Introduction
The burden of disease attributed to cervical cancer is the leading cause of disability-adjusted life years lost in women in forty-nine countries and is also associated with premature death in twenty-three countries (Reference Ginsburg, Bray and Coleman1). Global estimates reveal that in 2020, cervical cancer was the eighth most incident cancer in individuals above 15 years old, with approximately 600,000 new cases, and was the fourth most incidence among women, with an estimated 19.3 cases per 100,000 (Reference Sung, Ferlay and Siegel2).
The World Health Organization (WHO) launched the global strategy for the Elimination of Cervical Cancer in 2020, aiming to have 70 percent of women worldwide regularly screened with high-performance tests, in addition to other goals related to vaccination, diagnosis, and treatment of identified precursor lesions. The HPV-DNA test is recommended as the preferred screening method, replacing cytology and visual inspection with acetic acid, which are the most used tests in developing countries (3).
Cytology is also the most used test in the triage phase due to its high specificity, although it has limited sensitivity and variability in interpreting results, besides false-negative results caused by sampling error or detection error (Reference Nanda, McCrory and Myers4).
Another well-known challenge faced in the follow-up phase is the low adherence to screening programs. The rate of inadequate follow-up ranges from 28.3 to 75 percent in cytology-based triage (Reference Martinez-Gutierrez, Chima and Boyd5).
No international consensus exists about the better triage strategy, especially in developing countries, where women with atypical squamous cells of undetermined significance (ASC-US) or low-grade intraepithelial lesions (LSIL) undergo repeat cytology testing between 6 and 12 months depending on age and type of lesion (6,Reference Fernández-Deaza, Caicedo-Martínez and Serrano7). The main clinical guidelines recommend HPV tests alone or co-testing with cytology for these cases (Reference Perkins, Guido and Saraiya8).
The use of a triage test aims to identify women with a higher probability of being carriers of precursor lesions or cancer and reduces the health system burden. Molecular tests with high specificity avoid unnecessary referrals for colposcopy or treatment (Reference Santos, Derchain, Sarian, Campos, dos and Fonsechi-Carvasan9).
For triaging women with ASC-US/LSIL cytology, tests based on HPV-DNA and RNA detection, detection of oncoproteins such as p16 and Ki-67, or DNA methylation can be used (10). Colposcopy can also be used as a triage test, but it implies higher costs and over-referral of healthy women to biopsy or treatment, misplacing attention from the most probable carriers of precursor lesions or cancer.
In Brazil, in the last 5 years, the proportion of exams with ASC-US/LSIL results represents 2.2 percent of the total number of exams performed and 69.5 percent of altered results (11), which would represent referring more than twice as many women per year for colposcopy, compromising the offer of exams for other women with cytological abnormalities more likely to represent cancer or its precursor lesions.
Given the diversity of technologies for reducing morbidity and mortality attributed to cervical cancer, high-quality evidence-based decisions are necessary for more efficient and effective screening programs, minimizing costs and increasing benefits to women’s health and quality of life. Health economic evaluations support such decisions and assist in defining public policies for the control and prevention of cervical cancer, providing economic evidence of triage strategies that should be analyzed together with clinical evidence (Reference Tantitamit, Khemapech, Havanond and Termrungruanglert12–Reference Pedersen, Burger, Sy, Kristiansen and Kim16).
Decision models are mathematical structures that reproduce the clinical and economic outcomes of alternative strategies in a simplified way, based on information collected from the context being evaluated. When developing models according to the natural history of the disease and diagnostic and treatment guidelines, in addition to the evaluation period or time horizon and the types of events that will be modeled, it is also important to consider the perspective of the analysis, which dictates the types of costs that will be reproduced in the model. Among the existing perspectives, those of the healthcare system, society, and patient are highlighted (Reference Soárez, Soares and Novaes17).
The results of economic evaluations in healthcare can be indicated by a ratio of health outcomes in natural units, such as life years gained or number of lives saved, and monetary units, producing the incremental cost-effectiveness ratio (ICER) (Reference Soárez, Soares and Novaes17). An alternative for better comparison of healthcare interventions in different contexts is the adoption of quality-adjusted life years (QALY) (18).
This parameter is strongly related to health utilities measured by several methods being the most successful approach to apply psychometric instruments in a questionnaire format, such as EQ-5D, to a representative sample of people to assess their preferences about life duration with life quality (18).
