Economic Evaluations of Cervical Cancer Screening Methods: a Systematic Review


 Background: Cervical cancer screening is an important public health priority with the potential to improve the detection of pre-cancerous lesions in high-risk females for early intervention and disease prevention. Test performance and cost-effectiveness differs based on the specific screening method used across different platforms. There is a need to appraise existing economic evaluations of cervical cancer screening. The objective of the present systematic review was to identify primary and model-based economic evaluations of cervical cancer screening methods, and to provide a contextual summary of associated outcomes associated with screening modalities. Methods: The review considered primary-based and model-based full economic evaluations of cervical cancer screening methods. Such evaluations methods include cost-effectiveness analysis, cost-utility analysis, cost-minimization analysis, cost-benefit analysis, and cost-consequence analysis. We searched the following databases for full economic evaluations of cancer screening methods globally: SCOPUS, Pubmed, National Health Economic Evaluation Database (NH EED), Cochrane, and Health Economic Evaluation Database (HEED). No date restrictions were applied. Model-based and primary-based full economic evaluations were included. A critical appraisal of included studies was performed by the main investigator, while a second independent reviewer assessed critical appraisal findings for any inconsistencies. Data were extracted using a standardized data extraction tool for economic evaluations. Data extracted from included studies were analysed and summarised to answer the study objective using the Joanna Briggs Institute (JBI) Dominance Ranking Matrix (DRM).Results: Out of 671 screened studies, 44 met the study inclusion criteria. Forty-three studies were cost-effectiveness analyses while two studies reported cost-utilities of cervical cancer screening methods.HPV DNA testing was reported as a dominant standalone screening test by 14 studies, while 5 studies reported VIA as a dominant standalone screening test. Primary HPV screening strategies were dominant in 21 studies, while three studies reported Cytology-based screening strategies as the dominant screening method. Conclusions: Evidence indicates that HPV-based and VIA testing strategies are cost-effective, but this is influenced by setting. Our review suggests the limited cost-effectiveness of cytology-based testing, which may be due in part to the need for specific infrastructures and human resources. Systematic Review Prospero Registration: CRD42020212454. The review protocol may also be found on Prospero.


Introduction
Cervical cancer is a common form of malignancy and a leading cause of cancer-related mortality in females (1). It is also an important contributor to disease burden in Sub-Saharan Africa, with an estimated 75 000 new cases documented each year, as well as approximately 50 000 new deaths annually (2). The highest incidence of cervical cancer in the world is reported in Eswatini, with 6.5% of women developing the disease before the age of 75 (3). Countries in Western, Middle, and Southern Africa are hardest hit by cervical cancer-related deaths, with world age standardised mortality rates (ASMR) of 23.0%, 21.1.%, and 20.0%, respectively (3). The economic burden of cervical cancer is also substantial. For example, Wu et al (2020) reported that, in the Henan province of China, costs associated with cervical cancer, from diagnosis to one year after discharge, ranged from $8,066 to$22,888 per patient (4).
Cervical cancer is caused by infection with high-risk serotypes of the Human Papilloma Virus (HPV) (5). Infection can lead to the development of pre-cancerous lesions and malignancy if left untreated (6). Since neoplastic transformation can take years or even decades to develop, early detection and treatment of pre-cancerous lesions provides an important opportunity for therapeutic intervention (7). Indeed, the World Health Organization (WHO) has identi ed cervical cancer as a potentially eliminable form of cancer (7). That being said, cervical cancer remains under-diagnosed in clinical settings, particularly for developing countries (8).
Evidence however indicates that adequate screening is associated with reduction in cervical cancer-related deaths (9). The WHO estimates that prevention through vaccination, screening and early treatment, applying the 90-70-90 targets, could reduce median cervical cancer incidence rate by 10% by 2030 (10).
Several approaches have been developed to assist clinicians in screening for cervical cancer. These include unaided visual inspection with acetic acid (VIA) as well as assisted cytological (e.g. Papanicolaou (PAP) smear) and molecular (e.g. HPV DNA testing methods) (11) (12). A Pap test is a liquid cytology-based test in which cervix cells are analysed (13). Unaided VIA is carried out by observing cervix cell colour changes in response to acetic acid exposure(14). These screening methods differ in terms of their diagnostic value, accuracy, and associated costs to the user and healthcare system (15).
Health economic evaluations (16) can help inform the clinical application and nancing of appropriate screening methods for cervical cancer. Health economics research could ensure that the potential positive impact of health interventions are weighed against their use of limited resources particularly in lower-to middle-income healthcare settings (17). Economic evaluations might thus provide a possible framework to assist decision makers in providing much needed interventions based on available clinical evidence leveraged against the cost to the healthcare sector (18).
Several economic evaluations have examined the nancial aspects of cervical cancer screening in resource-limited healthcare environments. Studies conducted in developing countries including India (19) and South Africa (20) suggest that the costeffectiveness of individual modalities might differ based on setting. Findings from other studies conducted in developed nations such as Portugal (21) also support combining different screening methods depending on individual factors (21). This approach might however not be applicable in settings where health resources including access to molecular testing are already limited.
Health economic evaluations focused on cervical cancer screening are however limited by the use of different methodologies, and generalization across prior studies is often not possible. This highlights the need for a methodical approach to exploring systematic difference across various such economic evaluations conducted to date. At least three prior systematic reviews (22) have provided evidence supporting the cost-effectiveness of cervical cancer screening. However, Nahvijou et al (2014) limited their systematic review to cost-effectiveness analyses of cervical cancer screening methods (23). In 2015, Mendes et al (24) published a review whose aim was to identify mathematical models used to evaluate the impact of cervical cancer screening strategies. Though critical insights were gleaned from this review, restricting the study type to mathematical modelling resulted in the exclusion of primarybased economic evaluations. In their more recent review, Mezei et al. (2017) also limited their review to cost-effectiveness analyses, with a focus on lower-to middle-income countries. Furthermore, the authors selected only model-based economic evaluations (25). A large body of economic evaluation evidence is founded on randomized controlled trials and primary cost-effectiveness studies. Authors did not carry out an appraisal of the methodological quality of studies, which reduced the validity of the results. Lastly, authors focus on the cost-effectiveness of screening methods. This review builds on ndings by Nahvijou et al (2014), Mendes et al (2015) and Mezei et al (2017), by reviewing all full economic evaluation methods including cost-utility, cost-bene t, costminimization, and cost-consequence analysis.
The aim of the present systematic review was to appraise the evidence on full economic evaluations of cervical cancer screening methods for improving pre-cancerous lesion detection in women from the payer and/ societal perspectives.

