The 2019 ASCCP Consensus Guidelines address clinical management of patients with abnormal screening results, patients who are under surveillance after colposcopy, and patients who are undergoing surveillance after treatment. The new guidelines separate the process of setting clinical action thresholds (risk thresholds) from the actual risk assessment that is based on a patient's cervical test results and other factors such as screening history.
The uniform biology and well-understood natural history of cervical cancer has led to development of a variety of new assays, including both human papillomavirus (HPV)-based and non-HPV tests, that are very effective for screening, triage, and management of cervical precancers.1–4 Current cervical cancer screening and management relies on HPV testing and cytology from cervical samples. In the United States, any test that is used for clinical management requires regulatory approval by the Food and Drug Administration. Several tests have been approved for primary screening, including liquid-based cytology (LBC), HPV testing alone, and HPV testing in combination with cytology (co-testing). The large effort and costs associated with conducting regulatory trials, balanced against the relatively small target populations, makes it unlikely that these trials will address management questions.5 Therefore, tests for management are typically used “off-label,” because they have not been approved for the specific management indications, only recommended because of other studies and expert opinion.
It is in the purview of clinical practice guidelines to evaluate posttest risk estimates in relation to clinical action thresholds for management indications and make recommendations for clinical use. Two major data sources were used to develop the 2019 guidelines. Clinical databases, most importantly from Kaiser Permanente Northern California, were used to calculate risk estimates integrating HPV testing, cytology, and screening history. Because these databases do not include all currently available assays and strategies, other data sources are needed to supplement risk estimates of clinically relevant outcomes. Therefore, to inform the ASCCP consensus guidelines on management, we conducted a comprehensive systematic review and meta-analysis of published studies with thorough quality assessment (Clarke et al., in this issue) to evaluate tests for postcolposcopy and posttreatment surveillance.
We conducted this systematic review and meta-analysis following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA; see Figure 1) guidelines. The literature search focused on identifying articles reporting on tests/assays for cervical cancer screening, triage, postcolposcopy surveillance and posttreatment surveillance. We searched English-language, peer-reviewed studies published since 2012, when the guidelines were last updated, in the MEDLINE database via PubMed and Embase using search terms defined in the Supplemental Methods, http://links.lww.com/LGT/A153. We also reviewed reference lists of articles identified in the primary search for additional relevant studies. For the main literature search, results were limited to articles in English language published between January 1, 2012, and January 28, 2019. One specific question related to the ASCCP consensus guidelines was to evaluate HPV alone versus HPV-cytology co-testing posttreatment. During the period of the systematic review, only 2 studies directly compared HPV-cytology co-testing to HPV alone for posttreatment management. In 2012, a systematic review was published that included 8 additional studies published between 2004 and 2011,6 we included these articles to address the HPV versus co-testing question.
Titles and abstracts of identified articles were equally divided among working group members to be screened for inclusion. Articles not relevant to tests/assays for cervical cancer screening, triage, postcolposcopy surveillance, and posttreatment surveillance were excluded. Full-text versions of eligible articles were reviewed to determine eligibility; for these articles, the indication and assay type were recorded. Data abstraction was conducted in 2 phases for posttreatment surveillance and postcolposcopy surveillance to address the aims of the ASCCP consensus guidelines. We abstracted data on study location, study design, treatment modality (if applicable), assay/test, study inclusion criteria, follow-up algorithms, testing intervals, and the number of cases and noncases by various test results.
For this meta-analysis, articles were included if they evaluated the clinical performance of assays/tests for postcolposcopy surveillance and/or posttreatment surveillance. For studies of postcolposcopy surveillance, tests had to be conducted among individuals who underwent colposcopy and biopsy, without an indication for treatment. Studies that predominantly evaluated risk in women who had an indication for treatment, but were not treated for various reasons, were not considered since they do not reflect the typical postcolposcopy population. For studies of posttreatment surveillance, tests had to be conducted after individuals were treated predominantly for histologically confirmed CIN 2+ or CIN 3+. If a study evaluated more than one assay on the same patients, we prioritized assays that were used more commonly to allow pooling of data. If a study evaluated more than one assay on different subsets of patients, we included both sets of results in the analysis.
