Uptake Rates of Risk-Reducing Surgeries for Women at Increased Risk of Hereditary Breast and Ovarian Cancer Applied to Cost-Effectiveness Analyses: A Scoping Systematic Review

Simple Summary For women who have tested positive for BRCA mutations, the decision to make use of preventive surgical options, such as risk-reducing mastectomy (RRM) or risk-reducing bilateral salpingo-oophorectomy (RRSO), depends on the women’s personal preferences and the cultural/social context. Among others, the cost-effectiveness of RRM and RRSO can be affected by the uptake rate of these preventive surgical options. Uptake rates of surgery should be given more attention in the conceptualization of health economic modeling studies for RRM and RRSO. Prospective multicenter studies are recommended to reflect regional and national variations in women’s preferences for preventive surgery. Abstract The cost-effectiveness of genetic screen-and-treat strategies for women at increased risk for breast and ovarian cancer often depends on the women’s willingness to make use of risk-reducing mastectomy (RRM) or salpingo-oophorectomy (RRSO). To explore the uptake rates of RRM and RRSO applied in health economic modeling studies and the impact of uptake rates on the incremental cost-effectiveness ratios (ICER), we conducted a scoping literature review. In addition, using our own model, we conducted a value of information (VOI) analysis. Among the 19 models included in the review, the uptake rates of RRM ranged from 6% to 47% (RRSO: 10% to 88%). Fifty-seven percent of the models applied retrospective data obtained from registries, hospital records, or questionnaires. According to the models’ deterministic sensitivity analyses, there is a clear trend that a lower uptake rate increased the ICER and vice versa. Our VOI analysis showed high decision uncertainty associated with the uptake rates. In the future, uptake rates should be given more attention in the conceptualization of health economic modeling studies. Prospective studies are recommended to reflect regional and national variations in women’s preferences for preventive surgery.


Introduction
In recent decades, genetic testing and counseling have evolved to become an essential part of hereditary breast cancer (BC) and ovarian cancer (OC) prevention. Women who are carriers of germline BRCA1 and/or BRCA2 mutations can be offered risk management strategies that can significantly reduce the risk of BC/OC and cancer-related mortality. Risk-reducing mastectomy (RRM) has been shown to decrease the risk of BC and to provide an overall survival benefit for BRCA1 mutation carriers [1]. Risk-reducing salpingooophorectomy (RRSO) decreases the risk of OC and improves cancer-related and overall survival [2], while there are inconsistent results on the impact of RRSO on BC risk in BRCA mutation carriers [2][3][4]. When opting for one or both risk-reducing surgeries, a woman has to weigh the benefit of reducing the cancer risk against potential negative consequences of these procedures, such as the loss of fertility, premature menopause, or psychological and physical suffering [5,6]. Hence, some women might opt either for delaying preventive surgery or for intensive surveillance instead.
The reimbursement of risk management strategies for BRCA mutation carriers depends on their clinical effectiveness for preventing cancer as well as on their economic consequences. To evaluate the lifelong health economic impact of preventive strategies, models are usually applied with several input parameters. These parameters include data on the course of the disease, such as the cancer incidence, the impact of preventive surgeries in reducing cancer risk, costs, and utilities (i.e., health state preference values).
A recent systematic review of health economic modeling studies concluded that targeted screening followed by risk-reducing strategies might be cost-effective. However, the input parameters applied to these models often differed, notably the uptake rates of surgeries (i.e., the women's choice for RRM and RRSO) [7]. The uptake of surgery among BRCA mutation carriers varies substantially around the world. According to data from an international database from 10 countries, the rate of RRM was highest in the United States (50%) and lowest in Poland (4.5%). The uptake of RRSO was highest in France (83%) and lowest in China (37%) [8]. Thus, to some extent, the heterogeneity of uptake rates could be explained by cultural differences across countries [7]. However, the study designs used for measuring women's uptake rates or the strength of recommendation for prophylactic surgeries in the clinical management of BRCA carriers might also explain the observed differences. In addition, there is some evidence that uptake rates of surgeries are sensitive parameters, resulting in potential uncertainty for the model outcomes [7].
To better illuminate how uptake rates are depicted in health economic modeling studies for preventing BRCA-induced cancer and to investigate the relevance of uptake rates for model results, we conducted a scoping review. The objectives of this review were (i) to systematically assess the sources of input data and assumptions for applying uptake rates of surgery within cost-effectiveness modeling studies and (ii) to assess the degree of uncertainty in the model outcomes that may result from different uptake rates in different settings. In addition, we conducted a value of information (VOI) analysis, based on one previously published model, to exemplify the decision uncertainty that results from uncertainty in the model outcomes [9].

