Preferences for Genetic Testing to Predict the Risk of Developing Hereditary Cancer: A Systematic Review of Discrete Choice Experiments

Background Understanding service user preferences is key to effective health care decision making and efficient resource allocation. It is of particular importance in the management of high-risk patients in whom predictive genetic testing can alter health outcomes. Purpose This review aims to identify the relative importance and willingness to pay for attributes of genetic testing in hereditary cancer syndromes. Data Sources Searches were conducted in Medline, Embase, PsycINFO, HMIC, Web of Science, and EconLit using discrete choice experiment (DCE) terms combined with terms related to hereditary cancer syndromes, malignancy synonyms, and genetic testing. Study Selection Following independent screening by 3 reviewers, 7 studies fulfilled the inclusion criteria, being a DCE investigating patient or public preferences related to predictive genetic testing for hereditary cancer syndromes. Data Extraction Extracted data included study and respondent characteristics, DCE attributes and levels, methods of data analysis and interpretation, and key study findings. Data Synthesis Studies covered colorectal, breast, and ovarian cancer syndromes. Results were summarized in a narrative synthesis and the quality assessed using the Lancsar and Louviere framework. Limitations This review focuses only on DCE design and testing for hereditary cancer syndromes rather than other complex diseases. Challenges also arose from heterogeneity in attributes and levels. Conclusions Test effectiveness and detection rates were consistently important to respondents and thus should be prioritized by policy makers. Accuracy, cost, and wait time, while also important, showed variation between studies, although overall reduction in cost may improve uptake. Patients and the public would be willing to pay for improved detection and clinician over insurance provider involvement. Future studies should seek to contextualize findings by considering the impact of sociodemographic characteristics, health system coverage, and insurance policies on preferences. Highlights Test effectiveness and detection rates are consistently important to respondents in genetic testing for hereditary cancer syndromes. Reducing the cost of genetic testing for hereditary cancer syndromes may improve uptake. Individuals are most willing to pay for a test that improves detection rates, identifies multiple cancers, and for which results are shared with a doctor rather than with an insurance provider.


Introduction
Efficient allocation of resources is fundamental to the success and sustainability of health services.Concurrently, there is increasing recognition of the value in incorporating public (taxpayers or insurance premium payers) and patient (care recipient) preferences into health care decision making. 1 Including these preferences in the design and configuration of health services could maximize utility, patient satisfaction, adherence, and uptake while simultaneously improving the efficiency of services. 2,3ereditary cancer is caused by the inheritance of genetic variants that confer a higher risk of developing malignancies. 4High-penetrance alleles have been implicated in the pathogenesis of several hereditary cancer syndromes including hereditary breast and ovarian cancer, 5,6 Lynch syndrome, 7 and familial adenomatous polyposis. 8For individuals with a family history suggestive of a hereditary cancer syndrome, testing for high-risk alleles can guide treatment management decisions such as screening frequency or risk-reducing surgery decisions. 9,10The decision to undergo genetic cancer testing comes with several challenges, including frequent co-occurrence with a recent cancer diagnosis, resulting implications for family members, and implications for childbearing.Understanding public and patient preferences is therefore of particular importance given the complexity of the decision-making process.
The consensus is that screening at-risk individuals for high-penetrance alleles associated with hereditary cancer syndromes improves health outcomes, can increase the opportunity for testing in wider family members, and is perceived as a relevant and useful clinical intervention.Nonetheless, the estimated uptake of predictive genetic testing in populations at risk for a hereditary cancer syndrome is only 50%. 11Reasons cited for opting out of testing in high-risk populations include lack of knowledge regarding test availability, distrust of test results, belief that testing is unlikely to alter health outcomes, sense that the negative effects of testing will outweigh the potential benefits, and concerns regarding insurance discrimination. 12,13xisting research highlights the importance of ensuring all at-risk individuals are offered testing and of informing eligible candidates of the potential testing benefits to facilitate informed decision making.It is widely recognized that clinical outcomes affect medical decision making; however, the impacts of process attributes, such as test characteristics and service characteristics, are less well understood. 14Understanding individuals' preferences for relevant predictive genetic testing attributes is vital to stakeholders and service providers to maximize patient and service utility.
Discrete choice experiments (DCEs) are increasingly used in health service research to understand public and patient preferences for health care interventions and in some cases test characteristics (attributes). 15,16In a DCE, respondent preferences are obtained through a series of hypothetical choice sets, each describing an intervention, service, or test with varying attribute levels. 17Respondents' choices are then analyzed to understand the tradeoffs and participant preferences for the intervention, service, or test attributes assessed. 17DCE findings can be used to predict the probability of uptake and (if cost is included as an attribute) provide an indication of willingness to pay (WTP) for services. 17his review adds to the literature and extends recent work by Ozdemir et al. 18 through focused consideration of genetic testing in cancer.It aims to identify the factors affecting the decision to undergo genetic testing for hereditary cancer syndromes by reviewing relevant DCEs, comparing attributes and levels used in each study, and ultimately investigating patient and public preferences for predictive testing for hereditary cancer.