Research design
The present review of studies examining the economic evaluation of cervical cancer screening methods was conducted using the Joanna Briggs Institute methodology for systematic reviews of economic evaluation evidence (26).

Selection criteria
This review focused on primary-and model-based full economic evaluations of cervical cancer screening methods. Inclusion criteria were: 1) studies published in English, 2) studies which considered female patients screened for cervical cancer using visual (VIA), cytological (Papanicolou smear) or molecular (HPV DNA testing) methods. Exclusion criteria were: 1) studies not available in English, 2) other systematic reviews and meta-analyses. No date restrictions were applied. Exclusion criteria were: full-text studies not available or not published in English, 2) systematic reviews and meta-analysis

Outcomes
Full economic evaluation methods of interest included cost-effectiveness (CEA), cost-utilities (CUA), cost-bene t (CBA), costminimization and cost-consequence (CC). Measures of interest included ICERS of cost/year lives saved (YLS), cost/death averted, cost/CIN2 detected, cost/QALY gained, cost/life-year (LY), marginal cost/case detected, cost/life year gained (LYG). Since the focus was on economic evaluations of global screening methods, there were no speci c socio-demographic or cultural factors considered as outcomes of interest. Incremental cost-effectiveness ratios (ICERS) were converted to international dollars using a base year of 2020. Original costs were converted to the local currency of the study market using market exchange rate data (27). Adjustment for in ation was carried out by multiplying ICERs by a GDP de ator, which were obtained from the World Bank.

Search strategy
The selection of suitable studies and the procedures for their identi cation was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (26). The main investigator (TS) performed a formal search of the academic literature over a 5-month period (September 2020 -September 2021) using the following databases: SCOPUS, Pubmed, National Health Economic Evaluation Database (NH EED), Cochrane, and Health Economic Evaluation Database (HEED). All searches and screening of suitable studies were duplicated by another researcher to ensure unanimous selection of appropriate economic evaluations for this review. The search terms and all logical synonyms and iterations thereof depending on the database used are summarized in Appendix I. The reference lists of selected studies were also screened for the identi cation of article citations of possible interest for the present research. All relevant citations identi ed using these criteria were then collated and uploaded into a Microsoft Excel template and duplicates removed. Titles and abstracts were then screened by two independent researchers. The full text versions of eligible studies were assessed. The reasons for exclusion of studies were also documented and reported. The methodological quality of suitable studies was then assessed using the JBI standardized critical appraisal instrument (26) as well as the Drummond's checklist for assessing economic evaluations (18). Model-based studies were appraised using a model assessment checklist (28).