Literature Review
The reporting of this scoping review is in accordance with the preferred reporting items for systematic reviews and the meta-analyses extension for scoping reviews (PRISMA-ScR) checklist [10]. The protocol for this scoping review was not pre-registered. A literature search was conducted in MEDLINE (via PUBMED) on 21 September 2022 and the Centre for Reviews and Dissemination (CRD) database to search for health economic modeling studies that addressed women who were offered RRM and/or RRSO after screening for germline BRCA mutations. In addition, we screened the studies included in the most recent systematic reviews published on the topic [7,[11][12][13][14]. The search strategy is provided in the Supplementary Materials (File S1). Two reviewers screened the titles/abstracts of studies and selected potential studies for full text reading. The study selection and data extraction were carried out independently, and, in case of disagreement, consensus was achieved by discussion.
We included cost-effectiveness modeling studies that (i) targeted women at high clinical or familial risk for carrying BRCA mutations or known carriers of BRCA mutations and provided genetic testing for inheritable germline mutations including but not limited to BRCA mutations, (ii) evaluated risk management strategies based on RRM and/or RRSO, and (iii) presented the model outcomes as incremental cost-effectiveness ratios. Studies were excluded if the reporting of the uptake rates was insufficient, or if the uptake rates were based on an assumption of perfect adherence. There was no language restriction.
From the selected models, we extracted the study characteristics (e.g., strategies used for comparison, model population) and the applied rates of the uptake of surgery (and respective age) in the case of a positive gene test result. In addition, we extracted results from deterministic sensitivity analyses to assess the impact of varying uptake rates of surgery on the incremental cost-effectiveness ratio (ICER). In order to assess if the uptake rates were appropriate for the models' target population, the cited sources were retraced, from which we extracted data with regard to the study design, setting, number of participants, and time of follow-up.

Value of Information (VOI) Analysis
Based on a model developed and previously published by our institution [9], we conducted a VOI analysis to estimate whether the costs of additional evidence (e.g., conducting a new study) for reducing decision uncertainty associated with model outcomes are worthwhile. The model, on which the VOI analysis was based, assessed the cost-effectiveness of screen-and-treat strategies for German women at risk of hereditary BC and OC versus no testing. The model had a lifelong time horizon and included the health states 'well', 'breast cancer without metastases', 'breast cancer with metastases', 'ovarian cancer', 'death', and two post (non-metastatic) breast or ovarian cancer states. The perspective of the German statutory health insurance (SHI) was adopted, and input data were predominantly taken from German sources. While the input data concerning uptake rates are reported in Table 1 (i.e., Müller 2018), all input data are reproduced in the Supplementary Materials (Tables S1 and S2) [9].
The expected value of perfect information (EVPI) is computed as the difference in terms of the net monetary benefit (NMB) between the expected value of a decision made with perfect information and the expected value of the decision based on the current evidence [15]. While the EVPI shows the overall uncertainty, the expected value of partial perfect information (EVPPI) determines which parameters are highly related to decision uncertainty and the potential value of reducing that uncertainty by collecting more data on these specific parameters [16].
The 10,000 iterations generated in the probabilistic sensitivity analysis from our model were entered into the Sheffield Accelerated Value of Information (SAVI), which consists of a regression-based method for the EVPI and EVPPI calculations [17]. The value of eliminating parameter uncertainty associated with the uptake rates was quantified in comparison to three sets of other relevant model parameters-utilities (i.e., the qualityadjusted life year values), cancer incidence on BRCA mutation carriers, and risk reduction of preventive surgeries. These parameter sets were chosen due to their relevance in deterministic sensitivity analyses [9]. The NMB, which indicates the value of an intervention in monetary terms, was calculated for a hypothetical willingness to pay EUR 10,000.

