A risk prediction model to allow personalized screening for cervical cancer

Predictive models for personalized precision medical intervention in spontaneous regression stages of cervical precancerous lesions

BACKGROUND

During the prolonged period from Human Papillomavirus (HPV) infection to cervical cancer development, Low-Grade Squamous Intraepithelial Lesion (LSIL) stage provides a critical opportunity for cervical cancer prevention, giving the high potential for reversal in this stage. However, there is few research and a lack of clear guidelines on appropriate intervention strategies at this stage, underscoring the need for real-time prognostic predictions and personalized treatments to promote lesion reversal.

METHODS

We have established a prospective cohort. Since 2018, we have been collecting clinical data and pathological images of HPV-infected patients, followed by tracking the progression of their cervical lesions. In constructing our predictive models, we applied logistic regression and six machine learning models, evaluating each model's predictive performance using metrics such as the Area Under the Curve (AUC). We also employed the SHAP method for interpretative analysis of the prediction results. Additionally, the model identifies key factors influencing the progression of the lesions.

RESULTS

Model comparisons highlighted the superior performance of Random Forests (RF) and Support Vector Machines (SVM), both in clinical parameter and pathological image-based predictions. Notably, the RF model, which integrates pathological images and clinical multi-parameters, achieved the highest AUC of 0.866. Another significant finding was the substantial impact of sleep quality on the spontaneous clearance of HPV and regression of LSIL.

CONCLUSIONS

In contrast to current cervical cancer prediction models, our model's prognostic capabilities extend to the spontaneous regression stage of cervical cancer. This model aids clinicians in real-time monitoring of lesions and in developing personalized treatment or follow-up plans by assessing individual risk factors, thus fostering lesion spontaneous reversal and aiding in cervical cancer prevention and reduction.

Human Leucocyte Antigen Class II Risk and Protective Alleles in Women with Cervical Intraepithelial Neoplasia

Background

Persistent human papillomavirus (HPV) infection is a necessary cause for development of cervical precancerous lesions and cervical cancer, however, only a small percentage of women progress to cervical cancer. The local immune response, determined, among other factors, by Human Leucocyte Antigen (HLA) genes, is thought to be significant. Still the results of genome studies are inconsistent and differ between ethnical populations. The aim of the study was to assess an association between HLA-DQA1*; DQB1*; DRB1* allele's genetic variants between women with cervical precancerous lesions and healthy controls in Latvia.

Materials and methods

From January until April 2017 we enrolled 84 consecutive patients referred for colposcopy to Riga East University Hospital (Latvia) due to abnormal cervical cytology results. 57 women who came for a regular check-up and had normal cytology smears were included in the control group. Material from the cervix was taken for subsequent HLA genotyping of 13 DRB1*, 8 DQA1*, and 12 DQB1* alleles. Colposcopy was performed on all participants. In case of visual suspicion for CIN cervical biopsy was done.

Results

There were 57 "no CIN" patients, 23 histologically proven CIN 1 and 61 CIN2+ cases in the study population. CIN2+ was more often associated with DQA1*0401 (OR 6.68, 95% CI 1.47-30.29, p=0.014), DRB*15 (OR 2.99, 95% CI 1.22-7.39, p=0.017), DQB1*0401 (OR 2.91, 95%CI 1.11-7.68, p=0.03), DQA1*0103 (OR 2.72, 95% CI 1.02-7.21, p=0.045), DRB1*11 (OR 2.42, 95% CI 1.10-5.33, p=0.029) and DQB1*0301 (OR 1.94, 95% CI 1.12-3.38, p=0.018). Women with "no CIN" more often had DQB1*0501 (OR 0.17, 95% CI 0.04-0.81, p=0.026), DRB1*16 (OR 0.21, 95% CI 0.06-0.78, p=0.019), DQA1*0301 (OR 0.35, 95% CI 0.14-0.87, p=0.024) and DRB1*14 (OR 0.59, 95% CI 0.01-0.46, p=0.007).

Conclusions

In the current study we have demonstrated a strong association with risk and protective HLA class II alleles that are determined by the HLA-DRB1*; DQA1*; DQB1*.

