Background
Cervical cancer (CxCa) is one of the most common cancers in women, with nearly 500,000 new cases being diagnosed each year worldwide [
1]. Its prevalence represents relevant costs for patients, their families and countries. Insiga et al. [
2] reported that 75% of CxCa-diagnosed women died before sixty and 25% before turning forty. Researchers also estimated that 29% of them would be professionally active in the year they died and, based on their salaries’ projections, represented a revenue loss of 1.3 Billion USD, a figure superior to the direct costs associated with CxCa in the US. The introduction of cervical cytology as a screening method in the mid-twentieth century contributed to a decrease in the rate of CxCa, but its low sensitivity for CIN2+ requires a frequent repetition of the secreening process [
3]. As a result, there is need for more efficient and cost-effective screening methods [
4]. Therefore, actual knowledge lead to the definition of new strategies of prevention and practice management that include Human Papillomavirus (HPV) testing and prophylactic vaccination. Portugal started the first organized cervical cancer screening program in the Centre Region in the late 90’s, extending it to more than half of the country nowadays. The HPV quadrivalent vaccine was introduced in the national vaccination program in 2008 extending the coverage to almost 90% of women. Regarding screening, the Portuguese Society of Gynaecology consensus document considers Pap Cytology with ASCUS HPV triage every three years as adequate. Nevertheless, also point out the primary high-risk HPV (hrHPV) testing with cytology triage every five years as the recommended screening algorithm [
5], based on the superior sensitivity of the HPV assay, validated by prospective clinical trials. Ronco G et al. (2014) [
3] point out that HPV-based screening provides 60–70% greater protection against invasive cervical carcinomas when compared to Pap cytology. Following this recommendation a law decret was published in 2017 confirming HPV as the primary screening test with 16/18 genotyping as a triage test for direct colposcopy and Pap cytology as a triage for other 12 hrHPV types [
6], which is determining a change in the screening algorithm. Despite this fact, only some of the organized screening programs implemented the project. In this context, a comparison of the clinical and budget impact of different screening strategies will help to clarify the health care gains obtained with the adoption of the new screening algorithm. This is the first budget impact analysis to date on this subject evaluating the best scenarios for Portugal.
Discussion
Overall, the output of the decision-tree and Markov model suggest the replacement of the current cytology-based screening with any of the two primary HPV screening strategies. It improves the detection of CIN2–3 and CxCa, displays better clinical outcomes and creates substantial cost-savings for the Portuguese healthcare system. The results are in line with similar health economic outcome research (HECON) studies [
36‐
39] and support the current clinical consensus on the move towards molecular screening for cervical cancer not just in Portugal but also other European countries. Nevertheless, as this study is based on a computer-based model cohort, a randomized clinical trial or real-world studies with randomized samples and appropriate methodologies are needed as further research to generalize these results to the population.
The major goal of cervical cancer screening is to detect its precursor lesions and treat them before they become invasive. According to the findings of the meta-analysis of 4 European randomized trials (Swedescreen (Sweden) [
40], POBASCAM (Netherlands) [
41], ARTISTIC (England) [
42] and NTCC (Italy) [
43], reinforced by the recent findings of the COMPASS (Australia) [
44] and FOCAL (Canada) [
45]), HPV-based screening provides 60–70% greater protection against invasive cervical carcinomas compared with Pap cytology and that screening intervals may be extended to at least five years [
3]. In the US, the Kaiser Permanente Northern California (analysing data from more than 330,000 women) [
46] and the ATHENA trial (recruiting around 47,000 women) [
28] confirmed the vantage of the HPV test over pap cytology as a primary screening method.
