You Are Here:
A Systematic Review for the U.S. Preventive Services Task Force
Release Date: October 2011
By Evelyn P. Whitlock, MD, MPH; Kimberly K. Vesco, MD, MPH; Michelle Eder, PhD; Jennifer S. Lin, MD, MCR; Caitlyn A. Senger, MPH; and Brittany U. Burda, MPH.
The information in this article is intended to help clinicians, employers, policymakers, and others make informed decisions about the provision of health care services. This article is intended as a reference and not as a substitute for clinical judgment.
This article may be used, in whole or in part, as the basis for the development of clinical practice guidelines and other quality enhancement tools, or as a basis for reimbursement and coverage policies. AHRQ or U.S. Department of Health and Human Services endorsement of such derivative products may not be stated or implied.
This article was first published in Annals of Internal Medicine on October 18, 2011 (http://www.annals.org).
Background: Screening programs using conventional cytology have successfully reduced cervical cancer, but newer tests might enhance screening.
Purpose: To systematically review the evidence on liquid-based cytology (LBC) and high-risk human papillomavirus (HPV) screening for the U.S. Preventive Services Task Force to use in updating its 2003 recommendation.
Data Sources: MEDLINE, Cochrane Central Register of Controlled Trials, and PsycINFO from January 2000 through September 2010.
Study Selection: Two independent reviewers selected fair- to good-quality English-language studies that compared LBC or HPV-enhanced primary screening with conventional cytology in countries with developed population-based screening for cervical cancer.
Data Extraction: At least 2 independent reviewers critically appraised and quality-rated studies and used standardized abstraction forms to extract data about test performance for detecting cervical intraepithelial neoplasia (CIN) and cancer and screening-related harms.
Data Synthesis: On the basis of 4 fair- to good-quality studies (141,566 participants), LBC had equivalent sensitivity and specificity to conventional cytology. Six fair- to good-quality diagnostic accuracy studies showed that 1-time HPV screening was more sensitive than cytology for detecting CIN3+/CIN2+ but was less specific. On the basis of 2 fair- to good-quality randomized, controlled trials (RCTs) (120,533 participants), primary HPV screening detected more cases of CIN grade 3 or cancers in women older than 30 years. Four fair- to good-quality diagnostic accuracy studies and 4 fair- to good-quality RCTs showed mixed results of cotesting (HPV plus cytology) in women aged 30 years and older compared with cytology alone, with no clear advantage over primary HPV screening. Incomplete reporting of results for all screening rounds, including detection of disease and colposcopies, limits our ability to determine the net benefit of HPV-enhanced testing strategies.
Limitation: Resources were insufficient to gather unpublished data, short-term trial data showed possible ascertainment bias, and most RCTs used protocols that differed from current U.S. practice.
Conclusion: Evidence supports the use of LBC or conventional cytology for cervical cancer screening, but more complete evidence is needed before HPV-enhanced primary screening is widely adopted for women aged 30 years and older.
Primary Funding Source: Agency for Healthcare Research and Quality.
Cervical cancer screening programs that use conventional cytology every 1 to 5 years have demonstrated reductions in both cervical cancer incidence and mortality over time (1). Conventional cytology, however, is imperfectly sensitive and labor-intensive, leading to keen interest in new screening technologies serving as alternatives or adjuncts (2). Liquid-based cytology (LBC) offers potentially improved test specimen collection that can support cotesting (HPV plus cytology), but its effect on screening test performance remains uncertain (2). Other, newer technologies have been spurred by the scientific establishment of the causal role of various high-risk human papillomavirus (HPV) types in cancer of the cervix and other tissues (3).
Currently, 3 tests for high-risk HPV—Digene Hybrid Capture 2 (Qiagen, Germantown, Maryland), Cobas 4800 HPV (Roche Diagnostics, Indianapolis, Indiana), and Cervista HR HPV (Hologic, Bedford, Massachusetts)—are approved by the U.S. Food and Drug Administration (FDA) for patients with atypical squamous cells of undetermined significance (ASC-US) on cytology to determine referral for colposcopy, and for co-testing women aged 30 years and older as a risk assessment or patient management tool (4, 5). A fourth test, Amplicor HPV (Roche Diagnostics), is awaiting FDA approval (6).
