You are here

Article Text

Article menu
Screening
Original article
Direct-to-consumer pharmacogenomic testing is associated with increased physician utilisation
  1. Cinnamon S Bloss1,
  2. Nicholas J Schork1,2,
  3. Eric J Topol1,2,3
  1. 1Scripps Genomic Medicine, Scripps Translational Science Institute, and Scripps Health, La Jolla, California, USA
  2. 2Department of Molecular and Experimental Medicine, The Scripps Research Institute, La Jolla, California, USA
  3. 3Scripps Clinic, La Jolla, California, USA
  1. Correspondence to Dr Cinnamon S Bloss, Scripps Translational Science Institute, 3344 N. Torrey Pines Court, Suite 300, La Jolla, California 92037, USA; cbloss{at}scripps.edu

Abstract

Background Direct-to-consumer (DTC) genomic testing has generated controversy, however the actual impact of testing on consumer behaviour has been understudied, particularly for pharmacogenomic (PGx) testing.

Methods We recruited a sample of adults who purchased a DTC genomic test and had previously received their genomic test results for complex disease risk. All participants additionally underwent PGx testing. At follow-up, to assess the impact of PGx testing on consumer behaviour, healthcare utilisation and psychological status were compared between approximately a third of participants who had received their PGx results and the remaining two-thirds of participants who were still awaiting results. The PGx test included genetic testing for drug effectiveness or risk of side effects for 12 medications.

Results At follow-up, there were 481 PGx test recipients and 844 non-recipients still awaiting results. PGx test recipients had more physician visits (p=0.04) and were more likely to share their results with their physician (p=0.001). Both groups showed a decrease in anxiety symptoms from baseline to follow-up, with a trend for PGx recipients to show less of a decrease compared with non-recipients (p=0.10). PGx recipients were more likely to report that their physician ordered additional tests (p=0.01) based on their genomic test. There were no group differences in follow-up test-related distress (p=0.67).

Conclusions DTC PGx risk profiling among a selected sample of individuals was associated with increased physician utilisation and did not result in any adverse changes in psychological health or follow-up test-related distress.

  • direct-to-consumer
  • genetic testing
  • personalized medicine
  • genomic risk assessment
  • consumer genomics

https://doi.org/10.1136/jmedgenet-2013-101909

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Direct-to-consumer (DTC) genomic testing companies are known for providing consumers with personal risk estimates for common complex diseases, but the majority of leading companies that offer this type of testing also provide pharmacogenomic (PGx) profiling services. Such services range from testing a handful of SNPs for information about drug response for selected medication-gene pairs,1 to more expanded offerings from some companies that encompass genetic assessments not strictly related to clinical pharmacogenomics, such as for caffeine and alcohol metabolism.2 Notably, DTC PGx testing leverages information derived from genome-wide association studies, which have been markedly more successful for identifying large effect variants associated with drug response3–5 versus the typical small effect variants discovered for complex disease susceptibility.

    Some have argued that PGx testing of variants with high predictive value does offer clear opportunities for clinical actionability.6 DTC PGx testing may also serve to raise public awareness of genome analysis and its relevance to therapeutic drug response. Further, individuals taking relevant medications may be able to avoid suboptimal or adverse responses to these medications by leveraging their genetic information to guide drug selection and or dosing. In contrast, however, DTC PGx testing has also been met with criticisms that are now familiar to the broader DTC genomic testing model. These include the fact that delivery does not involve a healthcare intermediary, that consumers may be confused or have poor psychological responses to the information provided, and that testing could potentially result in costly and unnecessary healthcare utilisation with little relative benefit. On this latter point, if care use is increased appropriately (eg, a patient finds they have a variant relevant to a drug they use), then this is presumably a good outcome. If, however, a patient makes appointments to simply share results with providers or out of worry or confusion without a clear clinical need, then this could be construed as a more negative outcome. Finally, and perhaps most importantly, there are a lack of studies demonstrating the clinical validity, clinical utility and cost-effectiveness of DTC PGx risk profiling. Despite these criticisms, however, there is a growing population of consumers who are purchasing these tests, which are becoming cheaper at less than $99 per sample7 and therefore more widely available. Thus research to determine the impact of testing on consumers and healthcare providers is critical.

    The Scripps Genomic Health Initiative (SGHI) was originally designed to measure the impacts of consumer genomic testing for common complex disease risk.8 ,9 All participants additionally chose to undergo PGx testing, however at follow-up, approximately a third of participants had received their PGx results whereas the remaining two-thirds of participants were still waiting for results. This enabled a secondary case-control analysis of response to PGx testing among our study participants, all of whom had already received genomic susceptibility testing for common disease.

