A review of the published and grey literature was conducted to identify relevant economic evaluations that assessed the cost-effectiveness of various HPV screening strategies. Given the high number of published economic evaluations identified on this topic, the Economic Review adopted a similar approach to the Clinical Review by focusing only on studies that were conducted in countries with a health care context comparable with Canada’s. With this inclusion criteria, 25 unique economic evaluations were identified that addressed the cost-effectiveness of at least two of the three screening approaches of interest (i.e., primary cytology, primary cytology with HPV triage, or primary HPV with cytology triage).23,68–90
Appendix 10 provides details on each economic evaluation.
Although most of the economic evaluations reviewed all three screening approaches of interest, there was considerable variation observed in the screening strategies that were evaluated between studies. Differences included the targeted age range for programmatic screening, the frequency of screens, the criteria for triage and colposcopy referral, and the management algorithms for abnormal screening findings. The age at which screening started ranged from ages 21 to 30 and screening intervals in-between tests ranged from one to ten years. Where most economic evaluations were similar was in the commencement of HPV testing at age 30 when it was included as part of the strategy, and in some cases, the screening strategy incorporated primary cytology prior to that age. Several studies in non-Canadian jurisdictions examined the impacts of incorporating vaccinations to various screening strategies on cost-effectiveness.23,68,73,75–77,83,91
Four economic evaluations were conducted in a Canadian setting, two of which adopted a national scope,80,84 while the other two were province-specific.72,89 All studies applied the public health care payer perspective and all but one compared the three screening approaches of interest. However, none fully captured all screening strategies that were of interest to this review due to variations in targeted age range for screening and the screening frequency. The modelling approach between the studies varied, with two studies employing a cohort-level state-transition model,80,92 one study using a dynamic event-based microsimulation,84 and the final study using a patient-level state-transition model.89 The study by Popadiuk et al. (2016)84 using a dynamic event-based microsimulation incorporated HPV transmission within the model, but the model only followed a cohort of females for 30 years. None of the Canadian models evaluated an HPV-vaccinated population. Strategies incorporating HPV testing, either as a primary test or as a triage following equivocal cytology results, appeared on the efficiency frontier in all four studies.
Thus, existing economic evaluations do not fully address the screening strategies of interest to this review and it remains unclear how the economic value of screening may differ between an unvaccinated and vaccinated population. Because of these gaps, a de novo economic analysis on the cost-effectiveness of different programmatic screening strategies for the Canadian population (pre-vaccinated and partly vaccinated) was conducted as part of the Economic Review. The economic models identified from the literature provided insights in conceptualizing and developing the model structure and in determining appropriate model assumptions.
Primary Economic Evaluation
Methods
A primary economic evaluation was conducted to assess the lifetime costs, health outcomes, and cost-effectiveness of HPV testing compared with cytology as the primary screening tool, with or without triage, as part of an organized cervical cancer screening program within a Canadian population eligible for screening. A protocol for the economic evaluation was written a priori and followed in the conduct of this review.93
Type of Analysis
Given the broad implications of implementing a population-level screening program, a cost-utility analysis was conducted. Health outcomes were expressed as quality-adjusted life-years (QALYs) to capture both the mortality and morbidity impacts related to detecting precancerous cervical lesions and cervical cancer. The primary outcome was the incremental cost per QALY gained, commonly referred to as the incremental cost-utility ratio (ICUR).
Target Population and Setting
Canadians eligible for cervical cancer screening represented the target population. Of particular interest were the age ranges of nine to 69 years as the lower bound matched the age in which Canadians would be eligible for HPV vaccination while the upper bound reflected the current recommended age for screening cessation.22 Single birth-year cohorts were defined and analyzed separately to better understand the potential impact of clinical heterogeneity on cost-effectiveness due to an individual’s age and potential vaccination history.94 Separate age cohorts that were tested in the model included a cohort of individuals aged 9 (i.e., “future incidence cohort” in which individuals entering the model are younger than the screening program start age), a cohort of individuals aged 20 (i.e., “incident cohort” in which individuals entering the model are at the start age of the screening program), and a prevent cohort defined as aged 30 at the start of the model. The proportion of vaccinated individuals eligible for screening within an age cohort was further considered within the analysis. Publicly funded HPV vaccination programs were introduced in Canada in 2006 with all provinces offering vaccination to pre-adolescent girls at the age of nine.95
At the start of the model, all individuals are clear of an infection and have no prior history of cervical cancer. Upon entry into the model, individuals are assigned to a level of sexual activity ranging from low (0) to high (3) (i.e., l ≡ [0,1,2,3] that corresponds to the number of lifetime sexual partners (i.e., 0 to 1, 2 to 10, 11 to 39, and 40 or more lifetime partners). This parameter impacts the age of onset of sexual activity96 and therefore, the age in which individuals in the model become at risk of acquiring a high-risk HPV infection. The proportion of females in each sexual activity level was based upon the Psychosocial Impact of Cervical Screening and Condylomas: An Epidemiological Study conducted in Canada.96
The setting in the model reflected the Canadian health care system. It was assumed that access to all screening tests would be available.
Time Horizon
Given that the impact of screening is long term in terms of reducing the lifetime risks of developing cervical cancer, a lifetime horizon was defined. The model followed a cohort of Canadians eligible for cervical cancer screening up to their life expectancy with screening offered in accordance to the screening strategy being evaluated. The model cycled yearly with costs and benefits discounted at 1.5%, adhering to the latest Canadian guidance.97 Sensitivity analyses were conducted using a 0% and 5% discount rate.97
Interventions
The Economic Review compared the cost-effectiveness of HPV and cytology screening tests in the context of an organized screening program given that these screening tests would be offered as part of the existing Canadian programmatic screening for cervical cancer. By taking a programmatic approach, the economic analysis could further assess the optimal screening frequency and screening age range. The screening program of interest to this review can be broadly categorized into three approaches (i.e., primary cytology, primary cytology with HPV triage, primary HPV with cytology triage), and can be further subcategorized by the frequency and targeted age range for screening. Although some non-Canadian screening programs now offer co-testing, high-quality studies that reported harms, safety, and long-term outcomes did not often compare co-testing with primary HPV tests; therefore, co-testing was not considered as an intervention in the review. outlines the 11 screening programs evaluated as part of this review. In addition, a no screening strategy was also included as a control to validate the economic model.
List of Cervical Cancer Screening Programs Evaluated.
There are different assays and techniques associated with each screening test. In past economic evaluations, different cytological methods (e.g., conventional Pap smear and LBC) were found to be comparable98 and, in this economic model, primary cytology refers to both by considering these methods interchangeable. In triage-based strategies (i.e., strategies B and C), the cytological method was assumed to be LBC. Per the implementation section of this review, a separate sample would need to be taken for HPV test under conventional cytology, whereas, with liquid-based preparations, the same smear sample can be used for both cytology and HPV test. This assumption has important implications to costs and convenience. In terms of costs, the screening costs would be lower than if the cytological method was based on Pap smear given that only a single physician visit would be required to collect the cervical sample, obviating the need for a repeat physician visit. This would also be more convenient to patients, thereby reducing the risk of non-participation that can arise if a second visit was required for further testing. Of note, alternative approaches to collect samples (i.e., self-sampling) were not considered in the economic model given the paucity of evidence from the Clinical Review in terms of diagnostic test performance.
