"Medicine, Faculty of"@en . "Pharmaceutical Sciences, Faculty of"@en . "Non UBC"@en . "DSpace"@en . "BMC Health Services Research. 2013 Oct 31;13(1):454"@en . "Najafzadeh et al.; licensee BioMed Central Ltd."@en . "Najafzadeh, Mehdi"@en . "Johnston, Karissa M."@en . "Peacock, Stuart J."@en . "Connors, Joseph M."@en . "Marra, Marco, 1966-"@en . "Lynd, Larry D."@en . "Marra, Carlo A."@en . "2015-12-19T02:24:17"@en . "2013-10-31"@en . "Background:\r\n The extent to which a genomic test will be used in practice is affected by factors such as ability of the test to correctly predict response to treatment (i.e. sensitivity and specificity of the test), invasiveness of the testing procedure, test cost, and the probability and severity of side effects associated with treatment.\r\n \r\n \r\n Methods\r\n Using discrete choice experimentation (DCE), we elicited preferences of the public (Sample 1, N\u00E2\u0080\u0089=\u00E2\u0080\u0089533 and Sample 2, N\u00E2\u0080\u0089=\u00E2\u0080\u0089525) and cancer patients (Sample 3, N\u00E2\u0080\u0089=\u00E2\u0080\u008938) for different attributes of a hypothetical genomic test for guiding cancer treatment. Samples 1 and 3 considered the test/treatment in the context of an aggressive curable cancer (scenario A) while the scenario for sample 2 was based on a non-aggressive incurable cancer (scenario B).\r\n \r\n \r\n Results\r\n In aggressive curable cancer (scenario A), everything else being equal, the odds ratio (OR) of choosing a test with 95% sensitivity was 1.41 (versus a test with 50% sensitivity) and willingness to pay (WTP) was $1331, on average, for this amount of improvement in test sensitivity. In this scenario, the OR of choosing a test with 95% specificity was 1.24 times that of a test with 50% specificity (WTP\u00E2\u0080\u0089=\u00E2\u0080\u0089$827). In non-aggressive incurable cancer (scenario B), the OR of choosing a test with 95% sensitivity was 1.65 (WTP\u00E2\u0080\u0089=\u00E2\u0080\u0089$1344), and the OR of choosing a test with 95% specificity was 1.50 (WTP\u00E2\u0080\u0089=\u00E2\u0080\u0089$1080). Reducing severity of treatment side effects from severe to mild was associated with large ORs in both scenarios (OR\u00E2\u0080\u0089=\u00E2\u0080\u00892.10 and 2.24 in scenario A and B, respectively). In contrast, patients had a very large preference for 95% sensitivity of the test (OR\u00E2\u0080\u0089=\u00E2\u0080\u00895.23).\r\n \r\n \r\n Conclusion\r\n The type and prognosis of cancer affected preferences for genomically-guided treatment. In aggressive curable cancer, individuals emphasized more on the sensitivity rather than the specificity of the test. In contrast, for a non-aggressive incurable cancer, individuals put similar emphasis on sensitivity and specificity of the test. While the public expressed strong preference toward lowering severity of side effects, improving sensitivity of the test had by far the largest influence on patients\u00E2\u0080\u0099 decision to use genomic testing."@en . "https://circle.library.ubc.ca/rest/handle/2429/55958?expand=metadata"@en . "RESEARCH ARTICLE Open AccessGenomic testing to determine drug response:measuring preferences of the public and patientsusing Discrete Choice Experiment (DCE)Mehdi Najafzadeh1, Karissa M Johnston3, Stuart J Peacock3,6, Joseph M Connors3,6, Marco A Marra4,Larry D Lynd2,5 and Carlo A Marra2,5*AbstractBackground: The extent to which a genomic test will be used in practice is affected by factors such as ability ofthe test to correctly predict response to treatment (i.e. sensitivity and specificity of the test), invasiveness of thetesting procedure, test cost, and the probability and severity of side effects associated with treatment.Methods: Using discrete choice experimentation (DCE), we elicited preferences of the public (Sample 1, N = 533 andSample 2, N = 525) and cancer patients (Sample 3, N = 38) for different attributes of a hypothetical genomic test forguiding cancer treatment. Samples 1 and 3 considered the test/treatment in the context of an aggressive curablecancer (scenario A) while the scenario for sample 2 was based on a non-aggressive incurable cancer (scenario B).Results: In aggressive curable cancer (scenario A), everything else being equal, the odds ratio (OR) of choosing a test with95% sensitivity was 1.41 (versus a test with 50% sensitivity) and willingness to pay (WTP) was $1331, on average, for thisamount of improvement in test sensitivity. In this scenario, the OR of choosing a test with 95% specificity was 1.24 times thatof a test with 50% specificity (WTP = $827). In non-aggressive incurable cancer (scenario B), the OR of choosing a test with95% sensitivity was 1.65 (WTP= $1344), and the OR of choosing a test with 95% specificity was 1.50 (WTP = $1080). Reducingseverity of treatment side effects from severe to mild was associated with large ORs in both scenarios (OR= 2.10 and 2.24 inscenario A and B, respectively). In contrast, patients had a very large preference for 95% sensitivity of the test (OR = 5.23).Conclusion: The type and prognosis of cancer affected preferences for genomically-guided treatment. In aggressive curablecancer, individuals emphasized more on the sensitivity rather than the specificity of the test. In contrast, for a non-aggressiveincurable cancer, individuals put similar emphasis on sensitivity and specificity of the test. While the public expressed strongpreference toward lowering severity of side effects, improving sensitivity of the test had by far the largest influence onpatients\u00E2\u0080\u0099 decision to use genomic testing.Keywords: Pharmacogenomics, Genomic medicine, Personalized medicine, Genetic testing, Discrete choice experiment,Conjoint analysis, Preference elicitation, Cancer treatmentBackgroundTreatment options for cancer are mainly chosen basedon the classification of the tumor and are usually basedon the best knowledge of histogenesis, histological type,and stage of disease [1]. However, these criteria often failto accurately differentiate among distinct subtypes oftumors, especially with respect to likelihood of responseto treatment, forcing clinicians and patients to chooseempirically. Thus, many patients end up experiencingsignificant side effects of chemotherapy without receiv-ing clinical benefit [2].Recent advances in genomics have created hope thatgenomic testing may help to identify patients who willlikely respond to a particular drug and/or experienceside effects. This information is valuable both for pa-tients and physicians when choosing among possibletreatment options and trading off between risks and* Correspondence: carlo.marra@ubc.ca2Faculty of Pharmaceutical Sciences, University of British Columbia,Vancouver, BC, Canada5Centre for Health Evaluation and Outcome Sciences (CHEOS), St. Paul\u00E2\u0080\u0099sHospital, 1081 Burrard Street, Vancouver, BC, CanadaFull list of author information is available at the end of the article\u00C2\u00A9 2013 Najafzadeh et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of theCreative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use,distribution, and reproduction in any medium, provided the original work is properly cited.Najafzadeh et al. BMC Health Services Research 2013, 13:454http://www.biomedcentral.com/1472-6963/13/454benefits. For example, panitumumab, a drug for thetreatment of colon cancer, was initially shown to be ef-fective only in 10% of cases. However, genomic testingrevealed that response rates were much higher in thosewithout a KRAS mutation in their tumor [3]. Other ex-amples are HER2 expression in breast cancer patients,which predicts response to trastuzumab [4] and theBCR-ABL genotype in chronic myeloid leukemia, whichpredicts response to imatinib mesylate [5].Despite some clear advantages for the use of genomictests to predict response to therapy, there are also somelimitations. Genomic test