Ellen M. G. Olthof

Under which realistic circumstances is hrHPV self‐sampling increasing cervical screening effectiveness in a partly vaccinated population? A modelling study

Abstract High‐risk Human Papillomavirus self‐sampling can increase attendance rates for screening. However, observed lower sensitivity and loss to follow‐up of self‐sampling could reduce programme effectiveness when attenders of clinician‐collected sampling switch to self‐sampling. We determined the tipping point for effectiveness (based on life years gained [LYG]) of self‐sampling and the consequences for cost‐effectiveness, taking into account waste by comparing full opt‐out (no waste) to no opt‐out (waste from unused self‐sampling kits). We used the STDSIM‐MISCAN‐Cervix microsimulation model to simulate a population of Dutch women born in 2000 (50% vaccinated [sensitivity analysis: 0–100%], 70% screening attendance [sensitivity analysis: 60–80%]). Self‐sampling deployment strategies (e.g., direct‐mail) were varied by the percentage of original attenders switching to self‐sampling and the percentage of new attendance from non‐attenders. Main outcome measures were LYG and cost‐effectiveness (cost per quality adjusted [QA] LY gained) compared to the current programme. We found that if self‐sampling does not reach non‐attenders, life years cannot be gained. When reaching 10% or 30% of non‐attenders, the tipping point lies at ≤40% and ≤100% switchers to maintain effectiveness, respectively (+4 LYG, +10 LYG). Scenarios were cost‐effective (<€50,000/QALY gained) if at least 10% of non‐attenders were reached. Full opt‐out improved cost‐effectiveness substantially. So, in a partly vaccinated population, self‐sampling deployment strategies need to reach at least 10% of non‐attenders to maintain programme effectiveness and cost‐effectiveness. A well‐functioning opt‐out system further improves cost‐effectiveness by preventing waste.

Improving Efficiency in Dutch Cervical Screening by Genotyping: An Analysis of Real World Program Data

ABSTRACT High‐risk human papillomavirus (hrHPV)‐genotype specific risk stratification may improve cervical screening efficiency. This study evaluates the risks of cervical intraepithelial neoplasia (CIN), cancer and unnecessary referrals by hrHPV‐genotype in cytology‐positive (ASCUS+) women, using data from the Dutch population‐based cervical screening program. Data from hrHPV+/ASCUS+ women screened between January 2017 and March 2018 were analyzed using the Dutch Screening and Pathology databases. Risks for CIN2+/3+, cancer, and unnecessary referral (i.e., without CIN2+) were evaluated by hrHPV‐genotype (HPV16, HPV18, hrHPV‐other (i.e., non‐16/18 hrHPV), or mixed HPV16/18) using logistic regression, adjusted for age, laboratory (as proxy for region), sampling method (self‐ vs. clinician sampling), and stratified by age (< 50/≥ 50 years). HPV16+ women had 3.7 (CI: 3.42–3.95) and 4.6 (CI: 4.24–4.99) times higher risks of CIN2+ and CIN3+, respectively, compared to hrHPV‐other. HPV18+ women had 1.6 (CI: 1.43–1.79) and 1.9 (CI: 1.68–2.18) times higher risks. The cervical cancer risk was tenfold higher for both HPV16 (OR: 9.85, CI: 6.50–14.95) and HPV18 (OR: 10.27, CI: 6.33–16.68). Women with HPV16 had 70% and HPV18 40% lower risks of unnecessary referral, compared to hrHPV‐other. All risk differences between HPV16 or HPV18 and hrHPV‐other were statistically significant in both age groups (< 50 and ≥ 50 years). Given the significantly higher risk of CIN2+/3+ and cancer associated with HPV16 and HPV18 and the reduced likelihood of unnecessary referrals compared to hrHPV other, these findings support the use of genotype‐based colposcopy referrals in cervical screening to enhance screening efficiency.