In the case of HPV testing, the Clinical Review found little evidence on the comparative DTA between different commercial assays (e.g., Cobas, HC, Aptima) and techniques (e.g., partial genotyping, full genotyping). In fact, the Cochrane review40 reported a pooled sensitivity and specificity that combined all commercial HPV tests together. Although a subgroup analysis was available within that study that pooled only the DTA data of HC2, the results of the subgroup analysis were similar to the original analysis that combined all HPV tests together. In consulting with the clinical experts involved in this review, it was noted that it would be less meaningful to evaluate HC 2 separately in the economic analysis as numerous other HPV tests have since been commercialized. Specific commercial assays were therefore not explored further in the Economic Review as it was assumed that HPV tests were broadly interchangeable.
Management of Primary Cytology (Strategy A)
Under this set of strategies, cytology is first conducted on individuals eligible for screening. The management of cytology outcomes in terms of follow-up and treatment was modelled to reflect Canadian clinical practice guidelines for cervical cancer screening.10,22,99
highlights the screening algorithm captured in the economic model and highlights where variation in clinical practice exists between Canadian provinces.22
Management of “Strategy A” Outcomes — Primary Cytology. ASCUS= atypical squamous cells of undetermined significance; ASC-H= Atypical squamous cells, cannot exclude HSIL; HSIL= high-grade squamous intraepithelial lesion; LSIL= low-grade (more...)
Cytological results are classified based on the Bethesda system in which squamous cell abnormalities can be classified into ASCUS, atypical squamous cells — cannot exclude HSIL (ASC-H), LSIL, or HSIL. In primary cytology, findings of ASCUS or LSIL during routine screening would result in triage with repeat cytology at six months. If individuals are found to have persistent abnormal cytological findings at six months, they are referred to colposcopy management. If under the age of 30, individuals with a corresponding normal cytology result during repeat testing would have cytology repeated at six months and return to routine screening following two consecutive negative results. If over the age of 30, individuals would return to routine screening upon negative findings in their repeat cytology. At any point of screening, those with ASC-H, HSIL, or carcinoma would be immediately referred for colposcopy examination (i.e., colposcopy with or without biopsy) for histological assessment of the cervix.
Management of Primary Cytology With HPV Test Triage (Strategy B)
The addition of HPV reflex testing for equivocal cytology results was similar to Strategy A with the exception to how ASCUS findings would be managed (). An ASCUS result would result in HPV triage. Patients who test positive for high-risk HPV following an ASCUS result would be referred immediately for colposcopy examination whereas, patients who test negative for high-risk HPV would return to routine screening.99
Management of “Strategy B” Outcomes — Primary Cytology With HPV Tests for Equivocal Results. ASCUS = atypical squamous cells of undetermined significance; ASC-H = Atypical squamous cells, cannot exclude HSIL; hr = high-risk; HSIL (more...)
Management of Primary HPV Test With Cytology Triage (Strategy C)
The management of primary HPV testing with cytology triage reflects existing Canadian and international guidelines ().100 HPV testing is first conducted to identify those with an existing high-risk HPV infection who would be triaged for immediate cytology. In those with abnormal cytological findings, immediate referral for colposcopy would be made, whereas in those with a negative cytological finding, they would undergo repeat screening by HPV test at 12 months. If the repeat test returned as negative, patients would return to routine screening; if persistent high-risk HPV was detected in the repeat test, individuals would be referred for colposcopy to rule out the possibility of a high-grade lesion.101
Management of “Strategy C” Outcomes — Primary HPV With Cytology Triage in HPV-Positive Results. ASCUS = atypical squamous cells of undetermined significance; ASC-H = Atypical squamous cells, cannot exclude HSIL; hr = high-risk; (more...)
Although the clinical experts consulted in this review noted that the management may differ in screening strategies that also include HPV genotyping, this was not modelled in the current analysis given that the Clinical Review found limited clinical data on its DTA.
Perspective
The perspective of a Canadian Ministry of Health was adopted, consistent with CADTH guidelines for the conduct of economic evaluations.97 As such, direct and indirect medical costs were captured, including the cost of laboratory and diagnostic tests, emergency visits, in-patient visits, and medical services. Indirect non-medical costs, such as productivity losses and out-of-pocket costs, were not considered in this analysis.
Decision-Analytic Model
Given that the benefit of cervical cancer screening is to detect patients with precancerous cervical lesions who can be treated before it progresses to cervical cancer, the economic model covered the full clinical spectrum from screening to diagnosis to treatment. A hybrid model was developed with two components: 1) a state-transition microsimulation that reflects the natural history of disease and 2) a decision tree that captures the impact of screening and modified the disease pathway.
Natural History Submodel (Epidemiologic Submodel)
The natural history submodel was loosely adapted from an existing Canadian decision-analytic model.72 Although the modelling approach in the original publication was a Markov cohort model, this was converted to a microsimulation to permit more flexible modelling of how an individual’s clinical history and past screening results can impact the natural history and epidemiology of HPV infection, cervical lesions, and cervical cancer, and how they are managed within a cervical cancer screening program. Furthermore, deviating from the original model, only high-risk HPV infections (i.e., oncogenic strains) were modelled in alignment with the scope of this review. In the original model, low-risk and high-risk HPV infections were considered independent and mutually exclusive; this was felt to not align with current evidence in which coinfection by both low-risk and high-risk HPV strains is possible.
Although the protocol for this study stated interest in two types of cervical cancer (i.e., SCC and adenocarcinoma), the Clinical Review found limited literature supporting the DTA of HPV and cytology tests in detecting precursor lesions of adenocarcinomas. The original scope of the project was therefore narrowed to focus solely on the impact of screening on SCC, which is estimated to represent from 70%6 to 90%102 of all cervical carcinomas. This meant that the potential cytological outcomes of atypical glandular cells, which is a precursor lesion to adenocarcinoma, was largely ignored in terms of how it could potentially influence patient management.
The epidemiological submodel captured the natural history of HPV infection and the potential development of cervical carcinoma. Distinct health states were defined that represented HPV infection, precancerous cervical changes, and cervical cancer (). At the start of the model, all individuals were clear of an infection and had no prior history of cervical cancer (defined as healthy). Each year, age-dependent probabilities for death (all cause) and total hysterectomy unrelated to cervical dysplasia were applied. If either of these events occurred, the individual would not be considered at risk of developing cervical cancer. In the case of those who had undergone total hysterectomy, the model would then estimate the expected life expectancy of that individual.
Disease States and Allowed Transitions for the Natural History Component of the Cervical Cancer Epidemiology Model. CIN= cervical intraepithelial neoplasia.