    There have been concerns raised in the literature that DTC genomic profiling may increase healthcare utilisation. Further, there is a natural link between tests pertaining to prescription medications and the typical role of physicians in providing access to and ongoing care and follow-up surrounding the use of such medications. Based on these factors, we aimed to test the hypothesis that there may be increased physician-related healthcare utilisation among PGx test recipients, particularly among PGx recipients at high genetic risk. Further support of this rationale comes from similar studies that have suggested a correlation between DTC advertising of prescription drugs and increased healthcare utilisation, particularly physician utilisation.10 Since it has also been suggested that DTC genomic testing has the potential to cause adverse psychological responses, we additionally compared psychological outcomes between groups.

    Materials and methods

    This study was approved by the Scripps Office for the Protection of Research Subjects and Institutional Review Boards. Informed consent was obtained electronically from each study participant. Details of our methods have been published elsewhere.8 ,9

    Study design and instruments

    The SGHI was designed as a longitudinal cohort study in which all participants were administered baseline (pre-risk-disclosure), as well as 3-month and 12-month follow-up (post-risk-disclosure) health assessments using the web-based survey tool, SurveyMonkey.11 PGx testing was administered to a subset of study participants between the 3-month and 12-month follow-up assessments and therefore survey data from the 12-month follow-up was used in the current study.

    Healthcare utilisation and psychological response

    Our baseline and follow-up assessment batteries have been previously described.8 ,9 ,12 Assessment of psychological status included measurement of state anxiety13 and test-related distress.14 In terms of assessment of healthcare utilisation, at baseline we asked participants to report the number of physician visits they had completed in the 12-month period prior to completion of the baseline assessment, and at 12-month follow-up we asked them to report the number of physician visits they had completed since receiving their results. There are previous studies that have shown strong correlations between self-reported and medical records-based healthcare utilisation rates, including physician visits and utilisation.15 At follow-up we also asked participants to report the number of visits to non-physician healthcare providers, as well as whether or not the participant had shared their results with their physician and whether their physician ordered additional tests based on the genetic test results. We further queried participants as to whether or not they had spoken to a Navigenics genetic counsellor about their results.

    Perceptions of DTC genomic testing

    We also surveyed participants with respect to their attitudes and perceptions of DTC genomic testing. Specifically, participants were presented with items to gauge literacy (‘Do you understand your test results?’), perceived utility of the genomic test (‘In general, do you feel your genetic test results are useful to you?’) and privacy concerns (‘Are you concerned about privacy issues or that you may be discriminated against based on your genetic test results?’). We further included items pertaining to social outcomes of the genomic testing. Specifically, we queried participants as to whether they shared their results with any family members, as well as whether they would recommend testing to family members and/or non-family members such as friends and colleagues.

    Procedures

    Participants were recruited primarily from health and technology companies. All participants either paid for their test directly, or their employer paid on their behalf, thus they are all considered ‘consumers’ as well as research participants. Study procedures pertaining to enrolment and administration of the baseline and 3-month health assessments have been described in detail elsewhere.8 ,9 On the 365th day after participants received their genomic disease susceptibility test results, an email was sent to each participant requesting that they complete the 12-month follow-up health assessment. The 12-month survey was closed to completion in March 2011.

    All participants received their PGx results in May 2010 or shortly thereafter. However, for some individuals, depending on when they originally enroled in the study, this time point was prior to completion of their 12-month follow-up and for some it was after completion of their 12-month follow-up. Thus, participants were not prospectively assigned to receive PGx testing at different time points; rather this was determined by when they happened to enrol in the study. Participants did not consent again or provide a new sample for PGx testing as this component of the genomic test was included in the initial study consent.

    Genomic test and safety monitoring

    We examined the impact of PGx testing with the ‘Medications’ component of the Navigenics Health Compass,1 a commercially available test at the time the study was conducted16 (see online supplementary figures 1 and 2). The PGx test included personal genomic results pertaining to drug effectiveness or risk of side effects for a total of 12 medications, including abacavir hypersensitivity, β blocker response, carbamazepine adverse side effects, clopidogrel efficacy, floxacillin toxicity, fluorouracil toxicity, irinotecan adverse side effects, statins (response and risk of myopathy), pseudocholinesterase deficiency, azathioprine/6-mercaptopurine toxicity and warfarin sensitivity. Genetic counselling provided by Navigenics’ staff of board-certified genetic counsellors was made available at no charge to SGHI participants.