Acceptability of risk-based triage in cervical cancer screening: A focus group study

BACKGROUND

Compared to the previous cytology-based program, the introduction of primary high-risk human papillomavirus (hrHPV) based screening in 2017 has led to an increased number of referrals. To counter this, triage of hrHPV-positive women in cervical cancer screening can potentially be optimized by taking sociodemographic and lifestyle risk factors for cervical abnormalities into account. Therefore, it is essential to gain knowledge of the views of women (30-60 years) eligible for cervical cancer screening.

OBJECTIVE

The main goal of this qualitative study was to gain insight in the aspects that influence acceptability of risk-based triage in cervical cancer screening.

DESIGN

A focus group study in which participants were recruited via four general medical practices, and purposive sampling was used to maximize heterogeneity with regards to age, education level, and cervical cancer screening experiences.

APPROACH

The focus group discussions were transcribed verbatim and analyzed using reflexive thematic analysis.

PARTICIPANTS

A total of 28 women (average age: 45.2 years) eligible for cervical cancer screening in The Netherlands participated in seven online focus group discussions. Half of the participants was higher educated, and the participants differed in previous cervical cancer screening participation and screening result.

KEY RESULTS

In total, 5 main themes and 17 subthemes were identified that determine the acceptability of risk-stratified triage. The main themes are: 1) adequacy of the screening program: an evidence-based program that is able to minimize cancer incidence and reduce unnecessary referrals; 2) personal information (e.g., sensitive topics and stigma); 3) emotional impact: fear and reassurance; 4) communication (e.g., transparency); and 5) autonomy (e.g., prevention).

CONCLUSION

The current study highlights several challenges regarding the development and implementation of risk-based triage that need attention in order to be accepted by the target group. These challenges include dealing with sensitive topics and a transparent communication strategy.

Development of a multivariable prediction model for anal high-grade squamous intraepithelial lesions in persons living with HIV in Puerto Rico: a cross-sectional study

Background

Persons living with HIV (PLWH) are at high risk of developing anal high-grade squamous intraepithelial lesions (HSIL). We aimed to develop a prediction model for anal HSIL based on individual characteristics of PLWH.

Methods

Cross-sectional study of PLWH aged ≥21 years who attended the Anal Neoplasia Clinic of the University of Puerto Rico Comprehensive Cancer Center from 2016 to 2022. The primary outcome was biopsy-confirmed anal HSIL. For each sex, relations between potential predictors and HSIL were examined using univariate (ULRM) and multivariable (MLRM) logistic regression models. Risk modelling was performed with MLRM and validated with bootstrapping techniques. The area under the ROC Curves (AUC) was estimated with 95% CI.

Findings

HSIL was detected among 45.11% of patients, 68.48% were males, and 59.42% were ≥45 aged. Multivariable analysis showed that, in women, the only significant predictor for HSIL was having a previous abnormal anal cytology (p = 0.01). In men, significant predictors for HSIL were having a previous abnormal anal cytology (p < 0.001) and a history of infection with any gonorrhoea (p = 0.002). Other suggestive predictors for HSIL among women were obesity and smoking. No association between smoking and HSIL among men was observed (p < 0.05). The AUC estimated among women (0.732, 95% CI: 0.651-0.811) was higher than in men (0.689, 95% CI: 0.629-0.748).

Interpretation

Our results support that the inclusion of individual characteristics into the prediction model will adequately predict the presence of HSIL in PLWH.

Funding

This work was supported by the NCI (Grants #U54CA096297, #R25CA240120), the NIGMS (Grant #U54GM133807), and the NIMHD (Grant #U54MD007587).