Within the epidemiological submodel, age-dependent risk of acquiring high-risk HPV infection was applied once an individual was sexually active. Over annual cycles, high-risk HPV infections could be transient as the infection can clear spontaneously (i.e., returns to the healthy state) or persist and develop into precancerous abnormalities of the cervix. The severity of precancerous lesions reflected histological classification. Although both a two- and three-tiered classification system exist in clinical practice, the two-tier classification (i.e., CIN1, CIN2+) was selected given growing concerns regarding the poor differentiation in the diagnosis of CIN2 and the growing belief that CIN2 is not a distinct clinical entity but rather a heterogeneous mix of CIN1 and CIN3 lesions.103 Furthermore, this reflected current treatment guidelines in which clinical management of precancerous lesions is based upon the two-tiered system.101 High-risk HPV infection can progress to either CIN1 or CIN2+. These lesions may spontaneously regress to a lower severity, clear completely, or progress to more serious abnormalities. Clearance of a CIN lesion, either spontaneously or through treatment, may lead to the development of HPV-immunity whereby the individual is not at future risk of acquiring HPV infections. Lesions were assumed to be detected only by screening; undetected and untreated CIN2+ lesions can progress toward cervical cancer.
Once cervical cancer developed, regression was not possible. The natural progression from cervical cancer was described by four stages based upon the International Federation of Gynecology and Obstetrics staging system (stage I = local; stage II-III = regional; stage IV= distant).104 Cancer progression was assumed to be sequential and unidirectional with asymptomatic cancer possibly developing into more severe stages. Cancer detection was either made possible from the presence of symptoms or through the outcome of routine screening and, upon diagnosis, patients would receive cancer treatment in alignment to existing clinical practice guidelines depending on their cancer stage.105,106 Treated cases were tracked during the first five years post-cancer given the increased mortality risk of these patients compared with a general population.107 Those who remained alive at five years post-cancer entered a cancer survivor health state. In this health state, individuals were assumed to have a life expectancy identical to an age-matched general population (i.e., mortality rates of cancer survivors beyond the first five years of treatment were assumed identical to those of a normal population).
Details on the value of the clinical inputs to the natural history of the condition can be found in the Clinical Parameters section.
Screening Model
The screening model was applied to the epidemiological model when an individual was eligible and participated in screening. Eligibility to participate in programmatic cervical cancer screening was based on the screening algorithm being evaluated (i.e., age range and screening frequency) and certain criteria in existing Canadian clinical guidelines.22,99 Specifically, patients who had undergone total hysterectomy unrelated to cervical dysplasia or who had not engaged in sexual activity were assumed ineligible for screening.
The screening algorithm reflected the screening strategy described under Interventions. Individuals who did not participate in screening (i.e., missed) would continue to be modelled in the epidemiological model but may return at any time to routine screening before their next scheduled screening visit.
Progression through the screening model is dependent on an individual’s health state within the epidemiological model at the time of screening (e.g., healthy, precancerous lesion, cervical cancer), the diagnostic performance of the screening tests and the individual’s adherence to the clinical management associated with screening (e.g., proportion of positive screens not lost to follow-up). For instance, in an individual with no histologic abnormalities (i.e., less than< CIN1), a result of ASCUS or worse on cytology would be considered a false-positive. However, in subsequent years, they may be infected with HPV and develop CIN2+ lesions. If screened again, a result of ASCUS or worse on cytology would be considered a true-positive.
As noted in both the Clinical Review and in the description of the screening strategies, results of cytology were based on the Bethesda classification system. shows how histological health states used in the epidemiological model were mapped to cytology outcomes.103
Correspondence of Cytology and Histological Diagnostic Terms.
In terms of the clinical management of a cervical abnormality identified from screening, this may include repeat screening or histological assessment. Repeat screening was modelled similarly to the abovementioned, but reflected the increased frequency of screening. In the case of cytology only (strategy A) and cytology with HPV triage (strategy B), screening occurred every six months and routine screening would resume after two consecutive negative results. In the case of HPV with cytology triage (strategy C), a repeat screen would be given a year after and routine screening would resume if the repeat screen produced negative findings.
Differences in the test characteristics and the order in which the screening tests were applied, alongside the natural epidemiology of an individual, therefore permitted the model to generate a different set of costs and health outcomes based on the screening algorithm that formed the basis of the comparative analysis.
Representation of the Decision Tree Capturing the Screening Algorithms. CC = cytology and colposcopy; AGC = atypical glandular cell; ASCUS = atypical squamous cells of undetermined significance; CIN = cervical intraepithelial neoplasia; LSIL = low-grade (more...)
Management of Abnormal Screening Outcomes
Based on the screening strategy, individuals with an abnormal cervical screening test may be referred for colposcopy to determine individualized management. Most published economic models have assumed that colposcopy and/or biopsy are the diagnostic gold standard for confirming the presence and grade or severity of CIN and cervical cancer and this was similarly assumed in this model. Biopsy-confirmed cervical disease was defined based on the histological classification of the cervical lesions or cancer (). Within the model, performing a colposcopy on an individual who in fact has no histologic abnormalities as per the epidemiological model (i.e., representing a false-positive screening test) would result in negative colposcopy findings and lead to appropriate workup. However, in an individual with CIN2+ lesions (i.e., representing a true-positive screening test), colposcopy and biopsy would lead to positive colposcopy and biopsy and subsequently determine the individual’s appropriate clinical management.
The clinical pathway of colposcopy and/or biopsy were modelled according to current clinical practice guidelines ().99,101 Referral to colposcopy in which the biopsy results were CIN1 or less would be managed conservatively with annual follow-up visits where colposcopy and cytology would be performed over three subsequent years. In individuals with cytology findings greater than LSIL during follow-up care, a biopsy would be performed to determine appropriate clinical management. Otherwise, individuals would be eligible for discharge from colposcopy following persistent normal, ASCUS, or LSIL cytological findings at the last follow-up visit. Prior to exiting colposcopy, high-risk HPV testing would be performed during the last follow-up visit to provide an objective risk assignment to inform the screening frequency upon discharge from colposcopy. Individuals with a negative HPV test were considered low risk and discharged to routine screening, whereas individuals with a positive HPV test were considered at elevated risk and discharged for annual surveillance in primary care for another three years.108
Management of Individuals With Abnormal Cytology Smears Who Proceed With Colposcopy/Biopsy. CC = cytology and colposcopy; CIN = cervical intraepithelial neoplasia; colpo = colposcopy; cyto = cytology; LEEP = loop electrical excision procedure; mo = month. (more...)
The clinical pathway to manage abnormal colposcopy with biopsy-confirmed CIN2+ findings was dependent on the individual’s age. Those 25 years or older would undergo loop electrosurgical excision procedure (LEEP). In the case of a successful LEEP, this would change the individual natural history within the Markov epidemiological model as they would return to a healthy state given the removal of the cervical dysplasia. Post-treatment SIL management would entail colposcopy and cytology six months after with both tests repeated a year thereafter. Similar to the outcomes of follow-up visits for less than CIN1, individuals with cytology findings greater than LSIL during these follow-up visits would have a biopsy performed to reassess appropriate clinical management. Otherwise, if cytological findings are equal or less than LSIL throughout the follow-up visits, individuals would be discharged from colposcopy. Prior to exiting colposcopy management, high-risk HPV testing would be performed at the last follow-up visit to guide the screening interval in primary care in the same manner as above.108 Conservative management with biannual colposcopy for two years would be offered to individuals with abnormal colposcopy with biopsy-confirmed CIN2+ findings if they were under the age of 25 years. If spontaneous resolution is observed at end of these follow-up visits, individuals would be discharged from colposcopy. If the individual turns 25 years of age during the follow-up period, clinical management would be reassessed based on the findings of the last colposcopy procedure (i.e., less than CIN1 would result in more frequent screening, while CIN2+ would receive LEEP). 108
While an individual is under management by colposcopy, if invasive cancer is detected, the individual would exit colposcopy management and enter clinical management by regional cancer programs.