The quantification of utility, denoted by a numerical scale ranging from 0 to 1 where “0” signifies the most adverse health condition and “1” denotes optimal health, is a customary practice in health-related assessments. These utility values are conventionally employed in the computation of QALYs through their multiplication by the duration of time spent in specific health states. This systematic approach allows for a standardized representation of health outcomes, facilitating a comprehensive evaluation of the overall impact on quality of life (Reference Virgili, Koleva, Garattini, Banzi and Gensini19). In a comprehensive search of the literature, four systematic reviews of cost-effectiveness evaluations of cervical cancer screening strategies were identified (Reference Mezei, Armstrong and Pedersen20–Reference Nahvijou, Hadji and Marnani23), but none evaluated cost-effectiveness studies of triage tests for women with ASC-US or LSIL cytology results.
In this study, we conducted a systematic review of primary studies that sought to estimate the cost-effectiveness of triage alternatives after primary cytology with ASC-US or LSIL results in different population contexts.
Methods
This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Reference Page, McKenzie and Bossuyt24). The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO), with the registration code CRD42021237140, before the start of this review.
Eligible studies were complete economic analysis (cost-effectiveness and cost-utility) with any analytic time horizon, according to the following question:
-
• Population: Adult women who started screening and presented ASC-US/LSIL in primary cytology.
-
• Intervention: Repeat cytology for referral to colposcopy or treatment.
-
• Comparator: Any other triage strategy
-
• Outcome: ICER
Cost analysis and economic evaluations without a comparative result of costs and benefits expected were excluded.
Articles published up to January 2024 were searched in the following databases: Medline/Pubmed, Lilacs, Embase, and The NHS Economic Evaluation Database. In addition to the databases, a hand-search in the references list of the included studies was also conducted. The sample search strings can be found in Supplementary Material.
The search strategy was constructed using controlled vocabulary terms (Health Descriptors—DECS/Medical Subject Headings—MESH), synonyms, truncation, and variations in the search fields of the databases combined with boolean operators to prevent the loss of any relevant publications. No date or language restrictions were applied.
Using the Rayyan® software, two reviewers (IM and LB) screened the titles and abstracts of studies identified in the electronic search. Full-text articles whose titles and abstracts were considered relevant by at least one reviewer were retrieved. Disagreements were solved by a third reviewer (MP).
The same two independent reviewers performed data extraction filling out a standardized data collection form on Microsoft Word® and assessed the methodological quality of included studies using the Quality of Health Economic Studies (QHES) tool. It corresponds to a set of criteria with weights assigned according to their relevance in economic evaluation studies. The quality score is determined by summing up the points assigned to each “yes” response for answered questions (Reference Chiou, Hay and Wallace25).
The QHES employs a scoring system ranging from 0 (indicating the lowest quality) to 100 (reflecting the highest quality). This continuous scale can be dichotomized, categorizing studies into high-quality (75–100 points) and not high-quality (0–74 points) based on their respective scores (Reference Spiegel, Targownik and Kanwal26).
The profile of studies and their characteristics were presented in tables, allowing for a comparison of economic and health outcomes.
To enable comparison of the ICER reported in each study, values were monetarily updated to 2024 and converted to U.S. dollars (US$). The costs were updated by inflating the local currency using local inflation rates between the date of publication and January 2024, and then exchanging to US$.
Results
Study selection and characteristics
After excluding duplicates, 126 records were screened. Of these, thirteen studies were selected for full-text reading, and five cost-effectiveness analyses were included (Figure 1).
Figure 1. Flowchart for study selection.
The comparators adopted to repeat cytology were HPV testing, immediate colposcopy, and cytology (conventional or liquid-based) with reflex HPV testing. Three studies have two comparators (Reference Kulasingam, Kim and Lawrence27–Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30), and one has three comparison groups (Reference Hughes, Glazner, Barton and Shlay31).
The included studies (Table 2) were published between 2005 and 2016 in Brazil (Reference Vanni, Legood and Franco28), Thailand (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30), USA (Reference Kulasingam, Kim and Lawrence27,Reference Hughes, Glazner, Barton and Shlay31), and Sweden (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29). Two evaluations are trial-based (Reference Kulasingam, Kim and Lawrence27,Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29), one is based on retrospective cohort data (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30), and the remainder follows a hypothetical cohort (Reference Vanni, Legood and Franco28,Reference Hughes, Glazner, Barton and Shlay31).
