Background
A number of retrospective and prospective studies have documented substantial rates of regression in cervical intraepithelial neoplasia grade 2 lesions in young women. Initial observational management of cervical intraepithelial neoplasia grade 2 is increasingly accepted as appropriate for women under 25 years of age with screen-detected abnormalities and is included in a number of clinical guidelines. However, there has been a paucity of large prospective studies on observational management with strict inclusion criteria. A number of important questions remain, specifically regarding the clinical variables that are associated with the risk of progression or persistence of disease. To investigate these factors and to ensure that young women with cervical intraepithelial neoplasia grade 2 undergoing observational management were being managed in a well-monitored and an appropriately informed fashion, we conducted a large, multicenter prospective study on observational management of cervical intraepithelial neoplasia grade 2 in women under 25 years.
Objective
This study aimed to determine the regression rates and clinical, cytologic, and pathologic predictors of regression of cervical intraepithelial neoplasia grade 2 in women under 25 years undergoing observational management over 24 months.
Study Design
This study was a multicenter prospective study on observational management of cervical intraepithelial neoplasia grade 2 (ie, repeat colposcopy, cytology, and cervical biopsy every 6 months) for up to 24 months. A total of 615 consenting women under 25 years with newly-diagnosed, biopsy-proven cervical intraepithelial neoplasia grade 2 were recruited (from 2010 to 2016) through 16 hospital-based colposcopy units in New Zealand and Australia.
Results
At completion, 326 women had confirmed regression, 156 had persistent high-grade cervical intraepithelial neoplasia grade 2 or 3 or adenocarcinoma in situ, and 24 had unconfirmed regression (ie, first regression at the 24-month follow-up). A total of 109 women did not complete the protocol (41 because of delayed follow-up, 41 lost to follow-up, 22 elected treatment, 4 refused a biopsy, and 1 died of an unrelated cause). Confirmed regression was observed in 53% (326 of 615) of all women enrolled in the study and, when missing data were imputed, it was estimated that 64% of women (95% confidence interval, 60%–68%) would have experienced regression. Similarly, lesions regressed in 64% (326 of 506) of women who completed the observational protocol. Based on a multivariable analysis, detection of human papillomavirus 16 in a liquid-based cytology sample at the time of initial colposcopy decreased the chance of regression by 31% (risk ratio, 0.69; 95% confidence interval, 0.56–0.86; P <.001). In addition, at initial colposcopy, low-grade or normal colposcopic impression, later year of diagnosis, low-grade or normal cytology, and being a nonsmoker were all independently associated with an increased chance of regression.
Conclusion
More than half of women under 25 years with cervical intraepithelial neoplasia grade 2 will regress to cervical intraepithelial neoplasia grade 1 or normal within 24 months without destructive treatment. The absence of human papillomavirus 16 is the most important predictor of regression.
Introduction
Persistent infection of the cervix with high-risk human papillomavirus (hrHPV) is associated with the development of high-grade squamous intraepithelial cervical lesions (HSIL), which, if left untreated, may develop into cervical cancer. , The detection of HSIL through a screening program and treatment thereof are associated with a reduction in the incidence of cervical cancer. , In contrast, it is now well accepted that most low-grade squamous intraepithelial cervical lesions (LSIL) are associated with transient viral infections and do not require treatment.
Why was this study conducted?
Cervical intraepithelial neoplasia grade 2 (CIN2) is known to regress without treatment for many young women. However, there are limited prospective data that enable us to predict for which women these lesions will regress.
Key findings
Around 64% of women aged <25 years with CIN2 will regress without treatment within 2 years. CIN2 lesions associated with human papillomavirus 16 had a 31% decreased chance of regression in women aged <25 years. Baseline factors associated with an increased chance of CIN2 regression include low-grade or normal colposcopic impression, low-grade or normal cytology, and being a nonsmoker.
What does this add to what is known?
Our prospective data confirmed the findings of retrospective studies in which a high proportion of CIN2 lesions in women aged <25 years will regress without treatment. Baseline factors that can help us estimate the likelihood of regression were identified.