Colposcopy may lead to a diagnosis of cervical cancer and result in more timely treatment management at an earlier stage of the disease. Re-referral to colposcopy would be based on routine screening results.
The model was developed in Microsoft Excel 2010.
Clinical Parameters
Natural History
Epidemiology of HPV Infection, Precancer Lesions, and Cervical Cancer
Natural history parameters on HPV infection, precancerous lesions, and cervical cancer are described in .
Natural History Parameters — Annual Values Unless Otherwise Specified.
The simulated population was assigned to a level of sexual activity from low (0) to high (3) corresponding to the expected number of lifetime partners. Proportion of individuals by sexual activity level were based on a calculation of PISCES data performed by Brisson et al96 with the total proportions (by summing the proportions in the four levels) scaled to 1. The age of onset of sexual activity was determined based on an age and sexual activity level-specific rate of sexual activity initiation among females. The rate of onset was determined by fitting to the data from the Canadian Community Health Survey on the percentage of girls who had ever had sex ().96
Sexual Activity Parameters — Proportion of Individuals by Sexual Activity Levels and Rate of Onset of Sexual Activity.
Since estimates on the test performance of an HPV test are conditioned on underlying histology rather than the HPV’s strain, the model made no distinction between different strains of high-risk HPV. Incidence, progression, and regression estimates therefore represent averages for all viral types.
Estimates on the annual incidence of high-risk HPV infection were based on an epidemiological modelling study prepared for the U.S. Preventive Services Task Force.109 In that study, annual age-specific incidence rates were back-calculated in order to produce incidence rates that aligned with several reported HPV-prevalence studies conducted prior to the introduction of HPV vaccination.109 Although a study of a longitudinal cohort of women aged 15 to 49 years whom were recruited from physician practices in Ontario could have provided Canadian estimates,110 this study was not selected for a number of reasons. This study followed up on 253 of 500 previously HPV-negative women recruited from a prior prevalence survey. Incidence estimates were derived from a small sample size and the incidence of HPV infection beyond the studied age range is not clear. However, as this is one of few Canadian studies identified that reported age-specific annual HPV incidence rates, a sensitivity analysis was conducted with these numbers. The incidence of HPV infection was independent of the individual’s sexual activity level. Clearance of high-risk HPV infections was based on the abovementioned US model in which the annual probability was calibrated based on Surveillance, Epidemiology and End Results data.109
It is difficult to directly estimate the progression and regression between high-risk HPV, CIN1, and CIN2+ from published literature given the variation and differences between study designs, follow-up intervals, performance of screening in detecting cervical lesions, and protocols to manage abnormal results. As such, the progression and regression of CIN lesions were taken from a recent US model that based these values on both a review of the literature and calibration of data to observed clinical event rates. The estimates reflect an average for all types of high-risk HPV strains. Although CIN has been historically viewed as a continuum with progression from HPV infection to CIN1, CIN2, and CIN3 assumed to occur slowly over decades, recent understanding of the disease suggest that, among younger women, a different disease progression may be more appropriate given higher disease burden. Specifically, younger women can develop a CIN2+ lesion within a short period of time (i.e., less than 2 years) with most regressing and only a small proportion progressing. Some of the parameters on progression and regression of high-risk HPV, CIN1, and CIN2+ therefore were age-specific.109 The only parameter that differed from that model was the progression from CIN2+ to cancer as their estimate was specific to CIN3 health state. The annual rate of progression was instead taken to be 0.18%.111 The values were confirmed by consultation with the clinical experts involved in this review.109 The transition from high-risk HPV infection to CIN1, CIN2+, and cervical cancer stage were obtained from previous economic models and are reported in .
The effectiveness of LEEP was based on meta-analysis112 that further incorporated the rates of success reported from a clinical study that was not part of the original meta-analysis.113
Canadian age-specific rates of hysterectomy unrelated to cervical cancer (i.e., hysterectomy for reasons other than cervical cancer) were applied.96 In these individuals, no further screening was assumed necessarily (i.e., not at risk of developing cervical cancer); therefore, these individual do not accumulate further costs related to screening and the model estimated their overall life expectancy.
Once an individual develops cervical cancer, asymptomatic or undetected cancer can progress to more severe stages of the disease. As reported in past economic evaluations, there is limited direct clinical data to inform the rate of progression from localized cervical cancer to distant cancers and the proportion of cervical cancers that presents symptomatically. We therefore adopted an approach taken in past economic evaluations78,114 whereby the distribution of cervical cancer cases, by disease stage, in an unscreened population was assumed to be a function of both the rate of disease progression and the probability of symptomatic presentation. The progression rates between cancer stages and the probability of symptomatic presentation was varied to calibrate against reported distribution of cervical cases, by stage, in cervical cancer patients who have never been screened.115–118
Mortality
Baseline mortality rates were informed by female age-specific mortality rates from Statistics Canada’s lifetables122 and adjusted to remove age-specific cervical cancer mortality.123 With respect to cervical cancer patients, stage-specific cervical cancer mortality rates were applied based on the reported five-year observed survival post-diagnosis from the Surveillance, Epidemiology and End Results database.107 As the reported data were based on the TNM staging system, this was mapped to the International Federation of Gynecology and Obstetrics classification as follows: localized cervical cancer corresponded to stage I, regional cervical cancer corresponded to stages II and III and, lastly, distant cervical cancer corresponded to stage IV within the model. It was assumed that there would be no cancer-related mortality after five years post-diagnosis and baseline mortality rate would be appropriate in these survivors. As per the original model, it was assumed that individuals with asymptomatic and untreated cervical cancer had 1.03 times the risk of death compared with individuals who were diagnosed and treated for their cervical cancer. Parameters relating to mortality are summarized in .
Diagnostic Accuracy
The characteristics of each screening test (e.g., sensitivity and specificity) were taken from the Clinical Review. In brief, diagnostic test accuracies were based on pooling sensitivity and specificity based on a bivariate model that assumed perfect reference standards (Appendix 9). The output of the analysis included the hierarchical summary receiver operating characteristics curve, which described the joint distribution between sensitivity and specificity in order to support probabilistic analysis while preserving the correlation between these two DTA parameters ().
Diagnostic Test Accuracy.
Given the low proportion of unsatisfactory samples, the model disregarded non-confirmatory outcomes given that such a finding would mean a return for a repeat screen in which the cost of an additional screen would be minimal.
The distribution of cervical abnormalities among Canadian women undergoing cytology was inferred from pooling the findings of multiple published clinical studies and reported in . For instance, within the general population (i.e., primary cytology and cytology triage screening strategies), individuals with CIN1 in whom abnormal cervical lesions were observed, 40.5% would be categorized as an ASCUS, 42.6% would be categorized as LSIL, and 16.8% would be categorized as HSIL by cytology.124–126 The distribution of cervical abnormalities by cytology would be different within a subset of high-risk HPV+ individuals. For instance, among high-risk HPV+ individuals with CIN1 in whom abnormal cervical lesions were observed, 38.1%, 50.7%, and 11.2% would be categorized by cytology as an ASCUS, LSIL, and HSIL, respectively.58,127
Distribution of Cytological and Histological Findings.