results often have a probabilisticrelationship with drug response \u00E2\u0080\u0093 a certain genotype inthe tumor may increase (or decrease) the probability oftreatment response but this relationship is rarely absolute[3]. This prediction error in genomic testing may lead tothe misclassification of those that will respond (i.e. sensi-tivity and specificity of tests are not perfect). In practice,the extent that an imperfect genomic test will be used isaffected by multiple factors. Patients and physicians con-sider various factors such as invasiveness of the testingprocedure, probability and severity of associated side ef-fects of the treatment, and the overall costs before decid-ing about the usefulness of a genomic test [6,7].The other important challenge is the impact of gen-omic testing on health care costs. An increasing numberof diagnostic and predictive tests as a result of advancesin genomics are creating increasing pressure on alreadysoaring health care costs. There are ongoing debatesabout added clinical and economic value of these newtechnologies and appropriate methods for measuringthose potential benefits [8,9]. New genomic tests, even ifproven to deliver clinical benefit, are rarely cost saving.Thus, the decision about their overall value should bemade based on the appropriate balance between clinicalbenefits and the costs of these technologies. In this con-text, it is important to determine which attributes of agenomic test are of more importance for patients whendeciding about their treatment options. In general, ap-proval and use of genomic tests varies widely across dif-ferent jurisdictions and for different populations.Publicly (or privately) funded health care benefit pro-viders are often interested in learning about tax payers\u00E2\u0080\u0099(or privately insured populations\u00E2\u0080\u0099) opinion about thevalue of these genomic tests. Knowledge about thesepreferences will enable health benefit providers to selectgenomic tests with the highest perceived value whenmaking funding decisions. This information can be usedto prioritize future research areas and suggest aspects ofgenomic testing where improvement will have the mostvalue to patients. Finally, this investigation may offer fur-ther insight about perceptions of patients who have dir-ectly experienced the disease and about their evaluationof different aspects of testing for cancer treatment. Thisinformation can potentially help physicians to offertreatment options that better match patients values andpreferences [10].Using a discrete choice experiment (DCE), we exploredthe relative impacts (i.e. relative preference weights) of dif-ferent attributes of a genomic test on individuals\u00E2\u0080\u0099 decisionto use the test for guiding cancer treatment. We investi-gated whether these relative impacts are influenced bytype of cancer and its prognosis. Finally, we investigatedhow these relative impacts may differ between cancer pa-tients and the public. Our knowledge about these relativepreference weights can offer a value-based framework [11]for evaluating and comparing new genomic tests.MethodsStudy sampleTwo samples from the public (sample 1 and sample 2)and a sample of current or former cancer patients partici-pated in this study. The samples from the public (sample1 and sample 2) were recruited by Ipsos Reid (Vancouver,British Columbia) and were representative of the Canadiangeneral population in terms of demographics and socio-economic characteristics. The third sample (sample 3)consisted of current or former lymphoma patients whohad voluntarily agreed be contacted about research pro-jects in British Columbia (BC), Canada.All subjects were invited to participate in this web-based study through email. All participants were at least19 years old and were able to read and write in English.In the initial letter, we provided a brief description of thestudy and invited individuals to participate. Once theyagreed, each participant provided informed consent andthen followed a web link to the online questionnaire.Participants could choose not to answer any of the ques-tions or withdraw at any point. The protocol for thisstudy was reviewed and approved by the University ofBritish Columbia - British Columbia Cancer Agency(BCCA) Research Ethics Board.Study procedureAt the beginning of the DCE questionnaire, we de-scribed one of two possible scenarios to the participants.We asked participants to imagine a situation where theyhave been diagnosed with either an aggressive curablecancer (scenario A) or a non-aggressive incurable cancer(scenario B) and that they have the option to choose agenomic test that can predict the likelihood of their re-sponse to a new chemotherapy (Table 1). We explainedthat the genomic test had limited accuracy, which mightresult in false negative (misclassifying responders asnon-responders) and false positive (misclassifying non-responders as responders) predictions. Finally, we ex-plained the attributes and levels in the DCE questionnaire(Table 2) and asked participants to complete 16 choiceNajafzadeh et al. BMC Health Services Research 2013, 13:454 Page 2 of 12http://www.biomedcentral.com/1472-6963/13/454tasks [12,13]. We used the same choice questions for allthree samples, but varied the underlying form of cancerdescribed for one of the samples from the public: thepreamble in the questionnaire described an aggressivecurable cancer (scenario A) to participants in the firstsample from the public and the sample from patients,and a non-aggressive, incurable cancer (scenario B) asthe scenario for the second sample from the public. Thedesign of the DCE questionnaire has been explained inthe next section and a sample choice task has been pre-sented in Table 3.The extent to which a genomic test will be used in prac-tice is affected by the perceived benefits, risks and costs ofusing the genomic test. As such, in the DCE questionnaireparticipants needed to make a trade-off between the con-sequences of not taking the new chemotherapy when infact it was beneficial, experiencing additional side effectsof new chemotherapy without receiving any clinicalTable 1 Scenarios for DCEScenario A: Scenario B:Aggressive curable Cancer Non-aggressive incurable CancerImagine that you have recently been diagnosed with a fast-acting butcurable form of cancer. Currently, approximately 50 out of 100 (50%) ofpatients are cured after the first round of chemotherapy. If you are curedby this initial treatment, you will have a normal life expectancy; otherwiseyour life expectancy is approximately 1 year. In this case you will be giventhe second round of chemotherapy but your chance of being cured isabout 10 out of 100 (10%).Imagine that you have recently been diagnosed with a slow-acting butincurable form of cancer. This means that the spread of the disease isusually slow, but treatments are only able to slow the spread further, andcannot cure the disease. Your life expectancy after being diagnosed withthis type of cancer is approximately 10 to 13 years. You will receive treat-ment after you start experiencing symptoms, which may take severalyears after your initial diagnosis. Even if your treatment is successful, youare likely to experience numerous relapses, in which the disease returnsafter a period of improvement. These relapses will be treated until alloptions for treatment have been exhausted.By adding a new medication to the first round of chemotherapy the curerate increases from 50 out of 100 (50%) to 75 out of 100 (75%) . However,only some of individuals can benefit from the new medication(responders) and other individuals receive absolutely no benefit fromadding the new medication to the standard chemotherapy(non-responders).By adding a new medication to the first round of chemotherapy your lifeexpectancy can be increased by 2 years on average. However, only someof individuals can