Identification of Studies
Information specialists searched to August 31, 2022, in Medline, Embase, PsycINFO, HMIC, Web of Science, and EconLit using pretested published terms for discrete choice studies. 15DCE hits were combined into a single database in EndNote X9, which was searched using terms related to named adult hereditary cancer syndromes, general terms for malignancy, and genetic testing, as detailed in Appendix A. Forward and backward citation chasing was conducted on included studies.

Study Selection
Titles and abstracts were independently screened by 2 reviewers (from a selection of S.D., N.M., T.S.) to identify DCEs investigating public or patient preferences related to genetic testing in oncology.This included genetic testing for the purposes of risk prediction, pharmacogenetics, and diagnosis.Studies meeting these initial inclusion criteria were reviewed at full text by 2 of the 3 available reviewers (S.D., N.M., T.S.) using the criteria below.
DCEs included at full text were to investigate patient or public preferences related to genetic testing for hereditary cancer syndromes; investigate genetic or genomic testing of constitutional DNA to determine whether the patient is at higher than average risk for developing cancer due to genetic variation; relate to a single gene or polygenic risk score, provided the genes were hereditary; and consider predictive genetic testing (currently unaffected individuals) or diagnostic genetic testing (people with preexisting cancer).
Exclusion criteria were the following: genome testing for personalized cancer treatment conjoint analyses (excluded at both screening stages), and non-English language studies.
Results were summarized in line with PRISMA guidelines. 19ta Extraction and Quality Appraisal Data were extracted on study characteristics (population, cancer type, survey administration), respondent characteristics (age, gender, income, education), DCE attributes and levels, methods of data analysis and interpretation, and key study findings.Reporting quality was assessed using a quality appraisal checklist published by Lancsar and Louviere. 17This checklist does not include a scoring system but considers the transparency of reporting regarding each stage of the DCE without including judgments on whether particular methods improve or worsen validity.Data extraction and quality appraisal were conducted by 1 of 2 reviewers (S.D., N.M.), with all studies checked by a third reviewer (A.M.L.).

Analysis and Synthesis
In each study, the attribute with the greatest impact on test utility was assigned a preference weight of 1, and all other attributes were scaled relative to this attribute by dividing their relative importance score by that of the most preferred attribute, to facilitate cross-study comparison.WTP and test uptake costs were converted to 2022 US$ using International Monetary Fund (IMF) purchasing power parity values.Prices were inflated from the year of data collection or, where not reported, inflated from the year prior to publication.Results are summarized in a narrative synthesis supported by crosstabulation.

Results
Electronic database searches identified 768 studies, of which 621 were unique records following deduplication.Of these, 584 were excluded during title and abstract 254 Medical Decision Making 44 (3)     screening and 27 excluded at full-text review.The 10 studies included at full text reported the results of 7 unique discrete choice studies.Three studies [20][21][22] were excluded to prevent double counting of data as they used the same study sample and attributes as other included studies, leaving 7 studies summarized in a narrative synthesis.Further details are presented in the PRISMA diagram (Figure 1).