Benefits from screening rely primarily on histologic diagnosis and treatment of cervical intraepithelial neoplasia (CIN) (7) during the long preclinical period typical of cervical cancer (8, 9). Although there are varying levels of CIN (1, 2, and 3), CIN3 is considered the only truly precancerous lesion because it includes carcinoma in situ (10, 11) and is more likely to progress to invasive cervical cancer (12). Although CIN2 is the usual treatment threshold, it is heterogeneous, equivocal in cancer potential, and more likely to regress than CIN3 (10, 11). Histologic diagnoses (CIN or cancer) are made from a biopsy specimen taken during colposcopy. In the United States, the cytologic threshold for immediate colposcopy referral is generally a low-grade squamous intraepithelial lesion (13). For abnormal screening test results that do not meet the immediate referral threshold, retesting at shorter intervals is recommended; colposcopy referral should be triggered for persistent or progressively abnormal results on retesting (13, 14).
In 2003, the U.S. Preventive Services Task Force (USPSTF) recommended cervical cancer screening in sexually active women with a cervix (grade A recommendation), but concluded that the evidence was insufficient to recommend for or against the routine use of LBC or HPV testing as alternatives or adjuncts to cytology screening. In support of its updated recommendation, the USPSTF commissioned a targeted systematic review (15) and a separate modeling exercise comparing the benefits and harms of various screening strategies (16). We summarize the evidence from our full report here and in our companion paper (17).
This article addresses the following questions:
- To what extent does LBC improve sensitivity, specificity, and diagnostic yield and reduce indeterminate results and inadequate samples compared with conventional cervical cytology?
- What are the harms of LBC?
- What are the benefits of using HPV testing as a screening test, either alone or in combination with cytology, compared with not testing for HPV in women aged 30 years and older?
- What are the harms of using HPV testing as a screening test, either alone or in combination with cytology in women aged 30 years and older?
We followed a standard protocol; search, selection, assessment, and synthesis methods, with evidence tables, are detailed in our full report (15). This article summarizes the evidence about primary HPV screening in women aged 30 years and older. The full report also details the evidence for HPV screening in younger women and for HPV triage of ASC-US or low-grade squamous intraepithelial lesions on cytology (15).
We initially searched for systematic reviews, meta-analyses, and evidence-based guidelines on cervical cancer screening listed in the Database of Abstracts of Reviews of Effects, the Cochrane Database of Systematic Reviews, PubMed, and the Health Technology Assessment Database from 2000 through 2007. Two systematic reviews addressing LBC screening (18, 19) were used to identify primary studies before 2003. No systematic reviews on HPV testing that met our inclusion criteria were identified. We considered all studies in the previous USPSTF review (20) and conducted literature searches from 2003 through September 2010 by using MEDLINE, the Cochrane Central Register of Controlled Trials, and PsycINFO.
We conducted a targeted search for any studies related to the trials included in our review (published from September 2010 to 3 August 2011 in PubMed) to ensure that all relevant studies were captured in our previous literature searches. In addition, selected experts in the field were queried on 8 August 2011 to identify relevant publications. We found 9 additional studies including no additional reports from trials included the review; 4 contextually relevant (21, 22) or unrelated (23, 24) reports from previously identified cohorts, 1 performance study of a new HPV test (25), 2 unrelated reports from trial authors (26, 27), and 2 public health reports (28, 29). None added primary results to our key questions, but most added to our discussion (21-23, 26, 27).
We evaluated 4262 abstracts and 641 full-text articles (Figure 1). Two reviewers evaluated abstracts and articles against prespecified inclusion criteria. Discrepancies were resolved by consensus. We included fair-to-good-quality studies that provided evidence regarding test performance for detection of CIN2+ (CIN2, CIN3, or cancer) or CIN3+ (CIN3 or cancer), as well as harms. Included studies met design-specific quality standards that minimized the effect of verification bias; and were conducted in routine screening populations in countries with developed population-based screening for cervical cancer. For question 3, we evaluated the evidence regarding the use of HPV testing in screening scenarios: primary screening with HPV testing alone, primary HPV testing with cytology triage of positive HPV (reflex cytology), primary HPV plus cytology screening (cotesting), and cytology testing with HPV triage of ASC-US or low-grade squamous intraepithelial lesion on cytology (reflex HPV). Cytology with reflex HPV is covered in our full report (15).
