Main findings
The nationwide implementation of primary high-risk HPV DNA screening in the Netherlands has been successful, with the programme now fully implemented and results generally as expected, apart from a lower than anticipated participation rate. In the first year, we observed a participation rate of 61%, which was lower than observed in the previous cytology-based programme (64%). Screen positivity was higher in the hrHPV-based programme. The cytology programme recommended observation of ASC-US/LSIL results, while the hrHPV-based programme recommended colposcopic referral for hrHPV-positive, ASC-US/LSIL results. As expected, this increased both the number of colposcopic referrals and CIN2+ lesions detected.
Factors influencing participation rates
The introduction of self-sampling had been expected to increase participation, as a previous Dutch study (PROHTECT) found that screening non-attenders who were offered self-sampling were more likely to be screened than non-attenders [22]. While 8% of screened women used self-sampling, this did not increase overall participation, suggesting that switching is occurring. Information about switching was not publicly reported in the 2017 official monitoring report [23], and further research is needed into the characteristics of women who choose for self-sampling to provide reliable estimates of this indicator. One important difference between PROHTECT and the real-world implementation was that women needed to opt in to self-sampling in the screening programme. Secondly, the 4-month waiting period for the self-sampling kit may have delayed uptake of screening amongst women who opted in. The self-sampling kit may be used by women who find it more convenient than attending the GP; one of the main reasons identified in a Dutch study for using a self-sampling kit [24]. Finally, although self-sampling is generally acceptable to women [12], 23% of self-sampling kits requested by the 2017 cohort have not yet been returned (as of December 2018; personal communication, RIVM, 21 December 2018). Although the return of these kits would not have a large effect on overall participation, the reasons for not returning them should be further investigated.
Organisational factors, such as the phased roll-out of the new programme and changes in the invitation process, may also have resulted in lower participation. Due to the phased roll-out of the new programme over the first quarter of 2017, women had less time to take up their screening invitations compared with the cytology-based programme, although we still observed a lower participation rate when calculating it based on 18 months of data. If the phased implementation is the cause of lower participation, we would expect participation to increase in the coming months. In the cytology-based programme, GP practices could invite patients for screening, rather than women receiving an invitation from the regional screening organisation. Women who received invitations sent from GP practices were more likely to participate in the cytology-based programme than women who received invitations from screening organisations [25]. Discontinuing the involvement of GP practices in the invitation and reminder process may have led to a decline in participation, as invitations are now sent from organisations that may be unfamiliar to women; this needs further investigation.
Comparison with other studies
The hrHPV positivity rate was higher than anticipated at 9.1%, as a previous population-based Dutch study (DuSC) found a hrHPV positivity rate of 8% amongst women of screening age [26]. This difference may be explained by differences in sociodemographic characteristics of women participating in the programme overall and the women included in DuSC. It could also be that there has been an increase in the incidence of hrHPV infections over time. The higher than expected hrHPV positivity rate may explain differences between the estimated referral rate of 3.4% (based on modelling) [6] and the observed referral rate of 3.9%.We found 48.2% CIN2+ detection in all women with histologically confirmed diagnosis, which was higher than the rate predicted by modelling (45%), which may be due to differences in the assumed test characteristics and the real-world performance of the hrHPV test [6].
One surprising finding was that hrHPV positivity was lower in self-samples than in the clinician-collected samples, contrary to previous Dutch studies. One population-based study found higher hrHPV positivity in self-samples than in clinician-collected samples [12], and one randomised non-inferiority trial (IMPROVE) found equivalent hrHPV positivity between the two test types, although IMPROVE used a different clinician-collected test than is used in the screening programme [27]. Despite this, we found higher CIN2+ detection in self-sampling than in clinician-collected sampling. This may indicate that the self-sampling test has a higher CIN2+ specificity than the clinician-collected test, in contrast to results from IMPROVE, which reported CIN2+ specificity of the self-test was non-inferior (relative accuracy of 1.00) [27]. Further analysis of the self-sampling kit within the screening programme is needed, controlling for background risk and population factors.