Cost‐Effectiveness of Computer‐Assisted Cytology in a Primary hrHPV‐Based Cervical Cancer Screening Programme

ABSTRACTBackgroundComputer‐assisted screening (CAS) shows equal performance compared to manual screening, although results are heterogeneous. Furthermore, using CAS may save costs through a potentially increased screening productivity of technicians, therefore also offering a solution for temporary and structural capacity shortage. We evaluated the circumstances under which CAS will be cost‐effective compared to manual cytology triage in a primary HPV‐based cervical screening programme.MethodsMicrosimulation model MISCAN‐Cervix was used to evaluate 198 different CAS scenarios with varying probabilities to detect cervical intraepithelial neoplasia grade 1 (CIN1) and CIN3 and cost reductions per test, compared to manual cytology triage. Cost‐effectiveness was evaluated by costs per (quality‐adjusted) life year ((QA)LY) gained.ResultsCAS will be cost‐effective in all scenarios, except for the following combinations: (1) no cost reduction and an increased probability of detecting CIN1, (2) a cost reduction of €2 per test and an increased probability of detecting CIN1 from 4% onwards or (3) a cost reduction of €4 per test and an increased probability of detecting CIN1 from 6% onwards, compared to manual cytology triage. All CAS scenarios with any reduction in the probability of detecting CIN1 (i.e., increased CIN2+ specificity), or a reduction in costs from €6 per test onwards suggested a more cost‐effective strategy compared to manual cytology triage.ConclusionAs we based our analysis on a realistic range in costs and test performance, the implementation of CAS is likely to be cost‐effective. Our results can be used as a guideline to advise when to choose CAS instead of manual cytology triage.

The impact of loss to follow‐up in the Dutch organised HPV‐based cervical cancer screening programme

AbstractLoss to follow‐up (LTFU) within cervical screening programmes can result in missed clinically relevant lesions, potentially reducing programme effectiveness. To examine the health impact of losing women during the screening process, we determined the proportion of women LTFU per step of the Dutch hrHPV‐based screening programme. We then determined the probability of being LTFU by age, screening history and sampling method (self‐ or clinician‐sampled) using logistic regression analysis. Finally, we estimated the number of missed CIN2+/3+ lesions per LTFU moment by using the CIN‐risk in women compliant with follow‐up. Data from the Dutch nationwide pathology databank (Palga) was used. Women eligible for screening in 2017 and 2018 were included (N = 840,428). For clinician collected (CC) samples, the highest proportion LTFU was found following ‘referral advice for colposcopy’ (5.5% after indirect referral; 3.8% after direct referral). For self‐sampling, the highest proportions LTFU were found following the advice for repeat cytology (13.6%) and after referral advice for colposcopy (8.2% after indirect referral; 4.3% after direct referral). Self‐sampling users and women with no screening history had a higher LTFU‐risk (OR: 3.87, CI: 3.55–4.23; OR: 1.39, CI: 1.20–1.61) compared to women that used CC sampling and women that have been screened before, respectively. Of all women LTFU in 2017/18, the total number of potentially missed CIN2+ was 844 (21% of women LTFU). Most lesions were missed after ‘direct referral for colposcopy’ (N = 462, 11.5% of women LTFU). So, this indicates a gap between the screening programme and clinical care which requires further attention, by improving monitoring of patients after referral.

Resilience of the Dutch HPV-based cervical screening programme during the COVID-19 pandemic

Organisation of a screening programme influences programme resilience to a disruption as COVID-19. Due to COVID-19, the Dutch human papillomavirus-based cervical screening programme was temporarily suspended. Afterwards, multiple measures have been taken to catch-up participation. This study aimed to investigate programme resilience by examining the effect of COVID-19 and programme measures taken on participation in cervical screening. Observational cohort study. Data from the national screening registry and Dutch nationwide pathology databank (Palga) were used on invitations and follow-up in 2018/2019 (pre-COVID) and 2020 (COVID). Sending invitations, reminders and self-sampling kits were suspended from March to July 2020. Main outcome measures include distribution of participant characteristics (age, region and screening history), participation rates by age and region, time between invitation and participation (i.e. response time) and self-sampling use per month. Participation rate was significantly lower in 2020 (49.8%) compared to 2018/19 (56.8%, P < 0.001), in all ages and regions. Compared to 2018/19, participation rates decreased most in women invited from January to March 2020 (-6.7%, -9.1% and -10.4%, respectively). From August, participation rates started to recover (difference between -0.8% and -2.7%). Median response time was longer in February and March (2020: 143 and 173 days; 2018/19: 53 and 55 days) and comparable from July onwards (median difference 0-6 days). Self-sampling use was higher in 2020 (16.3%) compared to 2018/19 (7.6%). The pandemic impacted participation rates in the Dutch cervical screening programme, especially of women invited before the programme pause. Implementation of self-sampling in national cervical screening programmes could increase participation rates and could serve as an alternative screening method in times of exceptional health care circumstances, such as a pandemic. Due to the well-organised programme and measures taken to catch-up participation, the impact of COVID-19 on the screening programme remained small.