Screening for the prevention and early detection of cervical cancer: systematic reviews to inform an update to recommendations by the Canadian Task Force on Preventive Health Care.

PURPOSE

To inform updated recommendations by the Canadian Task Force on Preventive Health Care (Task Force) on screening in primary care for the prevention and early detection of cervical cancer in individuals with a cervix who are 15 years or older who have been sexually active and have no symptoms of cervical cancer. We systematically reviewed evidence from Very High Development Index countries of the following: screening effectiveness (focusing on ages to start and stop) and comparative effectiveness (strategies and intervals); comparative test accuracy (of comparisons without comparative effectiveness data); informed individuals' values and preferences, and effectiveness of interventions to improve screening rates among the under-/never-screened. Two existing systematic reviews provided evidence regarding adverse pregnancy outcomes associated with the conservative management of cervical intraepithelial neoplasia (CIN).

METHODS

We searched Medline, Embase, and Cochrane Central for effectiveness and accuracy questions; Medline, Scopus, and EconLit for patient preferences (to Sept/Oct 2023 for screening effects and preferences and March 2019 for accuracy and interventions to increase uptake) and reference lists of included studies and relevant systematic reviews. Two reviewers independently screened studies and assessed risk of bias. Most data were extracted by one reviewer with verification by another; outcome data for screening effectiveness were extracted in duplicate. We performed meta-analysis where suitable. Absolute effects were expressed as events among 10,000 individuals. Two reviewers appraised the certainty of evidence using GRADE. The Task Force chose outcomes of importance and determined thresholds for the certainty about comparative effectiveness. Due to limited data and certainty about the critical outcomes of mortality and incidence of CIN 3 and invasive cervical cancer (ICC) during follow-up, the Task Force used screen-detection of precursors (CIN 2 and CIN 3, respectively) as indirect measures when examining evidence on comparative effectiveness.