Study Characteristics
The earliest included study was published in 2009, 23 while all other studies have been published more recently, from 2015 onward.Three studies were conducted in the United States [24][25][26] and 1 each in the United Kingdom, 23 the Netherlands, 27 Singapore, 28 and Australia. 29][29] Most studies considered a single hereditary cancer: colorectal, 24,25,27 breast, 28 or ovarian. 26One study considered polygenic risk score, 29 and 1 study evaluated the risk of developing genetic breast, ovarian, or colorectal cancer. 23CEs were predominantly administered online. 24,25,27,29ne study opted for a mailed questionnaire, 23 1 From Moher et al. 19 Available from http://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1000097. y The 10 studies that met the inclusion criteria for this review described the results of 7 different discrete choice experiments; therefore, 3 studies were excluded from the narrative review, at full-text screening, to prevent double counting of data.

Morrish et al.
conducted a face-to-face survey, 28 and 1 did not report the mode of administration. 26More details are described in Table 1.

DCE characteristics.
Studies included between 4 and 8 attributes, with a minimum of 2 and maximum of 6 levels.These attributes covered service design, cost, and results/impact.Further detail of individual attributes and levels is provided in Table 2. Respondents were presented with between nine 25,27 and twenty-three 23 choice sets.Four studies used a blocked design, [24][25][26]29 of which 1 randomly assigned respondents to 1 of 4 versions of the survey. 25 ll DCEs were designed using a fractional factorial design with main effects.29 One study used dummy coding of attributes, 29 while all others effectscoded attributes (n = 6), of which 2 also included continuous coding for a selection of attributes.24,28 Data were analyzed using mixed-logit specification (n = 6), 1 of which also used latent class analysis 29 or multinomial logit regression (n = 1).23 Mixed logit specification relaxes the assumption of independent errors, allowing unrestricted patterns of substitution and random parameters account for preference heterogeneity across individuals. 30 Th random parameters approach is particularly appropriate where data show considerable preference heterogeneity; however, it is limited by the need for a larger sample size, which was not necessarily achieved by the included studies.
Respondent characteristics.Respondent characteristics are presented in Table 3.The sample size ranged from 94 respondents 26 to 1,045 respondents, 27 although it clustered around either 100, 300, or 1,000 respondents.The response rate varied from 50.2% 28 to 83% 29 where reported.Participants were generally of older and middle age with mean age ranging from 45 y 23 to 63 y. 25 Studies contained either close to equal distribution of male and female respondents, or a fully female population.Six studies recorded educational attainment.In one study 90% of participants had some form of further education, 24 while in another only 9% had college level training 28 both of which are in contrast with the other studies which were more equally split across different academic levels.

Quality Appraisal
Quality appraisal focused on conceptualization, design, data collection, and analysis of the DCE as guided by the   Two versions of the survey were administered to respondents: in half of these surveys, the risk attribute was either positively framed (survival) or negatively framed (mortality).
quality checklist. 17Further details of the quality appraisal can be found in Appendix B.