benefit from the new medication (responders) andother individuals receive absolutely no benefit from adding the newmedication to the standard chemotherapy (non-responders).The downside of adding the new medication to the standardchemotherapy is that it increases the likelihood and severity of treatmentside-effects.The downside of adding the new medication to the standardchemotherapy is that it increases the likelihood and severity of treatmentside-effects.Table 2 Attribute and levels included in the DCE questionnaireAttribute LevelsUntreated responders&: 5%, 20%, 35%, 50%Proportion of patients who could be cured by the new medication (responders) but will not receive itas a result of inaccurate genetic test result.Unnecessary treatment of non-responders\u00E2\u0080\u00A0: 5%, 20%, 35%, 50%Proportion of patients who would not benefit from the new medication (non-responders) but will receiveit as a result of wrong genetic test result.Severity of side effects: Severe, Moderate, MildThe new medication may be associated with side effects such as nausea, hair loss, skin rash and fatigue.The potential levels of Side Effect Severity were:Likelihood of side effects: 5%, 50%, 95%The side effects described in Attribute 3 will not necessarily occur for all individuals. Instead, they willoccur with a particular percentage chance. Possible levels were:Genetic test turnaround time: 2 days, 7 days, 12 daysThe time required to obtain the genetic test results, after the test has been performed.Genetic test procedure: Mouth swab, Blood sample,Tumor biopsy, Bone marrowbiopsy, Liver biopsyType of the procedure that is needed for doing the genetic test.Genetic test cost: $50, $500, $1000, $1500Please assume that you would be paying only for the genetic test out-of-pocket.&1-Sensitivity.\u00E2\u0080\u00A01-specificity.Najafzadeh et al. BMC Health Services Research 2013, 13:454 Page 3 of 12http://www.biomedcentral.com/1472-6963/13/454benefit, the invasiveness of the genomic testing procedure,the test turnaround time, and the cost of the genomic test.The descriptions at the beginning of the questionnaireexplicitly stated that in the absence of a genomic test, allpatients would be offered the new chemotherapy. As such,choosing the \u00E2\u0080\u009Cneither\u00E2\u0080\u009D option in a choice task implied arespondent\u00E2\u0080\u0099s preference for opting-out from genomic test-ing and taking the new chemotherapy regardless of thelikelihood of response. We did not specify the type of can-cer, treatment, and the associated genomic test to increasethe generalizability of the results. Nonetheless, the sampleof patients in this study were former and current lymph-oma patients in British Columbia, and the disease descrip-tions provided in the DCE questionnaires were similar toaggressive and non-aggressive types of lymphoma.Questionnaire designDiscrete choice experiment is a method to elicit individ-uals\u00E2\u0080\u0099 strength of preferences for different aspects of ahealth intervention (or a product in general). The conceptof DCE is based on Random Utility Theory and the as-sumptions that: 1) a health care intervention (or any prod-uct or service in general) can be characterized by severalattributes; and 2) individuals choose among availablehealth interventions (or products or services) by evaluat-ing and comparing their attributes [12-14]. These attri-butes can describe health outcomes (e.g. test accuracy,likelihood or severity of treatment side effects) or inter-vention process (cost, test procedure, or turnaround time).In this study, we assumed that a genomically-guidedcancer treatment could be described by seven attributes(Table 2). Considering that a large number of (hypothet-ical) treatment options can be generated by using theseattributes and all various combinations of their levels(i.e. full factorial design), we implemented a fractionalfactorial design where we selected 10 versions of theDCE questionnaire each consisting of only 16 choicetasks. Therefore, each respondent had to complete arandomly assigned version of the DCE questionnairethat contained 16 choice tasks. In each choice task,respondents had to choose between two treatment op-tions and a neither option. A sample choice task hasbeen presented in Table 3 and the complete DCE ques-tionnaire can be found in Additional file 1. The effi-ciency of our fractional factorial design was assuredusing simulation of responses. We generated large num-ber of possible designs and then selected the design thatprovided the most precise coefficient estimates (i.e.smallest standard errors) and a better D-efficiency giventhe sample size [14,15]. The statistical design of thequestionnaire ensured that a random selection of re-sponses would result in preference weights that are notstatistically different from zero (i.e. non-informative co-efficient estimates).Several sources were used for selection of attributesincluding published literature, physicians\u00E2\u0080\u0099 opinion, andfeedbacks that we received from three pilot surveys. Weidentified several studies that had investigated character-istics of pharmacogenomic testing and their impact onpatients\u00E2\u0080\u0099 and physicians\u00E2\u0080\u0099 decisions for utilizing them[16-18]. We compiled a list of attributes based on the re-sults of these studies and discussed this list with physi-cians who were in direct contact with cancer patients inthe BC cancer agency. We then selected the seven attri-butes deemed to have the greatest influence on patient\u00E2\u0080\u0099sdecisions about treatment options. These attributes andlevels were then tested in a pilot study where 7 formercancer patients and 50 individuals from the public com-pleted the preliminary version of the DCE questionnaire.By analyzing the data in the pilot phase, we examinedrationality and consistency of the responses and whetherthe estimated coefficients conformed to our prior ex-pectation in terms of direction and sign. Our prior ex-pectation was based on the assumption that individuals\u00E2\u0080\u0099preferences (and willingness to pay) decrease by decreas-ing sensitivity and specificity of the test, and by increas-ing severity and likelihood of side effects, turnaroundtime, cost, or invasiveness of the testing procedure.Using this approach, we ensured that the respondentsunderstood the content of the DCE questionnaire andour instructions for completion of choice tasks. Further-more, we used the comments provided by respondentsat the end of the questionnaires to hone the preamble,descriptions, attributes, and levels used in the final ver-sion of the questionnaire.Two out of 16 choice tasks in the DCE questionnairecontained a clearly dominant option. By checking answersto these fixed choice tasks, we tested if respondents actu-ally read and understood the DCE questionnaire. Thesefixed choice tasks are usually part of the DCE question-naire design in order to verify consistency and rationalityof responses. We also included the \u00E2\u0080\u009Cneither\u00E2\u0080\u009D option in thechoice tasks to provide the possibility to opt-out whenevernone of the presented alternatives was adequatelyTable 3 A sample choice taskAttributes Option 1 Option 2 NeitherUntreated responders 50 out of 100 5 out of 100 0 out of 100Unnecessary treatmentof non-responders50 out of 100 5 out of 100 100 out of 100Severity of side effects Moderate Mild SevereLikelihood of side effects 50 out of 100 5 out of 100 95 out of 100Cost of genetic test $1000 $500 $0Genetic test turnaroundtime7 days 2 days 0 daysGenetic test procedure Bone Marrow Biopsy Blood Sample NoneWhich option youwould choose?O O ONajafzadeh et al. BMC Health Services Research 2013, 13:454 Page 4 of 12http://www.biomedcentral.com/1472-6963/13/454attractive to the respondent. Thus, we avoided forcing non-demanders to choose an alternative and ensured estimationof unconditional