Data Extraction and Quality Assessment
At least 2 investigators critically appraised and independently quality-rated all eligible studies by using criteria based on the USPSTF methods, supplemented by the National Institute for Health and Clinical Excellence criteria for quality of systematic reviews and the QUADAS (Quality Assessment of Diagnostic Accuracy Studies) tool (30-32). Good-quality studies generally met all design-specific criteria, whereas fair-quality studies did not meet all the criteria but had no fatal flaws in study design. Poor-quality studies had substantial flaws or lack of reporting that implied bias affecting interpretation of study results and were therefore excluded after agreement among reviewers. One investigator abstracted data from included studies into evidence tables, and a second reviewer verified these data.
Data Synthesis and Analysis
We performed qualitative data synthesis because heterogeneity in the samples, study designs, screening protocols, and instruments did not allow for quantitative synthesis. We synthesized results from diagnostic accuracy studies (to evaluate 1-time test performance) separately from randomized controlled trials (RCTs). For RCTs of HPV screening, we report results for each round of screening, as well as cumulative results. In these RCTs, results were generally reported for women screened (rather than an intention-to-screen analysis). For consistency, we report the results for women screened (denominator) unless otherwise noted. We also synthesize results for both CIN2+ and CIN3+ even though many CIN2 lesions will regress.
Role of the Funding Source
The Agency for Healthcare Research and Quality funded this work, provided project oversight, and assisted with internal and external review of the draft evidence synthesis, but had no role in the design, conduct, or reporting of the review. The authors worked with 8 USPSTF members, who helped set the review scope and provided input into methodological issues during the conduct of the review.
Benefits and Harms of LBC Compared With Conventional Cytology
We identified 1 fair- and 1 good-quality RCT (33, 34) comprising 134,162 women exclusively or predominately aged 30 to 60 years. These trials compared relative detection of CIN3+ and CIN2+ and relative positive predictive value after a single screening with LBC or conventional cytology. We identified 2 fair-quality observational studies (35, 36) of 7404 similarly aged women that reported absolute sensitivity and specificity of both tests in primary care applicable settings (Table 1).
On the basis of these studies, LBC and conventional cytology did not differ substantially in relative detection or absolute sensitivity or specificity for detection of CIN2+/CIN3+ at any cytologic threshold (Table 1). Although the fair-quality NTCC (New Technologies for Cervical Cancer Screening) trial reported lower relative positive predictive value for LBC than for conventional cytology (33), its findings are inconsistent with the good-quality NETHCON (Netherlands ThinPrep versus Conventional Cytology) trial (34) and with both observational studies (35, 36). The limitations of the NTCC trial, including the newness of LBC reading in many centers and lack of blinding, could have influenced these results. In terms of specimen adequacy, most of the evidence indicated a lower proportion of unsatisfactory slides for LBC than for conventional cytology (0.4% vs. 1.1% in NETHCON; 2.6% vs. 4.1% in NTCC). Technical issues probably explain disparate findings in smaller observational studies (15). Although we found no studies that directly addressed harms of LBC testing, we would not expect to find differential patient effects because LBC differs from conventional cytology primarily in specimen preparation and handling (37, 38).
Benefit and Harms of HPV Testing in Women Aged 30 Years and Older as an Alternative or Adjunct to Conventional Cytology Screening
We included 6 diagnostic accuracy studies, 6 RCTs of comparative effectiveness, and 4 studies on psychological harms of HPV screening. The volume (and quality) of evidence varied among 3 HPV screening strategies (Table 2 and Table 3 and Appendix Table) (39-51). For primary HPV screening compared with cytology, we found 1 RCT (NTCC phase 2; 49,196 participants) (39) and 6 diagnostic accuracy studies (comprising 37,431 participants) (36, 40-44); for HPV screening followed by cytology triage compared with cytology alone, we found 1 Finnish RCT (71,337 participants) (45); and for HPV and cytology cotesting, we found 4 RCTs (NTCC phase 1, POBASCAM [Population Based Screening Study Amsterdam Program], Swedescreen, and ARTISTIC [A Randomised Trial in Screening to Improve Cytology]; comprising 127,149 participants) (46-49) and 4 diagnostic accuracy studies including 21,739 participants (36, 41-43).
