aged 19–29) with LSIL cytology, disease progression was closely associated with only a specific subset of onocogenic HPVs (Fig. 3). The absolute risk of CIN 3 within the next 5 years was 16.9% (95% CI: 6.2–36.2%) in young women positive for G1HR types and 0% (95% CI: not determined) in young women negative for G1HR types.

Figure 3.

HPV genotype-specific risk of persistence and progression of LSIL/CIN1-2 lesions in relation to histology and women's age. In each group of histology and age, detection of HPV16/18/31/33/35/45/52/58 denoted at least a twice higher risk of CIN3 within the next 5 years. Similarly, cervical lesions with these genotypes were likely to persist for at least twice as long.

To examine whether a type-specific HPV test for subtypes 16 and 18 may be useful for identifying women at an increased risk of LSIL persistence and progression, we distinguished HPV 16 and 18 from high-risk HPV genotypes. The cumulative probability of progression to CIN 3 within 5 years was higher among women with HPV 16 or 18 (22.4%; 95% CI: 11.7–40.3%) than women with the other high-risk types (14.2%; 95% CI: 9.7–20.4%; Table 2). However, this difference was not statistically significant (p = 0.47; Fig. 2a). The cumulative probability of regression within 2 years among women with HPV 16 or 18 was similar to that among women with the other high-risk HPV types (62.1% vs. 56.6%; p = 0.99; Fig. 2a).

Several studies have reported the poor reproducibility of cervical histologic interpretations even among well-trained observers.24, 25 For this reason, we analyzed the follow-up data for CIN 1 and CIN 2 separately (Tables 1 and 2). These analyses did not change the findings, although the number of women with CIN 2 was small (n = 91).

The IARC-pooled study classified 15 types as high-risk types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73 and 82) and three types as probable high-risk types (26, 53 and 66).11 Although HPV 26, 53, 66, 73 and 82 were classified as low-risk types in our study, women with these types did not show progression during the follow-up. The 2-year regression rate in women with HPV 26, 53, 66, 73 and 82 (77.1%; 95% CI: 54.7–93.6%) was similar to that in women with other low-risk types (79.6%; 95% CI: 66.5–90.2%).

Discussion

In clinical management of women with LSIL cytology or CIN 1-2 lesions, the role of HPV genotyping is not established. In our study, women negative for high-risk types were at the lowest risk (1.7%) of subsequent CIN 3 within the next 5 years. This was consistent with previous clinical follow-up studies of women with mildly abnormal cervical cytology.17, 26 More importantly, our prospective data showed that risk of LSIL persistence and progression varies considerably even among high-risk genotypes. In a Japanese population, the risk of progression to CIN 3 in the next 5 years was at least twice as high in women with HPV 16, 18, 31, 33, 35, 52 and 58 compared to that in those with HPV 39, 51, 56, 59 and 68, suggesting that genotype-specific HPV testing will enable us to characterize a woman's risk more precisely than high-risk HPV testing that does not distinguish individual genotypes. Although HPV 45 is rarely detected in East Asia,27, 28 HPV 16, 18, 31, 33, 35, 45, 52 and 58 are the eight most common genotypes detected in women with cervical cancer worldwide, accounting for >90% of cases of cervical cancer in each area.11, 12, 22, 23 Therefore, testing for these eight genotypes may help estimate risks of disease persistence and progression as an adjunct to colposcopy in all regions of the world. For instance, monitoring women with CIN 2 lesions without immediate treatment may be acceptable if they are negative for the eight highest risk types, because their risk of CIN 3 within the subsequent 5 years (8.3%) was similar to that in women with CIN 1 who are usually followed without immediate treatment (9.7%). In contrast, women who are positive for the eight highest risk types with CIN 1 may require long-term careful follow-up, because (i) the median time to regression was 21.4 months and (ii) their risk of CIN 3 within 5 years was relatively high (16.6%).

A prospective cohort study has shown that women who are positive for HPV 16 or 18 with a normal Pap smear have an 18–21% 10-year risk of developing CIN 3, compared to a risk as low as 1.5% among those with other high-risk types.16 Therefore, a type-specific HPV test for subtypes 16 and 18 is recommended for women with a positive high-risk HPV test and negative cytology.29 However, our data suggest that separate detection of HPV 16 and 18 may not be very useful in the management of women with LSIL cytology because women who were positive for other oncogenic types with LSIL smears also had a high risk (14.2%) of developing CIN 3 within the next 5 years.