    Although they provided the DTC genomic test used in the study, as well as genetic counselling services for participants, Navigenics was not involved in the data analysis or manuscript preparation for this study.

    Outcome measures

    Primary outcomes were differences between baseline and follow-up in the number of physician visits and degree of anxiety symptoms for PGx recipients versus non-recipients. Having shared results with a physician at follow-up was also a primary outcome between study groups. Secondary outcomes were all assessed at follow-up and included whether or not the physician ordered additional tests based on the genomic results (for physician sharers), the number of visits to non-physician healthcare providers, having shared results with a genetic counsellor and level of test-related distress. Perceived utility, understandability and privacy concerns were also compared between groups, as were social outcomes, including whether results were shared with family members and whether participants would recommend genomic testing to family members and/or non-family members.

    Some previous studies have suggested that the magnitude of genetic risk disclosed can influence behavioural responses to genetic testing.17 Therefore, we also conducted analysis of the same set of outcomes between individuals for whom there was at least one instance of receiving a PGx high-risk result for a medication that they reported either current or previous use of (‘PGx high risk’) versus individuals for whom there were no such instances (‘PGx low risk’).

    Data analysis

    All statistical analyses were conducted using the statistical software package SPSS V.14.0 and VasserStats web utility for computing z-tests and CIs for proportions.18 Two-sided t-tests, Mann-Whitney U tests or χ2 tests were used to compare baseline variables between PGx recipients and non-recipients. We used repeated measures analysis of variance controlling for eight covariates (age, sex, education, ancestry (a dichotomous variable reflecting self-identified Caucasian: yes/no), income, health-related occupation, follow-up interval in days and completion of the 3-month assessment) to assess the extent to which receipt of PGx testing was associated with change in number of physician visits, as well as change in scores on the state-trait anxiety inventory between baseline and follow-up. Analysis of covariance was used to test the association between receipt of PGx testing and number of visits to non-physician healthcare providers and follow-up test-related distress, also accounting for our eight covariates. Logistic regression was used to assess the association between receipt of PGx testing and whether results were shared with a physician (yes/no), whether the physician ordered additional tests based on the genomic test (yes/no) and whether results were shared with a genetic counsellor (yes/no). Ordinal and logistic regression approaches were used to test the association between receipt of PGx testing and perceptions of DTC genomic testing and social outcomes related to testing. Analogous approaches were used to test the effect of receiving a high versus low PGx risk result among PGx recipients. All reported p values are uncorrected.

    Results

    Sample characteristics

    Figure 1 depicts enrolment numbers and outcomes. Descriptive statistics and comparisons between PGx recipients (n=481) and non-recipients (n=844) on demographic and baseline outcome variables are shown in table 1. Individuals who received PGx were younger, less likely to be in a health-related profession and more likely to have completed the 3-month follow-up assessment. Notably, however, there were no differences between recipients and non-recipients in baseline measures of either primary outcome variable (ie, number of physician visits and state anxiety). The average follow-up interval was 14.0 (SD=1.3) and did not differ between study groups. The same baseline comparisons were also performed between PGx recipients at high risk (n=41) versus low risk (n=434; see online supplementary table 1). Of note, nearly the entire cohort reported having health insurance (1321 out of 1325 individuals, or 99.7%).

    Figure 1
    Figure 1

    Enrolment, study groups and outcomes. Individuals who withdrew cited financial reasons, insurance concerns and change of mind. Non-compliance was when an individual had not completed the baseline health assessment after three email requests. Technical issues means that the individual submitted duplicate or triplicate surveys, and protocol deviation indicates that the individual's genetic results were unintentionally released to them prior to completion of the baseline health assessment. Twelve-month follow-up assessments were solicited from individuals who had completed the 3-month follow-up and individuals who had not completed the 3-month follow-up. PGx, pharmacogenomic; QC, quality control.

    View this table:
    Table 1

    Demographic and baseline behavioral characteristics for pharmacogenomic (PGx) recipients and non-recipients

    All individuals in the study had previously received personal genomic testing for risk of 28 common diseases. Therefore, we also compared these risk estimates between PGx recipients and non-recipients (see online supplementary table 2) to ensure that disease risk was not a confounding factor in our analysis. As shown, after correcting for multiple comparisons (ie, 28), there were no significant differences in the risk estimates between the groups.

    Impact of pharmacogenomic testing

    Healthcare utilisation

    PGx test recipients had an increase in physician visits at follow-up (p=0.04; see online supplementary figure 3) and were more likely to share their results with their physician (p=0.001; table 2) relative to non-recipients. Further, the within-subjects change (increase) in physician visits was statistically significant for the group that received PGx results (p=0.03) and non-significant for those awaiting PGx results (p=0.73).