Before the widely adopted recommendations of the Lower Anogenital Squamous Terminology (LAST) nomenclature report, HSIL was histologically classified as cervical intraepithelial neoplasia grade 2 (CIN2) or CIN3 in accordance with the Bethesda classification, whereas LSIL was classified as CIN1. The standard approach to histologically-confirmed CIN2 or CIN3 has been to ablate or excise the affected cervical transformation zone. This approach to management for young women has been questioned by many because of the risks associated with treatment, the low risk of cancer, and documentation that some high-grade lesions will resolve without destructive treatment.
Although it is clear that pathologists are limited in their ability to reliably differentiate between CIN2 and CIN3 diagnoses, CIN2 lesions have been associated with lower rates of progression to invasive carcinoma and higher rates of regression than CIN3. A number of retrospective and prospective studies have now documented significant rates of regression in CIN2 lesions. Many of these were summarized in a systematic review published by Tainio et al in 2018. In a subgroup of 1069 women under the age of 40 years from 4 studies, the overall documented regression rate was 60%. The review authors, however, urged caution in their interpretation of the review because of the heterogeneity in the reviewed studies. In particular, there was a paucity of large prospective studies with strict inclusion criteria. Despite this paucity of prospective data, the initial observational management of CIN2 is increasingly accepted as appropriate for women under 25 years and is included in a number of clinical guidelines. ,
A number of important questions remain. More specifically, these questions relate to the counseling of women on the longer term safety of observational management, what observational management should entail, what the selection criteria for observational management should be, and what clinical variables are associated with risk of progression or persistence of disease. Of particular importance to this article, the authors of a systematic review on rates of regression in CIN2 lesions concluded the following: “To date, no tests and biomarkers permit the prediction of CIN2 with true progressive potential.” To answer these important questions and to ensure that young women with CIN2 undergoing observational management are being managed in a well-monitored and an appropriately informed fashion, we conducted a large, multicenter prospective study on observational management of women under 25 years with CIN2.
The principal aims of this study were to determine the practicality and safety of observational management of CIN2 in women under 25 years and to document the regression, persistence, and progression of disease. The secondary aims were to determine the variables that are predictors of regression and persistence or progression.
Materials and Methods
New Zealand and Australia have well-organized national cervical screening programs. During recruitment for the study (2010–2016), New Zealand women aged 20 to 69 years were invited to undergo liquid-based cervical cytology (LBC) screens every 3 years and, in Australia, women aged ≥18 years were invited to undergo conventional cytology screens every 2 years. ,
The methods of this study have been described elsewhere, however, in brief, women aged <25 years with screen-detected cervical abnormalities who underwent a colposcopy and had histologically-confirmed CIN2 and who met the inclusion criteria were invited to join the study. If consent was obtained, the baseline data were retrieved from the initial colposcopy and, when possible, the LBC vial was retrieved for hrHPV testing. The screening cytology procedure that necessitated the colposcopy examination was recorded as the referral cytology procedure. Following a review of histologic specimens and cytology, consenting women underwent a follow-up colposcopy with cervical biopsy and cytology every 6 months for up to 2 years. Women with a subsequent histologic diagnosis of CIN3, a glandular abnormality, or clinical or cytologic concern for invasion or a glandular abnormality were recommended to undergo an excisional treatment. Similarly, women requesting treatment, those who were unable to comply with 6-monthly follow-ups, and those with a persistant CIN2 lesion at 24 months were recommended to undergo excisional treatment.
The inclusion criteria for the study were an age of <25 years at enrolment into the study, referral for a colposcopy for a screen-detected cytologic abnormality, histologically-confirmed CIN2, no previous high-grade cytology or biopsy (per their National Cervical Screening Program record), a cervical transformation zone and abnormality fully visualized at colposcopy, no colposcopic, cytologic, or histologic suspicion of a glandular abnormality or invasion, multidisciplinary case review by at least 1 colposcopist and expert gynecology pathologist, agreement by the woman to undergo a colposcopy every 6 months for 2 years, and informed consent. The exclusion criteria were pregnancy, immunosuppression, large complex lesions for which adequate sampling was difficult, type 3 transformation zone (previously described as unsatisfactory colposcopy), and concern by the treating colposcopist that the woman would not complete follow-ups.