Adherence and Coverage of Screening Programs
There are three sources of nonadherence within a screening program: 1) non-participation to programmatic screening, 2) screening less frequently than recommended (i.e., underscreening), and 3) loss to follow-up of abnormal results. The economic model captured all three aspects together.
For the first source of nonadherence, age-stratified screening rates were utilized and reflect the combined participation rate of cytology among those eligible for screening (i.e., corrected for hysterectomy), reported in the provinces of Manitoba and British Columbia.128 Although these rates reflect the years from 2011 to 2013, it was assumed that cervical cancer screening participation would remain stable as has been observed when comparing the participation rates from 2004 to 2006 against the rates reported from 2010 to 2012.129,130 In addition, the target participation rate (80%) set by the Canadian Task Force on Preventative Health Care was tested in sensitivity analysis.129
For the second source of nonadherence, among individuals who did not participate in the year they were supposed to be screened (i.e., missed screens), the model permitted these individuals to return to routine screening in between the time intervals of their next scheduled screen. As there were limited data on the rates of return to screening in those who had missed their scheduled screen, it was assumed that the rates of screening in these patients would be similar to the abovementioned age-stratified screening rates in the general population. Upon return to screening, their next screening period would be shifted by the time interval dictated by the screening program. Sensitivity analyses were conducted to explore alternative assumptions to the return of screening in individuals who missed screening.
Lastly, in terms of failure to follow-up, it was assumed that there would be no loss-to-follow-up when conducting the triage test (i.e., undergoing HPV with cytology triage or cytology with HPV triage). This assumption was based on the fact that, if LBC samples were collected, it would permit both tests to be performed without an additional clinical visit as an HPV DNA test requires only the residual liquid following extraction of the LBC sample. For patients whose follow-up procedure after Pap involved repeat testing (e.g., ASCUS or LSIL results), an SR noted lower adherence for additional repeat testing. In particular, in an RCT performed in the Netherlands, only 66.3% of individuals with ASCUS or LSIL were found to have completed repeat testing.131 Age-specific follow-up rates to direct referral for colposcopy immediately after cytology were taken from a national report that summarized the performance of cervical cancer screening in five Canadian provinces.132 In particular, the report noted that more than 70% of individuals in Canada had undergone a colposcopy within a year of an abnormal cytological examination (i.e., AGC, ASC-H, HSIL) except in individuals aged 60 to 69 years old (69%).
Vaccination
To model the potential impact of HPV vaccination in conferring immunity to certain HPV strains, and thereby reducing cervical cancer risks, vaccination was modelled as follows. Cohorts born from 1994 onward have been part of Canada’s publicly funded vaccination programs that was introduced in 2006.95 In these cohorts, the uptake of vaccination was assumed to reflect the reported average rate of 55.92%.95 It was further assumed that vaccination would confer lifelong immunity and reduce risk of HPV infections by 95.5%.133
Utilities
The health effects of cervical cancer screening programs were expressed in terms of QALYs. Baseline age-specific utility values from a general Canadian female population, based on EuroQoL 5-Dimensions-3-Levels questionnaires, were taken from Johnson et al.134
Given that, in the literature, no single measurement tool was found to have elicited utility values for all health states associated with screening, diagnosis and treatment of precancerous lesions and cervical cancer that are relevant to the current economic evaluation, two sets of health utility weights were considered (). The following assumptions were made in estimating utilities in the model. Individuals vaccinated or who have undiagnosed health conditions (i.e., HPV infection, cervical dysplasia or cancer) would have a similar utility weight to the general population. Disutility for short-term events such as repeat screening due to low-grade cervical dysplasia or undergoing colposcopy evaluation for a false-positive test results were not considered in the model. A sensitivity analysis was conducted that incorporated a utility weight to abnormal screen results that led to repeat screening or referral to colposcopy.
Description of Utility Weights Within the Economic Model, Estimated by Standard Gamble Technique.
In the model, relevant health state utilities from screening or from a diagnosed condition were adjusted by an age-specific general utility values using a multiplicative approach.
Utility Weight — Reference Case
Utility weights associated with histologically confirmed CIN1 or CIN2+ were based on an Australian study. In this study, utility weights associated with different screening outcomes in 43 women undergoing cervical cancer screening were elicited by a two-stage Standard Gamble technique.135 Median utility values reported in the publication were incorporated into this economic evaluation.
Few studies were identified from the literature search that have elicited utility for cervical cancer based on a Standard Gamble technique in female-only participants. As such, utility weights elicited from a general Korean population (including male and female) that used the Standard Gamble technique were taken. The study elicited utility weights for different treatment approaches of cervical cancer.136 To map the treatment approaches to cancer staging, it was assumed that patients with stage I cervical cancer would be managed by surgery only (i.e., cone biopsy or hysterectomy), patients with stage II to III cervical cancer would be managed by radical hysterectomy+radiotherapy±chemotherapy), whereas patients with stage IV cervical cancers had utility values corresponding to chemotherapy. Although this Korean study also reported utility weights relating to cervical neoplasia, it was not incorporated into the model given more appropriate utility values elicited specifically in females were available. However, incorporating utility weights from this study is likely to have a negligible change in the interpretation of the model findings as the median utility values for CIN1 and CIN2+ (both equal to 0.9) were similar to the utility weights that were tested in the subsequently described sensitivity analysis.
Utility Weight — Sensitivity Analysis
One conference proceeding was identified that reported health state utility values based on the time trade-off approach in US women.137 Utility weights for screening-related outcomes in the model were substituted with the mean utility weights for a variety of cytological and histological health states, where appropriate.
Costing
All costs were based on Canadian data and converted to 2017/2018 dollars using the general Consumer Price Index for the year of data collection.139 Based on the perspective of the analysis, only medical costs paid by the Ministry of Health were considered. Costs in the analysis are outlined in .
Cost Parameters in Economic Evaluation.
Direct screening costs included those for consumable supplies, office visits, outside hospital diagnostic procedures and professional services. The unit costs of cytology, HPV tests, and related fees were extracted from a variety of sources, including Ontario Schedule of Laboratory Fees and a previously published economic evaluation.84
The average costs of colposcopy with or without biopsy and LEEP included the physician’s professional fees and associated costs of the procedure, including any laboratory fee for the biopsy specimen. Physician fees, including those of related to pathology, were taken from the Ontario Schedule of Benefits,140 while procedure-related costs were obtained from the Ontario Case Costing Initiative that assumed these procedures would be performed in an ambulatory setting.141 It was assumed that biopsies, if required, would be conducted concurrently with the colposcopy procedure and this may have underestimated to the actual cost as, in some cases, biopsy is performed subsequently.