    View this table:
    Table 2

    Impacts of receiving pharmacogenomic (PGx) testing on study outcomes at 12-month follow-up (n=844 No PGx, n=481 Received PGx)

    PGx recipients were also more likely to report that their physician ordered additional tests (p=0.01) based on their genomic results, and were more likely to discuss their results with a genetic counsellor (p<0.0005). There were no group differences at follow-up in number of visits to non-physician healthcare providers (p=0.22).

    Timing of PGx results delivery

    For those individuals who received PGx results, we assessed the length of time these participants had had access to their results at the time they completed the 12-month follow-up assessment. The mean number of days with access was 68.8 (2.29 months). We further tested the extent to which the number of days with access to PGx results was correlated with the change in number of physician visits from baseline to 12-month follow-up and found a significant positive correlation (rs=0.120, p=0.009) characterised by a greater number of days with access to PGx results being associated with a greater increase in physician visits at 12-month follow-up.

    Psychological response

    Both groups showed a non-significant decrease in anxiety symptoms from baseline to follow-up, with a trend for PGx recipients to show less of a decrease compared with non-recipients (p=0.10; table 2). There were no group differences in follow-up test-related distress (p=0.67).

    Perceptions of DTC genomic testing

    The majority of individuals in both study groups indicated that their results were useful (No PGx: 56.7% yes, 38.1% somewhat; Received PGx: 70.0% yes, 27.1% somewhat; see online supplementary figure 4) and that they understood their results (No PGx: 69.9% yes, 28.5% somewhat; Received PGx: 82.6% yes, 16.6% somewhat), though individuals in the PGx group were more likely to report utility (p<0.0005) and understandability (p<0.0005; table 3). Although the majority of individuals in both study groups also indicated that they did not have privacy concerns (No PGx: 69.6% no; Received PGx: 62.5%), individuals in the PGx group were more likely to report privacy concerns (p=.02).

    View this table:
    Table 3

    Perceptions and social outcomes of the genomic test at 12-month follow-up in pharmacogenomic (PGx) recipients and non-recipients

    Although a greater percentage of PGx recipients versus non-recipients indicated that they shared their results with a family member (No PGx: 84.8%; Received PGx: 89.2%) this difference was not statistically significant (p=0.08; table 3). The majority of individuals in both groups also indicated that they would recommend the test to others, however, PGx recipients were more likely to indicate that they would recommend it to family members (p=0.002) and non-family members (p=0.002; table 3).

    Impact of high versus low pharmacogenomic risk

    Among PGx recipients, the median number of PGx risks returned was 2 out of a total of 12 that were tested (Mean=2.4, SD=1.2). Further, of those who received PGx, there were a total of 41 individuals who had at least one instance of receiving a PGx risk for a medication they were either currently taking or had previously taken (see online supplementary table 3).

    Individuals at high PGx risk had a greater increase in physician visits (p=0.006; see online supplementary figure 3), were more likely to share their results with their physician (p=0.04), and were more likely to speak with a genetic counsellor about their results (p=0.02) at follow-up relative to individuals at low PGx risk (table 4). The within-subjects change (increase) in physician visits among those individuals at low PGx risk was significant (p=0.02), however, among those at high PGx risk this effect did not reach statistical significance (p=0.85), likely due to low power to detect the effect with such a small sample size (n=41).

    View this table:
    Table 4

    Impacts of receiving pharmacogenomic (PGx) testing on study outcomes at 12-month follow-up in high risk versus low risk PGx recipients (n=434, Low Risk PGx, n=41 High Risk PGx)

    Individuals at high PGx risk did not differ from those in the low risk group in terms of whether or not their physician ordered additional tests based on their genomic results (p=0.40) or in their number of visits to a non-physician healthcare provider (p=0.74) at follow-up.

    Psychological response

    Individuals at high PGx risk were not found to differ from low risk individuals in anxiety symptoms between baseline and follow-up (p=0.99), and there were no group differences in follow-up test-related distress (p=0.41).

    Perceptions of DTC genomic testing

    Individuals at high PGx risk did not differ from low risk individuals with respect to their perceptions of DTC genomic testing or with respect to any of the social outcomes assessed (see online supplementary table 4).