Conceptualizing the DCE.
Six studies reported a literature review, clinical expert consultation, and interviews or focus groups with the target population or patient groups used to identify and select attributes.Despite this methodological similarity, the coverage of attributes varied widely across studies.The seventh study selected 4 attributes based on a single published study and reported that the additional 2 attributes, cost and availability of testing, were included to assess the aims of the study. 23ll included studies presented generic alternatives within their choice tasks.Within each choice set, 3 studies included only the testing alternative, 26,28,29 1 presented 2 testing alternatives and an opt-out, 24 2 used forced-choice design followed by an opt-out option, 25,27 and 1 offered an indifference choice between the pair of testing alternatives presented. 23CE piloting was poorly reported across all studies and not reported at all in 2 cases. 23,24Where reported, it was used to check understanding and/or complexity.Only 2 studies reported the use of piloting to check coverage of the attributes and levels. 25,27No study explicitly reported how long it took to complete the DCE during piloting.DCE design.Assessment of experimental design quality was limited due to poor reporting across studies.All included studies used fractional factorial designs and investigated main effects.Two also considered interaction terms. 28,29Only 1 study reported conducting sample size calculations (based on Orme's Rule of Thumb). 28This study was also the only study to provide an incentive to enhance response rates.Three studies reported use of software packages to generate a D-efficient design, 24,25,27 of which 2 used D-optimal algorithms to maximixe efficiency, 24,25 however none of the studies reported the efficiency of the design.One study used an orthogonal array in which each choice set included a constant comparator profile that described current practice at the genetic testing service, which was compared against another alternative. 23nother used orthogonal array and partial profiles with overlapping attributes to improve complexity and efficiency. 29The remaining 2 studies were unclear in their reporting of design efficiency. 26,28ta collection.Three studies recruited patient populations at risk for a hereditary cancer syndrome based  The completion rate was 92% (total number of participants who completed the survey divided by total number who started the survey).
Morrish et al.
on personal and/or family history 23,24 or a diagnosis of epithelial ovarian cancer. 26Four studies recruited participants from the general population necessitating hypothetical contemplation of risk status and testing.This was contextualized to participants through an explanation of gene testing prior to completing the DCE, 28 description of a hypothetical scenario before presenting choice tasks, 29 or incorporation of a hypothetical level of colorectal cancer risk as an attribute to provide a baseline context for each choice. 25,27ta analysis and interpretation.6][27] One study effects coded only individuals with a genetic variant detected, 24 and another effects coded all but the cost attribute. 28The final study dummy coded all attributes. 29One study reported that model specification was selected based on model fit tests. 27A further 2 studies calculated R 2 statistics for goodness-offit and/or calculated Akaike and Bayesian information criteria; however, neither study used them to inform model specification. 23,29One study tested the internal validity of the DCE; however, in this study, responses that violated random utility theory axioms were not excluded. 23ctors Affecting the Utility of Testing Relative importance of attributes.26][27]29 To facilitate between-study comparison, the relative importance of attributes was calculated in the remaining studies as the maximum difference in attribute level utility rescaled so the largest is 1. 23,28 Relative attribute importance is displayed in Figure 2, which reflects the relative distance of all attributes to the most important attribute on a scale of 0 to 1, where 1 is the most important attribute.
Where attributes for cost, service design, and results/ impact were all included, either cost or attributes related to the test result (accuracy, survival, detection) were considered most important by participants.Cost was considered most (or next to most) important in 5 studies [24][25][26]28,29 but conversely of middling or least importance, depending on patient risk level, in another. 23 Liewise, test accuracy, being the chance of a correct (or incorrect) test result, was considered least important in 1 study 25 but of greatest importance in another.29 Regarding results and impact, participants were overall less concerned about attributes related to impact (resulting cancer risk, impact on life insurance, impact on future cancer screening) than those related to direct test results (detection, accuracy, survival).One study considered only cost and service design attributes, 23 identifying wait time for results to be the most important attribute; otherwise, where results/impact attributes were also evaluated, wait time was deemed of low relative importance in the included studies.24,26 The least preferred attributes for service design were observed as test location, 28,29 number of tests, 24 and sample type.26,28 Four studies observed effects across different subgroups as illustrated in Appendix C. Veldwijk et al. 27 considered the effect of negative (mortality) versus positive (survival) framing of the effectiveness outcome. A milar trend in preference was observed across both types of framing; however, survival was more important than colonoscopy frequency under positive framing, whereas colonoscopy frequency was more important than mortality in negative framing.Little variation in relative preference was observed by Davidson et al. 26 when comparing multigene testing to nonmultigene or no testing; however, multigene testing did place greater importance on test sensitivity for mutations and probability of detecting uncertain variants.Griffith et al. 23 found that the relative importance of attributes varied considerably when comparing across different risk groups.All groups valued wait time for results as most important.Meanwhile, the least important attributes varied across risk groups: test availability (high risk), counseling staff (moderate risk), and cost (low risk).Finally, Venning et al. 29 considered relative importance across 3 latent classes.As expected, the ''more is better'' group placed most value on the cancer type attribute, ''aim for accuracy'' considered test accuracy to be most important, and ''choose cheaper'' respondents most valued the test cost attribute.29 Monetary value. Five tudies calculated patient WTP for genetic testing, as presented in Table 4. [24][25][26]28,29 These were in some cases described as WTP, although the naming was inconsistent and 3 different methods were used.Marginal WTP (mWTP) is obtained as the marginal rate of substitution between a single nonmonetary attribute and monetary attribute (cost) on the utility scale (i.e., it is given by the ratio of part-worth utilities).Moneyequivalent values (MEV) are also obtained by translating utility differences into monetary terms by using the marginal utility of cost, but these utility differences result 260 Medical Decision Making 44(3) from differences across multiple attributes and give monetary values to comparisons of alternatives.Both mWTP and MEV can also be estimated by fitting the choice model in WTP space, which can lead to more reasonable distributions of mWTP when preference heterogeneity is included in the model. 32Compensating variation is a measure of WTP that applies when the choice set is changed (due to entry or exit of choices or changes in the quality or price of existing choices); the compensating variation is the amount of money that can be removed from an individual following a change in the choice set to return them to their utility level prior to the change. 33f the included studies, 1 calculated WTP using compensating variation, 24 while 4 calculated mWTP, 25,26,28,29 of which 1 also calculated MEV. 25 Regarding service design, participants most valued a test for multiple cancers compared with a test for a single cancer ($103.03) 29or a specialist doctor compared with a nurse educator ($37.04). 28The results and impact of genetic testing suggest participants place the highest value on a large improvement in probability of detecting deleterious mutation from 60% to 88% ($949.42) 26and on health professionals seeing the results rather than insurance companies ($1,242.08 and $1,104.07). 25Knight et al. 25 suggested who sees the results, being health professionals rather than insurance providers, is of greater monetary value to participants than test accuracy is.Conversely, however, Venning et al. 29 concluded that greater value is associated with increased accuracy, while the smallest value concerns the impact on life insurance eligibility or premiums.
Where studies consider mWTP for both service design and results/impact, 25,29 attributes affecting results/impact hold the greatest monetary value, being whether the insurance company saw the test results 25 or the test accuracy increasing from 60% to 90%. 29 Values are not, Values reflect the relative distance of all attributes to the most important attribute on a scale from 0 to 1 (1 indicating the most important attribute).Where multiple subgroups were reported, only 1 was selected for Figure 2 based on sample representation and positive framing (less emotional response).z Genetic testing preference was not included as an attribute in the discrete choice experiment but was calculated by comparing the utility associated with no test to the alternative of the average test.however, consistently lower for service design attributes across the studies, as, for example, participants also place a high monetary value on the decision to test and whether to test for multiple cancers.