The quality of selected studies was independently evaluated by N.W. and M.A.C. using adapted Quality Assessment of Diagnostic Accuracy Studies (QUADAS)-2 revised tool criteria for quality assessment of included articles (Clarke et al., in this issue). Briefly, each article was evaluated in 4 areas (population, index test, reference test, flow and timing) using standardized signaling questions adapted to studies of cervical cancer screening and management. For each question, the risk of bias was evaluated, in categories of yes/no/unclear.
We used data abstracted on cases (CIN 2+ and CIN 3+) and noncases (<CIN 2) to estimate pooled absolute risks and 95% CI using multilevel logistic-normal random effects models. Between-study variation was quantified using the τ2 statistic. A lower τ2 value is an indicator of lower between-study variance and hence less heterogeneity. We visualized variation in study-specific estimates using forest plots. We evaluated pooled risk estimates for the following assay classes/types: HPV (all assays), cytology, and co-testing. For HPV assays, we further evaluated pooled risks for Hybrid Capture 2 (HC2), the most widely used test in the articles reviewed. For each assay/test, we calculated pooled risks of CIN 2+ or CIN 3+ among the full study population (i.e., baseline risk), the pooled risk in the test positives (corresponding to the positive predictive value), and the pooled risk in the test negatives (corresponding to the complement of the negative predictive value). All analyses were performed in Stata, Version 15.1 (StataCorp, College Station, TX).
Role of the Funding Source
The guidelines effort received support from the National Cancer Institute and ASCCP. Participating organizations supported travel for their participating representatives. All participating consensus organizations, including the primary funders, had equal and balanced roles in the consensus process including data analysis and interpretation, writing of manuscript, and decision to submit for publication. No industry funds were used in the development of these guidelines. The corresponding authors had final responsibility for the submission decision.
The PRISMA diagram summarizes the systematic review process (see Figure 1). A total of 2,862 articles were identified through the search, of which 168 non-English articles were excluded. Title and abstract review removed 2,301 articles, leaving 393 articles for all indications, including 50 articles for postcolposcopy and 66 articles for posttreatment indications, respectively. For the postcolposcopy surveillance indication, 50 articles underwent full-text review and 57–11 were included for data abstraction. Given that too few studies were available to conduct a meta-analysis per assay/test, we performed a qualitative synthesis of posttest risks, but all studies were included in a meta-analysis of baseline risk, which is independent of the assay used. For the posttreatment surveillance indication, 66 articles underwent full-text review and 2312–34 of these studies were included in the meta-analysis.
Summary of Assays
Among 28 articles included in the evaluation of postcolposcopy and posttreatment surveillance, data on 14 different assays were reported (see Table 1). These included LBC, dual stain cytology, p16 histology, 9 HPV DNA assays, and 2 HPV mRNA assays. The HC2 was the most widely used HPV DNA assay. With few exceptions, most assays used in the studies are available as commercial kits, and several of the HPV assays have regulatory approval for screening either alone or in co-testing.
Only 5 articles included data that could be abstracted to evaluate test performance for postcolposcopy surveillance. The baseline postcolposcopy risk of CIN 2+ in all included studies was 11% (95% CI = 8%–15%) ranging from 7%–17% (τ2 = 0.13) in individual studies (Supplemental Figure 1, http://links.lww.com/LGT/A148). There were not enough studies available to conduct a meta-analysis for each assay type. Two studies evaluated HPV testing, one using PCR (GP5+6+),8 and the other evaluating both HC2 and an HPV DNA Chip.10 In these studies, the risk of CIN 2+ after a negative HPV test ranged from 1.3% to 8.0%. Among those with a positive HPV test, the risk of CIN 2+ ranged from 11.1% to 16.3%. In the 2 studies that evaluated HPV 16/18 genotyping, one with GP5+6+8 and the other with HPV GenoArray,9 the risks among individuals testing negative were 7.0% and 13.0%, respectively, and the risks among individuals testing positive were 26.8% and 34.0%, respectively. In the one study that evaluated risk of CIN 2+ after p16 histology testing,7 the risk among individuals testing negative was 5.5% and 17.6% among individuals testing positive (see Figure 2).