Strategies Being Compared
A screen-and-treat intervention comprising BRCA genetic testing (i.e., full sequencing of BRCA genes) followed by RRM and/or RRSO was compared with a no prevention strategy by 10 studies [9,20,21,[23][24][25]30,32,33,35]. A reference model developed by NICE compared testing vs. no testing; in this model, a proportion of women received riskreducing surgery independent of the provision or outcome of testing [18]. In addition, in seven models, risk-reducing surgery was offered to both intervention and controls with differences between the compared strategies: Two models compared two testing strategies, namely full sequencing of BRCA genes versus a 7-or 14-gene panel [19,28], while five studies compared testing women based on familial/clinical risk versus different populational criteria [22,26,27,29,34]. While most studies provided immediate surgery for women who had tested positive, five studies modeled a woman's option to delay surgery [19,22,26,30,33].

Study Population
In eight models, the model population was composed of index patients (i.e., the first person in the family diagnosed with a BRCA mutation after a diagnosis of either BC or OC), followed by cascade testing of first-and second-degree healthy relatives [18,19,21,24,25,30,34,35]. Nine models addressed healthy women at increased risk for BRCA mutations due to familial risk [18,19,21,24,25,30,34,35]. Whereas some studies limited the population to first-or second-degree relatives of women affected by cancer with BRCA mutations [20,27,32,33], others defined the population by an established familial risk (with or without a known mutation in the family) [9,22,23,28,29]. Kwon et al. included only index patients diagnosed with BC at different ages [26].
Women entered the models at different ages, varying from 10 years [35] for siblings and children to 55 years [19] for healthy women and from 40 [26] to 55 years [19] for index patients. In most models, risk-reducing surgery was offered immediately after entering the model, while in five studies, the possibility of delaying surgery was accounted for [19,22,26,30,33].

Uptake Rates Applied to the Models
The uptake rate of RRM applied to the included health economic models ranged from 6% [9] to 47% [34], while those of RRSO varied between 10% [19] and 88% [21]. Figure 2A,B illustrates the variability in the uptake rates of RRM and RRSO for two different age groups, whereas the actual rates are described in Table 1. Whereas in some studies, the uptake rates were obtained from local centers or smaller departments [9,18,21,25,28,30,34], in others, the information was obtained from multicentric studies or registries reflecting larger regions of a country [19,24,27,32,33,35]. The rates were obtained from a single study [19,23,25,28,30,32,33,35] or from multiple studies [18,24,26,29,34]. Uptake rates without providing a reference or based on unpublished data were found in three studies [20,21]. Most studies considered country-specific evidence, with the exception of four [25,29,32,33]. Considerations with regard to the appropriateness of the selected uptake rates for the models' target population were missing in all studies.
With regard to RRSO, the highest uptake was reported by Chai et al. (i.e., 86% of BRCA1 and 70% of BRCA2 mutation carriers under 50 years) [42]. In this study, all women were unaffected by cancer. In contrast, the lowest uptake of RRSO (26%) was reported by a retrospective single-center study that addressed women with previous BC to prevent a recurrent or contralateral BC [51].