Data Extraction and Synthesis
All studies selected for inclusion underwent data extraction and synthesis (to the furthest extent possible), regardless of their methodological quality. Data were extracted from studies selected for inclusion in the review using the standardized data extraction tool from JBI SUMARI by one reviewer. A second independent reviewer assessed extracted data for inconsistencies and discrepancies. The JBI SUMARI tool was augmented by a data extraction tool developed by Wijnen et al. (29). Extracted information were: 1) descriptive data about cervical cancer screening studies including, research perspective, geographical setting, population characteristics, and methods, 2) measures for CEA, CUA, CB and CC, 3) conclusions about factors which drive (impede) the costeffectiveness of cervical cancer screening.
Extracted data were then analysed and summarized using the JBI Dominance Ranking Matrix (DRM) (ref). This tool measures three potential outcomes for the cost of an intervention of interest against the health outcome(s) of interest, i.e. a) strong dominance (characterized by decisions distinctly favouring either the intervention or comparator, from a cost or clinical effectiveness standpoint), b) weak dominance (data favours either costs or effectiveness), and c) non-dominance (characterized by a less effective or more costly intervention).

Study inclusion
From a total of 671 titles and citations screened following removal of duplicates (n=16), 80 abstracts were screened, and 74 studies selected for full-text review ( Figure 1). Following exclusion of ineligible studies, a total of 44 were included in this review.

Methodological Quality of Studies Selected for Inclusion in Review
The overall methodological quality was high, with most studies identifying all important and relevant cost outcomes, as well as using relevant types of sensitivity analyses (univariate, multivariate, probabilistic, deterministic). Model-based studies generally reported structural assumptions as well as cervical cancer states and pathways. In certain cases, cost and outcome measures were not credible (Campos et al., 2012;Vassilakos et al., 2018), while in others, the procedures for valuation of costs and outcomes were not clear (Campos et al., 2012(Campos et al., , 2015, while in others still (n=5), it was uncertain whether adjustment for differential timing was performed. Several studies (n=, 7%) did not establish prior clinical effectiveness of screening, and two studies did not report sensitivity analyses. All 44 studies were however included for review. Decision rules for inclusion and exclusion were not created and all studies, regardless of their methodological quality, underwent data extraction to the furthest extent possible.

Description of Studies Included in Review
From a total of 671 titles and citations screened following removal of duplicates (n=16), 80 abstracts were screened, and 74 studies selected for full-text screening. Following exclusion of ineligible studies (Figure 1), a total of 44 were included in this review. In general, studies that were excluded during full-text selection compared health technologies beyond the scope of the research question.
The nal selection included 44 studies published in English between 2004 and 2021. The characteristics of these studies are summarized in Appendix IV. A total of 38 studies (88%) were model-based and thus focused on hypothetical female cohorts as eligible participants. Studies were conducted across different locations, including South Africa, India, Greece, Lebanon and Nicaragua. Though studies assumed varied names to characterise perspectives, perspectives can be broadly categorised into two modalities, namely, payer and societal perspectives. A total of 14 (33%) studies assumed a societal approach, while 18 (42%) used a payer perspective.

Economic Evaluation ndings from Cost-Effectiveness Studies
The most common economic evaluations examined cost-effectiveness (n=43; 97%). VIA was reported as a dominant screening method in 5 studies, while HPV DNA testing was reported as dominant by 14 studies. No study reported Cytology testing as a dominant standalone screening technology. A total of 20 (45%) cost effectiveness studies reported singular screening methods as dominant, while 26 cost effectiveness studies reported screen and treat strategies as dominant. Table 1 summarizes dominant screening technologies reported by studies.  Table 2 outlines cost-effectiveness outcome measures associated with dominant screening methods and strategies. Estimated outcomes were: ICERS of cost/year lives saved (YLS), cost/death averted, cost/CIN2 detected, cost/life-year (LY), marginal cost/case detected, cost/life year gained (LYG). Studies which analysed both cost-effectiveness and cost-utility included cost/QALY gained as an outcome measure. Costs were reported in international dollars, using a base year of 2020.