A risk prediction model mediated by genes of APOD/APOC1/SQLE associates with prognosis in cervical cancer

Cervical cancer is one of the most common gynecological malignancies. Due to the high heterogeneity of cervical cancer accelerating cancer progression, it is necessary to identify new prognostic markers and treatment regimens for cervical cancer to improve patients' survival rates. We purpose to construct and verify a risk prediction model for cervical cancer patients. Based on the analysis of data from the Gene Expression Omnibus database (GEO) and The Cancer Genome Atlas (TCGA), differences of genes in normal and cancer samples were analyzed and then used analysis of WGCNA along with consistent clustering to construct single-factor + multi-factor risk models. After regression analysis, the target genes were obtained as prognostic genes and prognostic risk models were constructed, and the validity of the risk model was confirmed using the receiver operating characteristic curve (ROC) and Kaplan-Meier curve. Subsequently, the above model was verified on the GSE44001 data validation followed by independent prognostic analysis. Enrichment analysis was conducted by grouping the high and low risks of the model. In addition, differences in immune analysis (immune infiltration, immunotherapy), drug sensitivity, and other levels were counted by the high and low risks groups. In our study, three prognostic genes including APOD, APOC1, and SQLE were obtained, and a risk model was constructed along with validation based on the above-mentioned analysis. According to the model, immune correlation and immunotherapy analyses were carried out, which will provide a theoretical basis and reference value for the exploration and treatment of cervical cancer.

Evaluating the Feasibility of Machine-Learning-Based Predictive Models for Precancerous Cervical Lesions in Patients Referred for Colposcopy

Background: Colposcopy plays an essential role in cervical cancer control, but its performance remains unsatisfactory. This study evaluates the feasibility of machine learning (ML) models for predicting high-grade squamous intraepithelial lesions or worse (HSIL+) in patients referred for colposcopy by combining colposcopic findings with demographic and screening results. Methods: In total, 7485 patients who underwent colposcopy examination in seven hospitals in mainland China were used to train, internally validate, and externally validate six commonly used ML models, including logistic regression, decision tree, naïve bayes, support vector machine, random forest, and extreme gradient boosting. Nine variables, including age, gravidity, parity, menopause status, cytological results, high-risk human papillomavirus (HR-HPV) infection type, HR-HPV multi-infection, transformation zone (TZ) type, and colposcopic impression, were used for model construction. Results: Colposcopic impression, HR-HPV results, and cytology results were the top three variables that determined model performance among all included variables. In the internal validation set, six ML models that integrated demographics, screening results, and colposcopic impression showed significant improvements in the area under the curve (AUC) (0.067 to 0.099) and sensitivity (11.55% to 14.88%) compared with colposcopists. Greater increases in AUC (0.087 to 0.119) and sensitivity (17.17% to 22.08%) were observed in the six models with the external validation set. Conclusions: By incorporating demographics, screening results, and colposcopic impressions, ML improved the AUC and sensitivity for detecting HSIL+ in patients referred for colposcopy. Such models could transform the subjective experience into objective judgments to help clinicians make decisions at the time of colposcopy examinations.

Towards a data-driven system for personalized cervical cancer risk stratification

Mass-screening programs for cervical cancer prevention in the Nordic countries have been effective in reducing cancer incidence and mortality at the population level. Women who have been regularly diagnosed with normal screening exams represent a sub-population with a low risk of disease and distinctive screening strategies which avoid over-screening while identifying those with high-grade lesions are needed to improve the existing one-size-fits-all approach. Machine learning methods for more personalized cervical cancer risk estimation may be of great utility to screening programs shifting to more targeted screening. However, deriving personalized risk prediction models is challenging as effective screening has made cervical cancer rare and the exam results are strongly skewed towards normal. Moreover, changes in female lifestyle and screening habits over time can cause a non-stationary data distribution. In this paper, we treat cervical cancer risk prediction as a longitudinal forecasting problem. We define risk estimators by extending existing frameworks developed on cervical cancer screening data to incremental learning for longitudinal risk predictions and compare these estimators to machine learning methods popular in biomedical applications. As input to the prediction models, we utilize all the available data from the individual screening histories.Using data from the Cancer Registry of Norway, we find in numerical experiments that the models are strongly biased towards normal results due to imbalanced data. To identify females at risk of cancer development, we adapt an imbalanced classification strategy to non-stationary data. Using this strategy, we estimate the absolute risk from longitudinal model predictions and a hold-out set of screening data. Comparing absolute risk curves demonstrate that prediction models can closely reflect the absolute risk observed in the hold-out set. Such models have great potential for improving cervical cancer risk stratification for more personalized screening recommendations.