Studies of HPV-enhanced primary cervical cancer screening primarily evaluated Hybrid Capture 2, whereas a few used polymerase chain reaction testing. We report results for women aged 30 years and older to reflect the age bracket for FDA-approved use of Hybrid Capture 2 as an adjunct to cytology (4, 5) and the reduced prevalence of high-risk HPV in women as they age (17, 52). For results in younger women, please go to our full report (15).
Primary HPV Screening Alone Compared With Cytology Alone
In 6 fair- or good-quality diagnostic accuracy studies, 1-time HPV testing was more sensitive but less specific than cytology. For CIN3+ outcomes, point estimates for sensitivity ranged from 86% to 97% for HPV testing versus 46% to 50% for cytology at a colposcopy referral threshold of ASC-US. For CIN2+ outcomes, sensitivity ranged from 63% to 98% for HPV testing versus 38% to 65% for cytology (Table 2). However, specificity for CIN2+ and CIN3+ was consistently 3 to 5 percentage points lower for HPV testing than for cytology (Table 2).
In phase 2 of the NTCC, a fair-quality Italian RCT comparing Hybrid Capture 2 HPV screening with cytology in 35,471 women aged 35 to 60 years, about twice as many cases of CIN3+ were detected in the HPV testing group after a single round, with relatively fewer cases detected in the second screening round (relative detection ratio, 0.23 [95% CI, 0.07 to 0.82]) (Table 3) (39). After the second screening round (using cytology only in both groups) and a median of 3.5 years of follow-up from baseline, the cumulative relative detection of CIN3+ still increased in the HPV testing group (1.57 [CI, 1.03 to 2.40]). Because women with a positive HPV result or ASC-US on cytology were immediately referred for colposcopy, many baseline colposcopies were done overall but many more were done in the HPV testing group than in the cytology group (5.8% vs. 2.5%).
Trial investigators pooled cumulative cases of invasive cancer from the primary Hybrid Capture 2 screening strategy (NTCC phase 2) (39) with Hybrid Capture 2-cytology cotesting strategy (NTCC phase 1) (46), citing insignificant statistical heterogeneity between the trials. Pooled results suggested decreased cumulative cases of invasive cancer after HPV screening, compared with cytology (6 vs. 15; P = 0.052) in women aged 35 years or older. These findings are preliminary, since these cancer outcomes were based on pooling non-comparable screening strategies and also did not reflect similar opportunities for diagnosis in both strategies. More valid studies would ensure or control for similar delivery of colposcopy or provide longer follow-up with registry linkages to allow disease ascertainment outside the screening program.
Phase 2 of NTCC referred many women for colposcopy who would instead have been retested in the United States. The Appendix Table details other interpretation and quality issues with NTCC phase 2. Determination of benefits and burdens or harms of HPV testing and cytology screening is impossible because neither cumulative colposcopy results nor cumulative relative positive predictive value over both screening rounds were reported.
Primary HPV Screening Followed by Cytology Triage Compared With Cytology Alone
A large, fair-quality Finnish trial (59,757 women aged 35 to 65 years) compared primary Hybrid Capture 2 screening (with cytology triage for positive HPV tests) to cytology screening alone (45). Women with minimally abnormal results had repeated testing recommended. After a single screening round, Hybrid Capture 2 testing with cytology triage compared with cytology increased relative CIN2+ detection (1.36 [CI, 0.98 to 1.89]); effects on relative detection of CIN3+ were less clear because of the small sample size and wide CIs (1.38 [CI, 0.81 to 2.36]) (Table 3). Colposcopy referrals were modest in women older than 35 years and similar between HPV screening (0.9%) and cytology alone (1.0%); however, these probably include only immediate colposcopy referrals, because retesting was recommended for slightly more women who received HPV testing than who received cytology (7.2% vs. 6.6%). Extended follow-up (mean, 3.3 years; maximum, 5.0 years) with linkage to registry data in 38,670 screened women aged 30 to 60 years found significantly increased relative detection of CIN3+ (and other cancers) after a single round of HPV screening (1.77 [CI, 1.16 to 2.74]) (51). Among women with positive results in either group, most were retested to confirm abnormalities before colposcopy referral. Women with negative results on initial HPV testing tended toward a lower cumulative 5-year CIN3+ rate than women with negative results on initial cytology, although the CI for this estimate was wide (0.28 [CI, 0.04 to 1.17]) (data not shown).