Morrish et al.
One study considered test compared with no test using MEV, yielding a value of $773.35. 25 Only 1 study used compensating variation to estimate WTP, which it did by modeling the impact of introducing MPS testing in 3 scenarios to an existing choice set of traditional testing and no testing. 24These analyses revealed that participants were willing to pay for better detection rate (results) even with longer wait time for results (service design). 24A greater increase in WTP was observed between 60% and 80% detection rate than between 80% and 90%; however, the latter could be confounded by the increased wait time for results from 3 wk to 1.5 mo. 24n additional article by Griffith et al. 23 did not calculate monetary values but observed that when an alternative with a cost of $6,570 (3,000 GBP in 2002) was presented, it was sometimes selected, indicating some participants were willing to pay this much (or more) sometimes.This study did not include an opt-out option, and the alternatives presented always incurred a cost of at least $3,285, with 1 of the alternatives (a constant comparator) always having a cost of $5,475.These methodological choices likely render any WTP estimates from the study unfit for purpose.
Predicted uptake of testing.Three studies predicted uptake of testing, as presented in Table 5. 24,25,28 The highest predicted uptake was for a scenario for a hereditary colorectal cancer test in which the chance of a falsenegative test result was 0%, cost of testing was $621.66 ($500 in 2015) and test results were shared with a primary care doctor.In this scenario, the predicted uptake of testing was 97% (95% confidence interval: 95299), 25 as determined by a probability-based sample of adults presented with a hypothetical level of colorectal cancer risk provided for context.In comparison, the least preferred scenario had an uptake of 3.7%, describing a ''realistic base case'' of buccal swab type, in a hospital, with a specialist doctor, costing $228.33 (S$175 in 2018), while the pricier option costing $391.42 (S$300 in 2018) had 0.0% uptake. 28This test was a single-nucleotide polymorphism gene test for breast cancer with preferences Most preferred alternative = 88.9% All costs inflated to 2022 US$ rates using International Monetary Fund purchasing power parity values.Prices are inflated from the year of data collection or, where not reported, inflated from the year prior to publication.FN, false negative; TP, true positive.