Upon the presentation of symptomatic cancer, it was assumed that a colposcopy examination would be performed to confirm diagnosis. The average cost of treatment for cervical cancer were derived from a cost-analysis conducted in British Columbia that reviewed resource patterns of 563 patients between January 2004 and December 2009 in terms of patient-level resource patterns from diagnosis to death or five-year discharge.105 Cancer-related medical costs were applied up to the cancer survivor’s lifetime.142
Statistical Analyses and Sensitivity Analyses
The reference case reflects the probabilistic results based on running 10,000 individuals through the model over 10 runs. Three specific single birth cohorts (i.e., ages 9, 20, and 30) were evaluated. The probabilistic results characterize the extent to which parameter uncertainty impacts the cost-effectiveness estimates in the model. Standard distributional forms were taken to describe the probability distribution functions relating to input parameters: transition probabilities and relative risks were characterized by beta and normal distributions, utilities were characterized by beta distribution, and costs were characterized by gamma distributions. Where possible, the diagnostic test accuracies of the screening tests (i.e., sensitivity and specificity) were sampled from the joint distribution function described by the hierarchical summary receiver operating characteristics curve.
The ICUR was calculated according to convention and, in most cases, the sequential ICUR was presented unless otherwise specified. Strategies that were dominated (i.e., another strategy that has lower expected costs and higher expected QALYs) or “extended dominated” (i.e., at least one possible combination of two treatment strategies exist that would be less costly and result in higher QALYs) were identified. Results of the probabilistic analysis are presented on a cost-effectiveness acceptability curve that highlights the screening programs on the efficiency frontier (i.e., the set of optimal strategies that, for varying costs, produce the highest health benefits). This graph presents the probability that each screening program is optimal given different willingness-to-pay values for an additional QALY gained. In addition, the model’s predicted impact in terms of health care resources (e.g., number of colposcopies performed) required under each specific screening program was estimated and presented. Similarly, clinical outcomes associated with each screening program were reported.
Further sensitivity analyses were conducted to evaluate the degree to which uncertainty in the model parameters (i.e., parameter uncertainty) and uncertainty in its assumptions (i.e., structural uncertainty) would impact the results. These include:
Vaccination uptake: The uptake rate of vaccination came from a pooled estimate from a Canadian review. However, within the same review, it was noted that the uptake rate can range broadly within Canada ranging from 12.40% to 88.20%.95 Sensitivity analyses were conducted across this range.
Discount rate: The reference case was based on a discount rate of 1.5%,97 with sensitivity analyses conducted by applying a higher discount rate of 5% and an undiscounted scenario (i.e., discount rate = 0%).
Incidence rate of HPV infection: Incidence of acquiring an HPV infection was based on an Ontario study by Sellors et al.110
Missed screening: The reference case analysis assumed patients who miss screening in the index year can return to screening between screening intervals according to reported age-specific screening rates.128 A sensitivity analysis was conducted that assume patients would not return to screening until their next scheduled screening period.
Screening participation rate: Screening participation rate was based on observed data. A sensitivity analysis was conducted in which participation rates were set to the current targeted rate of screening (80%) that was set by the Canadian Task Force on Preventative Health Care.1
Alternative utility weights based on a different elicitation tool: The reference case’s health state utilities were elicited by the Standard Gamble approach. A sensitivity analysis was conducted in which utilities weight for diagnosed cervical lesions (i.e.,CIN1, CIN2+) were elicited by the time trade-off method.137
Disutility from abnormal screening results requiring repeat testing or false-positive findings: In the reference case, no disutility was associated with screening results that led to repeat testing or additional follow-up visits if no cervical abnormalities were detected by colposcopy and biopsy. Rather, age-adjusted baseline utility values were applied in such instances. A sensitivity analysis was conducted that applied a lowered utility weight in patients with screening test outcomes that led to repeat testing or entry into colposcopy management regardless if the initial screening tests was a true-positive or false-positive finding. Mean utility score for the following screening outcomes were applied: cytology findings equal to or under LSIL (0.9996), cytology findings equal to or under LSIL with normal colposcopy (0.9985), HPV-positive with normal cytology (0.9986), HPV-positive with normal colposcopy (0.9987), CIN1 (0.9989), and CIN2+ (0.9983).135 These were applied in the year of the screening results.
HPV costs: Current costs for HPV testing were estimated from a Canadian economic evaluation in which the difference in lab costs between HPV and cytology was $41.74. However, as no real-world Canadian data were found that accurately estimated the associated lab fees for HPV testing per patient, several sensitivity analyses were conducted. A previously published economic evaluation conducted under the province of Quebec suggested that, based on their own personal communications, HPV lab fees could be only an additional $9 more compared with conventional cytology.89 A sensitivity analysis was therefore performed in which HPV lab fees were assumed to be $16.52. Furthermore, a threshold analysis was performed to determine the cost of HPV lab tests whereby the cost-effectiveness of HPV-based screening would be under $50,000 per QALY.
Validation
The model structure and inputs were presented to two Canadian clinical experts to ensure that the model, its parameters, and its assumptions reflected Canadian clinical practice and the available body of literature (i.e., face validity). Internal validity was assessed by ensuring that the mathematical calculations were performed correctly and were consistent with the model specification, and that logical discrepancies were assessed by evaluating the model under hypothetical and extreme conditions. The model further underwent external technical peer review. External validation was conducted by comparing model outputs against independently published studies.123,144–146
Assumptions
lists the assumptions in which the reference case of the economic analysis was based on.
Assumptions Used to Populate the Economic Model.
Results
Validation
A series of external validation tests were conducted, comparing results with independent studies that had not informed the development of the economic model, to assess to what extent the model was able to predict observed outcomes. summarizes the key findings from this exercise.
The deterministic model predicted age-specific onset of sexual activity within a plausible range (). Assuming the current screening involving cytology every three years between the ages of 21 to 69 and real-world adherence to screening, the model predicted an absolute lifetime risk of developing cervical cancer of 0.61%, which is aligned with the reported Canadian lifetime risk of cervical cancer between the ages of 0 to 74 (i.e., 0.6% to 0.66%).147 Age-specific incidence is mostly aligned within the range expected.123 However, the peak of cervical cancer was predicted to occur later in the model, with a median age of 54 years old for cervical cancer.89,128 This is higher than Statistics Canada reports, which is that individuals in their early forties were the highest risk age group for cervical cancer, with a median age of diagnosis at 47 years old.123
Modelled and Reported Proportion of Sexually Active Women, by Age. CCHS = Canadian Community Health Survey.
Hysterectomy rates (18.6%) in the model were underestimated. There may be a variety of causes for this. First, the model specifically captured hysterectomy due to non-benign conditions, whereas the reported data represents all-cause hysterectomy. Furthermore, the available hysterectomy rates were from the province of Quebec, which has one of the lowest rates of hysterectomy.144
Reference Case
The analyses reflect 10 Monte Carlo simulations of 10,000 individuals each.
The model was used to generate data regarding the disease history of HPV infection and cervical cancer in Canada, and to profile the effectiveness of the current cervical cancer screening program. Assuming that the current risk factors for HPV infection remain stable and there is no programmatic screening (i.e., screening coverage in the model is set to 0), the absolute lifetime risk of cervical cancer would be 2.56% (). These projections would reflect the risk of cervical cancer in women who are not involved in any programmatic screening for cervical cancer, a subpopulation that continues to exist despite existing routine screening programs, as noted in the Implementation Review. Without participation in routine screening, cervical cancer lesions can only be detected if cervical cancer is symptomatic. As a result, in the “no screening” strategy, the expected cost only includes the medical costs associated with cervical cancer treatment (including diagnosis upon symptomatic cervical cancer) and cancer survivorship.