In the Finnish trial, issues with interpretation and quality primarily reflect its incomplete reporting and implementation to date and the attributes of a pragmatic trial. As with many other trials, data for cumulative colposcopies, compliance with colposcopy, and retesting referrals for the entire first screening round are not yet reported. A second screening round at 3 years is planned. As more data from this trial are reported, differences with practice in the United States will also need to be considered.
Combination HPV and Cytology Screening (Cotesting) Compared With Cytology Alone
Four diagnostic accuracy studies (comprising 21,739 participants) reported the absolute test performance of HPV–cytology cotesting (Table 2). Two studies reporting sensitivity and specificity for Hybrid Capture 2 cytology cotesting among 17,885 women aged 30 to 60 years (36, 41) used a positive result from either test so that all HPV-positive patients met the threshold. For the detection of CIN3+/CIN2+, Hybrid Capture 2 testing plus cytology (either test positive) was more sensitive but less specific than cytology alone (Table 2). The combination of Hybrid Capture 2 plus cytology did not differ in performance from Hybrid Capture 2 alone. Two smaller studies (36, 43), comprising 3852 participants, reported positive cotesting results only if results of both tests were positive, unless a relatively high cytology threshold (i.e., high-grade squamous intraepithelial lesion) was met, similar to some cotesting trials. Wide CIs limit sensitivity comparisons, although specificity with this type of cotesting was clearly better than that of Hybrid Capture 2 alone (Table 2).
Four large, fair-quality RCTs (46-49)—NTCC phase 1, POBASCAM, Swedescreen, and ARTISTIC (comprising 82,390 participants)—compared cotesting with cytology screening alone in European women aged 30 to 64 years (Table 3). Cumulative CIN3+ detection was the same between cotesting and cytology alone after 2 screening rounds in all 4 RCTs, even though most cotesting trials also reported differences in round-specific relative CIN detection (Table 3). Cumulative invasive cancer detection was similar or slightly higher for cytology alone than for cotesting in 3 trials (46, 47, 49). However, ARTISTIC (48) had the opposite result: more cases of cancer were found after 2 screening rounds in the cotesting group (8 total), compared with cytology (4 total). Mixed round-specific and cumulative results among the trials may reflect between-trial differences in colposcopy referral and retesting protocols, as well as incomplete reporting of results.
Only 1 trial, NTCC phase 1 (46), found a relative increase in any cumulative CIN measure after cotesting. This test, however, used a lower threshold for immediate colposcopy than the other trials. Women aged 35 years or older were referred for colposcopy with either a cytology threshold of ASC-US or for a positive HPV result regardless of cytology. This strategy increased detection of both CIN2+ and CIN3+ after 1 screening round and cumulative CIN2+ detection overall (RR, 1.50 [CI, 1.13 to 1.98]) compared with cytology alone; however, it did not substantially decrease cases of CIN3+ in the second round or affect cumulative CIN3+ detection. More cases of invasive cancer occurred in the cytology-only group than in the cotesting group (10 vs. 2). On the basis of indirect comparisons between NTCC phases I and 2, cotesting offers no additional CIN3+ detection above primary HPV screening alone but may yield more false-positive results.
In the other 3 trials, high-grade squamous intraepithelial lesion was the referral threshold for colposcopy, with colposcopy referral for HPV-positive results only after repeated testing revealed persistent HPV positivity or abnormal cytology (47-49). These trials have not completely reported second-round detection outcomes for a substantial proportion of trial participants (47), the complete follow-up period (49), or both (48). Data from a third screening round reported in 2011 from ARTISTIC do not correct all of these reporting deficiencies but provide 6-year cumulative rates of CIN2+ and CIN3+ development by baseline screening test results (53).