Quality assessment
Based on the QHES tool, three studies were considered to have high quality (Reference Kulasingam, Kim and Lawrence27–Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29), and two not high quality [(Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30,Reference Hughes, Glazner, Barton and Shlay31); Table 1]. Common reasons for score downgrading were not presenting a clear declaration of the assumptions and a discussion of the magnitude and direction of potential bias.
Table 1. The main characteristics of the included studies
Abbreviations: HPV, human papillomavirus
Hughes, 2005 (Reference Hughes, Glazner, Barton and Shlay31) received the lowest score due to several limitations—no clear description of the methodology for data abstraction and estimation of resource quantities; insufficient time horizon; the model structure was not presented nor described, either the components of the numerator and denominator were displayed clearly; authors did not discuss the direction of potential bias; lack of sensitivity analysis; and finally, there was not a statement disclosing the funding source for study.
Study perspective
Four studies adopted the healthcare system perspective (Reference Kulasingam, Kim and Lawrence27–Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29,Reference Hughes, Glazner, Barton and Shlay31), and one both the payer and the patient perspectives (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30). Most of the studies that adopted the perspective of the healthcare system modeled public and private providers’ investment in triage programs, (Table 1). The payer perspective refers to a specific health provider in Thailand, the King Chulalongkorn Memorial Hospital (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30).
Decision model
Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27) and Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) chose a decision tree, whereas Vanni, 2010 (Reference Vanni, Legood and Franco28) chose a Markov model with a 6-month cycle length. Hughes, 2005 (Reference Hughes, Glazner, Barton and Shlay31) did not specify the type of model.
Vanni, 2010 (Reference Vanni, Legood and Franco28) adopted a lifetime horizon applying a five percent discount on costs and benefits yearly, Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27), a 2-year horizon, and the others a 1-year horizon (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29–Reference Hughes, Glazner, Barton and Shlay31).
All authors adopted modelling assumptions, as expected, but it was considered sufficiently described by Vanni, 2010 (Reference Vanni, Legood and Franco28) and Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27). The most common health outcome was the number of CIN2+ cases detected. Other outcomes were the number of true positive results and life years saved (Table 3).
Test characteristics and clinical parameters
The sensitivity and specificity values for the triage tests are shown in Table 2. When available, value ranges were also reported. Ostensson, 2010 (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29) extracted data from a Swedish randomized clinical trial (Reference Andersson, Dillner, Elfgren, Mints, Persson and Rylander32) and Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27) from the ALTS trial (Reference Schiffman and Solomon33), and the detection rates measured were used to populate the model.
Table 2. Test performance estimates reported in the included studies
Abbreviations: CIN, cervical intraepithelial neoplasia; HPV, human papillomavirus
Table 3. Cost-effective strategy and incremental cost-effectiveness ratio reported in the included studies
Abbreviations: CIN, cervical intraepithelial neoplasia; HPV, human papillomavirus
Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) extracted data from medical records of women assisted at the King Chulalongkorn Memorial Hospital and collected the detection rates to populate the model. Hughes, 2005 (Reference Hughes, Glazner, Barton and Shlay31) employed various sources to extract main estimates and ranges for the different tests analyzed, mainly observational studies.
Vanni, 2010 (Reference Vanni, Legood and Franco28) extracted data from two published meta-analyses. One regarding data for HPV testing (Reference Arbyn, Buntinx, Van Ranst, Paraskevaidis, Martin-Hirsch and Dillner34) and the other for colposcopy data (Reference Mitchell, Schottenfeld, Tortolero-Luna, Cantor and Richards-Kortum35). Both meta-analyses provided aggregated accuracy ratios at a threshold of ASCUS or LSIL for an outcome of CIN2+ in women with abnormalities in cytology smears. Cytology accuracy was derived from an epidemiological mathematical model (Reference Goldhaber-Fiebert, Stout, Ortendahl, Kuntz, Goldie and Salomon36).
Clinical parameters included in the cost-effectiveness models were screening coverage, adherence to screening strategy, loss to follow-up, prevalence of low-grade and precursor lesions, HPV infection, and cervical cancer.
