Epidemiology

This real-world study group was selected to be representative of the local public sector cervical cancer screening population in terms of age range, screening history, and the HIV-treatment and disease control of HPW. The high cervical disease and hrHPV prevalence is typical of the South African public health service population; underlying causes and implications were discussed before [26]. Similar high disease prevalence is reported in several other subpopulations in the region and in sub-Saharan Africa [14, 30,31,32,33,34]. Contrary to this high disease burden, comparative study populations from the Global North are typically heavily pre-screened and many study groups are enriched with referral populations to address a low disease prevalence [10,11,12, 20, 35, 36].

In addition to true epidemiological differences, the majority of screening studies omit histology in screen-negatives and report disease prevalence based on the assumption that negative screening tests are true negatives [10,11,12, 23]. In this project, about half of screen-negative women has biopsy data, with enough unexpected positive histology results to influence the calculated test performance. Further investigation of the unexpected high number of CIN2 + histology is underway in the form of histology review, immunohistochemistry, and extended genotyping to understand the role of non-high-risk types, and correlation with methylation markers.

Screening strategy criteria

As far as we could establish universal criteria to evaluate, compare and interpret test performance and select screening strategies do not exist – usually the “best” option is selected per population [37]. The ideal balance of sensitivity and specificity depends on the management algorithm, available health infrastructure, screening frequency, tolerance for missed cases and overtreatment, etc. In South Africa, and similar societies desperate to address high disease burdens, the best possible sensitivity is needed due to long screening intervals, but in test-and-treat programmes high test specificity is also needed to limit overtreatment.

In addition to having excellent test performance, the strategy should be as simple as possible; the simplest is a dual-result approach where positives are treated without further testing and negatives are referred back to the next screening round. If that is not possible, the size of the intermediate-risk group should be as small as possible. Similarly, having a universally applicable screening strategy is preferable, or at least a universal screening test, with the option of different management algorithms [22]. We chose as a standard of acceptability the best universally achievable sensitivity (68% for CIN3 +), and to limit overtreatment to 15% (specificity of 85% for < CIN2 +).

Single test strategies

The sensitivity of cytology and visual inspection among HPW was significantly better than among HNW, as has also been reported by many others [14, 15, 17, 31]. In HPW these tests performed similar or even better than found in other studies from China, Sub-Saharan Africa, and South Africa, possibly partially explained by the fact that most HPW in this study were screened at a single facility and by a single, experienced, and well-trained nurse colposcopist [14, 15, 17]. In spite of our well-established cytology service, the sensitivity of cytology among HNW was poor, and only cytology at the low cut-off of “ASCUS + ” came near acceptability with a sensitivity of 59%. Visual inspection among HNW was done at several study centres by different investigators and performed poorly.

As expected, and found by other researchers, hrHPV(any) had the highest overall sensitivity, which was better among HPW than HNW, while the specificity was higher among HNW [17, 18, 38]. Immediate referral of women with HPV types 16/18 (and 45) is already widely recommended [12, 30, 39]. Here we showed that the hrHPV (any) had sufficient specificity for immediate treatment without further genotyping in HNW but not HPW. In this study we could, however, not duplicate the widely reported ultra-high sensitivity of HPV-tests in either of the cohorts [14, 40, 41]. Probable contributing factors include differences in epidemiology, dysplasia associated with low-risk HPV types, underdiagnosis in other studies due to lack of histology among screen-negatives and histological overdiagnosis in the current study. Further investigation of hrHPV-negative CIN2 + lesions are underway.

Unfortunately, among single test strategies with sensitivity near the selected standard of 68%, the best universally achievable specificity was 75% for < CIN2 (25% overtreatment). South Africa and other nations with large HIV-positive populations, are therefore forced to consider combination test strategies in order to achieve a specificity of 85% or to have separate screening tests for HPW vs. HNW.

WHO recommendations

When considering the recommendations made in the WHO 2021 guidelines [13], this study confirms recommendations nr. 1, 2 & 3 for general populations or HNW, namely that they should be screened with HPV-testing rather than VIA or cytology (poor sensitivity) and that a second test is not strictly needed before treatment (good specificity). The data presented here, however, showed high referral numbers based on hrHPV only, and that a second test would reduce referral and treatment burdens. On the other hand, recommendation nr.21, stating that HPW should be screened with an HPV-test in favour of cytology, is not supported by the current study. In the HIV-positive cohort cytology performed better than HPV-screening, as was also found by others [15, 17, 18]. Our data and calculations support the recommendation (nr. 22) that HPV-screening requires a second test when used in HPW (due to poor specificity).