    Discussion

    Receipt of a commercially available DTC PGx risk test was associated with increases in physician utilisation among a selected sample of individuals. Specifically, PGx test recipients were 1.5 times more likely to share their genomic results with their physician than non-recipients. PGx recipients also showed a statistically significant increase in the number of physician visits from baseline to follow-up relative to non-recipients, though importantly, the average increase, as well as average difference from non-recipients, was less than one visit. Our data and sample size do not permit us to assess the cost-effectiveness of DTC PGx testing for the panel of drug-gene pairs for which our participants received results, or whether testing resulted in improved precision in prescriptions leading to avoidance of major side effects or assurance of efficacy or optimal dosing. It is worth noting, however, that while the Agency for Healthcare Research and Quality reported that in 2006, an office physician visit cost $180 on average,19 costs associated with adverse drug events are known to be at least an order of magnitude higher, reflecting the need for hospitalisations, and some are associated with fatal outcomes.20 Our findings further highlight the need for more research, including cost-effectiveness studies, on the impact of DTC genomic risk testing on patients, physicians and healthcare systems.

    The largest consumer genomics company, 23andMe, has now sold their genomic testing panel to over 200 000 individuals. The panel costs $997 and includes PGx testing for over 20 medications,2 including the majority of the drug-DNA variant interactions in the present study. Accordingly, cost-effectiveness can be assessed in the context of the diminishing cost for a fairly extensive PGx panel.

    An emerging literature documenting the lack of physician education, training and knowledge in genomics is relevant to our findings. For instance, a recent survey found that only 10% of physicians felt he or she had the necessary training and knowledge in genomics to provide adequate care in this area to their patients,21 and other studies have produced similar findings.22 ,23 Fortunately there are a handful of recently initiated efforts24 to bridge this gap, but significant work remains. This is underscored by our findings that a large fraction of DTC genomic test consumers, and an even larger fraction of PGx test recipients specifically, do turn to their physicians spontaneously with their results, presumably for additional advice and counsel regarding how their genomic findings may impact their health generally and prescription drug use specifically. Notably, PGx recipients in our sample were also more likely to discuss their results with a genetic counsellor, a service that was offered to all participants free of charge. While this finding raises questions regarding the possible importance of making genetic counselling services available to individuals undergoing PGx risk testing, it also exposes potential problems stemming from the very low numbers of trained genetic counsellors in the USA.

    Importantly, our sample consisted of a self-selected group of individuals, likely representative of the current population of consumers of DTC genomic testing, who chose to undergo and pay out of pocket for testing. In comparison, another study of a young (age 25–40 years), healthy, insured population25 found lower levels of baseline utilisation, as well as no difference in the use of health services before and after free multiplex genetic risk testing for eight common diseases (the study did not offer or provide PGx testing to participants). In addition to the fact that PGx results were not provided, other factors that may account for differences in findings between these two studies are the inclusion of individuals in our study who were much older (up to age 80+ years) and therefore perhaps more likely to have higher rates of baseline utilisation, the possibility that cost-sharing for the test in our study motivated increased healthcare seeking behaviours, as well as the relatedly high income and education levels of our sample. Taken together, observations from each study raise several hypotheses, including that responses to genetic testing may be test specific (ie, testing for PGx vs complex disease risk), cost structure specific (ie, free vs co-pay), and somewhat population and demographic specific. Further work is needed to better understand the impacts of DTC PGx risk testing on representative samples of individuals in USA and other countries in which consumers are purchasing such tests.

    PGx testing was not associated with changes in psychological health or increases in test-related distress. In addition, while the large majority of participants in both groups reported that the DTC genomic test was useful, understandable, and that they did not have privacy concerns, PGx test recipients were somewhat more likely to perceive their test results as useful, understandable and to indicate privacy concerns (14% of PGx recipients vs 11% of non-recipients). This latter finding regarding privacy concerns is notable, and it is unclear what may be driving this observation. One possibility is that PGx information may be more likely to be used by physicians and or added to an individual’s medical record, which may be concerning for some individuals who choose to undergo testing. Consistent with this, we and others have found evidence that some groups of DTC genomic test consumers may be hesitant to share or discuss their genetic results with a healthcare provider precisely because of privacy concerns.26 ,27 Further research will be needed to better understand this result and possible explanations. The finding that PGx test recipients were also more likely to rate the test as more understandable is somewhat counterintuitive. It is possible that PGx recipients were answering this question with the PGx results in mind while those who were still awaiting their results had the more uncertain complex disease risk results in mind.