From 2013 and when possible to obtain the LBC vial, partial hrHPV genotype testing was performed by the local laboratory (via Roche or Abbott polymerase chain reaction assay) on the cervical cytology sample taken at the initial visit or first follow-up visit and then repeated at 12 months. Partial genotype testing included detection of 14 hrHPV types––the genotyping results for HPV 16 and HPV 18 were reported individually and 12 other hrHPV types (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) were reported in a pooled category as “other hrHPV” without being individually specified. These HPV categories were analyzed individually.
Routinely collected clinical and colposcopy data, including a subjective assessment of lesion size, were included in the study data. The histology was not reviewed centrally, however, it was always reviewed in conjunction with the cytology by a specialist gynecologic pathologist routinely participating in a regional colposcopy multidisciplinary review session. Histologic reporting according to the LAST nomenclature was primarily on the basis of morphology with liberal use of immunohistochemistry (primarily p16 and ki67) when the diagnosis was uncertain.
Regression was only confirmed if cytology and colposcopy (and histology, if a lesion was present) indicated CIN1 or less on 2 consecutive colposcopy visits at least 6 months apart. Regression was not confirmed if cytology or colposcopy findings were not available or if a low-grade colposcopic lesion had been visualized but not biopsied. Women with confirmed regression could be referred back to the smear taker for cytology follow-up if deemed appropriate by the treating clinician.
The primary outcome measure of this study was the proportions of study participants with confirmed regression or persistent CIN2 or progression of disease over 24 months of follow-ups. A key aim was to determine the association of the clinical variables (eg, age, smoking, ethnicity, combined oral contraceptive use, colposcopic impression, lesion size) and cytologic and pathologic variables (eg, referral cytology, cytology at initial visit, HPV type at initial visit, and immunohistochemical markers at initial visit) with regression or progression of disease or persistent CIN2. The proportion of study participants who were unable to attend regular follow-ups or were lost to follow-up (>9 months) during the study was also investigated.
A recruitment target of 600 women was chosen to allow for an estimated 10% loss to follow-up and requested treatment (leaving an anticipated 540 women for primary analyses). Based on previous literature, the rate of regression was expected to be at least 33% and thus with a sample of 540 women, the precision of estimates was calculated to be within ±5% (95% confidence interval [CI], 29–37). From our retrospective study of 452 young women with CIN2, we identified 1 woman with a microinvasive carcinoma. This gave an estimated probability of invasive cancer in the population of 0.22% (95% CI, 0.039%–1.24%). A larger Australian retrospective registry study of young women with CIN2 provided an estimated probability of invasive cancer with similar confidence intervals (ie, not higher than 0.67% [95% CI, 0.036%–1.23%]). Thus, we estimated the probability of invasive cancer in our study group to be 0.036% to 1.24%. A sample size of 540 women was calculated to be sufficient to provide >99% power to detect an event rate of 1% or higher and 93% power to detect an event rate of 0.5% or higher. Carcinoma in this cohort is rare but does occur, and to identify any safety issues, a sample of at least 600 women was required.
Clinical, demographic, and colposcopy data were collected on site by means of case report forms and collated centrally. Ethnicity was reported as the total combined responses for ethnicity and thus every ethnicity recording for a woman is counted independently. Hence, the total number of ethnicity responses will usually exceed the study population. Pathology reports were reviewed centrally. Data from case report forms were entered into an Excel database and statistical analysis was performed using R version 3.5.3 (R Core Team, 2019, Vienna, Austria).
To compensate for the presence of missing data in baseline predictors and clinical endpoints for women who did not complete the observation management protocol, multiple imputation by chained equations was performed using logistic, proportional odds, and polytomous logistic regression models for binary, ordinal, and categorical data, respectively. All reported baseline variables and study outcomes (confirmed regression, unconfirmed regression, or persistent high-grade) were used in all prediction models in addition to high-risk HPV status recorded at the first follow-up. A total of 50 imputed data sets were created each with a burn-in phase of 20 iterations. Trace lines were inspected to verify convergence of the Markov chain Monte Carlo algorithm.