Only 2 trials (47, 48) have reported cumulative colposcopies. These were slightly higher in the cotesting group than in the cytology group of POBASCAM (3.4% vs. 2.8%), although both groups received HPV testing with polymerase chain reaction in the second round. For women aged 30 to 64 years, cumulative colposcopy referrals after 2 screening rounds in ARTISTIC were 6.0% in the cotesting group compared with 4.9% in the LBC-only group (48). However, ARTISTIC varied somewhat from other trials in several round-specific findings, so the relative colposcopy requirement between groups is probably not applicable to trials with different protocols and CIN detection results. Although the interpretation and quality issues vary between cotesting trials, reporting on colposcopy referrals, compliance, referrals for retesting, CIN treatments, and related harms was insufficient across all trials. Cotesting trials also generally represent approaches to managing abnormal screening results that differ from current U.S. recommendations.
Harms of HPV Testing
Human papillomavirus testing could increase harms relative to cytology by increasing the number of unnecessary colposcopies and downstream consequences related to diagnosis and treatment. These concerns cannot be completely addressed due to incomplete reporting, but are considered further elsewhere (9, 21). To evaluate the potential psychological effects of HPV testing, we found 4 fair-quality observational studies (54-57) that used mailed questionnaires to examine the immediate and short-term effects of HPV testing in 4104 women in the United Kingdom or Australia. Levels of immediate anxiety and distress were increased in women who tested positive for HPV compared with those who tested negative. These differences, however, were resolved by 6-month follow-up. Data on other psychosocial outcomes and longer-term follow-up were sparse.
Substantial new evidence has become available since the previous USPSTF review and recommendation and continues to accrue. Large RCTs clearly establish that for cytology-based screening, LBC does not differ from conventional cytology in sensitivity, specificity, or relative CIN detection but may yield a lower proportion of unsatisfactory slides. Cost, overall screening strategy, and other considerations may also pertain to local decisions on which approach to use for conducting cytological screening.
Numerous studies have confirmed that HPV testing is more sensitive than cytology, but with a tradeoff in terms of reduced specificity. Thus, although HPV-enhanced screening strategies offer a potential cancer prevention benefit compared with cytology alone, test performance studies alone are insufficient to justify substituting HPV testing for cytology (58). Diagnostic work-up for false-positive results and diagnosis of regressive or nonprogressive histologic predisease could result in harms from unnecessary procedures or overtreatment. Understanding the tradeoff from reduced specificity is critical, particularly given the relatively low incidence of cervical cancer and the established practice of repeated cervical cancer screening (17). Thus, experts agree that large, pragmatic, comparative RCTs of repeated screening rounds are necessary, with increasing emphasis on the need to confirm the effect not just on surrogates (such as CIN) but also on cancer incidence and mortality (27).
On the basis of large RCTs, primary HPV screening seems very promising, particularly when coupled with reflex cytology to triage positive results before colposcopy. Screening with HPV testing enhances the detection of CIN3+ compared with cytology alone but also increases CIN2+ detection and immediate colposcopy referrals. All CIN lesions, even CIN3, have some potential for overdiagnosis and therefore potential harms (26). Thus, the net effect of primary HPV screening needs to be determined through the completion of ongoing trials and more detailed reporting of potential harms, as well as benefits from completed trials. An ongoing trial in Canada will also provide new evidence that directly compares primary HPV screening and cytology triage with cytology screening and HPV triage in a protocol more similar to U.S. practice than the European trials (59); also expected is a collaborative pooling of European trial results (60).
