Comparative cost analysis of cervical cancer screening programme based on molecular detection of HPV in Spain

A cost-analysis de novo model was developed using Microsoft Excel^®. The design and structure of the model as well as the parameters required for the development of the analysis were defined by a multidisciplinary expert panel composed of 2 gynaecologists, 3 pathologists, 1 epidemiologist and 3 health economics specialists. A structured questionnaire that included the values identified in the scientific published literature was developed and filled by the expert panel. Subsequently, an expert meeting was carried out to validate and agree upon all the values used in the analysis.

Two scenarios were considered to estimate the costs associated with the CCS for all Spanish populations covered by the screening programme, including the first colposcopy. The initial analysis defined as the base case, was performed in women aged 35–65 years, where the primary screening was performed with HPV testing. An alternative analysis (alternative case) considered the subgroup of women 25–34, where the primary screening was carried out with LBC.

According to the recommendations of the current screening guidelines in Spain [13], and due to the clinical feature of the disease, it was decided to represent the evolution of patients by the CCS through a decision tree model, which was designed following the ISPOR recommendations for modelling a good research practices [21], for the two scenarios (Fig. 1).

The analysis started with the cohorts accessing the CCS. Along the time horizon of the simulation, women were transitioning between the different nodes according to the probability of event occurrence. For this economic analysis, the model was stops after performance of the colposcopy in those women requiring it. The decision tree nodes represented events derived from the findings of their HPV, LBC, colposcopy, and biopsy results.

Furthermore, LBC was considered in the analysis for sample collection, allowing the reflex test to be performed in both scenarios, thus avoiding the collection of a new sample.

The probabilities needed for model feeding were obtained from the available scientific epidemiological publications and clinical trials of HPV testing.

Following the recommendations of the Spanish guidelines for CCS [13, 22], the primary HPV CCS included the Spanish population of women aged 35–65 years (base case), and the cytological screening included women aged 25–34 (alternative case).

Figures of the different populations were obtained from the National Institute of Statistics for 2018 [23]. It was assumed that all women in the target population would be invited to participate in the CCS. In accordance with the AFRODITA study [24], it was considered that 70% of invited women would attend their screening appointment. Therefore, the final population assessed in the model comprised 7,263,529 Spanish women aged 35–65 years and 1,947,925 women aged 25–34 years, assuming that all of them were asymptomatic.

Based on the expert panel recommendations, three molecular detection HPV tests were assessed in the present analysis: one HPV mRNA test [Aptima human papillomavirus assay (Hologic, Inc., San Diego, USA)] and two HPV DNA tests [Hybrid Capture 2 high-risk HPV DNA test (Qiagen, Gaithersburg, MD, USA) and Cobas 4800 HPV test (Roche Molecular Diagnostics, Pleasanton, CA, USA)].

The Hybrid Capture 2 (HC2) high-risk HPV DNA test is considered the gold standard of HPV assays, as its performance was validated in many randomized controlled trials. HC2 collectively detects 13 high-risk (hr) HPV genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68). Aptima human papillomavirus (AHPV) is an in vitro nucleic acid amplification test for the qualitative detection of E6/E7 viral transcript mRNA from 14 h HPV types in cervical smear samples. The hr HPV types detected by the assay are 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68. Likewise, the Cobas 4800 test detects the same 14 h HPV genotypes as the AHPV assay. However, while the Roche assay detects hr HPV DNA, the Aptima assay detects hr HPV oncogenic mRNA expression and is designed to be more specific in identifying clinically significant hr HPV infections that are likely to lead to high-grade squamous dysplasia and neoplasia.

In the model, two different scenarios were compared: scenario 1, AHPV versus HC2; and scenario 2, AHPV versus Cobas 4800. Although some of these HPV tests give, in addition to the overall result for all hr HPV types including (positive/negative) provide the partial result for HPV 16 and HPV 18, this partial result was not considered in this analysis.

A literature review was conducted in Medline through PubMed to identify scientific publications regarding the clinical evidence available in this field and to extract the probabilities required for the model. Details of the search strategy applied for the literature review is shown in Additional file 1: Appendices 1 and 2.