RESULTS

We included 112 studies across questions (14 on ages to start and stop screening, 10 on screening intervals, 17 on comparative effectiveness between different strategies, 10 on comparative accuracy, 23 on patient preferences, and 44 on interventions to increase uptake). When reviewing evidence to help inform ages to start and stop screening, only observational studies on cytology screening were identified. There was very low certainty evidence for the effects in individuals 20-24, 25-29, and 30-34 years of age to prevent invasive cervical cancer (ICC) or mortality (all-cause and cervical-cancer specific). For individuals 60-69 years of age, screening with cytology is probably (moderate certainty) associated with reduced ICC (≥ 9 fewer per 10,000) and cervical-cancer mortality (≥ 0.19 to 0.29 fewer) over 10-15 years of follow-up among those who had no screening, abnormal, or inadequate screening in their 50s. A reduction for these outcomes among those 60-69 years who were adequately screened during their 50s is less certain. For persons aged 70-79 years, screening with cytology reduced ICC with low certainty for those with no, abnormal, or inadequate screening histories. Evidence for ICC for those adequately screened and on mortality overall was very uncertain. Very low certainty evidence was found for reduction in ICC and cervical-cancer mortality for cytology screening every 3 years versus 3-to-5 years. Sixteen of 17 studies examining comparative effectiveness of different screening strategies provided results for only one round of screening, and results reflect this context. Across 10 groups of comparisons between screening strategies (e.g., initial testing with cytology vs. high-risk human papillomavirus [hrHPV], different triage and/or sampling methods, different populations), we are very uncertain about any differential impacts on all-cause and cervical-cancer mortality and on overdiagnosis. Across the incidence outcomes, there was often low certainty evidence of little-to-no difference or very low certainty evidence. (i) Compared with cytology alone, hrHPV alone may (low certainty) make little-to-no difference for 25-59-year olds for incidence of CIN 3 + (hrHPV detecting 30.2 more CIN 2 + per 10,000) but is probably associated with more (possibly ≥ 600 per 10,000) referrals for colposcopy and false positives for CIN 2 + and CIN 3 + for those aged 25-29 years. (ii) hrHPV with triage to cytology versus cytology alone may reduce incidence of ICC (e.g., 24.2 more CIN 3 + detections) for those aged 29-69 years, though when adding a recall phase (with additional testing beyond the initial triage), there are probably more harms for 25-29-year olds. (iii) hrHPV with cytology triage versus cytology with hrHPV triage (over one round) probably reduces incidence of ICC (46 and 31.5 more CIN 3 + detected with and without using recalls) without added harm for those aged 30-59 years. For those aged 25-29 years, using recalls may allow for a reduction in incidence of ICC (111.8 more CIN 3 + detected) and CIN3 + (271 more CIN 2 + cases detected) but also considerably increases harms (≥ 800 false positives). One round of hrHPV with cytology triage versus two rounds of cytology with hrHPV triage, both strategies including recall, may reduce incidence of CIN 3 + (31.2 fewer) for those 25-69 years, and probably leads to similar harms for those 30-69 years (with uncertainty for those 25-29 years). Because the cumulative detection rates during screening were similar between strategies, the lower rates of CIN 3 + at follow-up clearly indicate early detection of these lesions in the HPV-based arm. (iv) The effects on incidence of ICC are uncertain from adding partial genotyping to these hrHPV and cytology triage strategies; for those aged 30-59 years, there may be little-to-no difference in incidence of CIN 3 + (no difference in CIN 2 + detection) and is probably no difference in harms. From studies only enrolling those aged 30-59 years, (v) when comparing hrHPV with cytology triage of negative tests versus cytology with hrHPV triage, both arms having recall, low certainty evidence found reduced incidence of ICC (36.0 more CIN 3 + detected) from the hrHPV strategy arm and little-to-no difference between strategies for incidence of CIN 3 + , with moderate certainty evidence that the hrHPV strategy results in more referrals to colposcopies and false positives (about 600 per 10,000); (vi.a) there was low certainty of little-to-no impact on incidence of CIN 3 + (from CIN 2 + detection) and moderate certainty of little-to-no difference in false positives between hrHPV self-sampling with cytology triage compared with hrHPV clinician-sampling with cytology triage; (vii) evidence was low certainty for little-to-no difference in incidence of CIN 3 + (from CIN 2 + detection) and in false positives for hrHPV self- versus clinician-sampling, each with triage to repeat hrHPV testing at 3-6 months. From studies examining hrHPV self-sampling strategies versus those using cytology with or without hrHPV triage among populations who were non-responders or underscreened, (vi.b and viii-x) evidence was of very low certainty across all reported outcomes (detection of CIN 2 + and 3 + and false positives). From comparative accuracy studies, adding cytology triage to hrHPV testing alone (via self- or clinician-sampling), or replacing the hrHPV test with one allowing partial genotyping with or without cytology triage, reduces the number of false positives (high certainty; > 300 fewer per 10,000 screened). There is probably little-to-no difference in false positives between hrHPV with partial genotyping (types 16/18) and hrHPV with cytology triage. hrHPV with partial genotyping (types 16/18) versus cytology alone may increase specificity (reducing false positives) at the expense of sensitivity, though the number of missed cases may be very small (e.g., up to 9 fewer cases of CIN 3 + detected). There was higher sensitivity and specificity from hrHPV with partial genotyping (types 16/18) with triage to cytology of non-16/18 types versus cytology alone (moderate certainty). Cytology with hrHPV triage versus cytology alone may make little-to-no difference for sensitivity or specificity for CIN 3 + detection. Self- versus clinician-sampling for hrHPV alone probably achieves similar specificity though misses a few cases of CIN 2/3 (e.g., 13 to 27 missed cases). In relation to adverse pregnancy outcome from treatment, findings from two existing systematic reviews of observational studies found very low certainty evidence about whether conservative management of CIN 2/3 is associated with total miscarriage rates, second trimester miscarriage, preterm birth (≥ 37 weeks' gestation), low birth weight (< 2500 g), or cervical cerclage. Despite findings that would lead to very small increases in some outcomes among the entire screening population, the evidence was considered indirect for current practices that use a more cautionary approach to treatment particularly for CIN 2 in individuals prioritizing a reproductive future. Findings from studies on patient preferences via measurement of the disutility (i.e., impact on participant's quality of life, values ranging between 0 [no impact] and 1 [similar to death]) of having one of the outcomes indicated that ICC (disutility of 0.11) may be at least twice as important as CIN 2/3 (0.05), and that both cervical cancer and CIN 2/3 are probably much more important than false positives that did not cause any disutility. Other studies on patient preferences about cytology screening indicated, with low certainty, that a large majority of individuals eligible for and informed about screening may weigh the benefits as more important than the harms of screening using cytology, but think it is important to provide information on benefits and harms for decision-making. Findings from a single study suggested that some individuals < 25 years may have intentions to screen even if informed that screening does not reduce cancer diagnoses or deaths for their age group and leads to overdiagnosis. Five types of interventions to improve screening rates for under/never-screened individuals were reviewed. All were found with moderate or high certainty to improve screening rates: written contact (relative risk [RR] 1.50, 95% CI 1.22 to 1.84; 619 more per 10,000, 95% CI 273 to 1041; 16 trials, N = 138,880); personal contact (RR 1.50, 95% CI 1.07 to 2.11; 797 more, 95% CI 1116 to 1770; 7 trials, N = 17,034); composite interventions (usually mixture of written and personal contact; RR 1.73, 95% CI 1.33 to 2.27; 1351 more, 95% CI 610 to 2350; 8 trials, N = 17,738); universal mail-out of hrHPV self-sampling kit (RR 2.56, 95% CI 2.10 to 3.12; 1534 more, 95% CI 1082 to 2085; 22 trials, N = 211,031); and opt-in to receive a hrHPV self-sampling kit (RR 1.56, 95% CI 1.19 to 2.03; 727 more, 95% CI 247 to 1338; 11 trials, N = 71,433).