rather than conditional preferences [14].The design of the web-based questionnaire, which facili-tated direct data entry into our secured server, was doneusing the Choice Based Conjoint (CBC) application ofSawtooth (Sawtooth software Inc, SSI web version 6.6.6).Statistical analysisAssuming the general framework used in random utilitytheory [14], given a set of options, the log odds ratio ofchoosing one of the options is proportional to a linearfunction of attributes of that option. Therefore, by gath-ering stated choice data using a DCE questionnaire witha known statistical design and by knowing attributes andlevels presented in each choice task, the coefficients ofattributes can be estimated using generalized linearmodels. These coefficients, also known as relative prefer-ence weights, reflect average impact of attribute levelson likelihood of being chosen as the preferred option.Also the ratio of coefficients can be interpreted as mar-ginal rate of substitution (MRS) between any two attri-butes. By inclusion of cost as an attribute in the DCEquestionnaire, the marginal rates of substitution betweeneach attribute and cost, also known as Willingness toPay (WTP) [14], can be calculated. WTP can provideuseful interpretations for estimated preference weightsas they indicate how much individuals on average arewilling to pay to receive a certain amount of change in oneof the attribute levels [14]. The odds ratios (OR) as-sociated with each attribute levels also were calcu-lated. These odds ratios suggest, given two optionswith the same attribute levels, how a change in oneof the attribute levels will affect the odds of becom-ing the preferred choice.The choice data were effect-coded for attributes withdiscrete values, with the exception of cost, which wasmodeled as a continuous variable [19]. Effect coding ofchoice data, instead of continuous coding, relaxes linear-ity assumptions and allows detecting non-linearity ofpreference weights in regards to different levels of an at-tribute. Also modeling cost as a continuous variableallowed us to estimate WTP values in a way that is easyto interpret. An alternative specific variable was dummycoded and indicated the situations where \u00E2\u0080\u009Cneither\u00E2\u0080\u009D waschosen [14,20]. The choice data was analyzed usingPROC MDC, SAS 9.2. We pooled the choice data fromtwo samples from the public who completed the ques-tionnaire under scenario A and scenario B and estimateda conditional logit model using choice as the dependentvariable. We defined a dummy variable that indicatedthe scenario in the pooled data. By including interactionterms between this dummy variable and attribute levelsin the regression analysis, we compared the estimatedpreference weights across two samples from the public.We also used the same approach to compare estimatedpreference weights in the samples from the public andpatients who had both completed the questionnaireunder scenario A. However, prior to doing this analysis,we used the propensity score method to select a sub-sample of the public that were similar to the sample ofpatients in terms of age, education, income, and havingdependent children. Considering that the characteristicsof patients in our sample were different from the public,using propensity scores was necessary to increase com-parability of the results across the samples from the pub-lic and patients in our analysis.There are a variety of statistical methods for the ana-lyses of DCE data that range from conditional logitmodels to Bayesian mixed logit models [21] and LatentClass Analysis (LCA) [22]. Critical assessment of thesemethods can be found elsewhere [23]. We chose condi-tional logit model for analyses of the DCE data in thisstudy. However, we verified the estimated results and ro-bustness of our findings by re-running the regressionsusing a mixed logit model.ResultsSample characteristicsInvitations were initially sent to 904 and 836 individualsfrom the public for participation in the study under sce-narios A and B, respectively. Although 588 (65%) indi-viduals in scenario A and 578 (69%) individuals inscenario B provided their responses to the question-naires, some of the questionnaires contained uncom-pleted choice tasks. To avoid potential bias as a result ofimbalanced frequency of responses, we decided to re-strict our analysis to the data from questionnaires withcomplete responses to all choice tasks (533 individualsin scenario A and 525 individuals in scenario B). Oursample of patients was limited to an email list providedby BC cancer Agency (BCCA). We initially contacted alist of 84 patients through email and 54 (64%) patientsagreed to participate in this study. However, after ex-cluding incomplete responses, we had choice data from38 patients for the final analysis.Table 4 has summarized the characteristics of the par-ticipants in the three samples. Mean age in the sampleof patients was 58.2 years, about 10 years higher than inthe samples from the public. Also 36.1% of individuals inthe sample from patients reported a household incomeof \u00E2\u0089\u00A5Can $125,000 (This rate was 6.6% and 5.5% in thesamples from the public). Patients who participated inthis study were also highly educated and 32.4% had amaster or doctorate degree (the proportions of individ-uals with master or doctorate degree were 2.5% and4.1% in the samples from the public under scenario Aand scenario B, respectively).Najafzadeh et al. BMC Health Services Research 2013, 13:454 Page 5 of 12http://www.biomedcentral.com/1472-6963/13/454Estimation resultsComparing preferences of the public under two scenarios Aand BThe estimated preference weights, odds ratios, and theWTP associated with the levels in each attribute havebeen reported in Table 5.The results suggested that in aggressive curable cancer(scenario A), the preference weight of the public for\u00E2\u0080\u009Csensitivity: 50%\u00E2\u0080\u009D was \u00E2\u0088\u00920.1686 (s.e. 0.466) and it in-creased to 0.1748 (s.e. 0.0266) for a test with \u00E2\u0080\u009Csensitivity:95%\u00E2\u0080\u009D (Table 5). Alternatively, the impact of test sensitiv-ity on respondent\u00E2\u0080\u0099s choice is evident in the reported ORsand WTPs. For example, everything else being equal, theodds of choosing a test with 95% sensitivity were 1.41times the odds of choosing a test with 50% sensitivityand they were willing to pay $1331 for increasing testsensitivity from 50% to 95%. However, they were willingto pay only $796 and $487 for increasing sensitivity to80% and 65%, respectively. In non-aggressive incurablecancer (scenario B), preference weights of \u00E2\u0080\u009Csensitivity:95%\u00E2\u0080\u009D and \u00E2\u0080\u009Csensitivity: 50%\u00E2\u0080\u009D were 0.2577 (s.e. 0.270) and\u00E2\u0088\u00920.2436 (s.e. 0.0479), respectively. Increasing sensitivityfrom 50% to 95% increased the odds of choice by 1.65times. Although this preference weight in scenario B waslarger compared to scenario A (0.2577 vs. 0.1748, differ-ence p-value = 0.0241), corresponding willingness to payvalues were comparable ($1331 vs. $1344 in scenario Band A, respectively). Preference weights and WTPs for atest with sensitivity of 80% or 65% in scenario B werenot significantly different from corresponding values inscenario A.In scenario A, the odds of choosing a test with 95% spe-cificity were 1.24 times the odds of choosing a test with50% specificity and the public was willing to pay $827 forthis amount of improvement in specificity level. The pref-erence weight for 95% specificity was more than two-foldlarger under scenario B compared to scenario A (0.2452,0.1008, difference p-value < 0.001). Therefore, under sce-nario B, the odds of choosing a test with 95% specificitywere 1.50 times the odds of choosing a test with 50% spe-cificity and the corresponding WTP was $1080. Also inscenario B, the preference weight of 