In this study, HPV 16/18 genotyping with cytology triage (comparator 2) provides better clinical impacts, while lowering the costs. Therefore, it is the most cost-effective CxCa screening strategy in the Portuguese context. Comparator 2 also reduces annual CxCa incidence and mortality and reduces the total annual screening costs in 24.0%, as well as the cost of detecting a CIN2+ in 19,2%. The screening performance of HPV (pooled) test with cytology triage (comparator 1) is lower when compared to comparator 2 in its detection rates but has similar costs. The improved clinical outcomes obtained through this HECON modelling result from an improved screening performance of both HPV strategies and an earlier detection of CIN2–3 and CxCa. On the other hand, the substantial cost savings are caused by the increased length of the routine screening intervals, allowed by the enhanced sensitivity and negative predictive value of HPV based screening. The performance of Pap cytology is low in the model, but consistent with data from an international meta-analysis showing a detection rate of 53,0% [
47].
Concerning the triage of HPV positive women, the authors of this study acknowledge the fact that pap cytology performance as a triage marker is affected by an expected increase in sensitivity and a decrease in specificity [
48,
49]. This means that the “real-life” sensitivity of the HPV based algorithms can be higher and specificity lower, positively impacting clinical results (CIN2+ detection) and increasing costs. To clarify this subject, a sensitivity analysis compared results from other studies that also addressed triage strategies for HPV positive women, such as POBASCAM [
32], PavDag [
33] and also from the ATHENA three years follow up results [
34]. The conclusion was that the cost difference towards cytology-based screening remained higher, consolidating the main findings on the cost savings created by the change towards HPV test-based screening.
The model also projects a difference in clinical outcomes between comparator 1 and 2 strategies: 5.9% in sensitivity for the detection of CxCa. The model results suggest that if the individual tests are used in a certain strategy or algorithm then the number of detected cases may differ. Although such differences were never reported from clinical trials, they can be partially explained by the fact that in this simulation, in comparator 1 all hrHPV positive / NILM cases (including HPV16 and 18 positives) are deferred and that 15.0% of these will actually miss follow-up.
This study has certain limitations. As any modelling study, the results are influenced by the input parameters and structure of the model. The performances of LBC and the HPV test (pooled and 16/18 genotyping) are based on data from a trial in the United States, not on Portuguese data. The clinical results are sensitive to changes in prevalence of HPV, HPV genotypes 16/18, CIN2 and CIN3. Also, the referral rates to colposcopy were determined in this model by the prevalence of the infection by the 14 hrHPV types in the Portuguese epidemiological study and these may differ if prevalence changes, as it is expectable in a vaccinated population. Another model limitation is that performance data reflect the baseline test (test performance among the general population) and that the same test in a re-testing situation among a previously hrHPV+ population may be different. The authors acknowledge that the model should allow two test performance inputs - one at baseline and another for re-test situation among previously hrHPV+ women. While HPV vaccination programs impact the cost of screening programs as shown in recently published real world data [
50], this model does not take that in consideration which may be considered also as a limitation.
The cost results are sensitive to changes in prices of cytology, HPV test and screening consultation. In the absence of a DRG reflecting the use of HPV as a screening test, the price of HPV tests was assumed the same as the cytology DRG. The future DRG price for an HPV test may be different. In this case, the authors tried to simulate the reality for Portugal as much as possible. In our country, there is no screening program at national level, and opportunistic and organized programs occur regionally with different levels of implementation. In the current setting, DRG prices are currently used as a reference for pricing both HPV and Cytology and the volume of tests would not allow for a cost reduction. A sensitive analysis varying the price of HPV test to either twice or half of that of cytology, shows that HPV based screening algorithms would remain cost-saving independently of the chosen scenario. Furthermore, the cost of screening can be lower if more suppliers compete in fair tenders. The guideline based on reproducibility and equivalent accuracy defined by Arbin et al. [
51] is a milestone in HPV-based cervical cancer screening. There are currently 8 HPV DNA assays fully matching these criteria, which can be recommended for HPV-based cervical cancer screening using clinician-collected cervical samples, half of which offer partial, extended or full genotyping. This means the offer is large and competitive, making public tenders able to implement either of the Comparator strategies. This is important not just for Portugal, but also for other countries. The time-horizon of this modelling study is limited to two screening cycles. Policymakers would like to understand to which level CxCa incidence and mortality decrease, after several screening cycles.