Transition probabilities
Given the type of decision model adopted in each cost-effectiveness analysis, only Vanni, 2010 (Reference Vanni, Legood and Franco28) applied transition probabilities to simulate the natural history of the lesions detected. They were derived from literature and whenever possible from studies conducted in Brazil or Latin America (Reference Muñoz, Méndez and Posso37–Reference Franco, Villa and Sobrinho39) as the target population of analysis was Brazilian women.
Cost inputs
As no study adopted a societal perspective, only direct cost parameters were included in the cost-effectiveness analyses. In addition to the cost of test kits and analysis, the authors also considered the costs of appointments, follow-up exams, and treatment of CIN2+ and cervical cancer.
Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27) and Hughes, 2005 (Reference Hughes, Glazner, Barton and Shlay31) retrieved unit costs from Medicare reimbursement values; Ostensson, 2010 gathered costs from three Sweden hospitals database: Stockholm South General Hospital, Danderyd Hospital, and Karolinska University (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29); Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) retrieved the main cost estimates from King Chulalongkorn Memorial Hospital, whereas the values for cost ranges in sensitivity analysis, were obtained from a survey of laboratories from various government and private hospitals in Thailand.
Vanni, 2010 (Reference Vanni, Legood and Franco28) retrieved most costs from the Brazilian Private Healthcare System standard procedure table (Classificação Brasileira Hierarquizada de Procedimentos Médicos), except for medical visits, nurse visits, and day hospital, which were provided by the Hospital de Clínicas de Porto Alegre. The HPV testing had not been reimbursed by the health system at the time when the study was published, so the authors assumed the same cost as a cytology smear because similar or lower prices were achieved in other settings.
Sensitivity analysis
Regarding sensitivity analysis, it was only not performed by Hughes, 2005 (Reference Hughes, Glazner, Barton and Shlay31). Except for Vanni, 2010 (Reference Vanni, Legood and Franco28), all the studies performed only one-way sensitivity analysis ranging the test accuracy parameters using the values of 95 percent confidence intervals from the literature estimates, and costs ranged on an arbitrary percentage.
Analyzing the difference in results with the variations, Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27) performed an interesting strategy to approximate the costs of all groups compared to the highest (immediate colposcopy) and lowest cost estimates (repeat conventional cytology).
Additionally, to one-way sensitivity analysis, Vanni, 2010 (Reference Vanni, Legood and Franco28) performed probabilistic and threshold analysis.
Study findings
In all cost-effectiveness analyses, repeat conventional cytology was less costly but less effective than comparators.
Comparing HPV testing and immediate colposcopy to repeat cytology, HPV triage presents a more cost-effective and equally or even more effective follow-up option for women aged 30 years or older. However, it may be dominated by immediate colposcopy in a further pairwise comparison depending on the unit cost of each test.
In Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) for provider and patient perspectives, and in Ostensson, 2010 (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29) for health system perspective, colposcopy was the most cost-effective strategy in comparison to HPV testing (Table 3).
Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) sensitivity analysis, showed the most cost-effective strategy shifted away from colposcopy as the costs of colposcopy increased or the costs of cytology decreased. Lowering the costs of HPV testing would be necessary to enhance the cost-effectiveness of the follow-up strategy, as demonstrated in Ostensson, 2010 (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29).
If liquid-based cytology were employed in primary screening, HPV triage would become significantly less expensive and more cost-effective compared to conventional cytology and immediate colposcopy.
When comparing liquid-based cytology to conventional cytology, there was an extra cost to detect each additional true positive result using the liquid-based method according to Hughes, 2005 (Reference Hughes, Glazner, Barton and Shlay31). The immediate colposcopy strategy was both more expensive and less effective than liquid-based cytology with reflex HPV testing, thus being dominated by the latter. The study also demonstrated that ICER increased with an increase in patients’ age.
Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27) and Vanni, 2010 (Reference Vanni, Legood and Franco28) found the HPV test to be the most cost-effective strategy compared to immediate colposcopy and repeat cytology. In Vanni, 2010, a threshold analysis study revealed that if the cost of HPV testing exceeded twice the cost of cytology, the optimal strategy would be immediate colposcopy and the discount rate did not change the conclusions of ICER magnitude (Reference Vanni, Legood and Franco28).
Discussion
This is the first systematic review aiming to investigate the cost-effectiveness of triage alternatives to replace repeat cytology in the management of women diagnosed with ASC-US and LSIL. This study fills a gap of robust economic evidence about this topic to be evaluated along with available clinical evidence in the updating of triage protocols.