    It is also notable that a very large proportion of our sample reported sharing their results with family members (over 80% in both groups), and that those who received PGx testing were statistically more likely to recommend testing to family and non-family members. This suggests that within the context of the DTC model, consumers also seek health-related counsel and support from other sources, including family. Testing may serve to engender health-related discussions within families and promote more accurate knowledge of family disease28 and prescription drug use histories, something that has been promoted by prominent public health groups, including the US Centers for Disease Control and Prevention.29

    Our study has some limitations. First, we studied a sample of convenience that consisted of individuals who elected to undergo DTC genomic testing for common disease risk. As such, though our sample has been shown to be representative of the current population of consumers of DTC genomic tests,9 it is not representative of the general population (ie, given the high education and income of the sample, and the fact that over 99% of individuals had health insurance), and therefore it is unknown how these findings may or may not generalise to a broader population. Second, though our sample size is relatively large as compared with typical behavioural studies of response to genetic and genomic information,25 ,30 ,31 it was not adequate to address some questions of interest. For instance, given the low frequency of some of the genetic variants included on the PGx test panel, coupled with the relatively low frequency of use of any of the medications on the panel among individuals in our sample, it does not allow us to document whether or to what degree DTC PGx testing resulted in improved precision in prescriptions leading to better drug-related outcomes. Third, although we did assess differences between PGx recipients and non-recipients in the number of non-physician visits at follow-up, we did not specifically evaluate the extent to which participants may have shown differences specifically in their utilisation of pharmacists, which would be an important question to pose in future studies.32 Fourth, we did not use a design in which participants were randomised to have either received PGx or be on a wait-list control. Finally, the data collected also do not distinguish how many physician visits occurred after receipt of PGx results and, similarly, data are not available on whether results disclosed to physicians were PGx results or other DTC results, or whether additional tests ordered or counsellor visits were PGx-related. Further, it is possible that the increase in the number of physician visits could be due to the effect of getting more information, rather than PGx-information-specific effects.

    Our study also relied on brief, web-based, self-report assessment of health behaviours. Although such assessments are not ideal as they can be less reliable and more prone to participant recall biases than inperson assessment,33 studies have suggested strong correlations between self-reported and medical records-based healthcare utilisation rates.15 Our findings are also based on a single follow-up assessment and do not speak to the effects of PGx testing over the course of multiple years, allowing for a timeframe in which some participants may develop health issues that may require the use of one or more medications included in the panel.

    In conclusion, DTC PGx risk profiling among a selected sample of persons was associated with statistically significant increases in physician utilisation. PGx testing did not result in any short-term changes in psychological health. Further evaluation of the utility and cost considerations for DTC PGx testing appear to be warranted.

    Acknowledgments

    The authors acknowledge the support of Laura Ornowski, MS of Scripps who assisted with data collection, as well as Vance Vanier, MD, Michele Cargill, PhD, and Elana Silver, MS of Navigenics, along with their genetic counsellors and other staff, who helped support the project. The authors also thank Burcu F Darst for her assistance with manuscript preparation.

