Factors predicting progression of conservatively managed cervical intraepithelial neoplasia grade 2: cytology and HPV integration

Study design

This study is a multicentre retrospective study, which enrolled a total of 2230 patients with CIN2 from the Department of Obstetrics and Gynecology of Tongji Hospital (786), Obstetrics and Gynecology Hospital of Fudan University (1089) and the First People’s Hospital of Anhui Province (355) between 2020 and December 2024. The inclusion criteria were as follows: (1) aged 18–50 years; (2) no previous history of cervical cancer or autoimmune diseases; (3) initial pathological diagnosis of CIN2, with no physical therapy or surgical treatment performed within at least half a year. The exclusion criteria were: (1) pregnant or lactating patients and (2) previous cervical conisation due to cervical lesions. All enrolled patients underwent P16 immunohistochemical staining, and the results were independently interpreted and diagnosed as CIN2 by two pathologists at the patient’s centre. By the end of the follow-up period, the outcomes of the patients’ lesions were analysed to explore the relevant clinical and molecular factors associated with the progression or regression of CIN2 in patients.

Thinprep cytologic test(TCT)

Pap samples were collected using a cytology brush and stored in tubes filled with ThinPrep PreservCyt Solution (Hologic). Slides were prepared, stained and processed on a Thinprep 5000 Processor (Hologic). All the slides were reviewed by experienced pathologists who had access to patients’ clinical history through computer-aided reading (Hologic). According to the Bethesda system, the cytologic results were broadly divided into the following categories in our study: Categories ≥ atypical squamous cells – cannot rule out HSIL (ASC-H) include ASC-H, atypical glandular cell (AGC) and high-grade squamous intraepithelial lesion (HSIL); categories＜ASC-H include negative for intraepithelial lesion or malignancy (NILM), atypical squamous cells of undetermined significance (ASC-US) and low-grade squamous intraepithelial lesion (LSIL).

HPV test

HPV test was conducted successively on each participant at the same time as the TCT. The HPV DNA assay was conducted using the Cobas 4800 system (Roche Molecular Diagnostics), which detects HPV16, HPV18 and non-HPV16/18 high-risk types (HPV 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68), or the Bioperfectus Human Papillomavirus Genotyping Real Time PCR kit, which detects 21 types of HPV (16, 18, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 26, 73, 82, 6, 11, 81).

HPV integration

The sampling method for the HPV integration test was the same as that of the HPV test, which used cervical exfoliated cells. After DNA extraction, 500 ng of genomic DNA was used from each patient for DNA library preparation. Based on our previous study, the HPV integration test was conducted using high-throughput viral integration detection.10 11 This technique detected 18 HPV integration types, including 13 high-risk HPV(HR-HPV) types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68) and 5 middle-risk HPV types (HPV26, 53, 66, 73 and 82). Low-quality bases of each read were trimmed using an average Q value <20. Window size=4 bases from left and right ends, and reads with length >100 bp were retained for downstream analysis. These clean reads were filtered using BWA(Burrows-Wheeler Aligner), and paired-end reads (partial read sequence aligned to the human genome and partial read sequence aligned to the HPV genome) were reserved. All integrated sites were verified using PCR and Sanger sequencing. Sequencing reads containing both HPV and human genome sequences (both not shorter than 20 bp in length) were labelled as one integrated read. Samples with more than five reads were considered HPV integration positive. Samples with discrepant HPV infection results between the HPV test and HPV integration test were excluded.

