CONSUMER ACCESS AND RESEARCH PARTICIPATION

There are several ways in which 23andMe is trying to make genetics more accessible to consumers, Tung said. The primary approach of the company has been to make the tests affordable, but an equally important step has been to make reports easy to understand. Toward that end, all written materials go through several rounds of internal testing and redesign to identify the key points that people need to take away from these reports. Internal testing has shown that there is greater than 90 percent comprehension of these key points across a wide demographic range, Tung said.

Making its tests available online and over the counter are additional ways in which 23andMe tries to broaden uptake, particularly for those individuals who may not have access to a clinical center that offers genetic testing or to providers who will order the genetic testing. Providing the kits and returning the results by mail can also help bring in individuals for whom privacy is a big concern.

At the time of the workshop, 23andMe had sold approximately 10 million kits, Tung said, and those ordering them have been primarily of European ancestry. While there is a fairly substantial cohort of Latinos, African Americans, and East and South Asians among the company's customers, Tung said that this group is smaller than would be expected based on U.S. demographics. The age distribution is bimodal, with one peak in the 30s and another in the 50s to 60s, and women are slightly more likely than men to participate. The 23andMe customer base skews toward those with a slightly higher income and more education than the average members of the U.S. population, Tung said, although there are a substantial number of individuals without a bachelor's degree who have been tested. Approximately 90 percent of 23andMe customers are in the United States.

Approximately 80 percent of 23andMe's customers consent to participate in research, Tung said, a figure that has been consistent through the company's history, even as the price of testing fell and the pool of customers grew more diverse. A theme frequently heard from 23andMe customers who participate in research is their desire to help other people, Tung said. Concerning research participation by ethnicity, there are some differences, but none of the differences are very large, she said; one difference is that East and South Asians participate slightly less than other ethnic groups. There are also few differences in research participation by age or sex.

It is well known that there is not enough diversity in human genetics research, Tung said, and 23andMe cannot provide information it does not have. While the company is trying to develop reports that represent non-European populations, the reality is that more topics have been studied in European populations, so that is the data the company has available to share. The lack of diversity in genetics databases also affects polygenic risk scores¹ that use common variants because of the smaller sample sizes available for non-European populations, Tung said. Without more data from underrepresented groups, she explained, it is difficult to create models that perform well in all ethnicities. For example, 23andMe examined the performance of a type 2 diabetes model that it trained across all ethnicities and found it was still performing better in Europeans than in groups of other ancestries.

23andMe is working in several ways to increase diversity in its genomics research, Tung said. The first initiative the company developed is called the Global Genetics Project,² which involves reaching out and engaging with customers who have four grandparents from underrepresented countries. A second initiative, the Population Collaborations Program,³ relies on partnerships around the globe to genotype people and develop reference populations from underrepresented communities. In addition, the company is investing in whole-genome sequencing of some of its African American and Latino customers as a means of improving its imputation panels that are used for research. Data from the African American sequencing study are now available through the National Institutes of Health's database of Genotypes and Phenotypes,⁴ Tung said, and 23andMe welcomes academic collaborations.⁵ 23andMe is also currently in talks to develop a partnership to create a large, non-European reference panel, Tung said.

Finally, the company is exploring novel methods for developing polygenic risk scores so that the data they do have can be better leveraged, both within and across populations, to develop better scores for non-Europeans. Recent work has shown that these meta-analytic methods are producing better results, Tung said, and the company will continue to try to increase the number of non-Europeans participating in genetics-based research.

DIVERSITY AND THE POTENTIAL FOR DISPARITIES

There is a skewed evidence base in human genomics research, and this matters for clinical genetic test performance and DTC genetic test performance, Malia Fullerton said. As a result of this skewed evidence base, she added, underrepresented populations are more likely to be affected, either from a lack of benefit because of uncertain results or from potential harms related to false positive or false negative results. Notwithstanding the ongoing efforts, Fullerton said, it is important to not wait for the evidence base to catch up to address this problem.

Awareness about DTC genomic testing varies across ethnic groups, Fullerton said, which suggests that improving the awareness of and access to DTC genomic testing would help address the currently recognized disparities in the uptake of these tests. However, she said, even if differences in awareness, education, marketing, and access were addressed in the near term, there would still be disparities in the clinical utility of DTC genomic tests. The reason, she added, is that the genomic research evidence base is markedly skewed toward individuals of northern and western European ancestry (Popejoy and Fullerton, 2016).

The vast overrepresentation of genomes from individuals of European ancestry in research databases matters to genomic discovery and translation because human populations vary genetically due to the evolutionary history of how humans dispersed across the planet, Fullerton said. Population differences become increasingly important when considering rarer variations—typically, gene variants with less than 1 percent minor allele frequency—that are exactly the sort of variation most likely to be involved in disease predispositions of high interest to precision medicine, she said.

As an example, the results of a study of 2,300 European Americans and 2,300 African Americans that was conducted by the National Heart, Lung, and Blood Institute's Exome Sequencing Project demonstrated that African Americans have, on average, a greater number of coding region variants than European Americans (Auer et al., 2016). In addition, a larger proportion of the rare variants were exclusive to African Americans than to European Americans in that study. Such differences suggest that when white and non-white individuals participate in DTC consumer genetic testing, particularly testing that focuses on the health impacts of rare genetic variation, the test results can vary in their quality and accuracy, even when exactly the same test is being used, Fullerton said.