    References

    1. ↵
      Navigenics. http://www.navigenics.com/ (accessed 1 Jul 2011).
    2. ↵
      23andMe. https://www.23andme.com/ (accessed 1 Jul 2011).
    3. ↵
      1. Ge D,
      2. Fellay J,
      3. Thompson AJ,
      4. Simon JS,
      5. Shianna KV,
      6. Urban TJ,
      7. Heinzen EL,
      8. Qiu P,
      9. Bertelsen AH,
      10. Muir AJ,
      11. Sulkowski M,
      12. McHutchison JG,
      13. Goldstein DB
      . Genetic variation in IL28B predicts hepatitis C treatment-induced viral clearance. Nature 2009;461:399–401.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. McCormack M,
      2. Alfirevic A,
      3. Bourgeois S,
      4. Farrell JJ,
      5. Kasperaviciute D,
      6. Carrington M,
      7. Sills GJ,
      8. Marson T,
      9. Jia X,
      10. de Bakker PI,
      11. Chinthapalli K,
      12. Molokhia M,
      13. Johnson MR,
      14. O'Connor GD,
      15. Chaila E,
      16. Alhusaini S,
      17. Shianna KV,
      18. Radtke RA,
      19. Heinzen EL,
      20. Walley N,
      21. Pandolfo M,
      22. Pichler W,
      23. Park BK,
      24. Depondt C,
      25. Sisodiya SM,
      26. Goldstein DB,
      27. Deloukas P,
      28. Delanty N,
      29. Cavalleri GL,
      30. Pirmohamed M
      . HLA-A*3101 and carbamazepine-induced hypersensitivity reactions in Europeans. N Engl J Med 2011;364:1134–43.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Chen P,
      2. Lin JJ,
      3. Lu CS,
      4. Ong CT,
      5. Hsieh PF,
      6. Yang CC,
      7. Tai CT,
      8. Wu SL,
      9. Lu CH,
      10. Hsu YC,
      11. Yu HY,
      12. Ro LS,
      13. Lu CT,
      14. Chu CC,
      15. Tsai JJ,
      16. Su YH,
      17. Lan SH,
      18. Sung SF,
      19. Lin SY,
      20. Chuang HP,
      21. Huang LC,
      22. Chen YJ,
      23. Tsai PJ,
      24. Liao HT,
      25. Lin YH,
      26. Chen CH,
      27. Chung WH,
      28. Hung SI,
      29. Wu JY,
      30. Chang CF,
      31. Chen L,
      32. Chen YT,
      33. Shen CY
      . Carbamazepine-induced toxic effects and HLA-B*1502 screening in Taiwan. N Engl J Med 2011;364:1126–33.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Brunham LR,
      2. Hayden MR
      . Medicine. Whole-genome sequencing: the new standard of care? Science (New York, NY 2012;336:1112–3.
      OpenUrlCrossRef
    7. ↵
      1. Brenner K
      . 23andMe wants to change the face of health care. Fortune 12 December 2012.
    8. ↵
      1. Bloss CS,
      2. Ornowski L,
      3. Silver E,
      4. Cargill M,
      5. Vanier V,
      6. Schork NJ,
      7. Topol EJ
      . Consumer perceptions of direct-to-consumer personalized genomic risk assessments. Genet Med 2010;12:556–66.
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      1. Bloss CS,
      2. Schork NJ,
      3. Topol EJ
      . Effect of direct-to-consumer genomewide profiling to assess disease risk. The N Engl J Med 2011;364:524–34.
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      1. Dorn SD,
      2. Farley JF,
      3. Hansen RA,
      4. Shah ND,
      5. Sandler RS
      . Direct-to-consumer and physician promotion of tegaserod correlated with physician visits, diagnoses, and prescriptions. Gastroenterology 2009;137:518–24, 24 e1–2.
      OpenUrlCrossRefPubMed
    11. ↵
      SurveyMonkey. http://www.surveymonkey.com/ (accessed 27 Jul 2011).
    12. ↵
      1. Bloss CS,
      2. Topol EJ,
      3. Schork NJ
      . Association of direct-to-consumer genome-wide disease risk estimates and self-reported disease. Genet Epidemiol 2011;36:66–70.
      OpenUrlPubMed
    13. ↵
      1. Spielberger CD,
      2. Gorsuch RL,
      3. Lushene PR,
      4. Vagg PR,
      5. Jacobs AG
      . Manual for the State-Trait Anxiety Inventory (Form Y). Palo Alto: Consulting Psychologists Press, Incorporated, 1983.
    14. ↵
      1. Creamer M,
      2. Bell R,
      3. Failla S
      . Psychometric properties of the Impact of Event Scale—Revised. Behav Res Ther 2003;41:1489–96.
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Ritter PL,
      2. Stewart AL,
      3. Kaymaz H,
      4. Sobel DS,
      5. Block DA,
      6. Lorig KR
      . Self-reports of health care utilization compared to provider records. J Clin Epidemiol 2001;54:136–41.
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      1. Lauerman J
      . Life Technologies Buys Navigenics for Genetic Diagnostics. Bloomberg News 16 July 2012.
    17. ↵
      1. Cameron LD,
      2. Sherman KA,
      3. Marteau TM,
      4. Brown PM
      . Impact of genetic risk information and type of disease on perceived risk, anticipated affect, and expected consequences of genetic tests. Health Psychol 2009;28:307–16.
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      VassarStats: Website for Statistical Computation. http://vassarstats.net/propdiff_ind.html (accessed 15 May 2012).