CONCLUSIONS

Screening for prevention or early detection of cervical cancer with cytology has been employed for decades and is probably effective for otherwise healthy persons with a cervix at least into their 60s. Whether to screen individuals younger than 35 years old using cytology was uncertain based on the need to rely on observational evidence without consistent reporting across age groups. Screening during one's 60s and 70s may have less effect for those adequately screened in their 50s. The effects of screening with cytology every 5 years versus 3 years are uncertain. The evidence provided very low certainty about any differential impacts between various screening strategies on mortality and overdiagnosis outcomes. Compared with one round of cytology alone or cytology with hrHPV triage, there was evidence of a small benefit from reducing ICC from one round of hrHPV with cytology triage though findings were less certain for those 60-69 years. We were not able to directly compare the effects between different recall approaches within the same overall comparison, but carefully choosing one that minimizes false positives may help improve the benefit-risk profile. Screening using hrHPV with triage to cytology every 4 years may reduce the incidence of CIN 3 + compared with cytology with hrHPV triage conducted every 2 years, though the effects compared with cytology alone were not examined. Furthermore, it is uncertain what the effects are on the incidence of CIN2 + , CIN3, or ICC from adding partial genotyping to the triage strategies. For those aged 30-59 years, moderate certainty evidence found little-to-no difference in false positives between hrHPV self-sampling with cytology triage compared with hrHPV clinician-sampling with cytology triage, and low certainty that there may be little-to-no impact on incidence of CIN 3 + . The comparative effectiveness studies did not examine all relevant comparisons, and thus, comparative accuracy data may help provide suggestions of possible alternative strategies with similar sensitivity and similar or higher specificity. Most of the studies on screening effects were undertaken in populations either in which HPV vaccination had not been implemented or carried out in a period when vaccination rates were low. For under- or never-screened individuals, the offer of self-sampling kits for hrHPV testing probably improves screening rates without missing an important number of CIN 2/3, but it is uncertain if findings apply in practice when triage to cytology is used because of the need for a clinic visit. ICC may be at least twice as detrimental on quality of life as is CIN 2/3, whereas a false positive result when using cytology alone does not have any impact; whether the lack of disutility of a false positive result applies to hrHPV testing is unknown. There was low certainty evidence that informed individuals eligible for screening think the benefits outweigh the harms from screening. Choices for screening strategies apart from cytology alone may result largely from contextual considerations such as access, acceptability, resources, and costs.