In total, 23 studies were included in the meta-analysis. The HC2 was evaluated in 13 studies, whereas the remaining 10 studies evaluated other HPV tests, with 3 of these studies comparing HC2 with another HPV test in the same population (see Table 1). Most studies were conducted in Europe (57%), followed by Korea (17%), China (13%), North America (9%), and Thailand (4%). A majority evaluated testing at 6 months after treatment; outcomes were ascertained for a range of follow-up periods between 6 and 121 months, with most studies ranging from 24 to 36 months.
The posttreatment risk of CIN 2+ in all studies was 4.8% (95% CI = 3.4%–6.8%), ranging from 0.4%–19.5% (τ2 = 0.57) in individual studies (Supplemental Figure 2, http://links.lww.com/LGT/A149). The risk was similar in studies that evaluated HC2 (5.0%, 95% CI = 2.8%–8.7%) and other HPV tests (4.2%, 95% CI = 2.5%–6.8%; p-Het = 0.6).
Among individuals testing negative for HPV posttreatment, the risk of CIN 2+ was 0.69% (95% CI = 0.3%–1.5%) in all studies, ranging from 0.0% to 8.6% (τ2 = 2.11) in individual studies. In studies evaluating HC2, the risk was 0.82% (95% CI = 0.3%–2.2%) and in studies evaluating other HPV tests the risk was 0.41% (0.1%–1.6%; p-het = 0.417). Among individuals testing positive for HPV posttreatment, the risk of CIN 2+ was 18.3% (95% CI = 12.1%–26.6%) in all studies, ranging from 2.0% to 59.5% (τ2 = 1.05) in individual studies. In studies evaluating HC2, the risk was 22.2% (95% CI = 12.6%–36.2%) and in studies evaluating other HPV tests the risk was 13.9% (95% CI = 7.6%–24.2%; p-het = 0.257; see Figure 3).
A total of 10 studies evaluated LBC testing after treatment. Most studies defined a positive cytology result as atypical squamous cell of undetermined significance or worse, with the exception of one study that used an low-grade squamous intraepithelial lesion cutoff.26 The posttreatment risk of CIN 2+ in all studies was 6.8% (95% CI = 4.7%–9.7%; Supplemental Figure 3, http://links.lww.com/LGT/A150). Among individuals with negative cytology, the posttreatment risk of CIN 2+ was 1.7% (95% CI = 1.0%–3.1%) and among individuals with positive cytology, the posttreatment risk of CIN 2+ was 36.6% (95% CI = 28.4%–45.7%; see Figure 4).
Human Papilloma Virus Alone Versus Co-testing in Management
A specific question related to the ASCCP consensus guidelines was to evaluate the posttest risk associated with a positive and negative test result for HPV alone versus HPV-cytology co-testing after treatment. During the period of the systematic review, only 2 studies directly compared HPV-cytology co-testing to HPV alone for posttreatment management.25,30 In 2012, a systematic review was published that included 8 additional studies published between 2004 and 2011.6 We conducted a meta-analysis pooling all 10 studies evaluating co-testing versus HPV and cytology testing for posttreatment. Overall, the risk of CIN 2+ in these studies was 9.5% (95% CI = 6.4%–14.0%; Supplemental Figure 4, http://links.lww.com/LGT/A151). Among individuals with negative co-test results, the risk of CIN 2+ was 0.68% (95% CI = 0.2%–2.0%) and the risks among individuals with negative HPV and cytology results were 1.4% (95% CI = 0.9%–2.1%) and 2.5% (95% CI = 1.4%–4.5%), respectively (p value for heterogeneity = .069). Among individuals with positive co-test results, the risk of CIN 2+ was 24.9% (95% CI = 19.8%–30.8%) and the risks among individuals with positive HPV and cytology results were 31.7% (95% CI = 24.1%–40.4%) and 32.2% (95% CI = 24.7%–40.7%), respectively (p value for heterogeneity = .228; see Figure 5).