Cervical Cancer in the Baltic States: Can Intelligent and Personalised Cancer Screening Change the Situation?

The three Baltic States (Estonia, Latvia, and Lithuania) are among the European Union countries with the highest incidence and mortality rates for cervical cancer. In order to tackle this public health challenge, there is an urgent need to implement more advanced and effective methods in cervical cancer prevention in Baltic countries.

Nationwide cervical cancer screening programs in the Baltic States commenced in 2004–2009. While the organized screening programs in these countries differ in some relevant details (target age groups, screening interval), the underlying principles and problems, barriers are universal. However, the outcomes of present screening programs are unsatisfactory. In addition, universal screening programs are extremely costly. There is a potential need for more intelligent and personalized cervical cancer screening program. In 2019 the project “Towards elimination of cervical cancer: intelligent and personalized solutions for cancer screening” (2020–2023) was developed with the main objective – to develop improved and personalized cancer screening methods within a sustainable health care system.

It is expected, that more sophisticated cervical cancer screening model will be implemented in Estonia, Latvia, and Lithuania, and will have a positive impact to epidemiology of cervical cancer and public health in general.

Development of models for cervical cancer screening: construction in a cross-sectional population and validation in two screening cohorts in China

BACKGROUND

Current methods for cervical cancer screening result in an increased number of referrals and unnecessary diagnostic procedures. This study aimed to develop and evaluate a more accurate model for cervical cancer screening.

METHODS

Multiple predictors including age, cytology, high-risk human papillomavirus (hrHPV) DNA/mRNA, E6 oncoprotein, HPV genotyping, and p16/Ki-67 were used for model construction in a cross-sectional population including women with normal cervix (N = 1085), cervical intraepithelial neoplasia (CIN, N = 279), and cervical cancer (N = 551) to predict CIN2+ or CIN3+. A base model using age, cytology, and hrHPV was calculated, and extended versions with additional biomarkers were considered. External validations in two screening cohorts with 3-year follow-up were further conducted (N_Cohort-I = 3179, N_Cohort-II = 3082).

RESULTS

The base model increased the area under the curve (AUC, 0.91, 95% confidence interval [CI] = 0.88-0.93) and reduced colposcopy referral rates (42.76%, 95% CI = 38.67-46.92) compared to hrHPV and cytology co-testing in the cross-sectional population (AUC 0.80, 95% CI = 0.79-0.82, referrals rates 61.62, 95% CI = 59.4-63.8) to predict CIN2+. The AUC further improved when HPV genotyping and/or E6 oncoprotein were included in the base model. External validation in two screening cohorts further demonstrated that our models had better clinical performances than routine screening methods, yielded AUCs of 0.92 (95% CI = 0.91-0.93) and 0.94 (95% CI = 0.91-0.97) to predict CIN2+ and referrals rates of 17.55% (95% CI = 16.24-18.92) and 7.40% (95% CI = 6.50-8.38) in screening cohort I and II, respectively. Similar results were observed for CIN3+ prediction.

CONCLUSIONS

Compared to routine screening methods, our model using current cervical screening indicators can improve the clinical performance and reduce referral rates.

Effect of Sequential Rounds of Cervical Cancer Screening on Management of HPV-positive Women: A 15-year Population-based Cohort Study from China