Morrish et al.
determined by women without history of breast cancer.
For both tests, the cost would be incurred by the patient.

Discussion
This review summarizes the relative importance and WTP for different attributes in genetic testing for hereditary cancer as identified across 7 included studies.Test effectiveness and detection rates were consistently important to respondents.Accuracy and cost, while also important, showed variation between studies.Studies also revealed patients and the public would be willing to pay for improved detection and clinician over insurance provider involvement.

Conducting a DCE
Attribute selection.Attribute selection is key for DCE design.Several previous studies have highlighted a lack of rigor in the process and emphasized how this can introduce bias, 34 which can be minimized by a multistep attribute-development process. 35Most studies included in this review used multiple stages to select attributes; however, only 1 reported the attribute selection process extensively enough to allow replication. 25Despite similar selection methods, the included attributes varied across studies.This may have been due to differing opinions expressed by respondents in focus groups or perhaps the different study aims and cancers targeted.For improved rigor and consistency, researchers conducting future DCEs in genetics should strive to follow the recommended 4-stage approach by Helter and Boehler 35 : raw data collection, data reduction, removing inappropriate attributes, and wording.
Study design.Included studies highlighted a number of design limitations.Five studies investigated only main effects, [23][24][25][26][27] meaning potential interactions between attribute levels were not acknowledged, potentially confounding parameter estimates. 36Second, a forced-choice design was implied in more than half the studies, [26][27][28][29] which, while beneficial in reducing missing data, could challenge external validity, overestimate preferences, and distort welfare measures. 36,37With regard dot genetic testing, opt-out or indifference options better reflect a real-world scenario in which an individual may choose not to test at all.Study design was also limited by lack of reporting on whether the variance covariance matrix was blocked diagonal, meaning identification was not checked, and parameter estimates may be confounded. 17In addition, piloting was poorly reported, if at all, and 1 study introduced design inefficiency by including a constant in each choice set, which discards information regarding the choice between attributes. 23Furthermore, we found an additional study reporting DCE design methods in genomic medicine for familial colorectal cancer; however, this was not included as no DCE results were reported. 38