We conducted a systematic review and meta-analysis of diagnostic assays addressing management of individuals in surveillance after colposcopy and after treatment. We identified several assays that were used for these indications, but we were only able to conduct a formal meta-analysis for posttreatment surveillance. For the posttreatment indication, there were enough data to pool HPV DNA tests; however, other assays, such as HPV mRNA testing, dual stain, or methylation, had limited data and could not be pooled. We had enough data to separately pool risk estimates for HC2 and combined other HPV tests. To provide evidence for the risk-based approach underlying the consensus guidelines effort, we pooled estimates of baseline risk, of risk in test negatives, and of risk in test positives. To inform recommendations for the consensus guidelines, absolute risk estimates from the systematic review were evaluated in the context of clinical management thresholds. Two factors are important to make recommendations for clinical management: (a) the risk estimate in relation to a clinical threshold and (b) the precision of the risk estimate, i.e., how wide the CIs for that estimate are.
For patients with negative HPV DNA testing after treatment, the risk was 0.69% for all HPV tests, and 0.82% and 0.41% for HC2 and other HPV tests, respectively. All of these risk estimates were clearly below the colposcopy referral threshold but had either point estimates above the 1-year return threshold, or wide CIs crossing the 1-year threshold, suggesting that a 1-year return is adequate for individuals evaluated with HC2 or with other HPV assays.
We conducted a thorough assessment of quality using adapted QUADAS-2 criteria (Clarke et al., in this issue). We identified risk of bias in many of the studies, related to all domains of the criteria. Almost all studies had a risk of bias in the patient selection domain, reflecting the wide variation in clinical practice and the lack of standards for conducting posttreatment studies. Several studies showed risk of bias in multiple domains (Supplemental Table 1, http://links.lww.com/LGT/A152). Because most studies used CIN 2+ and not CIN 3 as a primary outcome, we were only able assess risk of CIN 2 +.
Despite these limitations, several clear messages come from this effort: Our data confirm that HPV testing provides superior reassurance compared with cytology in management of individuals after treatment. We also demonstrate that HPV-cytology co-testing only provides a small risk reduction compared with HPV alone, at the cost of higher test positivity.
Pooling absolute risk estimates is a novel approach to summarize studies of diagnostic accuracy in a systematic review and meta-analysis. Most commonly, summary estimates are generated for assay accuracy measures (sensitivity and specificity). Some studies compare new tests to established standards and report relative accuracy measures. Assay performance measures such as sensitivity and specificity and absolute risks are directly connected by the disease prevalence (or pretest risk) in a specific population.35 For a given assay's performance, absolute risks will be higher in a population with higher disease prevalence. Therefore, absolute risk estimates from studies with possible bias in the patient selection domain may be more variable. We demonstrated that disease prevalence varies substantially across studies. We also observed substantial variation of absolute risk in test positives, as a consequence of variation in pretest risk. In contrast, the heterogeneity of risk in HPV-negative patients after treatment was low, with most studies confirming a low risk among HPV-negative individuals.
Despite the variation in risk observed, the risk of HPV-positive patients after treatment is clearly high enough for colposcopy referral but does not cross the threshold for immediate treatment. Conversely, the risk in HPV-negative individuals is consistent with recommendations for follow-up testing in 1 year.