An initial selection of 1408 localized references was performed by reading the title and abstract. Subsequently, 80 studies were considered relevant for this analysis and were selected and reviewed in full text. The probabilities of the different clinical data included in the analysis were obtained from those clinical studies considered the most relevants. Among all publications, the studies providing data from direct comparisons between the selected HPV tests were preferred and prioritized as sources for inputs.

In this sense, the prevalence of HPV for base case was obtained from a transversal head-to-head study about HPV tests in a screening population [25]. This study showed rates of 7.5%, 11.50% and 12.40% for AHPV, HC2 and Cobas 4800, respectively (Table 1).

Regarding the proportion of women with abnormal LBC after a positive HPV test result, several studies were identified [26‐28]. These transition probabilities are shown in Table 1.

For the alternative case, the prevalence of women with abnormal LBC was 6.5% [26]. Other transition probabilities for the different nodes considered in the decision tree model for this subgroup of women aged 25–34 years are shown in Table 1.

The analysis was carried out from the perspective of the Spanish National Health System (NHS); therefore, only direct health care costs were included, comprising HPV and diagnostic test costs (LBC, colposcopy, and biopsy costs). Figure 1 describes when and which of each of the health care resources were considered.

Unitary costs were obtained from the available scientific literature or regional public information [22, 29, 30] and from a national database of health care costs [31]. All costs included in the model were expressed in euros for 2019, updating the costs obtained from the literature based on the Spanish general consumer price index, if needed [32].

Table 1 shows the unitary costs of the direct health care resources included in the model.

Several sensitivity analyses (SA) were performed to incorporate the uncertainty into the analysis and to observe the effect of these modifications on the results. The following parameters were varied individually: 1) One-way SA was carried out considering a possible reduction in the women from 35 to 65 years of age who would attend their screening appointment, assuming that 36.3% of these women attended the CCS in private practice [24] and that 44.6% of first screening attendees used the public sector. In all cases, an organized CCS with an individual invitation to the target women was considered. 2) To represent a range for the most plausible results of the clinical data regarding women from 35 to 65 years with a positive HPV result, a univariate SA was performed with the studies showing the largest [16, 33] and smallest [34, 35] differences between the test considered in the analysis. 3) Finally, primary LBC, colposcopy and biopsy unitary costs were modified considering the alternative values identified in the literature [22, 36]. An additional scenario was performed with the upper and lower values of these health resources, obtained by applying the standard deviation (SD) to the mean value (Table 1).

The results of the different univariate SA are shown in Table 2. AHPV resulted the least costly option and reduced the number of tests performed in all the scenarios, confirming the base case results.

The one-way SA that had the highest influence on the results was considering alternative published evidence of the proportion of women with a positive HPV test result [34, 35], followed by the variation in cost based on literature evidence [22].

When the studies with the greatest differences in HPV positivity were used, AHPV saves up to € − 53,970,312 compared to HC2 and € − 78,647,307 compared to Cobas 4800. On the other hand, if we calculated cost savings based on the studies with the smallest differences in positive HPV results, AHPV showed savings of up to € − 17,232,932 and € − 21,935,110 when compared to HC2 and to Cobas 4800, respectively.

Regarding the reduction in women from 35 to 65 years of age who would attend their screening appointment in the public sector (n = 4,626,868 women), the use of AHPV provided a cost-saving of € − 2,537,896 versus HC2 and € − 26,669,032 versus Cobas 4800.

Based on the decrease in primary LBC, colposcopy and biopsy costs, when the upper value of the cost reported in the literature was applied, using AHPV would save up to € − 49,733,682 compared to HC2 and € − 52,257,582 compared to Cobas 4800 (Table 2). The savings are € − 22,865,273 versus HC2 and € − 23,342,515 versus Cobas 4800 when using the lower value of the reported cost (Table 2).

Regarding the SA calculated with the upper and the lower values obtained by applying the SD to the mean value of primary LBC, colposcopy and biopsy unitary costs, the savings ranged between € − 35,899,792 and € − 43,779,630 for AHPV versus HC2 and between € − 37,135,197 and € − 46,598,028 for AHPV versus Cobas 4800.