65% specificity wasnegative (\u00E2\u0088\u00920.1251) and statistically different (differencep-value = 0.0115) from its counterpart under scenario A(0.0051). The public perceived little value in increasingspecificity from 50% to 65% in scenario B.Reducing severity of treatment side effects from severeto mild was associated with large ORs in both scenarios(OR = 2.10 and 2.24 in scenario A and B, respectively).The public was willing to pay as much as $2882 and$2165 to receive a treatment with mild rather than severeside effects in aggressive curable cancer (scenario A) andnon-aggressive incurable cancer (scenario B), respectively.Furthermore, the odds of choosing a treatment with 5%likelihood of side effects were 1.62 and 1.75 times the oddsof choosing a treatment with 95% likelihood of side effectsin scenario A and B, respectively.Shortening test turnaround time from 12 days to ei-ther 7 days or 2 days had the smallest impact on prefer-ence weights, ORs, and WTPs under both scenarios. Incontrast, the level of invasiveness of the testing proced-ure had a large impact on estimated preference weights,Table 4 Characteristics of participantsThe public,scenario AThe public,scenario BPatients,scenario AN = 533 N = 525 N = 38Age (years) N = 512 N = 510 N = 37Mean (std) 48.2 (15.7) 47.6 (15.9) 58.2 (9.4)Education (%) N = 529 N = 514 N = 37Some high school 6.8% 6.6% 0%High school 42.2% 45.5% 2.7%College 36.1% 35.4% 27.0%Bachelor degree 12.5% 8.4% 37.8%Master degree 1.9% 3.5% 27.0%Doctorate 0.6% 0.6% 5.4%Gender N = 522 N = 516 N = 36Female 48.5% 50.6% 58.3%Male 51.5% 49.4% 41.7%Number of dependent children N = 532 N = 517 N = 36None 57.9% 56.5% 69.4%1 14.3% 15.5% 5.6%2 16.2% 16.1% 13.9%3 or more 11.6% 12.0% 11.1%Description of current healthsituationN = 526 N = 521 N = 37Excellent 8.9% 11.3% 10.8%Very good 29.9% 28.6% 16.2%Good 29.1% 33.6% 32.4%With some health problems 28.7% 23.8% 21.6%Having serious healthproblems3.4% 2.7% 18.9%If knew anyone diagnosed withcancerN = 528 N = 519 N = 36Yes, very closely 17.8% 16.2% 2.8%Yes 38.1% 41.6% 30.6%No 44.1% 42.2% 66.7%Household\u00E2\u0080\u0099s annual income (Can$) N = 516 N = 509 N = 36Less than 25000 15.5% 13.2% 2.8%25000 to 50000 30.2% 35.6% 27.8%50000 to 75000 23.1% 23.6% 11.1%75000 to 100000 16.9% 12.6% 13.9%100000 to 125000 7.8% 9.6% 8.3%More than 125000 6.6% 5.5% 36.1%Najafzadeh et al. BMC Health Services Research 2013, 13:454 Page 6 of 12http://www.biomedcentral.com/1472-6963/13/454ORs, and WTP values in both scenarios. For example,the public was willing to pay $2162 and $1474 for a gen-omic test that could be performed using a mouth swabrather than one involving a liver biopsy in scenario Aand B, respectively.Individuals from the public had negative preferenceweights for opting out from genetic testing (i.e. choosing\u00E2\u0080\u009Cneither\u00E2\u0080\u009D option). The preference weight was a largernegative number under scenario A compared to scenarioB (\u00E2\u0088\u00920.6323 in scenario A vs. -0.4967 in scenario B,Table 5 Estimated preference weights and Willingness to Pay (WTP) in samples from the publicThe public,scenario A(N = 533) The public,scenario B(N = 525) P-value fordifferenceCoefficient (s.e.) OR MWTP Coefficient (s.e.) OR MWTPUntreated responders&5% 0.1748 (0.0266)** 1.41 1,331 0.2577 (0.0270)** 1.65 1,344 0.024120% 0.0367 (0.0265) 1.23 796 0.0324 (0.0274) 1.32 740 0.848735% \u00E2\u0088\u00920.0429 (0.0276) 1.13 487 \u00E2\u0088\u00920.0465 (0.0285) 1.22 528 0.931550% \u00E2\u0088\u00920.1686 (0.0466) 1 Ref \u00E2\u0088\u00920.2436 (0.0479)** 1 RefUnnecessary treatment of non-responders\u00E2\u0080\u00A05% 0.1008 (0.0268)** 1.24 827 0.2452 (0.0270)** 1.50 1,080 0.000420% 0.0065 (0.0274) 1.13 461 0.0377 (0.0288) 1.22 524 0.669135% 0.0051 (0.0272) 1.12 456 \u00E2\u0088\u00920.1251 (0.0281)** 1.03 88 0.011550% \u00E2\u0088\u00920.1124 (0.0470) 1 Ref \u00E2\u0088\u00920.1578 (0.0485)** 1 RefSeverity of side effectsMild 0.3319 (0.0205)** 2.10 2,882 0.3295 (0.0211)** 2.24 2,165 0.3838Moderate 0.0798 (0.0210)** 1.63 1,905 0.1484 (0.0217)** 1.87 1,679 0.0621Severe \u00E2\u0088\u00920.4117 (0.0293)** 1 Ref \u00E2\u0088\u00920.4779 (0.0303)** 1 RefLikelihood of side effects5% 0.2490 (0.0204)** 1.62 1,861 0.2622 (0.0213)** 1.75 1,497 0.224550% \u00E2\u0088\u00920.0179 (0.0209) 1.24 826 0.0340 (0.0214) 1.39 885 0.226195% \u00E2\u0088\u00920.2311 (0.0292)** 1 Ref \u00E2\u0088\u00920.2962 (0.0302)** 1 RefGenetic test turnaround time2 days 0.1213 (0.0210)** 1.26 911 0.1266 (0.213)** 1.27 650 0.45337 days \u00E2\u0088\u00920.0076 (0.0208) 1.11 411 \u00E2\u0088\u00920.0107 (0.0218) 1.11 282 0.539612 days \u00E2\u0088\u00920.1137 (0.0296)** 1 Ref \u00E2\u0088\u00920.1159 (0.0305)** 1 RefGenetic test procedureMouth swab 0.2962 (0.0304)** 1.75 2,162 0.3045 (0.0311)** 1.73 1,474 0.9942Blood sample 0.2863 (0.0320)** 1.73 2,124 0.3882 (0.0332)** 1.88 1,698 0.5738Tumor biopsy \u00E2\u0088\u00920.0416 (0.0326) 1.25 853 \u00E2\u0088\u00920.0809 (0.0332)** 1.18 440 0.8152Bone marrow biopsy \u00E2\u0088\u00920.2792 (0.0321)** 0.98 \u00E2\u0088\u009268 \u00E2\u0088\u00920.3666 (0.0339)** 0.89 \u00E2\u0088\u0092325 0.6879Liver biopsy \u00E2\u0088\u00920.2617 (0.0636)** 1 Ref \u00E2\u0088\u00920.2452 (0.0675)** 1 RefNeither (No test) \u00E2\u0088\u00920.6323 (0.0379)** 0.53 \u00E2\u0088\u00922,451 \u00E2\u0088\u00920.4967 (0.0389)** 0.61 \u00E2\u0088\u00921,332 0.0169Genetic test cost \u00E2\u0088\u00920.00026 (0.00003)** Ref \u00E2\u0088\u00920.00037 (0.00003)** Ref 0.0091McFadden\u00E2\u0080\u0099s LRI 0.17Adjusted Estrella 0.33Log-likelihood Ratio 6216.5AIC 31050Schwartz Criterion 31329** p-value < 0.01.&1-Sensitivity.\u00E2\u0080\u00A01-specificity.Najafzadeh et al. BMC Health Services Research 2013, 13:454 Page 7 of 12http://www.biomedcentral.com/1472-6963/13/454difference p-value = 0.0169). The ORs of opting-out fromgenetic testing (vs. taking a test) were 0.53 and 0.61 inscenario A and B, respectively. The public had a largerWTP for having a test in aggressive curable cancer sce-nario ($2451) compared with non-aggressive incurablecancer ($1332). Finally, the preference weight for \u00E2\u0080\u009Cgen-etic test cost\u00E2\u0080\u009D was a larger negative number under sce-nario B compared to scenario A (\u00E2\u0088\u00920.00026 in A vs.-0.00037 in B, difference p-value = 0.0091), indicatingthat the public was more sensitive to price in scenario B.Comparing preferences of the public with preferences ofpatients under scenario AUsing propensity scoring we identified a subsample ofthe public (N = 83) who had similar characteristics to pa-tients (N = 38) in terms of age, education, income, andnumber of dependent children. Next we pooled the datafrom two samples (N = 121) and fitted a conditional logitmodel to estimate preference weights, ORs, and WTPsassociated with each attribute levels (Table 6).The preference weight of patients for \u00E2\u0080\u009Csensitivity: 95%\u00E2\u0080\u009Dwas significantly larger compared to the public (0.2480in the public vs. 0.8794 in patients, difference p-value <0.001). This large difference in preference weights for\u00E2\u0080\u009Csensitivity: 95%\u00E2\u0080\u009D also translated into large differences inWTP estimates ($2,658 for the public vs. $12,820 for pa-tients) and ORs (1.53 vs. 5.23, respectively). Patients\u00E2\u0080\u0099 hadconsistently larger preference weight for better sensitiv-ity and specificity of the test, as was evident based onORs and WTP values associated with different levels ofsensitivity. Among patients, the odds of choosing a genetictest that requires \u00E2\u0080\u009Cmouth swab\u00E2\u0080\u009D were 2.43 times the oddsof a test that needs liver biopsy. Patients also preferred atest that involves \u00E2\u0080\u009CBone marrow biopsy\u00E2\u0080\u009D instead of \u00E2\u0080\u009Cliverbiopsy\u00E2\u0080\u009D (OR = 1.76), while the public considered both typesof