We observed a scarcity of high-quality studies that address the important areas of management of individuals after colposcopy and treatment. The working group felt that it is very important to adhere to quality criteria when designing, conducting, reporting, and evaluating diagnostic studies. As part of this systematic review, QUADAS-2 criteria were adapted to questions related to cervical cancer screening and management. We have demonstrated that meta-analyses of studies with less risk of bias have less heterogeneity, emphasizing the importance of adhering to quality standards.
In summary, we conducted a systematic review and meta-analysis of diagnostic studies for postcolposcopy and posttreatment management. Despite many publications in these areas, only a limited number of studies had data that could be abstracted for a systematic review, limiting the meta-analysis to the posttreatment indication. Even among those studies that we included in the meta-analysis, a majority had a high risk of bias related to various factors, particularly in the domain of the patient selection. There are several new assays that have shown promise for improved detection of precancers but that have not been sufficiently evaluated for management settings. More high-quality studies are needed to properly evaluate these new assays and approaches.
1. Cuschieri K, Ronco G, Lorincz A, et al. Eurogin roadmap 2017: Triage strategies for the management
-positive women in cervical screening
programs. Int J Cancer
2. Wentzensen N, Arbyn M, Berkhof J, et al. Eurogin 2016 Roadmap: how HPV
knowledge is changing screening practice. Int J Cancer
3. Wentzensen N, Arbyn M. HPV
-based cervical cancer screening- facts, fiction, and misperceptions. Prev Med
4. Wentzensen N, Schiffman M, Palmer T, et al. Triage of HPV
positive women in cervical cancer screening. J Clin Virol
5. Wentzensen N, Silver MI. Biomarkers for cervical cancer prevention programs: the long and winding road from discovery to clinical use. J Low Genit Tract Dis
6. Kocken M, Uijterwaal MH, de Vries AL, et al. High-risk human papillomavirus testing versus cytology in predicting post-treatment
disease in women treated for high-grade cervical disease: a systematic review
and meta-analysis. Gynecol Oncol
7. Cortecchia S, Galanti G, Sgadari C, et al. Follow-up study of patients with cervical intraepithelial neoplasia grade 1 overexpressing p16Ink4a. Int J Gynecol Cancer
8. Gurumurthy M, Cotton SC, Sharp L, et al. Postcolposcopy management
of women with histologically proven CIN 1: results from TOMBOLA. J Low Genit Tract Dis
9. Ye J, Cheng B, Cheng YF, et al. Prognostic value of human papillomavirus 16/18 genotyping in low-grade cervical lesions preceded by mildly abnormal cytology. J Zhejiang Univ Sci B
10. Kang WD, Ju UC, Kim SM. Is human papillomavirus genotype important in predicting disease progression in women with biopsy-proven negative or CIN1 of atypical squamous cell of undetermined significance (ASC-US) cytology? Gynecol Oncol
11. Tverelv LR, Sorbye SW, Skjeldestad FE. Risk for cervical intraepithelial neoplasia grade 3 or higher in follow-up of women with a negative cervical biopsy. J Low Genit Tract Dis
12. Bruhn LV, Andersen SJ, Hariri J. HPV
-testing versus HPV
-cytology co-testing to predict the outcome after conization. Acta Obstet Gynecol Scand
13. Byun JM, Jeong DH, Kim YN, et al. Persistent HPV
-16 infection leads to recurrence of high-grade cervical intraepithelial neoplasia. Medicine
14. Ceballos KM, Lee M, Cook DA, et al. Post-loop electrosurgical excision procedure high-risk human papillomavirus testing as a test of cure: the british columbia experience. J Low Genit Tract Dis
15. Cubie HA, Canham M, Moore C, et al. Evaluation of commercial HPV
assays in the context of post-treatment
follow-up: Scottish Test of Cure Study (STOCS-H). J Clin Pathol