Women are anticipated to go through more than two rounds of cervical screening in their lifetime. Human papillomavirus (HPV) testing is increasingly used as the primary cervical cancer screening test. However, triage strategies for HPV-positive women were usually evaluated at baseline screening. We assessed the effect of sequential rounds of cervical screening on several algorithms for HPV triage. A total of 1,997 women ages 35-45 years were enrolled in 1999 in Shanxi, P.R. China and followed up three times at approximately 5-year intervals. Cervical intraepithelial neoplasia (CIN) grade 2 or worse (CIN2+) prevalence by prior HPV results and performance of 12 triage algorithms with cytology, genotyping, and prior HPV were examined among 229 HPV-positive women at the fourth round. CIN2+ prevalence varied from 56.5% (95% confidence interval, 36.8%-74.4%) following 15 years HPV persistence to 3.5% (1.2%-9.9%) with an incident HPV within 15 years. Triage with cytology (with threshold of atypical squamous cells of undetermined significance) yielded positive predictive value (PPV) of 21.4% (13.8%-29.0%), entailing immediate colposcopic referral, and negative predictive value (NPV) of 97.4% (94.6%-100%), permitting retesting at short intervals. Triage with genotyping (16/18/31/33/45/52/58) or prior HPV results showed comparable performance with cytology. Among 11 triage algorithms with similar NPV to cytology, triage with prior HPV results and reflex genotyping (16/18) achieved highest PPV of 28.9% (18.8%-39.1%) and lowest colposcopy referral of 33.2% (27.4%-39.5%). HPV persistence across rounds is an effective risk stratifier in HPV-positive women. Mainstream cytology and genotyping, with or without consideration of prior HPV results, remain effective for HPV triage at fourth round. PREVENTION RELEVANCE: The study highlights the sustained effectiveness of mainstream HPV triage methods, such as cytology and genotyping, after sequential rounds of cervical screening. It also suggests that use of HPV persistence across rounds can improve management of HPV-positive women in cervical cancer screening.

Risk factors for cervical cancer in women in China: A meta-model

Objectives:

Cervical cancer is a leading cause of cancer-related mortality in women in China. This analysis is a quantitative evidence synthesis pooling information about each cervical cancer risk factor.

Methods:

A meta-model was developed to estimate the risk of cervical cancer for a woman aged 18–85 years in Mainland China based on her risk profile at the time of assessment. The meta-model was built using findings of a systematic literature review that identified 21 case–control studies reporting data on 105 groups of cervical cancer risk factors in Chinese women. Extracted risk factors were ranked, and 17 were selected by Chinese clinical experts for inclusion in the meta-model. Risk equations were developed for each selected study. Predicted risks for each study were dependent on the risk profile under consideration and study-specific risks were pooled to an overall risk estimate using a random-effects meta-analysis. Sensitivity analysis was conducted using 100 artificial patient profiles (in the absence of patient data).

Results:

Predicted risks for the 100 profiles suggested that the model had good face validity and could differentiate between high and non-high cervical cancer risk profiles.

Conclusion:

This innovative meta-model approach assesses cervical cancer risk in Chinese women from a holistic perspective and could be adapted for other diseases and settings.

Risk prediction of cervical abnormalities: The value of sociodemographic and lifestyle factors in addition to HPV status

High-risk human papillomavirus (hrHPV) assessment as a primary screening test improves sensitivity but decreases specificity. Determining risk for cervical abnormalities and adapting policy accordingly may improve the balance between screening benefits and harms. Our aim is to assess the value of factors other than HPV in prediction of cervical abnormalities. Data from a Dutch prospective cohort were used. Women aged 18-29 years, not yet eligible for screening, were included in 2007. Data collection consisted of a questionnaire and a cervicovaginal self-sample. Linkage with PALGA (pathology database) was performed in 2017. The analyses included 1483 women. The full model, including sociodemographic and lifestyle factors, was compared to the null model, including baseline HPV only. The outcome of interest was cervical intraepithelial neoplasia 2 or worse (CIN2+). There were 86 women with CIN2+. Baseline hrHPV status was an important predictor (OR = 5.20, 95%CI = 3.27-8.27). The area under the ROC curve (AUC) of the null model was 0.67 (95%CI = 0.61-0.72). The full model had a slightly higher AUC of 0.73 (95%CI = 0.67-0.79). Bootstrap validation indicated that overfitting was present. This exploratory study has confirmed that a single hrHPV measurement is a strong predictor of cervical abnormalities, and additional risk factors in young women appeared to have limited added value. However, prediction based on hrHPV only does leave room for improvement. Future studies should therefore focus on women in the screening age range and search for other predictors to further enhance risk prediction. Adapting policy based on risk may eventually help optimise screening performance.