biopsies equally unfavorable (OR = 1.04). There was alarge difference between preference weight of the publicversus patients for opting-out from the test (\u00E2\u0088\u00921.0346 in thepublic vs. -0.1185 in patients, difference p-value = 0.0002).Consequently, the public was willing to pay as high as$6050 for having a genetic test while patients\u00E2\u0080\u0099 WTP forgenetic testing was only $919. This indicated patients hadsignificantly less aversion to opting out of genomic testing.DiscussionThis study shows the relative impact of different proper-ties of genomically-guided cancer treatment on test up-take. Change in severity and likelihood of side effects aswell as the test procedure have the largest influence onthe public\u00E2\u0080\u0099s decision to use genetic testing. In contrast,improving sensitivity of the test had a larger influenceon patients\u00E2\u0080\u0099 decision to use genomic testing.The type of cancer and its prognosis also influencedthe preferences of the public for different attributes ofgenomic testing. When we compared the results in thetwo samples from the public, we found that in aggressivecurable cancer, individuals emphasized the sensitivity ra-ther than specificity of the test. In contrast, for a non-aggressive incurable cancer, individuals put similar em-phasis on the sensitivity and specificity of the test andexpressed strong (positive and negative) preferences to-ward (high and low) specificity of the test. Furthermore,under this scenario (non-aggressive incurable cancer)the public also had a larger negative preference towardthe cost of genomic testing. Because for a non-aggressive incurable cancer the change in the survival isultimately small and is expected to be materialized after13 years, this lead to the public discounting the benefitsof new chemotherapy and becoming more selectiveabout accuracy of genomic testing in this scenario.Our study suggests that patients and the public havedifferent perceptions about the value of various aspectsof genomic testing to guide cancer treatment when fa-cing an aggressive curable cancer. Based on our results,patients were mostly concerned about improving sensi-tivity of the test (and presumably their survival chance),and in the absence of an adequately sensitive test theypreferred opting-out from genomic testing and takingthe treatment regardless of its side effects. Conversely,the public had a large negative preference weight foropting-out from genomic testing suggesting that theyare more inclined to use a test even with inadequate ac-curacy. This information may help physicians to tailortheir clinical advice considering type of cancer and pre-vious experience of their patient with cancer treatment.For example, if the prognosis of disease is expected tobe similar to our scenario for non-aggressive incurablecancer, then perhaps discussing false positive rates ofavailable tests can be of great importance for the averagepatient. Also, the observed differences between prefer-ences of patients and the public about different biopsyprocedures suggest that perhaps physicians can help pa-tients who have no prior experience of cancer treatmentin developing a more realistic perception about the rela-tive invasiveness of these procedures.There is a paucity of studies about preferences forcharacteristics of genomic testing. The increasing num-ber of new genomic tests ensuing from fast develop-ments in genomic sciences underlines the need forfurther investigations in this area. Knowledge aboutstrength of preferences toward different attributes ofgenomic testing can lead us toward value-based evalu-ation of these new technologies. In health care systemsthat rely on public funding resources, by consideringthese preference weights in funding decisions, genomictests with potentially higher value for a covered popula-tion can be determined. In addition, physicians can havebetter understanding about patients\u00E2\u0080\u0099 priorities given theNajafzadeh et al. BMC Health Services Research 2013, 13:454 Page 8 of 12http://www.biomedcentral.com/1472-6963/13/454type and prognosis of the disease. The differences inpreferences of patients and the public shown in ourstudy also suggests areas that physicians shouldemphasize when communicating with recently diag-nosed patients who presumably have no prior experienceof the disease. In a study conducted by Griffith et al.,willingness to pay for receiving breast cancer genomicservices was estimated by conducting a DCE on 242 in-dividuals with high, moderate, and low risk of develop-ing breast cancer [24]. Using a DCE and following aTable 6 Estimated preference weights and Willingness to Pay (WTP) in a propensity score matched subset of the publicand patientsThe public, scenario A (N = 83) patients, scenario A (N = 38) P-value fordifferenceCoefficient (s.e.) OR MWTP Coefficient (s.e.) OR MWTPUntreated responders&5% 0.248 (0.0687)** 1.58 2,658 0.8794 (0.1068)** 5.23 12,820 <.000120% 0.0528 (0.0676) 1.30 1517 0.0442 (0.1068) 2.27 6,346 0.608435% \u00E2\u0088\u00920.0942 (0.0702) 1.12 657 \u00E2\u0088\u00920.1492 (0.1133) 1.87 4,847 0.658250% \u00E2\u0088\u00920.2066 1 Ref \u00E2\u0088\u00920.7744 1 RefUnnecessary treatment of non-responders\u00E2\u0080\u00A05% 0.1867 (0.0679)** 1.39 1,919 0.1083 (0.1083) 1.59 3,580 0.966320% 0.0134 (0.0697) 1.17 906 0.2391 (0.1112)* 1.81 4,594 0.159835% \u00E2\u0088\u00920.0586 (0.0683) 1.09 485 0.0062 (0.1088) 1.43 2,789 0.65750% \u00E2\u0088\u00920.1415 1 Ref \u00E2\u0088\u00920.3526 1 RefSeverity of side effectsMild 0.2712 (0.0524)** 2.10 4,327 0.3084 (0.0839)** 2.09 5,716 0.7Moderate 0.1976 (0.053)** 1.95 3,897 0.1206 (0.0871) 1.73 4,260 0.3008Severe \u00E2\u0088\u00920.4688 1 Ref \u00E2\u0088\u00920.4290 1 RefLikelihood of side effects5% 0.2645 (0.0521)** 1.65 2,937 0.2735 (0.0838)* 1.60 3,650 0.811950% \u00E2\u0088\u00920.0267 (0.0529) 1.24 1,235 \u00E2\u0088\u00920.0762 (0.0864) 1.13 939 0.567895% \u00E2\u0088\u00920.2378 1 Ref \u00E2\u0088\u00920.1973 1 RefGenetic test turnaround time2 days 0.1673 (0.0531)** 1.40 1,978 0.0174 (0.087) 1.15 1,117 0.24677 days 0.0036 (0.0521) 1.19 1,021 0.1093 (0.0857) 1.27 1,829 0.525912 days \u00E2\u0088\u00920.1709 1 Ref \u00E2\u0088\u00920.1267 1 RefGenetic test procedureMouth swab 0.2591 (0.0783)** 1.75 3,258 0.3918 (0.1239)* 2.43 6,881 0.4303Blood sample 0.3382 (0.0825)** 1.89 3,720 0.1168 (0.1322) 1.85 4,750 0.2508Tumor biopsy \u00E2\u0088\u00920.04 (0.0828) 1.29 1,509 \u00E2\u0088\u00920.0829 (0.1342) 1.51 3,202 0.7787Bone marrow biopsy \u00E2\u0088\u00920.2593 (0.081)** 1.04 226 0.0702 (0.1286) 1.76 4,388 0.1121Liver biopsy \u00E2\u0088\u00920.2980 1 Ref \u00E2\u0088\u00920.4959 1 RefNeither (No test) \u00E2\u0088\u00921.0346 (0.1037)** 0.36 \u00E2\u0088\u00926,050 \u00E2\u0088\u00920.1185 (0.1382) 0.89 \u00E2\u0088\u0092919 0.0002Genetic test cost \u00E2\u0088\u00920.00017 (0.00007)* Ref \u00E2\u0088\u00920.00013 (0.00012) Ref 0.1901McFadden\u00E2\u0080\u0099s LRI 0.20Adjusted Estrella 0.35Log-likelihood Ratio 832.8AIC 3493Schwartz Criterion 3694** p-value < 0.01; * p-value < 0.05.&1-Sensitivity.