Impact of Varying Uptake Rates in Sensitivity Analyses
Most studies provided information about the impact of varying the uptake of surgery in a deterministic sensitivity analysis, except for three [25,27,32]. In all but one of these studies [21], higher uptake rates improved the incremental cost-effectiveness ratio (ICER). Table 3 summarizes the results of the deterministic sensitivity analysis reported by the models.  Varying the uptake rate of RRM from 0% to 50%, the latter reduced the ICER by 10%.
In seven studies, the impact of varying uptake rates on the cost-effectiveness was remarkable [9,21,23,24,26,33,35]. For instance, varying uptake rates changed the costeffectiveness ratio from 20% to 40% [21] to more than 70% [9]. A common aspect of these studies is that the strategies being compared comprise a screen-and-treat intervention versus a no-testing strategy (i.e., no surgery in the comparator arm). In contrast, in six of the modeling studies, the authors considered the impact of varying the uptake rates to be slight (≤10%) or negligible [18][19][20]22,29,34]. In all of these studies, a risk-reducing surgery was offered to both the intervention and the comparator arm [18,19,22,29,34].

VOI Analysis
The overall EVPI per person is estimated at EUR 1680, which is the value of acquiring perfect information (i.e., eliminating all uncertainty) about all parameters applied to the model (detailed in Tables S1 and S2). The EVPPI per person for the predefined parameter sets is shown in Figure 3. The EVPPI value indicates to what extent more information on these sets of parameters would reduce the decision uncertainty (i.e., the chance that the decision-maker incorrectly opts for the strategy with lower payoffs, which, in our model, was the no-testing strategy). The maximum return in terms of the net monetary benefit from removing uncertainty around the uptake rates was EUR 239 (standard error (SE): EUR 24), corresponding to 14% of the total EVPI. The second set of parameters with the highest EVPPI was cancer incidence in the BRCA mutation carriers (EUR 207, SE: 25), followed by the risk reduction of preventive surgeries (EUR 188, SE: 25), and the lowest were the utilities (EUR 154, SE: 27).