Triage of hrHPV-positive women
A higher CIN2+ detection rate was found in the hrHPV programme than in the cytology-based programme. This was expected based on the results of four large randomised trials of HPV screening [1]. However, in the new hrHPV screening programme, more referrals per screening round were needed to detect one CIN2+ lesion compared with cytology-based screening, mainly due to an increase in the number of referrals amongst women with ASC-US/LSIL cytology. This increase potentially leads to more harms for women, including anxiety for women unnecessarily referred [19] or potential overtreatment of low-grade lesions. Therefore, optimising triage to reduce unnecessary referrals should be a priority. Different triage strategies for hrHPV-positive screens have been proposed, including (but not limited to) p16/Ki67 dual staining, hrHPV genotyping, methylation, HPV E6 protein assays or combinations of these strategies [28]. Risk-based management could also be explored, in which risk factors (such as a woman’s screening history) are taken into account when triaging hrHPV-positive, ASC-US primary screens [29]. The performance of additional triage tests in the Dutch setting as well as the feasibility of implementation, any impacts on programme cost-effectiveness and the balance of harms versus benefits of the screening programme need to be considered prior to changing the triage algorithm. The harms-benefits ratio of the old cytology-based programme was considered acceptable in the Netherlands, and while in one round of screening the hrHPV-based screening programme had a more unfavourable balance, reducing the number of total screening rounds in the hrHPV-based programme (from seven to five for many women) will result in similar overall lifetime harms-benefits ratio to that of the cytology-based programme.
International comparisons
In several countries, hrHPV-based screening has been implemented, but published results are only available from Italy and Turkey. In Italy, HPV-based screening was implemented in 2012 in 19 screening programmes across ten regions. The direct referral rate from the Italian programme was comparable with the Dutch programme at 2.9% [10]. In 2014, primary HPV screening was implemented in Turkey; however, a direct comparison of results is difficult due to a low participation rate (36.5%) and incomplete histological follow-up data [9]. Neither study compared hrHPV-based screening with cytology-based screening. In general, the quality of a cytology-based programme influences such a comparison. In the Netherlands, the quality of the cytology-based programme was consistently high, with low rates of unsatisfactory smears and a high positive predictive value for CIN2+ lesions compared with other European countries [30]. In a country with a less highly performing cytology programme, the incremental effects of HPV-based screening versus cytology-based screening would be different.
Future implications for hrHPV screening in partly vaccinated cohorts
Given the increased sensitivity of hrHPV testing for CIN2+ lesions, detection rates are expected to be higher in the first round, as both prevalent and incident lesions are detected. As the programme reaches a steady state, and fewer prevalent lesions are detected, we expect that detection of CIN3+ lesions will decrease, as seen in the POBASCAM trial [31]. Therefore, it will be necessary to compare results from the first and subsequent screening rounds. In the Netherlands, hrHPV vaccination was offered in a catch-up programme to girls aged 13 to 16 years in 2009, meaning the first cohort of partly vaccinated women will be eligible for screening in 2023. This may necessitate changes to the programme, due to an anticipated reduction in HPV16/18 infections. Modelling has shown that with herd immunity levels greater than 50%, a reduction in the number of screening rounds may need to be considered to maintain programme cost-effectiveness in the Netherlands [32]. Finally, for full evaluation of the new screening programme, calculation of interval cancer incidence is essential to approximate the sensitivity of one screening round. Women are at highest risk of an interval cancer diagnosis 4 to 6 years after a negative screen [33], as the screening interval is 5 years. As such, the first opportunity for comparison of this indicator will come 5 years after the implementation of hrHPV-based screening.
Strengths and limitations of this study
This is the first study to report the results of the nationwide implementation of a hrHPV-based screening using prospectively collected cyto- and histopathological data. We have been able to compare this reliably with the previous cytology-based programme due to the nationwide coverage of PALGA. The large number of screens included in our study has allowed us to make statistically robust comparisons between indicators of the two programmes. Our study has some limitations. The follow-up time included in our study was shorter for the hrHPV-based programme than for the cytology-based programme, as the hrHPV-based programme was implemented more recently. We are unable to analyse characteristics of non-attenders to the programme, as characteristics of these women are not captured by PALGA. We are also unable to differentiate loss to follow-up after referral for colposcopy from cases where women attended colposcopy, but no cytology or histological diagnostic test was performed. This information is unavailable for both the hrHPV-based programme and the cytology-based programme. As such, we cannot investigate whether adherence to referral advice has changed over time. Furthermore, compliance to referral, used to differentiate cytology only and no follow-up with cytology or histology in Table 1, may have been underestimated for hrHPV screening due to the shorter follow-up time for the hrHPV-based programme; however, without data on colposcopies, the extent of this underestimation is unknown. The identifier used in PALGA to link records is non-unique (based on the first eight letters of a woman’s surname and her date of birth). This means that records from multiple women could be linked to one identifier (called an administrative fusion). It is unlikely that there is a difference in the number of administrative fusions between the two programmes, and therefore, we expect that this has not influenced our results. Finally, because the cytology-based programme recommended observation of ASC-US/LSIL results, while the hrHPV-based programme recommended colposcopic referral for hrHPV-positive, ASC-US/LSIL results, distinguishing the relative impact of the hrHPV test itself versus the lower threshold for referral on both unnecessary testing and CIN2+ detection is difficult.