\u00E2\u0080\u00A01-specificity.Najafzadeh et al. BMC Health Services Research 2013, 13:454 Page 9 of 12http://www.biomedcentral.com/1472-6963/13/454rigorous methodology, Hall and colleagues [25] exploredthe factors that influenced participation in genomic car-rier testing for Tay Sachs and cystic fibrosis among asample from the general community and a sample of theAshkenazi Jewish community. A recent study [26] alsoused DCE to estimate the tradeoffs among sensitivity,turnaround time, and cost of a postnatal genomic test topredict genomic abnormalities causing mental retard-ation in children. Finally, in a study done by Herbildet al. [27], they elicited preferences in the Danish generalpopulation for taking a pharmacogenomic test that couldimprove treatment of depression.Patients\u00E2\u0080\u0099 emphasis on sensitivity also has been shownin the context of using usual screening tests for colorec-tal cancer [28]. In exploring preferences of 1047 patientswith a history of colorectal cancer for different screeningmodalities, Marshall et al. used a DCE and estimatedhow likelihood of uptake may be affected by differentcharacteristics of the test. Similar to our results, theyfound that sensitivity of the test has the largest impacton the likelihood of uptake among these patients. Across sectional survey study by Haga et al. also showedthat primary care physicians consider the severity of sideeffects followed by predictive accuracy of a phramacoge-nomic test as the factors that have the largest influenceon their decision to prescribe it to their patients, whileturnaround times have a smaller influence on their deci-sion for using pharmacogenomics testing [16]. These re-sults, when considered in the context of our findings,suggest that perhaps neither the public nor physiciansshare patients\u00E2\u0080\u0099 highest priority for better test sensitivity.Direct comparison of physicians and patients prefer-ences about genomic testing can provide useful insightabout this matter and should be pursued further in fu-ture research.The distinct characteristic of our study is utilizingthree samples to demonstrate how the type of cancerand its prognosis affected preferences for a genomic test,and how preferences of patients differed from those ofthe public. Also, in contrast with previous studies, theresults of our study are applicable to most genomic testsfor guiding cancer treatment, as we did not specify thetype of cancer, treatment, or the associated genomic test.However, we acknowledge that in the absence of specify-ing the type of cancer, participants may make various as-sumptions about possible prognosis and potentialoutcomes. Therefore, this can be seen as a limitation ofour study as well. Throughout this study, participantsprovided their choices considering the following as-sumptions: 1) if they decided to opt-out from genomictesting, they would receive the new treatment regardlessof its effect, and 2) the new treatment was covered bytheir insurance policies. We acknowledge that under dif-ferent circumstances in terms of the effect of genomictesting on access to the new treatment, the current re-sults may not apply. The larger standard errors aroundthe estimated coefficients in patients suggested that thissample was slightly underpowered. However, the samplesize was restricted to a list of lymphoma patients in BCcancer agency\u00E2\u0080\u0099s contact list and willingness of thoseapproached to participate and thus could not be in-creased. Despite this limitation, all of the point estimatesin the sample of patients were in line with our prior ex-pectations in terms of the order of their magnitudes andcorresponding signs. Moreover, this sample was not anarchetypal sample of cancer patients in BC, as they hadhigh income, high education level, and were 10 yearsolder on average. Therefore, we used propensity scoringto find a subsample of the public with similar character-istics to increase comparability of the results. This issue,however, potentially limits the external validity of the re-sults based on these samples. Actual decisions that pa-tients or the public make in real life situations maydeviate from their stated preference in surveys like ours.This effect has been shown in the context of genetictesting as well [29]. However, several studies provide evi-dence suggesting strong correlation between stated andreal WTP [30] and preferences [31]. Answering DCEquestions can be a complex task and accuracy of re-sponses may eventually depend on participants\u00E2\u0080\u0099 numeracylevel (i.e. ability to interpret quantitative information) [32],language skills, familiarity of subject, and attentivenesswhile completing the questionnaire. We have used sev-eral standard approaches to assure quality of the databy including a fixed choice task to test rationality of re-sponses and by checking the time that each respondentspent on completing the questionnaire. Overall, giventhe directions and signs of the estimated preferenceweights, we believe that our results are robust and havenot been compromised by these potential problems.Finally, we acknowledge that the factors that can affectuptake of a genomic test are not limited to the sevenattributes that we have included in the current DCEdesign. We excluded several important aspects (e.g. riskinvolved in testing procedure) that individuals may takeinto account when making their actual decision aboutusing genomic testing. This selection was to use theminimum possible number of attributes and avoidoverly complex choice tasks [33].Our study demonstrates individuals\u00E2\u0080\u0099 preference strengthtoward characteristics of a genomic test when they arefaced with an aggressive but curable cancer versus a non-aggressive and incurable cancer. Additionally, these resultssuggest which characteristics of genomic testing have alarger potential value for society and patients. Physiciansmay find these average preferences as a benchmarkwhen providing treatment advice about pharmaco-genomics testing to cancer patients. These preferenceNajafzadeh et al. BMC Health Services Research 2013, 13:454 Page 10 of 12http://www.biomedcentral.com/1472-6963/13/454weights also can be used to inform funding decisionsby incorporating relevant populations\u00E2\u0080\u0099 valuation of dif-ferent aspects of genomic testing.ConclusionsWe explored the relative impact of different propertiesof genomically-guided cancer treatment on test uptake.We found that the type and prognosis of cancer affectedpreferences for genomically-guided treatment. Our re-sults also suggest that patients and the public have dif-ferent perceptions about the value of various aspects ofgenomic testing to guide cancer treatment. Physiciansmay find these average preferences as a benchmarkwhen providing treatment advice about pharmacogen-omics testing to cancer patients. These preferenceweights also can be used to inform funding decisions byconsidering relevant populations\u00E2\u0080\u0099 valuation of differentaspects of genomic testing.Additional fileAdditional file 1: The DCE questionnaire.Competing interestsAuthors have no financial or non-financial competing interests in relationwith the content of this study.Authors\u00E2\u0080\u0099 contributionsMN: Design, Data, Analysis, Interpretation, Drafting the Manuscript, and FinalApproval; KJ: Design, Drafting the Manuscript, and Final Approval; SP: Data,Drafting the Manuscript, and Final Approval. JC: Conception, Data,Interpretation, Drafting the Manuscript, and Final Approval; MM: Conception,Drafting the Manuscript, and Final Approval; LL: Interpretation, Drafting theManuscript, and Final Approval; CM: Conception, Design, Data, Interpretation,Drafting the Manuscript, and Final Approval. All authors read and approvedthe final manuscript.AcknowledgementsWe thank all participants who provided their opinions in the experiment,particularly lymphoma patients in British Columbia, who voluntarily acceptedto be part of this research. Mehdi Najafzadeh is grateful for the support fromCIHR (Fredrick Banting and Charles Best Canada Graduate Scholarship).Financial supportThis study was funded by Genome Canada/Genome BC.Author details1Department of Medicine, Harvard Medical School, Boston, MA, USA. 2Facultyof Pharmaceutical Sciences, University of British Columbia, Vancouver, BC,Canada. 3British Columbia Cancer Agency, Vancouver, BC, Canada. 4GenomeSciences Centre, Vancouver, BC, Canada. 5Centre for Health Evaluation andOutcome Sciences (CHEOS), St. Paul\u00E2\u0080\u0099s Hospital, 1081 Burrard Street,Vancouver, BC, Canada. 