Relative Importance of Attributes
Most studies included attributes for cost, service design, and results/impact.These studies considered a clinical attribute or measure of test effectiveness, such as detection or survival, to be of great importance in improving test utility. 24,26,27,29This is consistent with existing research in cancer treatment DCEs, in which treatment outcomes are considered most important to patients. 14][41] Meanwhile, in this review, accuracy was highly important in 1 study 29 but of least importance in another. 25ne factor influencing this discrepancy could be that while Venning et al. 29 described the chance of a truepositive result, Knight et al. 25 considered accuracy in terms of the risk of false negatives.Similar to attribute framing, which has been seen to influence WTP estimates in previous DCEs on colorectal cancer screening, 42 this different focus may influence participant perception of accuracy.In addition, preferences could be driven by different health care systems, with the study by Knight et al. conducted in the United States and that by Venning et al. in Australia.The differing relative importance could also be a construct of the other attributes included in the DCE.For example, Venning et al. included twice as many attributes, including aspects such as cancer type, risk reduction measures, and testing process, which may alter relative attribute importance.Particularly of note, Venning et al. evaluated the polygenic risk score and found cancer type to be a highly valued attribute, whereas Knight et al. considered colorectal cancer alone, thus suggesting the general population may place higher value on evaluating a wider range of cancer types.
Inconsistences were also observed regarding the cost attribute.0][41] Meanwhile, cost was of middle to low importance in the study by Griffith et al., 23 who considered only cost and service design attributes.There are multiple factors, besides co-occurring attributes, that may particularly influence the cost attribute.First, Griffith et al. conducted the DCE in Wales, where most health care is paid for by the NHS, making the concept, and therefore attribute, of paying for tests more hypothetical.Meanwhile the studies by Knight, Weymann, and Davidson were all conducted in the United States, where considerable fees can be associated with health care provision, even with a form of insurance, and Venning et al. 29 reported the general public in Australia to have concerns around the impact of genetic testing on life insurance. Secon, there could be cross-study differences in the financial status of the population and resulting ability to pay for testing.Third, as indicated in Griffith et al., 23 patient risk for developing cancer could be influential, with higher risk increasing the relative importance of service cost.Overall, however, this review suggests that participants in studies were sensitive to price; thus, lowering the cost of testing may lead to increased uptake.
Additionally of note, Griffith et al. 23 reported that the wait time for test results had the greatest impact on the utility of testing services across all risk groups evaluated.This was contrary to the findings in 2 other studies, 24,26 in which turnaround time was identified as of middle or low importance relative to other attributes.Griffith's conclusion could perhaps be a construct of the other attributes, none of which consider clinical effectiveness of the test through, for example, detection rate or test sensitivity.In addition, Griffith et al. not only considered test characteristics but included attributes related to genetic counseling, such as distance and duration, which are not covered by other studies and may influence preferences.
The impacts of framing, 27 genetic coverage, 26 and risk level 23 were explored through subgroup analysis.Veldwijk et al. 27 investigated the impact of framing the effectiveness attribute positively (survival) or negatively (mortality).Positive framing concluded effectiveness as the most important attribute, while negative framing suggested it was colonoscopy frequency. 27These findings emphasize the importance of effectively communicating risk to patients to facilitate informed decision making, which has been widely discussed in the literature. 43,44avidson, who considered genetic coverage, found little change in relative attribute importance; however, as would be expected, multigene testing valued test sensitivity for mutations and probability of detecting uncertain variants more than the nonmultigene or no-test group. 26

Monetary Value
Respondents were willing to pay for a test compared with no test at all 25 and for a test for multiple cancers compared with a test for a single cancer. 29They also valued having a specialist doctor over less-specialized professionals 28 and a private family clinic over a hospital, 28 although the monetary value was smaller.Considering the results and impact of the test, participants placed the highest value on a large improvement in mutation detection (from 60% to 88%) 26 and on health professionals rather than insurance companies seeing the test results. 25n addition, patients preferred traveling a greater distance to counseling, which authors suggest may be related to anonymity. 23elative WTP regarding accuracy and the involvement or impact on insurance remains unclear.Participants in one study placed greater monetary value on who sees the results (health professionals v. insurance providers) than on improved accuracy, 25 while respondents in another study conversely placed greater value on increased accuracy than on the resulting impact on life insurance. 29hese are all aspects that should be carefully considered by policy makers and clinicians in designing or implementing a genetic test for hereditary cancer.
Challenges also arise in making cross-study WTP comparisons from heterogeneity in attributes, levels, and analysis.WTP estimates evaluated either only a select number of attribute levels, with levels differing between studies, 25,26,28,29 or a combination of attributes to form a care program, 24,25 both of which prevented insightful comparison or a conclusion of dominant preference.Furthermore, only 1 of the included studies used the Hicksian compensating variation method, which better reflects the choice nature of the task in calculating WTP. 24Hicksian compensating variation provides a common metric that could facilitate cross-study comparison of relative attribute importance. 45Overall, if the same method for assessing monetary value had been used in all studies, further WTP conclusions may have been possible.
WTP estimates did not appear to be adjusted to reflect ability to pay, limiting the interpretation of the results.While income level, which could provide context, was reported in some studies, 24,25,28 the method of reporting was inconsistent.In addition, probability analysis was conducted in only 1 monetary valuation study, 28 making it impossible to ascertain the impact of the attributes studied, or of sociodemographic variables, on respondents' WTP.Lack of probability analyses hindered any opportunity to assess the impact of the attributes or sociodemographic characteristics on the predicted uptake of testing.