The FDA has already approved screening with cytology plus HPV testing (cotesting) in women aged 30 years and older (4, 5). Our report found that cotesting was much more sensitive than cytology alone but may represent a strategy that adds little when compared with primary HPV screening. On the basis of test performance data and indirect comparisons between trials, 1-time HPV–cytology cotesting was very similar to HPV testing alone for the detection of CIN2+ or CIN3+, with similar (or slightly reduced) specificity. However, incomplete reporting complicates the interpretation of cotesting trials, because most lack cumulative outcome reporting for their entire study populations, and all lack data on cumulative colposcopies and related harms. A large observational study (61) conducted in the United States (331,818 participants) reported high clinician and patient acceptance of cotesting in women aged 30 years and older, with rescreening deferred until 3 years after negative results in an HMO setting. The cumulative 5-year incidence of CIN3 and cervical cancer from this cotested cohort suggests that primary HPV testing, particularly if followed by cytology triage, would efficiently detect more cases of CIN3+ and cancers, particularly adenocarcinomas (78% vs. 15%), than cytology alone, with a very high negative predictive value for cancer after negative HPV test results (21). Data on the proportion of cumulative cases of cancer among HPV-negative/Papanicolaou-positive women and their relative stages at diagnosis would clarify any safety tradeoffs in moving away from cotesting.
A major benefit of HPV-enhanced primary screening could be identification of a low-risk cohort in whom a prolonged screening interval would be appropriate. As discussed in our full report, mounting evidence suggests that the cumulative risk for CIN3+ is very low for 5 or more years in women after negative results on HPV testing (15, 22, 47, 49, 51, 53, 62-64). Risk-stratifying approaches, whereby rescreening interval is prolonged on the basis of initial screening results, have not been directly incorporated into trials to date, and safety data on prolonged screening intervals in low-risk women based on baseline HPV testing (with or without cytology) seem promising but are still accruing. Such an approach could potentially reduce screening demands for many women; for example, in cotesting trials, 78% to 93% of tested women had negative results on both tests initially (15). Besides safety, feasibility or acceptability may affect adoption of a risk-stratified policy on cervical cancer screening because primary care physicians may not currently be extending the screening interval to 3 years after negative cotesting results (65). For women with positive as well as negative results on HPV screening, ongoing research into HPV subtypes (22), HPV-related biomarkers, and other factors (such as screening history ), will probably advance effective and efficient risk stratification necessary for appropriately targeted screening.
The most thoroughly studied HPV test for use in cervical cancer screening or triage is Hybrid Capture 2. In the absence of adequate RCT data, those planning substitution of other types of HPV testing in cervical cancer screening programs based on these studies should carefully consider clinical test performance when directly compared with Hybrid Capture 2, evidence of test–retest and interlaboratory test reliability, other quality control issues, and cost (66).
The main limitations of our review and of this body of evidence follow. Our search may have missed smaller European studies published in national journals only. Most studies used colposcopy or biopsy as the reference standard, neither of which is 100% sensitive for detecting preinvasive disease. Trials that do not have full or complete ascertainment for undetected disease can inaccurately reflect sensitivity or true disease detection. Longer follow-up after multiple screening rounds, ideally combined with methods of creating equal probability of cervical lesion detection in all participants, gives a truer picture of the relative effect of different screening strategies on disease (67). Linking screening trial results with outside data, including registries, can help overcome possible ascertainment biases that are particularly likely to distort screening comparisons with relatively short-term results. Most trials did not report results by using an intention-to-screen analysis, in which all women in the randomized group are in the denominator of all calculations.
Finally, the data from trials involving HPV testing are reported in many publications, with updated results being published over time. Despite our efforts to search for additional data from studies with incomplete reporting, some missing data may have been available through more extensive author requests (which were beyond our resource capabilities) or could soon be published. Thus, our findings will probably need rapid updating as more data from completed and ongoing trials become available. In addition, none of the trials included the effect of HPV vaccines on screening, which will be critical in the future.
Liquid-based cytology and conventional cytology perform interchangeably in terms of newer screening technologies for cervical cancer screening. Compared with cytology, HPV testing offers a tradeoff between increased sensitivity and decreased specificity. Because cervical cancer screening is repeated over time, results from RCTs should inform a proposed change in screening approach. Substituting a strategy of primary HPV screening (with or without cytology triage) for one of cytology alone in women aged 30 years and older is appealing, but important details remain unclear, including how much early disease detection is improved, whether such a strategy would have a beneficial effect on invasive cervical cancer, and what other effects it would have in terms of burden and diagnosis- and treatment-related harms.