Dual result combination strategies

In HPW all dual result strategies which combine cytology and hrHPV testing, showed acceptable and comparable test performance, with treatment groups from 28.1% to 34.7%. These performance and referral rates were similar to the best single test strategy cytology (LSIL), but better than hrHPV(all) alone.

In HNW the best dual result combination strategy was HPV-screening, followed by partial genotyping and cytology for non-16/18 HPV-positives (L) resulting in a treatment group of 13.9%. It was superior to the best single test strategy (hrHPV (all)) due to a smaller treatment group and better specificity. This strategy also had an excellent performance in HPW, which is in accordance with other reports [17, 21]. We found that this was the best dual result strategy to implement universally.

Triple results combination strategies

Among HPW all calculated triple result strategies reached both acceptability standards with similar test performance, but differences in treatment and follow-up group sizes. Our findings support the WHO recommendation (nr. 23) that partial genotyping, visual inspection, and cytology are all valid and similarly performing triage tests after hrHPV-screening (our strategies O2, R, P), but also validates cytology as primary test in a combination approach. Among HNW only combination test strategies starting with hrHPV (all) were acceptable and visual inspection, cytology, and partial genotyping with/without cytology performed well as secondary tests.

For universal implementation, the best triple result strategy appears to be the same combination as described as the preferred dual result combination strategy (L), but now with an intermediate-risk category (S). It results in the smallest intermediate risk group which consist of non-16/18 hrHPV positive women without cytologic abnormalities. The identification and management of the intermediate-risk group will increase the sensitivity by about 10 percentage points (from 73.6 to 82.1% in HPW; from 59.1 to 68.2% in HNW). Similar to the current study, others have also shown that treatment of HPV 16/18 combined with triage of other HPV types achieves an excellent balance of sensitivity and specificity among HPW [42]. Among HNW, follow-up of the intermediate risk group is recommended due to low sensitivity [43]. This strategy has high treatment numbers and is expected to have the fastest and biggest impact on invasive cancer prevalence, as treatment of all HPV16/18 should prevent at least 2/3 of cancer cases, and the addition of cytology for the other hrHPV positives will improve on this even without effective recall [44].

When calculating test performance, the allowance for an intermediate-risk group preserves the sensitivity of the initial test and increases specificity by refining the treatment group. Planning and implementing effective management for these intermediate-risk women will undoubtedly be challenging and expensive but is essential to realise the test performance quoted here. In the absence of effective recall, it is advisable to select a dual result strategy with a high specificity and smaller treatment groups, without the false promise of high sensitivity.

Single laboratory test

Using only one laboratory test, the best universal approach in this study was screening with hrHPV, followed by built-in partial genotyping as secondary test. HPV16/18 positives are referred to treatment, while non-16/18 positives are managed as an intermediate-risk group (R). We did not further investigate different management options for this latter group but presumed a zero lost-to-follow-up rate in calculating the sensitivity (increase from 37.7 to 82.1% in HPW and from 34.9% to 68.2% in HNW in comparison with strategy K). This strategy can be selected if it is expected that treatment facilities will be overwhelmed, or if capacity does not allow for the addition of cytology triage; treatment numbers will be half of that resulting from the triple result strategy which includes cytology (S) as described above.

Place for visual inspection

The only universally acceptable strategy employing visual inspection, was to use HPV as primary test and visual inspection as secondary (O2). Employing a triple result approach, this strategy resulted in treating 29.8 and 7.4% and following 18.6 and 16.1% of HPW and HNW respectively and can be a useful cheaper strategy using a cheaper non-discriminatory HPV test without the need for cytology. Alternatively, using a dual-result approach with the same two tests, acceptable test performance can be reached without the need for intermediate-risk groups. Using primary HPV-screening universally as primary test, it can be followed by two different strategies for the two HIV-subgroups: All hrHPW-positive HNW are directly referred for treatment (E), but positive HPW are called for VI, only double-positives are referred for treatment (H2). VILI should be preferred above VIA here due to superior sensitivity.

Relevance for screening policy

We previously discussed the high disease prevalence in our country which was confirmed here [26, 45]. While our study investigated different screening algorithms, it is acknowledged that the efficacy of secondary prevention will depend on improving on the current low treatment rates [46].