Discussion
According to this comparison, the uptake rates of risk-reducing surgeries applied in cost-effectiveness models are sensitive parameters. In the models' deterministic sensitivity analyses, there was a clear trend that a lower uptake rate increased the ICER and vice versa. Considering the vast potential of both RRM and RRSO for reducing the risk of cancer and cancer-related mortality, this is a little surprising. However, in one analysis, the authors reported a slightly higher cost-effectiveness ratio compared to the base case when the uptake of RRM was increased in a sensitivity analysis. The authors explained this counterintuitive effect with high costs for preventive treatment, which were not offset by survival gains (because of the high survival rates in women who do not undergo RRM) [21]. While deterministic analyses demonstrate the model's sensitivity to a single input parameter, the VOI analysis evaluates the uncertainty of multiple parameters simultaneously. By sampling each parameter several times from a given range at each iteration, a more reliable estimation of the uncertainty can be provided, especially in models with parameters that correlate to each other [15]. The VOI analyses indicate the potential NMB forgone by making the decision between two treatment alternatives with current (i.e., uncertain) parameters, in comparison to making the decision with perfect information. As a decision rule for VOI analyses, the cost of future studies to gather more information about uncertain model parameters should not exceed the NMB elicited in the VOI analysis [15].
The high EVPPI of uptake rates indicates that gathering more information about the uptake rates would have a slightly higher impact on reducing decision uncertainty than additional information about other parameter sets (i.e., cancer incidence on BRCA mutation carriers, risk reduction of preventive surgeries, or utilities). Although the VOI calculation reflects the uncertainty in the German model, this finding is likely to be replicated in similar models. The VOI analysis can be easily replicated using regression-based methods based on the iterations generated in the probabilistic sensitivity analysis [17].
The uptake rates applied to modeling studies varied substantially. To reflect the attitudes and preferences of the different target populations, different sources of input data have been chosen for models. This variability might be explained by several factors, such as (i) cultural differences, (ii) individual-related factors, (iii) age-dependent factors, and (iv) an improved acceptance of preventive surgeries over time.
(i) Cultural differences (e.g., perception of health and disease, femininity, autonomy) and the risk of financial and social discrimination might influence the preference for genetic testing and risk-reducing surgery [40]. In a previous systematic review, it was suggested that cultural differences between countries could explain the variability in uptake rates to a large degree [7]. However, according to our results, most models used country-specific data, and there was even considerable variability within countries. For instance, among studies conducted in the UK, the rates of RRM varied from 0.21 [46] to 0.43 [42], while those for RRSO varied from 0.26 [58] to 0.86 [42]. Similarly, studies conducted in the United States showed that the uptake rates of RRM varied from 0.36 [45] to 0.42 [42,50], and those for RRSO from 0.33 to 0.71 [48].
(ii) Individual-related factors are also prone to affecting the preferences of women towards RR surgeries. Individual factors that increase the uptake of RRSO include a personal history of BC [49,50,52], parity [44,46,52], and a woman's postmenopausal status [52], while the uptake of RRM tends to be higher among both parous women and those who have a first-degree relative with BC [44,46]. In addition, women who had a family history of OC were more likely to undergo any surgical option [50].
Moreover, many of these individual factors are (iii) age-dependent. While women who have tested positive for BRCA1/2 should consider an RRSO by the age of 35 or right after completion of childbearing [59,60], a prospective study shows that the usage of RRM and RRSO occurs later than recommended [42]. The proportion of women that opt for a risk-reducing surgery increases after age 40 probably because fertility is no longer a concern, and the cumulative risk of cancer is more paramount [42]. Accordingly, the uptake rates applied by the models were, in general, lower for women younger than 35 years, with the lowest uptake rate applied for RRSO (10%) in women younger than 35 [19].
Finally, recent evidence indicates (iv) an improved acceptance of preventive surgeries over time. A reason for this trend could be the improvement in genetic counseling protocols and the cross-center knowledge transfer [61]. Increased uptake rates of risk-reducing surgeries over time due to improved adherence have been observed for the uptake of RRM in women with BRCA mutations, while the uptake of RRSO remained stable [58]. Nevertheless, the trend for RRM was not confirmed by the modeling studies included in our review.
As a limitation of this literature review, it should be acknowledged that there was no protocol registration, and a critical appraisal within sources of evidence was not conducted. Furthermore, because sources of uptake rates were identified only if used for a costeffectiveness model, it is not possible to draw firm conclusions on temporal, regional, or cultural trends or individual factors. To evaluate these relationships more precisely, a comprehensive literature review of observational studies has to be performed. However, our review could demonstrate how sensitive models were when depicting the complexity inherent to the uptake rates.
The usage of outdated sources of evidence for decision-making carries substantial uncertainty regarding the payer's outcomes. The improved counseling for BRCA mutation carriers in recent years might have gradually reduced women's reluctance in opting for risk-reducing surgery, resulting in higher uptake rates. Hence, country-specific, prospective, multi-center studies including post-testing counseling with respect to age and subgroups should be performed to reflect the current status of women's preferences for or against surgical prevention. However, as long as updated evidence is not available, modelersat least in some countries-have to rely on data obtained from retrospective surveys, cross-sectional studies, or medical records without accounting for follow-up. In this case, assessing uncertainty associated with the uptake rates applied to the models is of utmost importance to provide the decision-maker with a realistic assessment of the economic consequences when adopting a screen-and-treat strategy for women with BRCA mutations.

Conclusions
The uptake rates of risk-reducing surgeries applied to modeling studies assessing the cost-effectiveness of screen-and-treat strategies vary considerably. Uptake rates of surgery are associated with high uncertainty, especially in modeling studies comparing a screenand-treat intervention versus a no-testing strategy. Country-specific and prospective studies including non-directive counseling should be performed to reflect women's preferences for or against surgical prevention and would provide a stronger evidence base for economic modeling studies.