Although the precision of the risk estimates was limited because of the small sample size, the risks for individual HPV types varied greatly, even among the eight high-risk types. For instance, the probabilities of progression to CIN 3 within the next 5 years ranged from 7.7% (for HPV 18) to 36.0% (for HPV 31). HPV16 had a 5-year cumulative risk of 26.4%. In another prospective study (ALTS) of women with ASCUS/LSIL cytology, the cumulative risks of CIN 3 diagnosed within 2 years was 39.1% for HPV 16, 6.1% for HPV 18 and 14.8% for HPV 31.17 The wide range of risks for individual HPV genotypes suggests that full genotyping may have an advantage over partial genotyping.

Several studies have reported that multiple HPV infections may be associated with an elevated risk of cervical cancer.30, 31 In our study, multiple infections did not increase risks of progression, while women with multiple infections were more likely to have persistent lesions (p = 0.07). Multiple infections may eventually increase the risk of disease progression in a long-term observation because women with persistent LSIL smears have more chances of progression to CIN 3. Alternatively, the difference in HPV detection methods may have affected our results. Different methods for HPV genotyping are known to provide different results on multiple infections.32, 33

In young women aged 18–29 years, early precursor lesions were more likely to regress to normal cytology, whereas women's age did not significantly correlate with disease progression after adjusting for CIN grade and HPV genotypes. The high regression rate observed in young women lends support to the clinical practice of monitoring young women with early lesions.8, 34 However, we note that detection of specific HPV genotypes conferred a significant risk of disease persistence and progression even in young women. The absolute risk of CIN 3 within the next 5 years was 16.9% in young women positive for HPV 16, 18, 31, 33, 35, 45, 52 or 58, and 0% in young women negative for these eight types. Although high-risk HPV testing as an adjunct to cytology in primary screening is not recommended for women younger than 30 years old,8 genotype-specific HPV testing might be applied to young women with cytological LSIL and histological CIN1-2 lesions for deciding how to manage them.

Cervical biopsy at enrollment may have affected the clinical course of CIN lesions. We performed colposcopy-directed punch biopsy because: (i) biopsy reduces a possible bias of disease misclassification at enrollment; (ii) follow-up of CIN 3 lesions was not ethically permitted in Japan; (iii) previous studies have reported that small cervical biopsies at baseline do not influence clinical outcomes of CIN lesions.35, 36

Our study has several limitations. First, this cohort study investigated the natural history of prevalent, not incident, CIN lesions. Therefore, the differences in time from CIN development to study enrollment could be a bias. This may have affected the time to regression and the number of women regressing. Second, our findings observed in women with LSIL cytology may not be generalizable to women with CIN 1-2 lesions underlying ASCUS or HSIL. Third, 110 women that had an insufficient number (<2) of follow-up visits were excluded from the analysis. Although the age distributions, initial biopsy results and baseline HPV genotypes in the excluded and included women were similar (data not shown), this loss to follow-up could have influenced the results. Finally, the reduced use of biopsies (only two small biopsies) can have led to misclassification of cervical lesions at entry. In addition, limitations in the sensitivity of colposcopy for finding precancerous lesions during follow-up may have affected the cumulative rates of LSIL progression reported in our study.37, 38

In conclusion, our data suggest that genotype-specific HPV testing helps stratify women with LSIL cytology according to risk of disease persistence and progression. Characterizing a woman's risk more precisely by partial or full genotyping may reduce the number of follow-up smears and colposcopy referrals. Although our data suggest the potential utility of HPV genotyping in clinical management of women with LSIL/CIN1-2 lesions, further studies will be warranted to validate our results.

Acknowledgements

The authors thank Dr. Tadahito Kanda (Center for Pathogen Genomics, National Institute of Infectious Diseases, Tokyo, Japan) for his comments on the study design and the article; Prof. Ian H. Frazer and Dr. Graham R. Leggatt (Diamantine Institute for Cancer, Immunology and Metabolic Medicine, University of Queensland, Australia) for their critical review of this article; Mr. Masafumi Tsuzuku (Department of Cytopathology, Cancer Institute Hospital, Japanese Foundation of Cancer Research, Japan) for his cytological review; many others who facilitated our study; and all the women that participated in the study. The authors declare that they have no conflict of interest relevant to this article. The supporting organization played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the article.

Ancillary