The primary endpoint was defined as the odds of having achieved confirmed regression within 24 months. The proportion of women who experienced this outcome was estimated for each level of the baseline predictor variables with 95% “Wilson” binomial confidence intervals. Unadjusted (univariable) and adjusted (multivariable) risk ratios with 95% CIs were calculated using log-binomial regression or Poisson regression with Huber-White sandwich standard errors. Estimates from multiple imputation data sets were pooled using Rubin’s rules.
This study was approved on April 14, 2010, by the Multi-Region Ethics Committee (ethics reference, MEC/09/07/079) and had site-specific local authorization. The Prediction of Regression in CIN2 study protocol is available at http://www.otago.ac.nz/christchurch/otago073807.pdf . An independent data safety monitoring board was in place and stopping rules were in place if an unexpected risk of cervical carcinoma was observed.
Results
A total of 615 women (598 women from 14 colposcopy units in New Zealand and 17 women from 2 colposcopy units in Australia) were recruited between 2010 and 2016. A Consolidated Standards of Reporting Trials flow diagram describes the participant study flow ( Figure ). The demographic and clinical features of the population are included in Table 1 .
| Baseline variable | Data completeness | Value | ||
|---|---|---|---|---|
| n | % | mean | SD | |
| Age (y), median (interquartile range) | 615 | 100.0 | 22 (21–23) | 1.5 |
| n | % | |||
| Ethnicity a | 589 | 95.8 | ||
| European | 509 | 86.4 | ||
| Māori | 95 | 16.1 | ||
| Pacific | 27 | 4.6 | ||
| Asian or other | 25 | 4.2 | ||
| Smoker | 598 | 97.2 | ||
| Yes | 185 | 30.9 | ||
| No | 413 | 69.1 | ||
| HPV vaccine | 370 | 60.2 | ||
| Full | 147 | 39.7 | ||
| Partial | 45 | 12.2 | ||
| No | 178 | 48.1 | ||
| Highest grade of previous abnormal cytology | 615 | 100.0 | ||
| None | 288 | 46.8 | ||
| Inflammation and repair | 1 | 0.2 | ||
| ASC-US | 59 | 9.6 | ||
| LSIL (HPV/CIN1) | 264 | 42.9 | ||
| Other | 3 | 0.5 | ||
| Referral cytology | 615 | 100.0 | ||
| ASC-US | 38 | 6.2 | ||
| LSIL (HPV or CIN1) | 212 | 34.5 | ||
| ASC-H | 138 | 22.4 | ||
| HSIL (CIN2 or 3) | 223 | 36.3 | ||
| Other | 4 | 0.7 | ||
| Cytology at baseline | 464 | 75.4 | ||
| Normal | 41 | 8.8 | ||
| ASC-US | 28 | 6.0 | ||
| LSIL (HPV/CIN1) | 146 | 31.5 | ||
| ASC-H | 77 | 16.6 | ||
| HSIL (CIN2/3) | 170 | 36.6 | ||
| Other | 2 | 0.4 | ||
| High-risk HPV genotype at baseline b | 343 | 55.8 | ||
| HPV 16 | 82 | 23.9 | ||
| HPV 18 | 25 | 7.3 | ||
| Other hrHPV | 271 | 79.0 | ||
| Not Detected | 23 | 6.7 | ||
| p16 | 329 | 53.5 | ||
| Negative | 33 | 10.0 | ||
| Positive | 296 | 90.0 | ||
a A woman may have >1 ethnicity recorded and thus percentages add to >100. Ethnicity was not recorded for women recruited in Australia
b A woman may have >1 hrHPV genotype and thus percentages add to >100.
