Strengths and limitations
A major strength of this study is the randomized design that contributes to making the detected differences between the tested interventions and usual care trustworthy. Another key strength is the effectiveness approach and the fact that the study was embedded in an ongoing routine cervical cancer screening program, which from an implementation point of view provides a reliable and representative estimate of the expected participation rates that could be obtained if the possibility of self-sampling together with the second reminder were to become an option. Furthermore, we minimized the risk of information bias and selection problems, firstly, by using data from the DPDB which has highly valid records on all pathology specimens for identifying outcomes and, secondly, by using a population-based design. Additionally, we used a combination of a clinically validated self-sample device and a clinically validated PCR-based HPV DNA test, resulting in a low proportion of invalid self-samples (< 0.5%).
This trial was not designed to estimate differences in the proportions of CIN2+ lesions between the intervention groups; thus, observed differences should be interpreted with caution.
Comparison with other studies
The participation rate in the directly mailed group (38.0%) was higher than in other comparable Dutch trials that achieved participation rates of 26.6 to 30.8% at 12 months [
38,
39]. This may be explained by the fact that the Dutch women offered self-sampling were not informed about the possibility of having regular cytology screening at their GP. In the opt-in group, the participation rate (30.9%) was also higher than in previous opt-in trials that report participation rates of 8.7 to 24.5% [
19,
20]. This may be due to different definitions of non-participants, differences in the time of participation assessment (range 3 to 12 months), and timely opt-in procedures which made it easier for women to participate. Recently, a Danish opt-in study that targeted women being unscreened for ≥4–6 years achieved a similar participation rate of 30.0%, assessed in a range of 7 to 18 months [
40].
Even though we found participation rates in the intervention groups to be slightly lower among the oldest women, the possibility of self-sampling increased participation in all age groups. This suggests that self-sampling is a suitable strategy across different age groups. It is expected that implementation of self-sampling in the Central Denmark Region screening program would increase the overall participation rate among invited women aged 30–64 years by an extra 2% or 5% for the opt-in or direct mail strategy, respectively. The latter figure is similar to the extra 5.2% achieved in a previous Dutch trial that used direct mailing of the self-sampling kit together with a second reminder [
38].
Almost one third (32.3%) of the women ordering a self-sampling kit in the opt-in group did not return it. This suggests that intention to be screened was present, but something made the women fail to return the sample. We have no clear explanation for this, but other opt-in studies have reported a similar tendency (range 11% to 39%) [
19,
40]. Thus, both mailing self-sampling kits directly and the opt-in procedure result in loss of kits. This should be taken into consideration when choosing the self-sampling invitation strategy and the self-sampling device. Using the webpage and sending a text message were more commonly used ways of ordering the kit than phone and e-mail. This result is in line with the Danish opt-in study, which reported that ordering via a webpage was considered far more acceptable than ordering by phone or e-mail (37% vs 1% vs < 1%, respectively) [
40].
Among women receiving the self-sampling offer, we also found that 18.6% of women in the directly mailed group still chose to approach a GP for a cervical cytology sample, while this was the case for even more women (22.6%) in the opt-in group. These findings suggest that self-sampling may be most effective if it is combined with other strategies, i.e. not used as the only option.
The effectiveness of self-sampling depends, among other things, on its capacity to recruit hard-to-reach women at increased risk of developing cervical cancer. Our results show that the direct mailing strategy was superior to the opt-in strategy and the standard second reminder in terms of higher participation among un- or underscreened women, but also among women who had not participated in previous screening rounds. This finding is supported by other trials that used the direct mail invitation strategy [
17,
18]; they report that underscreened women were more likely to participate when offered self-sampling (range 10 to 39%) than when sent an invitation/reminder for regular cytology screening (range 4.5 to 9.0%). Thus, self-sampling may have the potential to reduce the earlier documented social inequalities in cervical cancer screening participation [
41]. An up-coming registry study based on data from this trial will further explore this issue. These perspectives together with the cost-effectiveness of self-sampling should also be taken into account before planning a general rollout of self-sampling in the routine screening program.
An efficient self-sampling strategy also depends on a high level of compliance with follow-up among HPV-positive self-samplers. Compliance with cervical cytology triage at the GP after a HPV positive self-sample was high in this trial (90.7%); and when the relatively short follow-up measure was extended, the “long-term” follow-up was 95.8% within 180 days. This was higher than was found in previous Dutch trials that recorded follow-up of 89% to 90% at 18 months with a comparable triage protocol [
38,
39]. Most importantly, our results show that high compliance with follow-up could be achieved in a real-world screening setting without an intensive follow-up protocol as used in other trials [
19,
20]. HrHPV infections were detected in 13.0% of the self-sampling participants. This prevalence was higher than was found in an Australian trial (8.5%) that targeted never- or underscreened women aged 30–69 years using the same HPV test [
42]. The hrHPV prevalence among self-samplers (13.0%) was only slightly lower than, but not significantly different from, the 16.2% observed in a HPV screening study that included Danish women aged 30–65 years undergoing regular cytology screening using the Cobas® 4800 HPV test [
43].
Although our trial was not scaled to evaluate the effect of the self-sampling initiatives on the detection of CIN2+ lesions, we saw a clear tendency: the proportion of CIN2+ lesions per 1000 invited women was higher in the directly mailed and opt-in group than in the control group. This finding might be interpreted as an early indicator for the expected impact on cervical cancer prevention if self-sampling was to be introduced. It should be noted that there seemed to be a difference in CIN2+ detection between the directly mailed and opt-in group, even though four of the five women who were not followed up within 180 days were in the directly mailed group.
The higher detection of CIN2+ among self-sampling participants than among regular cytology participants in the control group, also shown in previous trials [
44,
45], may be explained by the fact that primary hrHPV testing is more sensitive than cytology testing for detecting CIN2+ [
46]. Assuming that the background risk for CIN2+ is increased among un-or underscreened women [
38], the improved coverage of these women when self-sampling was offered could be another explanation for the increased CIN2+ detection found among self-samplers.