Follow-up and management

In this retrospective cohort, patient management followed real-world clinical practice. After opting for conservative management, patients received baseline evaluation (history, examination, TCT and HR-HPV testing). Follow-up typically involved cotesting (TCT+HR-HPV) at approximately 6- to 12-month intervals. Referral for colposcopy, as documented, was generally prompted by (1) TCT≥ASCUS, (2) persistent HR-HPV positivity or (3) clinician/patient concern. Colposcopic biopsies were taken from suspicious lesions (acetowhite/iodine non-uptake areas) or randomly from the transformation zone if no lesion was visible (2–4 points). During the follow-up of patients, cervical treatment such as loop electrosurgical excision procedure(LEEP) or cold knife conization(CKC) was indicated if the colposcopy-guided biopsy result is CIN3, CIN2 persisted for more than 2 years or the patient actively requested treatment at any point during the follow-up period.

Clinical outcome

The clinical outcomes of the enrolled patients were categorised into the persistence or progression group and the regression group. The regression group included histological regression and HPV negativity. Patients with histological regression showed histological biopsy results of chronic cervicitis or low-grade squamous intraepithelial lesion during follow-up, and these patients might not achieve HPV negativity. Some patients did not undergo rehistological biopsy during follow-up but achieved HPV clearance, which is classified into the HPV negativity. The persistence or progression group included patients with persistent CIN2, who had histological biopsy results still indicating CIN2 during the 6-month or 12-month follow-up. A total of 101 patients were finally used to construct the model for the persistence or progression of CIN2. Patients with progressive CIN2 showed histological biopsy results of CIN3 during subsequent follow-up or pathological diagnosis of CIN2+ in the subsequent LEEP procedure.

Statistics

Statistical analysis was performed using SPSS 21.0 (IBM Armonk, New York, USA) with two-sided tests assuming a significance level of 5% (p<0.05). χ² tests or Fisher’s exact tests were used for categorical variables. Finally, univariate and multivariate analyses were used to study the effects of different categorical risk factors and the outcome CIN2+, adjusted for all significant univariate predictors. A prediction model for CIN2 persistent/progression was established based on those significant risk factors. The model was recurrently sampled using Bootstrap, and model calibration was evaluated through calibration plots to assess the accuracy of the model. Receiver operating characteristic (ROC) curves were prepared to verify the discriminative power of the model. All analyses were performed using R V.4.4.3 (R Foundation for Statistical Computing, Vienna, Austria), with the pROC, rms and dcurves packages.12 The association of risk factors and the outcome CIN2+ was studied with the OR with its 95% CI.

Baseline clinical characteristics of CIN2 patients receiving conservative management

This study ultimately included 147 patients diagnosed with CIN2 by colposcopic biopsy who opted for conservative management (figure 1). The study cohort predominantly consisted of women of reproductive age, with the vast majority (110 cases, 74.83%) aged 26–40 years. The distribution of patients’ basic clinical characteristics was shown in table 1. Overall, the characteristics of this study cohort reflect a typical profile of patients currently encountered in clinical practice who are diagnosed with CIN2, have fertility intentions or are managed conservatively, providing a representative sample base for subsequent analysis.

Relationship between demographic characteristics and disease outcomes

After 6–12 months of standardised follow-up, 110 out of 147 patients (74.83%) achieved lesion regression (105 patients’ pathological diagnosis was <CIN2, and 5 patients’ HPV tests were negative), while 37 patients (25.17%) had persistent or progressive disease (CIN2+). We first assessed the impact of conventional demographic and clinical factors on prognosis (table 2). Statistical analysis indicated no significant difference in the distribution of different age strata between the two outcome groups (p=0.660). Furthermore, contraceptive method (p=0.602), gravidity (p=0.354), parity (p=0.884), HPV vaccination history (p=0.666) and history of vaginitis within the past 3 months (p=0.608) all failed to show significant associations with disease outcome. These results collectively suggest that within the context of conservative management defined in this study, traditional demographic and general clinical factors are not core indicators for predicting CIN2 prognosis. The outcome of the lesions may depend more closely on the biological behaviour of the virus and specific characteristics of the host–virus interaction.