In summary, 615 of 1054 (58%) eligible women were recruited. The average age of the women at recruitment was 21.7 years (range, 16–24 years). The total response ethnicity was 86% European, 16% Māori, 5% Pacific, and 4% Asian or other. Of those in the included cohort, 31% were smokers. HPV vaccination status was not well reported (370 of 615), but for those recorded, 52% self-reported as having had the quadrivalent HPV vaccine. Although all women had a histologic biopsy diagnosis of CIN2, 59% had a high-grade (HSIL or atypical squamous cells) referral cytology and 53% had high-grade cytology at their initial colposcopy visit. Partial hrHPV genotyping was available for 343 (56%) women from cervical cytology specimens taken at the initial visit. Of the women with hrHPV testing, 82 (24%) tested positive for HPV 16, 25 (7%) tested positive for HPV 18, and 271 (79%) tested positive for other hrHPVs. Note that because a woman may have >1 hrHPV genotype, percentages add to >100. Infection with >1 hrHPV genotype was reported in 54 (16%) women. However, coinfection rates may have been higher than this because although types 16 and 18 were reported individually, the 12 other hrHPV types were pooled. Of those who were tested, 23 (7%) tested negative for hrHPV types. Expression of p16 was reported in 329 (53%) women from biopsies taken at the initial visit. For those that were reported, 90% of the samples were positive for p16 expression.
A total of 506 (82%) women completed the protocol, whereas 109 women (18%) did not complete the protocol. The reasons for this included delayed follow-ups (n=41, 7%), lost to follow-up (n=41, 7%), elected treatment (n=22, 4%), refused biopsy (n=4, 1%), and 1 woman died of an unrelated cause. Of the 41 women lost to follow-up, 13 had cytology follow-ups (but no treatment or colposcopy), whereas 28 were truly lost to follow-up (ie, no evidence of cytologic or colposcopic follow-up within 24 months of initial biopsy or 12 months after last follow-up visit, whichever was later). Women who did not complete the protocol did not differ from women who did in terms of age, ethnicity, grade of referral cytology, grade of cytology at baseline, lesion size, or colposcopic impression. However, a higher proportion of women who did not complete the protocol were smokers (X 2 1 , 8.5; P =.004) and had not had an HPV vaccination (X 2 1 , 3.9; P =.047). No women were diagnosed with cervical cancer during the study.
Of the 506 women who completed the observational protocol, 326 had confirmed regression, 24 women had unconfirmed regression (ie, they had their first regression at the final 24-month follow-up), and 156 completed with CIN3 (n=128), adenocarcinoma in situ (n=2), or persistent CIN2 (n=26). Of the women with confirmed regression, 309 (95%) women had at least 1 histology result of CIN1 or less to confirm regression. The remaining 17 (5%) women had no biopsy taken but no lesion visible and normal colposcopic opinion and low-grade or normal cytology.
This implies that the chance of confirmed regression is 53% (326 of 615; 95% CI, 49%–57%) in all enrolled women and, when the missing data were imputed, it is estimated that 64% of women (95% CI, 60%–68%) would have experienced regression. In alignment with this estimation, lesions were observed to regress in 64% (326 of 506; 95% CI, 60%–69%) of women who completed the observational protocol.
Table 2 shows the unadjusted (univariable) and adjusted (multivariable) analysis of the clinical and demographic variables at baseline and the chance of having confirmed regression within 24 months. Missing baseline data and outcomes for women who did not complete the observational management protocol were imputed. At the univariable level, low-grade or normal colposcopic impression, smaller lesion size, later year of diagnosis, low-grade or normal baseline cytology, hrHPV type (ie, HPV 16 negative), nonsmoking, and low-grade referral cytology were all associated with an increased chance of having confirmed regression within 24 months. HPV 18 status did not affect the likelihood of regression but, because of its rarity, this study was likely underpowered for this analysis. Expression of p16 did not affect the likelihood of regression. In the multivariable analysis, detection of HPV 16 in a cytology sample at baseline was associated with a 31% decreased chance of regression (adjusted risk ratio, 0.69; 95% CI, 0.56–0.86; P <.001). Baseline measures associated with an increased chance of regression were (1) low-grade or normal colposcopic impression (13% higher chance of regression than women with a CIN2 colposcopic impression and 27% higher chance of regression than women with a CIN3 colposcopic impression), (2) later year of diagnosis (6% increased chance per year), (3) low-grade or normal cytology (15% increased chance of regression than women with high-grade cytology), and (4) nonsmoking (15% higher chance of regression than smokers). There was weak evidence for the association between not using combined oral contraceptives and an increased chance of regression ( P =.052).