Results Comparing No Screening Programs With the Current Screening Program (Unvaccinated Cohort, Starting Age Nine).
Screening programs were found to reduce the burden of the disease. Compared with the most common screening program in Canadian jurisdictions (i.e., Pap cytology every three years between the ages of 20 to 69), the lifetime risk of cervical cancer reduced to 0.82 and this represented a 69.0% reduction in cervical cancer rates compared with a no screening strategy. This reduces the risk of developing cervical cancer from 1 in 40 (no programmatic screening) to 1 in 122 (with programmatic screening). Given incremental QALYs of 0.029 (), this would indicate that the current screening program could increase, on average, 10.6 days in perfect health (discounted) over an individual’s lifetime.
The current screening program was expected to cost $1,531 per person over their lifetime; and compared with no screening, the cost difference was $27 (). The expected lifetime costs associated with programmatic screening was composed of two elements: 1) screening and its associated diagnostic costs (e.g., routine screening, management of for abnormal results — e.g., frequent screening — colposcopy and/or biopsy) and 2) treatment costs for precancer lesions and cost of managing cervical cancer.
In considering all screening strategies of interest, primary HPV testing with cytology triage every five years from the ages of 25 to 69 was found to be the least costly but also the least effective strategy across all cohorts evaluated. In a future incidence cohort (i.e., population with a starting age of nine years), this strategy was associated with expected costs of $1,471 and resulted in 39.956 QALYs over a lifetime. The next strategy on the efficiency frontier (the set of optimal strategies that, for varying costs, produced the highest health benefits) was a screening strategy based on primary cytology every three years from the ages of 21 to 69. The primary cytology strategy would produce an additional 0.005 QALYs at an incremental cost of $551, resulting in an incremental cost-effectiveness ratio (ICER) of $112,717 per QALY gained. This strategy reflected the most intensive screening program among the screening strategies being evaluated as it was associated with the most frequent and longest screening duration. Indeed, compared with the reference strategy (C3) in which, on average, patients participated in 5.8 screens over their lifetime, this strategy was associated with an average of 11.5 screens over a lifetime (even when factoring participation and adherence to screening). All other strategies were either extendedly dominated (i.e., at least one possible combination of two treatment strategies would be less costly and result in higher QALYs) or dominated (i.e., another strategy has lower expected costs and higher expected QALYs). It is important to note that the incremental QALY between screening strategies were low (< 0.01). For instance, between the reference strategy (C3: primary HPV with cytology triage, every five years, from age 30 to 69) and the most clinically effective strategy (A1: primary cytology, every three years, from age 21 o 69), the difference in expected QALY over a lifetime was approximately 0.005, which equates to approximately 1.8 days of full health gained per patient.
The screening strategies on the efficiency frontier differed between the age cohorts evaluated. Different cohorts had different vaccination status. The future incidence and incidence cohort, with a starting age of 9 and 20, respectively, incorporated a partly vaccinated population based on current rates of participation in HPV vaccination programs in Canada whereas the prevalent cohort reflected an unvaccinated population with a starting age of 30. As such, a different set of strategies appeared on the efficiency frontier. However, in all cohorts evaluated, primary HPV testing with cytology triage remained the lowest costs and lowest QALYs strategy. However, the next screening strategy on the cost-effectiveness frontier would at least require a willingness-to-pay threshold greater than $88,163 per QALY gained to be considered cost-effective.
As the Clinical Review noted, HPV testing is more sensitive and less specific. In the economic analysis, this clinical utility translates to a lower lifetime risk of developing cervical cancer for strategies in which primary HPV testing is introduced to the broader population eligible for screening rather than implementing primary cytology screening. However, from a cost-only perspective, in comparing approaches to screening with all other characteristics of the screening program held constant (e.g., strategies A2, B1, and C2), strategies with primary cytology were found to be less costly than the equivalent strategies that involve primary HPV testing (e.g., C2). The higher costs associated with primary HPV with cytology triage were driven by the increased needs for repeat screening by HPV testing and/or cytology and the slightly higher rates for colposcopy.
With respect to the screening frequency, a trade-off was observed between costs and clinical benefits. This was most notably observed in the primary HPV with cytology triage strategies. Increasing the time interval between screens from a three-year to a five-year interval was found to lower costs (due to lower numbers of screening-related procedures performed) but resulted in a higher lifetime risk of cervical cancer as some cases of cervical cancers would not be detected by screening as screening became less frequent. The impact of extending the targeted age range less clear. The expected costs were identical in the prevalent cohort given that the majority of patients entering the model were eligible for programmatic screening at the model start whereas, in the future incidence cohort, patients would not be eligible for screening at the model start given the actual start age of screening would be conditional on the eligible start age for programmatic screening and the individual’s sexual activity status. Although the average number of programmatic screening tests were higher with a lower start age, it was not always clear whether this would translate to clinical benefits in terms of reducing the impact for repeat testing or averting cervical cancer.
Sensitivity Analyses
The results of the sensitivity analyses indicate that the scenarios and parameters in which the cost-effectiveness model responded most sensitively to differed by the population evaluated. The following results are therefore ordered by the population being analyzed.
Future Incident Population
The model was sensitive to the following sensitivity analyses that were conducted on the future incident population:
Rate of vaccination uptake: Although the screening strategies forming the efficiency frontier remained identical to the reference case, the expected vaccination uptake rate impacted the estimated ICER values. In cases when the rate of vaccination was set to the lower bound of the reported 95% CI (12.40%), the overall lifetime risk of developing cervical cancer increased. As such, the expected cost associated with each strategy increased while the expected QALYs reduced. The impact on incremental QALY was larger than the impact on incremental costs, resulting in the ICER for primary cytology (every three years, ages 21 to 69) decreasing to $60,345 per QALY gained (). The results highlight that, with lower rates of vaccination, a more intensive primary cytology screening program may be more appropriate. The contrary observation could be made when the vaccination uptake rates were higher than the reference case values.
Sensitivity Analyses Results for the Future Incident Cohort.
Discounting: If no discounting was applied, (i.e., neutral time preference with respect to present and future costs and benefits), the ICERs reduced for the strategies on the efficiency frontier with more intensive screening (i.e., reducing the frequency, extending the duration) become more economically attractive. Specifically, the ICER associated with strategy A1 (primary cytology, every three years, age range 21 to 69) reduced to $76,279 per QALY gained. When a higher discounting rate was set, primary cytology no longer formed part of the efficiency frontier. Rather, strategy A1 was found to be dominated by strategy C2 (primary HPV with cytology triage, three years, 25 to 69) (i.e., strategy A1 was more costly and less effective than strategy C2), which was associated with an ICUR of $318.294 per QALY gained ().