Strengths and Limitations
While including conjoint studies may have provided a greater depth of evidence, this review was restricted to DCEs to ensure greater comparability.It is also consistent with existing research, which found very few conjoint analyses to add. 18Furthermore, this study focused Morrish et al.
only on genetic testing for hereditary cancer syndromes and polygenic cancer risk, providing greater relevance and utility to the population under consideration.Including predictive genetic testing for other complex diseases with a strong hereditary component may have increased the evidence base but come at the expense of precision given the differing nature and impact on preferences and quality of life related to different diseases.While the studies included some similar attributes, they were defined differently and contained different levels giving rise to heterogeneity, a limitation noted by other reviews in the area. 18

Future Research
Evaluation of the impact of sociodemographic characteristics on preferences for testing is limited, and existing research has reported contradictory results regarding willingness to undergo testing across sociodemographic groups. 117][48][49] Future studies should endeavor to include sociodemographic covariates and other respondent characteristics within the main model to assess the impact of factors related to the individual on willingness to undergo testing.This would help contextualize the relative importance of attributes across studies and assist in providing ability-to-pay context for WTP estimates.
Attention should also be paid to lesser studied attributes such as insurance provider interest and involvement, particularly in countries without free health care or social health insurance.Understanding the tradeoffs between various attributes, such as effectiveness and cost, across different health systems with different priorities would be particularly informative to policy makers and health service providers.Findings could help maximize the utility of testing services by, for example, understanding how changes in test or service characteristics may compensate for poorer clinical outcomes or higher costs, which at present seem most important to patients and the public.

Conclusion
This review provides insight into the content and relative importance of attributes in genetic testing for hereditary cancer.In general, test result/impact is consistently important to both patients and the public; however, the impact of test accuracy needs further research to determine its relative impact.There is variation in the relative importance of cost, which we suggest could be dependent on health system, service cost, ability to pay, and cancer risk.Overall, participants were sensitive to prices; thus, lowering the cost of testing may lead to increased uptake.Finally, this review compared findings on WTP, which suggested individuals would be most willing to pay for a test that improves detection rates, identifies multiple cancers, and for which the results are shared with a doctor rather than with an insurance provider.Hence, these factors should be carefully considered by clinicians, researchers, and policy makers when attempting to assess the tradeoffs of genetic testing.

a
Defined as the number of participants starting the survey divided by the number of online panel members invited to the survey.b A total of 120 respondents returned the discrete choice questionnaire; however 5 of the questionnaires were incomplete.c A total of 114 patients consented.Of these, 20 were excluded: 4 pilot subjects, 4 screen failures, 1 duplicate, and 11 did not complete the survey.d

Figure 2
Figure 2 Relative importance of attributes.

Table 2
DCE Characteristics

Table 2
(continued) a Attribute levels included in the constant comparator scenario.bHalf of the questionnaires said $1,000, half $1,500.c

Table 3
Respondent Characteristics x, mean; NR, not reported; s, standard deviation; S$, Singapore dollar.

Table 4
Monetary Values (US$) All costs inflated to 2022 US$ rates using International Monetary Fund purchasing power parity values.Prices are inflated from the year of data collection or, where not reported, inflated from the year prior to publication.''v.''indicates ''relative to,'' that is, the testing alternative v. comparator or ''change to'' v. ''change from.''CI, confidence interval; GP, general practitioner; MEV, money-equivalent values representing the additional amount of money that subjects would pay for tests with more preferred features; MPS, massively parallel sequencing; mWTP, marginal willingness to pay; VUS, variant of uncertain significance; WTP, willingness to pay; WTP CV , willingness to pay calculated using compensating variation.a Test characteristics set to the mean values in the experimental design.Addition of the specified MPS test to an existing choice set of traditional diagnostic testing (40% detection rate, 3-mo wait for results) and no testing.

Table 5
Predicted Uptake of Testing