Discussion
This review sought to synthesise available evidence on cervical cancer screening methods and strategies to achieve optimal precancerous lesion detection and thus avert cervical cancer. We thus performed a critical appraisal of economic evaluation studies of cervical cancer screening methods (n=44). All studies supported the cost-effectiveness of cervical cancer screening. In particular, our results suggest that primary HPV DNA testing strategies are likely to be cost-effective in several settings. VIA may be cost-effective in some settings, including rural areas, but not others. Similarly, cost-utility ndings comparing Cytology and VIA associated VIA with higher utility. These ndings echo those reported in a systematic review by Mezei et al. (25) who concluded that HPV testing and VIA are the most cost-effective screening methods in LMICs. Pap testing is less frequent as a dominant method, but also cost-effective in the context of co-testing and triaging. Our results also suggest that cervical cancer screening modalities are most effective when applied within a broader context of treatment and intervention. This would include consideration of the health economics of cervical cancer, in addition to evidence for the clinical effectiveness of different established modalities. Our review further suggests that sample collection, screening sequence, algorithms, and coverage are important.
One factor thought to in uence cost-effectiveness of cervical cancer screening modalities is sample collection. Mezei et al. (32) compared self-collection followed by clinic-based VIA triage to clinic-based collection and triage in HPV-positive females in Uganda. Reduction in cervical cancer incidence and ICERs (USD/YLS) were used as measures of cost-effectiveness. Use of Monte Carlo modelling allowed the authors to show that self-collection was more cost-effective than clinic-based VIA triage based ICER outcomes. Using cytology-based screening as a comparator, Vassilakos et al (33) also reported that offering HPV self-testing is more cost-effective and associated with a reduction in cervical cancer cases and related deaths. A critical gain that both authors correlate to HPV self-testing is increased population coverage.
The method sequence could also affect cervical cancer screening cost-effectiveness. Jin et al. (34) compared the three screening methods for cervical cancer of interest in this review and found that these differed in their accuracies of detection. Co-testing was identi ed as more accurate, but less cost-effective. These ndings echo those of Campos et al. (35) where the authors compared different methods and interventions in theirs of lifetime risk reduction and ICERS (USD/YLS). These measures found HPV testing with intervention to be more cost-effective compared to cytology-based strategies. Using the Nicaraguan cost-effectiveness threshold (GDP per capita of US$2090) HPV-cryotherapy remained comparatively cost-effective, with an ICER of US$320/YLS (35).
Several studies included in this review underscored the importance of screening coverage. In Lebanon, results from a model-based cost-effectiveness analysis indicated that using cytology as a screening modality with a shift from the current 20% coverage to at least 50% would reduce cervical cancer incidence considerably (36). More gains would be achieved if HPV-testing was used as a screening modality at 50% coverage: resulting in a 23.4% cervical cancer incidence reduction (36). Modulating coverage for different strategies (50-80%) tend to favour the cost-effectiveness of HPV-based screening strategies (36) Our review had several limitations. First, it was di cult to account for uncertainty associated with heterogeneity and model structure across studies, meaning that internal or external model-consistency could not be guaranteed. Several model-based studies also used the same model, e.g. Campos et al. (37)(38) (39), and study ndings were not disparate. Second, critical appraisal and data extraction were performed by one reviewer. However, this limitation was offset by critical appraisal and extracted data being assessed for inconsistencies by another independent reviewer. Given the signi cant heterogeneity of studies, study results could not be pooled and meta-analysed, a limitation common in economic evaluation systematic reviews. This limitation underscores the need to further develop and standardize economic reporting globally. An interim measure which researchers can apply is sub-set group analysis. In other words, researchers should aim to pool and compare studies similar in setting, participants, and outcomes.

Conclusions
In conclusion, our review supports the general cost-effectiveness of HPV testing and VIA used to screen for cervical cancer. Compared to HPV testing and VIA, cytology testing is least cost-effective. However, researchers should keep in mind that health economic reviews are not intended to provide conclusive recommendations for routine practice, but rather to guide policy makers in developing optimized strategies for testing and intervention (26). A combined approach might also prove feasible, and clinicians might need to consider the order in which screening is performed in order to maximize cost-effectiveness. Furthermore, studies show that a large body of models and simulations targeted toward cervical cancer screening evaluation exist. Countries intending to introduce more relevant and improved cancer strategies can leverage on the existing body of knowledge by learning from documented best practices.
Future studies would do well to explore the economic evaluation of cervical cancer screening in relation to the test performance of screening modalities. Furthermore, parameters such as the order of screening methods, and its relationship to the screening intervention, screening coverage, screening modality, and the number of screening visits, could have important implications for care. Further research would do well to determine what treatment options are associated with ideal clinical and economic value. The ultimate success of cervical cancer screening and treatment could depend on a broader perspective in deciding which strategy is most appropriate for the individual patient and context. Availability of data and materials

List Of Abbreviations
The datasets during and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests
There is no con ict of interest in this project.

Funding
This work was supported by a grant from the Harry Crossley Foundation. The Harry Crossley Foundation did not play a role in data collection and analysis, or in the interpretation of the results.
Authors' contributions TS searched and downloaded articles using the team-approved search strategy. LN independently screened articles and abstracts. TS appraised study methodological quality, extracted and analysed data. LN was a major contributor in manuscript conceptualization and writing. All authors read and approved the nal manuscript.