6Faculty of Medicine, University of British Columbia,Vancouver, Canada.Received: 24 September 2012 Accepted: 28 October 2013Published: 31 October 2013References1. Torpy JM, Lynm C, Glass RM: JAMA patient page. Cancer: the basics. Jama2010, 304(14):1628.2. Mullighan CG: New strategies in acute lymphoblastic leukemia:translating advances in genomics into clinical practice. Clin Cancer Res2010, 17(3):396\u00E2\u0080\u0093400.3. Allison M: Is personalized medicine finally arriving? Nat Biotechnol 2008,26(5):509\u00E2\u0080\u0093517.4. Wolff AC: Liposomal anthracyclines and new treatment approaches forbreast cancer. Oncologist 2003, 8(Suppl 2):25\u00E2\u0080\u009330.5. Capdeville R, Silberman S, Dimitrijevic S: Imatinib: the first 3 years. Eur JCancer 2002, 38(Suppl 5):S77\u00E2\u0080\u0093S82.6. Najafzadeh M, Davis JC, Joshi P, Marra C: Barriers for integratingpersonalized medicine into clinical practice: a qualitative analysis. Am JMed Genet A 2013, 161A(4):758\u00E2\u0080\u0093763.7. Rogausch A, Prause D, Schallenberg A, Brockmoller J, Himmel W: Patients\u00E2\u0080\u0099and physicians\u00E2\u0080\u0099 perspectives on pharmacogenetic testing.Pharmacogenomics 2006, 7(1):49\u00E2\u0080\u009359.8. Conti R, Veenstra DL, Armstrong K, Lesko LJ, Grosse SD: Personalizedmedicine and genomics: challenges and opportunities in assessingeffectiveness, cost-effectiveness, and future research priorities. Med DecisMaking 2010, 30(3):328\u00E2\u0080\u0093340.9. Grosse SD, Wordsworth S, Payne K: Economic methods for valuing theoutcomes of genetic testing: beyond cost-effectiveness analysis. GenetMed 2008, 10(9):648\u00E2\u0080\u0093654.10. Wittink MN, Cary M, Tenhave T, Baron J, Gallo JJ: Towards patients-centered care for depression: conjoint methods to tailor treatmentbased on preferences. Patient 2010, 3(3):145\u00E2\u0080\u0093157.11. Sullivan R, Peppercorn J, Sikora K, Zalcberg J, Meropol NJ, Amir E, Khayat D,Boyle P, Autier P, Tannock IF, et al: Delivering affordable cancer care inhigh-income countries. Lancet Oncol 2011, 12(10):933\u00E2\u0080\u0093980.12. Louviere JJ, Hensher DA, Swait JD: Stated Choice Models. Analysis andApplications: Cambridge University Press; 2000.13. McFadden D: Econometric models for probabilistic choice amongproducts. J Bus 1980, 53(3):S13\u00E2\u0080\u0093S29.14. Lancsar E, Louviere J: Conducting discrete choice experiments to informhealthcare decision making: a user\u00E2\u0080\u0099s guide. Pharmacoeconomics 2008,26(8):661\u00E2\u0080\u0093677.15. Louviere JJ, Islam T, Wasi N, Street D, Burgess L: Designing discrete choiceexperiments: Do optimal designs come at a price? J Consum Res: AnInterdisciplinary Quarterly 2008, 35(2):360\u00E2\u0080\u0093375.16. Haga SB, Burke W, Ginsburg GS, Mills R, Agans R: Primary care physicians\u00E2\u0080\u0099knowledge of and experience with pharmacogenetic testing. Clin Genet2012, 82(4):388\u00E2\u0080\u0093394.17. Payne K, Fargher EA, Roberts SA, Tricker K, Elliott RA, Ratcliffe J, Newman WG:Valuing pharmacogenetic testing services: a comparison of patients\u00E2\u0080\u0099 andhealth care professionals\u00E2\u0080\u0099 preferences. Value Health 2011, 14(1):121\u00E2\u0080\u0093134.18. Issa AM, Tufail W, Hutchinson J, Tenorio J, Baliga MP: Assessing patientreadiness for the clinical adoption of personalized medicine. PublicHealth Genomics 2009, 12(3):163\u00E2\u0080\u0093169.19. Bech M, Gyrd-Hansen D: Effects coding in discrete choice experiments.Health Econ 2005, 14(10):1079\u00E2\u0080\u00931083.20. Kontoleon A, Yabe M: Assessing the impacts of alternative \u00E2\u0080\u0098Opt-out\u00E2\u0080\u0099 formats inchoice experiment studies: consumer preferences for genetically modifiedcontent and production information in food. J Agric Policy Res 2005, 5:1\u00E2\u0080\u009343.21. Train KE: Discrete choice methods with simulation. Cambridge UniversityPress; 2003:334. ISBN 0-521-81696-3.22. Greene WH, Hensher DA: A latent class model for discrete choice analysis:contrasts with mixed logit. Transp Res B Methodol 2003, 37(8):681\u00E2\u0080\u0093698.23. Louviere J: What you don\u00E2\u0080\u0099t know might hurt you: some unresolvedissues in the design and analysis of discrete choice experiments. EnvironResour Econ 2006, 34(1):173\u00E2\u0080\u0093188.24. Griffith GL, Edwards RT, Williams JM, Gray J, Morrison V, Wilkinson C, TurnerJ, France B, Bennett P: Patient preferences and National Health Servicecosts: a cost-consequences analysis of cancer genetic services. FamCancer 2008, 27:27.25. Hall J, Fiebig DG, King MT, Hossain I, Louviere JJ: What influencesparticipation in genetic carrier testing? Results from a discrete choiceexperiment. J Health Econ 2006, 25(3):520\u00E2\u0080\u0093537.26. Regier DA, Ryan M, Phimister E, Marra CA: Bayesian and classicalestimation of mixed logit: an application to genetic testing. J Health Econ2009, 28(3):598\u00E2\u0080\u0093610.27. Herbild L, Gyrd-Hansen D, Bech M: Patient preferences for pharmacogeneticscreening in depression. Int J Technol Assess Health Care 2008, 24(1):96\u00E2\u0080\u0093103.28. Marshall DA, Johnson FR, Phillips KA, Marshall JK, Thabane L, Kulin NA:Measuring patient preferences for colorectal cancer screening using achoice-format survey. Value Health 2007, 10(5):415\u00E2\u0080\u0093430.Najafzadeh et al. BMC Health Services Research 2013, 13:454 Page 11 of 12http://www.biomedcentral.com/1472-6963/13/45429. Sanderson SC, O\u00E2\u0080\u0099Neill SC, Bastian LA, Bepler G, McBride CM: What caninterest tell us about uptake of genetic testing? Intention and behavioramongst smokers related to patients with lung cancer. Public HealthGenomics 2010, 13(2):116\u00E2\u0080\u0093124.30. Bryan S, Jowett S: Hypothetical versus real preferences: results from anopportunistic field experiment. Health Econ 2010, 19(12):1502\u00E2\u0080\u00931509.31. Mark TL, Swait J: Using stated preference and revealed preference modelingto evaluate prescribing decisions. Health Econ 2004, 13(6):563\u00E2\u0080\u0093573.32. Woloshin S, Schwartz LM, Moncur M, Gabriel S, Tosteson AN: Assessing valuesfor health: numeracy matters. Med Decis Making 2001, 21(5):382\u00E2\u0080\u0093390.33. Reed Johnson F, Lancsar E, Marshall D, Kilambi V, Muhlbacher A, Regier DA,Bresnahan BW, Kanninen B, Bridges JF: Constructing experimental designsfor discrete-choice experiments: report of the ISPOR Conjoint AnalysisExperimental Design Good Research Practices Task Force. Value Health2013, 16(1):3\u00E2\u0080\u009313.doi:10.1186/1472-6963-13-454Cite this article as: Najafzadeh et al.: Genomic testing to determine drugresponse: measuring preferences of the public and patients usingDiscrete Choice Experiment (DCE). BMC Health Services Research2013 13:454.Submit your next manuscript to BioMed Centraland take full advantage of: \u00E2\u0080\u00A2 Convenient online submission\u00E2\u0080\u00A2 Thorough peer review\u00E2\u0080\u00A2 No space constraints or color figure charges\u00E2\u0080\u00A2 Immediate publication on acceptance\u00E2\u0080\u00A2 Inclusion in PubMed, CAS, Scopus and Google Scholar\u00E2\u0080\u00A2 Research which is freely available for redistributionSubmit your manuscript at www.biomedcentral.com/submitNajafzadeh et al. BMC Health Services Research 2013, 13:454 Page 12 of 12http://www.biomedcentral.com/1472-6963/13/454"@en . "Article"@en . "10.14288/1.0221457"@en . "eng"@en . "Reviewed"@en . "Vancouver : University of British Columbia Library"@en . "BioMed Central"@en . "10.1186/1472-6963-13-454"@en . "Attribution 4.0 International (CC BY 4.0)"@en . "http://creativecommons.org/licenses/by/4.0/"@en . "Faculty"@en . "Pharmacogenomics"@en . "Genomic medicine"@en . "Personalized medicine"@en . "Genetic testing"@en . "Discrete choice experiment"@en . "Conjoint analysis"@en . "Preference elicitation"@en . "Cancer treatment"@en . "Genomic testing to determine drug response: measuring preferences of the public and patients using Discrete Choice Experiment (DCE)"@en . "Text"@en . "http://hdl.handle.net/2429/55958"@en .