Disutility from abnormal screen: When a one-year disutility was applied for abnormal screen results (i.e., true-positive and false-positives), this was found to reduce the expected QALYs across all screening strategies. The overall impact of incorporating such a disutility was less for screening approaches that entailed primary cytology with HPV triage compared with other approaches to screening. This was expected as, even in the reference case, this approach was associated with the lowest rates of repeat screening (). Furthermore, given the small differences in QALYs between strategies, incorporating a minor disutility from abnormal screen results (< −0.001) could have an impact on which strategies formed the efficiency frontier. Although the reference strategy remained identical to the reference case analyses (i.e., primary HPV with cytology triage), primary cytology with HPV triage replaced all other screening strategies. In the future incidence cohort, the ICER associated with B2 (primary cytology with HPV triage, every three years, age range of 30 to 69) reduced to $19,547 per QALY gained whereas, in the prevalent cohort, the ICER associated with B1/B2 (primary cytology with HPV triage, every three years, age range of 25/30 to 69) was $14,681 per QALY gained ().
Probabilistic Base-Case Results.
Incident Cohort
The economic evaluation was found to be more sensitive to change under the incident cohort population compared with the future incident population. Across all sensitivity analysis performed, although the reference strategy (i.e., the strategy with the lowest expected costs) remained identical (i.e., strategy C3 — primary HPV with cytology triage, five years; 30 to 69 years old), the ICERs for technologies on the efficient frontier lowered in several of the analyses.
Many of the same previously noted trends in the future incident population could be applied to the prevalent cohort, although there were additional sensitivity analyses in which the model was sensitive to under this modelled cohort:
Discounting: If no discounting was applied (i.e., neutral time preference with respect to present and future costs and benefits), the ICERs reduced for the strategies on the efficiency frontier with more intensive screening (i.e., reducing the frequency, extending the duration) becoming more economically attractive. Specifically, the ICER associated with strategy A3 (primary cytology, every three years, age range 30 to 69) reduced to $58,830 per QALY gained.
Missed screening: Assuming that patients who missed their programmatic screening would not return to screening until their next scheduled screening period made more intensive screening programs appear more favourable. Although the strategies on the efficiency frontier remained identical, the ICER for strategy A1 (primary cytology, every three years, age range 21 to 69), primary cytology reduced to $56,073 per QALY gained ().
Sensitivity Analyses Results for the Incident Cohort.
Prevalent Cohort
The economic evaluation was found to be most sensitive to change under the prevalent cohort population. Across all sensitivity analysis performed, the reference strategy (i.e., the strategy with the lowest expected costs) remained identical (i.e., strategy C3/C4 — primary HPV with cytology triage, five years) and in most cases, strategy C1/C2 (primary HPV with cytology triage, three years) was the most expensive and most effective intervention, the efficiency frontier often included other screening programs. Many of the same previously mentioned trends in the future incident and incident population could be applied to the prevalent cohort, although there were additional sensitivity analyses in which the model was sensitive to under this modelled cohort:
Alternative incidence rate: With a different set of HPV infection incidence rates, all four screening programs emerged on the efficiency frontier (i.e., no screening program was dominated or extendedly dominated). Although the expected costs and QALYs for each screening program were similar to the reference case, this analysis highlights the sensitivity of this cohort to even minor changes in the expected results. Although strategy C1/C2 and strategy C3/C4 remained both the cheapest and least effective, and the most expensive and most effective strategies, respectively, primary cytology and primary cytology with HPV triage both emerged on the efficiency frontier in between the primary HPV with cytology triage strategies.
Utility from time trade-off: Due to the small incremental QALY difference between strategies in this cohort, minor changes to utility sources impacted which strategies formed the efficiency frontier. When utility weights were derived by the time trade-off method, primary cytology with HPV triage emerged to be on the efficiency frontier. Specifically, the ICER for B1/B2 (primary cytology with HPV triage, every three years) was found to be $43,789 per QALY gained, while the ICER for C1/C2 (primary HPV with cytology triage, every three years) remained similar at $144,978 per QALY gained.
Cost of HPV: When the cost of HPV testing lowered, primary HPV with cytology triage became increasingly attractive as the expected costs for these strategies reduced. These strategies therefore formed the efficiency frontier if the lab costs of HPV were to reduce. Under a willingness-to-pay threshold of $52,634 per QALY, strategy C3/C4 (primary HPV with cytology triage, every five years) would be preferred; above this value, strategy C1/C2 (primary HPV with cytology triage, every three years) would be preferred.
Sensitivity Analyses Results for the Prevalent Cohort.
Additional details on the sensitivity analyses in which the model findings were found to be robust can be found in Appendix 16.
Summary of Results
The economic evaluation presented herein demonstrates that programmatic screening of individuals with a cervix remains highly effective for the prevention of cervical cancer, with lifetime cancer risks estimated to reduce from one in 40 to one in 122 individuals in existing screening programs (i.e., primary cytology every three years, between the ages of 21 to 69). The Clinical Review concluded that HPV tests are associated with a higher sensitivity but lower specificity than cytology. At a programmatic level, comparing primary cytology with primary HPV with cytology triage over a cohort’s lifetime translates to a lowered risk of developing cervical cancer in strategies that involved primary HPV testing. Holding all other characteristics of a screening program constant (i.e., frequency, interval), lifetime costs were found to be higher for primary HPV testing than for the equivalent strategies that involves primary cytology, whereas QALY differences between these strategies were small. Expanding to consider other aspects of a screening program, the analysis found that, by reducing the frequency of screening from every three to every five years for primary HPV with cytology triage screening, incremental costs were lower than primary cytology every three years given that fewer programmatic screening tests would be performed while utilities remained comparable.
Although more frequent screening was found to better detect more lesions, there is a trade-off between over-screening and cancer prevention. Indeed, the lifetime QALY difference between screening strategies were found to be small, being at most 0.019 under the reference case (which represents an additional seven days of perfect health per individual) and sensitivity analysis found that the economic model was most sensitive to whether a disutility was applied to abnormal findings. Specifically, more frequent screening may become less favourable as it increases the number of abnormal findings that require clinical management and may lead to overtreatment. Indeed, in comparing primary HPV with cytology triage strategies in which the screening frequency was varied, the reference case found that screening every five years produced slightly lower QALYs than screening every three years; however, when a disutility was applied for abnormal findings, the contrary was observed (i.e., screening every five years produced slightly higher QALYs than screening every three years). This indicates that there are considerable differences between screening programs in the number of repeat testing that would be expected to be performed. Even assuming small one-time disutilities (> −0.001) for abnormal screening results can have a major impact in the economic findings.
Expanding the eligible age range for programmatic screening does avert cases of cervical cancer, although it comes at the cost of detecting transient infections that may lead to unnecessary clinical management and overtreatment. Indeed, the economic analysis only evaluated a starting age of screening of 21 for primary cytology; otherwise, the starting age evaluated for all other screening approaches ranged from 25 to 30 years old.
The Economic Evaluation reflected, as much as possible, Canadian guidelines on the management of screening outcomes, which was extensively validated by clinical experts involved in this review. However, variations in the management may impact the overall cost-effectiveness of a screening program. This was considered outside the scope of this review, which was more focused on comparing different types of tests. Furthermore, these analyses did not compare between different commercial assays of the HPV tests nor the impact from the increasing practice of HPV genotyping to inform clinical management.