    19. ↵
      1. Machlin SR,
      2. Kress M
      . Trends in Health Care Expenditures for Adults Ages 18–44: 2006 versus 1996. Agency for Healthcare Research and Quality, 2009.
    20. ↵
      1. Classen DC,
      2. Pestotnik SL,
      3. Evans RS,
      4. Lloyd JF,
      5. Burke JP
      . Adverse drug events in hospitalized patients. Excess length of stay, extra costs, and attributable mortality. Jama 1997;277:301–6.
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      1. Healy M
      . As genetic testing races ahead, doctors are left behind. Los Angeles Times 24 October 2009.
    22. ↵
      1. Scheuner MT,
      2. Sieverding P,
      3. Shekelle PG
      . Delivery of genomic medicine for common chronic adult diseases: a systematic review. Jama 2008;299:1320–34.
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      1. Haga S,
      2. Burke W,
      3. Ginsburg G,
      4. Mills R,
      5. Agans R
      . Primary care physicians’ knowledge of and experience with pharmacogenetic testing. Clin Genet 2012;82:388–94.
      OpenUrlCrossRefPubMed
    24. ↵
      1. Patay BA,
      2. Topol EJ
      . The unmet need of education in genomic medicine. Am J Med 2012;125:5–6.
      OpenUrlCrossRefPubMed
    25. ↵
      1. Reid RJ,
      2. McBride CM,
      3. Alford SH,
      4. Price C,
      5. Baxevanis AD,
      6. Brody LC,
      7. Larson EB
      . Association between health-service use and multiplex genetic testing. Genet Med 2012;14:852–9.
      OpenUrlCrossRefPubMed
    26. ↵
      1. Darst BF,
      2. Madlensky L,
      3. Schork NJ,
      4. Topol EJ,
      5. Bloss CS
      . Characteristics of Genomic Test Consumers Who Spontaneously Share Results With Their Health Care Provider. Health Commun 2013;29:105–8.
      OpenUrlPubMed
    27. ↵
      1. Roberts ME,
      2. Riegert-Johnson DL,
      3. Thomas BC
      . Self Diagnosis of Lynch Syndrome Using Direct to Consumer Genetic Testing: A Case Study. J Genet Couns 2011;20:327–9.
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Janssens AC,
      2. Henneman L,
      3. Detmar SB,
      4. Khoury MJ,
      5. Steyerberg EW,
      6. Eijkemans MJ,
      7. Mushkudiani N,
      8. Oostra BA,
      9. van Duijn CM,
      10. Mackenbach JP
      . Accuracy of self-reported family history is strongly influenced by the accuracy of self-reported personal health status of relatives. J Clin Epidemiol 2012;65:82–9.
      OpenUrlCrossRefPubMed
    29. ↵
      Public Health Genomics: Translating Genomics, Family History Public Health Initiative. Centers for Disease Control and Prevention, 2011. http://www.cdc.gov/genomics/famhistory/famhist.htm (accessed 9 Feb 2013).
    30. ↵
      1. Green RC,
      2. Roberts JS,
      3. Cupples LA,
      4. Relkin NR,
      5. Whitehouse PJ,
      6. Brown T,
      7. Eckert SL,
      8. Butson M,
      9. Sadovnick AD,
      10. Quaid KA,
      11. Chen C,
      12. Cook-Deegan R,
      13. Farrer LA
      . Disclosure of APOE genotype for risk of Alzheimer's disease. N Engl J Med 2009;361:245–54.
      OpenUrlCrossRefPubMedWeb of Science
    31. ↵
      1. Heshka JT,
      2. Palleschi C,
      3. Howley H,
      4. Wilson B,
      5. Wells PS
      . A systematic review of perceived risks, psychological and behavioral impacts of genetic testing. Genet Med 2008;10:19–32.
      OpenUrlCrossRefPubMed
    32. ↵
      1. Crews KR,
      2. Cross SJ,
      3. McCormick JN,
      4. Baker DK,
      5. Molinelli AR,
      6. Mullins R,
      7. Relling MV,
      8. Hoffman JM
      . Development and implementation of a pharmacist-managed clinical pharmacogenetics service. Am J Health Syst Pharm 2011;68:143–50.
      OpenUrlAbstract/FREE Full Text
    33. ↵
      1. Janssens AC,
      2. Kraft P
      . Research Conducted Using Data Obtained through Online Communities: Ethical Implications of Methodological Limitations. PLoS Med 2012;9:e1001328.
      OpenUrlCrossRefPubMed
    View Abstract

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Contributors CSB contributed to the design of the study, performed statistical analyses, drafted the manuscript and edited/revised the manuscript for critical scientific content. NJS and EJT contributed to the design of the study and edited the manuscript for critical scientific content. All authors have read and approved the final manuscript.

    • Funding This work was supported in part by a NIH/NHGRI R21 grant (1R21HG005747; PI: Cinnamon S Bloss, PhD), a NIH/NCRR flagship Clinical and Translational Science Award grant ( S/B 5UL1RR025774 and 1UL1TR001114; PI: Eric J Topol, MD) and Scripps Genomic Medicine Division of Scripps Health.

    • Competing interests None.

    • Provenance and peer review Not commissioned; externally peer reviewed.