In different contexts analyzed by five cost-effectiveness studies repeat conventional cytology was not the most efficient strategy. Strategies involving HPV testing as the main test, or as a reflex test was reported as cost-effective in three studies, whereas two studies found colposcopy to be more cost-effective.
Sensitivity analysis performed by all studies showed that ICER results were highly sensitive to variation in test unitary costs and accuracy estimates. Although a significant range in accuracy estimates can be observed through the studies due to the variability of literature sources employed in data extraction, the data follow the same tendency. Besides, these parameters were explored in sensitivity analysis in all studies, so it should not be considered a topic of high concern in future decision-making based on this systematic review.
The reason immediate colposcopy was cost-effective in Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) and Ostensson, 2010 but not in Kulasingam, 2006 (Reference Kulasingam, Kim and Lawrence27) and Vanni, 2010 (Reference Vanni, Legood and Franco28), might be related to differences in some clinical and cost inputs included in the first two studies.
Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) included loss to follow-up in HPV testing and cytology groups, with high rates of missed cases, but not in the immediate colposcopy group, which might have affected the effectiveness of HPV testing and cytology. Ostensson, 2010 (Reference Ostensson, Fröberg, Hjerpe, Zethraeus and Andersson29) considered that after cytology and HPV testing, there would be a referral to gynecological consultation or a follow-up visit to a midwife, raising the follow-up costs in these groups, whereas the colposcopy group would not undergo these steps.
Although the preferred health outcome for economic evaluations is QALY (18,Reference Drummond40), it was not evaluated by any of the five studies. Vanni et al., 2010 (Reference Vanni, Legood and Franco28) justify that the choice of life years saved as the outcome was due to the absence of health utilities for Brazilian women eligible for triage or with cervical cancer in the literature.
Tantitamit, 2015 (Reference Tantitamit, Termrungruanglert, Oranratanaphan, Niruthisard, Tanbirojn and Havanond30) also mentioned the decision for CIN2+ detection as an outcome of the lack of QALY data for the target population or a similar one. The detection of CIN2+, the main health outcome in the majority of studies included, has been considered a validated outcome to be used as a proxy for the occurrence of cervical cancer (Reference Arbyn, Ronco and Anttila41,Reference Ronco, Dillner and Elfström42).
Subsequent investigations should be prioritized to obtain QALYs of the women who were diagnosed with ASC-US/LSIL smear results after performing different strategies to allow the comparability and interpretability of their findings regarding cost-effectiveness. Decision makers may encounter challenges in drawing decisive conclusions when confronted with a body of evidence characterized by diverse metrics of health benefits.
The comparison between the included studies was limited by the variability of the analyzed tests, decision models, health outcomes, resources, costs, and time horizon. In cost-effectiveness studies, a new intervention is usually compared to the standard practice, which differs according to the analyzed context. The ICER is strongly influenced by the comparator due to the nature of the calculation that includes the benefits and costs of the new intervention versus the comparator, and thus, the comparison of reported ICERs was challenging (Reference Shields and Elvidge43).
Given the nature of economic evaluations related to the analyzed topic, the choice of the best cervical cancer triage strategies is highly complex due to the number of criteria to consider from the perspective of health policies (Reference Mendes, Bains, Vanni and Jit21).
The included cost-effectiveness analyses were published at least a decade ago. Not only there is limited economic evidence on the ASC-US/LSIL population but also, the more recent studies are already evaluating the cost-effectiveness of more advanced strategies to overcome HPV testing shortcomings (Reference Pedersen, Burger, Sy, Kristiansen and Kim16), emphasizing the urgent need for management updating in countries that still practice cytology-based triage.
In this sense, this systematic review can be useful by comparatively presenting parameters and assumptions of interventions, providing information on the application of decision analysis models, and guiding the health managers’ decisions on the implementation of new triage tests, considering the reality of each health system.
Thus, studies of this nature can contribute to other research and to actions aimed at expanding access to technologies and health services for the female population.
The simplification enabled by the alternative strategies in comparison to repeat cytology regarding the number of procedures required to complete triage, lower loss to follow-up, and increasing detection of true positives CIN2+, allows timely treatment that is turned into life years saved and efficient resource allocation.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0266462324000540.
Acknowledgments
We thank Rosângela Caetano and Flávia Miranda for the discussion on the study concept and preliminary results.
Funding statement
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing interest
None.
Open access