Univariate analysis of pathological, HPV infection characteristics and prognosis

To identify key biological indicators affecting prognosis, we analysed a series of pathological and virological factors (table 3). The study found that HPV integration status was the strongest factor distinguishing outcomes: among integration-positive patients, the proportion of disease progression was as high as 52.63% (10/19), whereas the progression rate was 15.85% (13/82) in integration-negative patients, a difference with extremely significant statistical meaning (p=0.002). This result is highly consistent with the existing consensus that the integration of viral DNA into the host genome is a key molecular event driving the progression of cervical lesions towards high-grade disease and malignant transformation.13 14 Second, baseline cytology results (≥ASC-H) were also associated with a higher risk of progression (58.82% vs 41.18%, p=0.002), which aligns with the more active management strategy assigned to high-grade cytological abnormalities in the American Society for Colposcopy and Cervical Pathology (ASCCP) risk-based management guidelines.15 However, whether the HPV type was 16/18 (p=0.117) and single or multiple infections (p=0.974) and HPV persistence time (p=0.355) showed no significant differences between the two groups, suggesting that at this specific CIN2 stage, the traditional risk stratification by viral genotype might be overshadowed by other stronger molecular biological markers.

Given the decisive impact of HPV integration status observed in the overall cohort, we further investigated whether it held consistent predictive value across different high-risk HPV genotype subgroups (online supplemental table 1). The analysis revealed that among patients infected with HPV types 16/18 (n=52), the lesion progression risk in the integration-positive group (53.85%) was significantly higher than that in the integration-negative group (15.38%, p=0.016). Among patients infected with non-16/18 high-risk HPV types (n=49), the progression proportion of HPV-integration-positive cases was higher than that in the HPV-integration-negative group (50.00% vs 16.28%, p=0.090).

Univariable and multivariable logistic regression analysed the predictors of CIN2 persistence or progression

To control for confounding factors and identify true independent predictors, we performed logistic regression analysis (table 4). A total of 101 patients undergoing HPV integration test were included in this analysis. In the univariate analysis, among numerous candidate factors, gravidity (OR 4.40, 95% CI 1.20 to 16.06), cytology (OR 4.44, 95% CI 1.36 to 14.44) and HPV integration (OR 5.90, 95% CI 2.01 to 17.33) were significantly associated with the risk of CIN2 persistence/progression. After including variables with a p value <0.1 from the univariate analysis (gravidity, parity, HPV integration and cytology) into the multivariate logistic regression model, the results further confirmed that cytology severe than ASC-H (adjusted OR 5.29, 95% CI 1.37 to 20.41, p=0.016) and HPV integration positivity (adjusted OR 6.77, 95% CI 2.04 to 22.46, p=0.002) were the independent strong risk factors for CIN2 persistence/progression. The molecular characteristics of the 101 patients included in the analysis were presented in the online supplemental table 2.

Selection of factors influencing CIN2 outcome prediction model and model discrimination

Based on the key risk factors identified above, we constructed a risk stratification model incorporating cytological and virological indicators to evaluate its clinical utility (figure 2). Cytological status and HPV integration status were selected as core predictors for the logistic regression model forecasting persistent or progressive CIN2 lesions. Stratified risk analysis revealed drastically elevated progression risks in combined subgroups: patients with TCT ≥ASC-H plus HPV integration positivity carried a 100% risk of CIN2+, followed by HPV integration positive with cytology <ASCH (43.75%), while the lowest risk (12.68%) occurred in those with TCT <ASC-H and negative HPV integration (online supplemental table 3). Single biomarkers alone demonstrated weaker risk stratification capacity, which supports the use of the combined two-factor model. ROC curve analysis yielded an AUC of 0.782 (95% CI 0.689 to 0.874), indicating moderate-to-good discriminatory power to distinguish patients at high risk of adverse CIN2 outcomes. Calibration curve and decision curve analysis further validated stable predictive performance and favourable net clinical benefit across common probability thresholds, supporting the reliable discriminative value of this combined cytology-HPV integration model for CIN2 risk stratification (figure 3).