目的

本研究旨在比较中国低卫生资源地区人群筛查中不同宫颈癌筛查策略的绩效指标，为优化低卫生资源地区的宫颈癌筛查策略提供来自真实世界的证据。

方法

研究设计 本研究为一项前瞻性与回顾性相结合的研究，本研究回顾性和前瞻性地收集了2018年1月至2024年12月期间山西省武乡县三家医院的所有宫颈癌筛查数据，并通过身份证号或门诊登记号匹配初筛与随访记录。筛查策略 武乡县2018~2024年采用了四种宫颈癌初筛策略：①基于人乳头瘤病毒（HPV）初筛并使用醋酸染色肉眼观察/卢戈氏碘液染色肉眼观察（VIA/VILI）分流（ST1）：HPV16/18阳性或HPV其他12型阳性且VIA/VILI异常者转诊阴道镜。②HPV单独筛查（ST2）：HPV阳性者转诊阴道镜。③基于液基细胞学（LCT）单独筛查（ST3）：LCT>无上皮内病变或恶性病变（NILM）者转诊阴道镜。④HPV+LCT联合筛查（ST4）：HPV阳性或LCT>NILM者转诊阴道镜。阴道镜检查结果≥低度鳞状上皮内病变（LSIL）的女性，需要在阴道镜指导下取可疑部位宫颈组织进行病理活检。主要指标 筛查绩效指标包括初筛指标（参与人数、初筛阳性率）、后续筛查指标（阴道镜转诊率、阴道镜转诊依从率、活检率）、病变检出指标宫颈上皮内瘤变（CIN）2+/癌症的检出率、CIN2+/癌症的阳性预测值）以及获益指标（每检出1例CIN2+或宫颈癌所需的阴道镜转诊人数（NNR））。并对居住地、年龄组、筛查来源做亚组分析。数据分析 所有分析通过STATA 17.0和R 4.4.2完成。采用倾向性评分加权以校正各组间的基线分布差异并进行敏感性分析。采用R×C卡方检验或Fisher精确检验比较不同策略间的绩效指标，显著性水平设定为0.05，两两比较中根据Bonferroni校正将显著性阈值调整为0.0083。采用Cochran-Armitage趋势检验探讨年龄与主要指标之间的关系。

结果

经倾向性评分（PSW）加权后，本研究共纳入29,931名女性的39,741条筛查记录，共有5,190例阳性检出，初筛阳性率12.95%。与其他三种策略相比，ST3初筛阳性率显著较低（ST1 14.37%，ST2 17.49%，ST3 8.51%，ST4 22.19%；ST3与ST1/2/4比较P＜0.001）。在后续筛查环节，共有3,720人需要做阴道镜（阴道镜转诊率9.28%），实做阴道镜2,642例（阴道镜转诊依从率71.02%）。与其他三种策略相比，ST1的阴道镜转诊率显著较低（ST1 5.22%，ST2 17.49%，ST3 8.51%，ST4 15.04%；ST1与ST2-4比较P<0.001），且具有不低的阴道镜转诊依从性（ST1 82.44%，ST2 59.18%，ST3 62.23%，ST4 82.10%）。本研究共检出CIN2+病例246人（检出率0.82%），宫颈癌病例24人（检出率0.08%）。在人群获益方面，ST1每检出1例CIN2所需的阴道镜检查次数显著较少（ST1 6，ST2 32，ST3 9，ST4 26；ST1与ST2-4比较P<0.0083）。

结论

研究结果表明，基于HPV的宫颈癌筛查在带来更大益处的同时也存在潜在的危害，引入VIA/VILI作为分流方法可以有效降低阴道镜转诊率和随访流失率。该研究为在低卫生资源地区选择宫颈癌筛查策略提供了真实世界参考。

Objective

This study aims to compare the performance indicators of different cervical cancer screening strategies in large-scale population screening in low-health-resource areas of China, providing real-world evidence for optimizing cervical cancer screening strategies in such settings.

Methods

Study Design Screening data were collected retrospective and perspective from three hospitals in Wuxiang County, Shanxi Province from January 2018 to December 2024. Primary screening and follow-up records were matched using ID numbers or outpatient registration numbers. Screening Strategies The four screening strategies (mentioned as ST 1 to 4) were defined as follows: ST1 (human papillomavirus (HPV) combined with visual inspection with acetic acid/visual inspection with Lugol's iodine (VIA/VILI) Triage): participants with HPV16/18 positivity or other 12 high-risk HPV types with abnormal VIA/VILI results were referred for colposcopy. ST2 (HPV-based): participants with HPV positivity were referred for colposcopy. ST3 (liquid-based cytology test (LCT)-based): participants with LCT > negative for intraepithelial lesion or malignancy (NILM) were referred for colposcopy. ST4 (HPV Combined with LCT): participants with HPV positivity or LCT > NILM were referred for colposcopy. Pathological biopsy was needed if the result was worse than low-grade squamous intraepithelial lesion (LSIL), which determine the reference standard. Main Indicators Screening performance indicators included primary test metrics (participation; positive rate), referral metrics (colposcopy referral rate; colposcopy compliance rate; biopsy rate), detection metrics (detection rate for cervical intraepithelial neoplasia (CIN)2+ or cancer; positive predictive values (PPV) for CIN2+ or cancer) and benefit metrics (number of colposcopies needed to detect one case (NNR) of CIN2+ or cancer). Sub-group analyses were performed based on residence, age group, and screening source. Data Analysis All analyses were performed using STATA 17.0 and R 4.4.2. Propensity score weighting (PSW) was used to adjust for characteristic distribution differences and sensitivity analysis was conducted to confirm the validation of the results. Performance indicators across strategies were compared using R×C chi-square tests or Fisher’s exact tests, with a significance level of 0.05 and adjusted to 0.0083 for pairwise comparisons using Bonferroni correction. The Cochran-Armitage trend test was used to explore the relationship between age and main indicators.

Results

After PSW, 39,741 screening records from 29,931women were included, with 5,190 positive cases (12.95%) detected. ST3 had a significantly lower primary screening positive rate compared to the other strategies (ST1 14.37%, ST2 17.49%, ST3 8.51%, ST4 22.19%; ST3 vs. ST1/2/4, P < 0.001). In the follow-up screening phase, 3,720 women (9.28%) were referred for colposcopy, and 2,642 (71.02%) compliance with the referral. ST1 had a significantly lower colposcopy referral rate (ST1 5.22%, ST2 17.49%, ST3 8.51%, ST4 15.04%; ST1 vs. ST2-4, P < 0.001) and relatively high colposcopy compliance (ST1 82.44%, ST2 59.18%, ST3 62.23%, ST4 82.10%). A total of 246 (0.82%) CIN2+ cases and 24 (0.08%) cervical cancer cases were detected. In terms of population benefit, ST1 required significantly fewer colposcopies per CIN2+ case detected (ST1 6, ST2 32, ST3 9, ST4 26; ST1 vs. ST2-4, P < 0.0083).

Conclusion

The results shew that while HPV-based screening offers greater benefits, it also carries potential risks. Introducing VIA/VILI as a triage method effectively reduces colposcopy referral rates and follow-up losses. This study provides real-world evidence for selecting cervical cancer screening strategies in low-health-resource areas.