16. Du R, Meng W, Chen ZF, et al. Post-treatment
human papillomavirus status and recurrence rates in patients treated with loop electrosurgical excision procedure conization for cervical intraepithelial neoplasia. Eur J Gynaecol Oncol
17. Fan A, Wang C, Han C, et al. Factors affecting residual/recurrent cervical intraepithelial neoplasia after cervical conization with negative margins. J Med Virol
18. Friebe K, Klapdor R, Hillemanns P, et al. The value of partial HPV
genotyping after conization of cervical dysplasias. Geburtshilfe Frauenheilkd
19. Gosvig CF, Huusom LD, Deltour I, et al. Role of human papillomavirus testing and cytology in follow-up after conization. Acta Obstet Gynecol Scand
20. Hansen J, Waibel J, Timme S, et al. Validity parameters of the human papillomavirus detection test Hybrid Capture 2 with and without cytology after laser destruction and large loop excision of the transformation zone treatment
of high-grade cervical intraepithelial neoplasia lesions. J Low Genit Tract Dis
21. Herfs M, Somja J, Howitt BE, et al. Unique recurrence patterns of cervical intraepithelial neoplasia after excision of the squamocolumnar junction. Int J Cancer
22. Innamaa A, Dudding N, Ellis K, et al. High-risk HPV
platforms and test of cure: should specific HPV
platforms more suited to screening in a ‘test of cure’ scenario be recommended? Cytopathology
23. Kalampokas E, Wilson J, Gurumurthy M, et al. Effect of high-risk human papillomavirus but normal cytology at test of cure on achieving colposcopy
standards. J Low Genit Tract Dis
24. Kang WD, Kim SM. Human papillomavirus genotyping as a reliable prognostic marker of recurrence after loop electrosurgical excision procedure for high-grade cervical intraepithelial neoplasia (CIN2-3) especially in postmenopausal women. Menopause
25. Khunamornpong S, Settakorn J, Sukpan K, et al. Application of HPV
DNA testing in follow-up after loop electrosurgical excision procedures in Northern Thailand. Asian Pac J Cancer Prev
26. Kong TW, Son JH, Chang SJ, et al. Value of endocervical margin and high-risk human papillomavirus status after conization for high-grade cervical intraepithelial neoplasia, adenocarcinoma in situ, and microinvasive carcinoma of the uterine cervix. Gynecol Oncol
27. Lubrano A, Medina N, Benito V, et al. Follow-up after LLETZ: a study of 682 cases of CIN 2-CIN 3 in a single institution. Eur J Obstet Gynecol Reprod Biol
28. Molloy M, Comer R, Rogers P, et al. High risk HPV
testing following treatment
for cervical intraepithelial neoplasia. Ir J Med Sci
29. Persson M, Brismar Wendel S, Ljungblad L, et al. High-risk human papillomavirus E6/E7 mRNA and L1 DNA as markers of residual/recurrent cervical intraepithelial neoplasia. Oncol Rep
30. Polman NJ, Uijterwaal MH, Witte BI, et al. Good performance of p16/ki-67 dual-stained cytology for surveillance of women treated for high-grade CIN. Int J Cancer
31. Ryu A, Nam K, Kwak J, et al. Early human papillomavirus testing predicts residual/recurrent disease after LEEP. J Gynecol Oncol
32. Torne A, Fuste P, Rodriguez-Carunchio L, et al. Intraoperative post-conisation human papillomavirus testing for early detection of treatment
failure in patients with cervical intraepithelial neoplasia: a pilot study. BJOG
33. Wu J, Jia Y, Luo M, et al. Analysis of residual/recurrent disease and its risk factors after loop electrosurgical excision procedure for high-grade cervical intraepithelial neoplasia. Gynecol Obstet Invest
34. Zhao C, Hong W, Li Z, et al. Human papillomavirus testing and cytologic/histopathologic "test of cure" follow-up results after excisional treatment
for high-grade cervical intraepithelial neoplasia. J Am Soc Cytopathol
35. Wentzensen N, Wacholder S. From differences in means between cases and controls to risk stratification: a business plan for biomarker development. Cancer Discov