There are different types of outcome disparities that can result from these evidentiary disparities. For example, given that sequence-based tests can detect variants that have not been observed previously and whose clinical significance is unknown, individuals who come from population genetic backgrounds that remain underrepresented in the current evidence database are more likely to receive uncertain genetic test results (Caswell-Jin et al., 2018).

Consumers from underrepresented backgrounds can be also be harmed by receiving results misclassified as pathogenic as a result of a failure to consider ancestry-matched controls in the course of variant interpretation, Fullerton said. For example, one retrospective analysis of patients who had undergone clinical genetic testing for a condition known as hypertrophic cardiomyopathy found that a number of variants initially returned as pathogenic were subsequently reclassified once data on their frequency in unaffected African ancestry individuals became available (Manrai et al., 2016). While there are fewer direct comparisons to draw on with regard to potential false negative findings, recent work that has been focused on the generalizability of polygenic risk scores, for example, suggests that scores which have been validated in European ancestry populations may be poor predictors of genetic risk in individuals from other ancestral genetic backgrounds (Martin et al., 2019).

For consumers, there is currently no obvious near-term remedy for these potential test outcome disparities, Fullerton said. The ability to obtain one's complete raw genotype data file as a direct download could, in theory, allow consumers to take their DTC genomic test results to a third-party interpretation service in an effort to identify additional information that was not available in the initial DTC genetic testing encounter, she said. The problem, she said, is that nearly all of these third-party interpretation tools draw on the same skewed genomics reference evidence base. Additionally, there are no good estimates of how many DTC genomic test customers actually choose to download their data and explore options for additional corroborations or interpretation. When one of Fullerton's recent projects surveyed more than 1,100 self-identified DTC genomic test customers, it found that 72 percent had downloaded their raw data file and taken it to a variety of third-party interpretation tools designed to provide information about health-related risks and ancestry (Nelson et al., 2019).

Regarding possible solutions, Fullerton said that DTC genomic testing services may be the easiest way in the near term for traditionally underrepresented populations, including those with uneven health care access or health insurance coverage, to benefit from the fruits of medical genetics and precision medicine. However, steps should be taken to ensure that disparities are not further exacerbated by continuing to use the same skewed data. In her opinion, she said, DTC genomic testing companies should consider exploring the use of analytic or algorithmic adjustment, a point that Tung had discussed earlier regarding the efforts at 23andMe to improve the predictive accuracy of polygenic risk scores. Having an awareness of biases in the genomics space is critical as new algorithms are designed that take into account the uneven nature of the data, Fullerton said.

The field should also consider adopting evidence standards for gene inclusion on multigene panels, Fullerton said, as a way of reducing the number of variants of uncertain significance that get returned to participants. At the same time, she said, there is a need to be more transparent to consumers about the limitations of test result interpretation, particularly for underrepresented groups whose members are going to be affected disproportionately by the problems she identified.

Ultimately, the ethnically skewed evidence base undercuts the value of all genetic testing, including DTC genomic testing, Fullerton said. Even with greater attention to access and education, underrepresented minority populations are going to benefit less often and more often be harmed as a consequence of these evidentiary biases, she said. “We need to be looking into the adoption of analytical approaches that can be implemented, even while we rectify the underlying data biases,” she said. “These corrections are urgently needed and cannot wait until the evidence base becomes more diversified.”

IMPLICATIONS OF GENETIC ANCESTRY TESTING FOR DIVERSITY AND COMMUNICATION ABOUT HEALTH RISKS

The 1977 television mini-series Roots inspired some African Americans to start asking deep, significant, and often painful questions about their ancestry and their past. In a way, Shawneequa Callier said, genetics-based ancestry testing has become a point of entry for African Americans into the consumer genomics marketplace. However, caution needs to be exercised to prevent overselling DTC ancestry tests as a way to identify exactly where someone's ancestors came from with precise geographic detail, she said. Furthermore, a recent advertising campaign from Ancestry designed to reach African Americans missed the mark, she said, and greater care and cultural sensitivity is required when reaching out to underrepresented, marginalized populations.

One problem with DTC ancestry tests serving as an entry point into the world of genetics for underrepresented populations is that consumers may not understand the complexity of genetic ancestry, Callier said (Royal et al., 2010). “By encouraging consumers to see themselves as a percentage of allegedly distinct ethnic groups, commercial DNA tests may re-inscribe notions of race and miscommunicate the complexity of ancestry,” Callier said. “Furthermore, the validity of these tests has come into question, which could impede trust in the technologies and possibly spill over into the clinical genetics testing realm.” For example, in 2019 investigators at Consumers' Checkbook sent DNA samples to three different DTC genomic testing companies, and the ancestry results for one African American staffer varied widely, ranging from 18 percent West African and 33 percent Central African to 87 percent West African and 3 percent Central African (Brasler, 2019).

Another challenge for DTC ancestry tests lies in the inconsistent way in which ancestry is discussed by consumers and researchers alike, Callier said. For example, the term “continental ancestry” or the idea that someone descends from “African Ancestry” or “European Ancestry” populations obscures the tremendous amount of diversity that exists on those continents and within the associated populations. In fact, Callier said, there may be a missed opportunity to use consumer genomics as a new way to engage in discussions about ancestry. As to the question of whether genetic ancestry tests are improving the way consumers think about race, two studies have found that consumers pick and choose the genetic ancestries they want to embrace (Roth and Ivemark, 2018; Shim et al., 2018). In other words, she said, test results had no effect on how people perceived who they are or the communities and people with whom they affiliated.

Currently, it is not clear whether DTC genomic testing can be useful for engaging with African American populations and if these tests should be used in the clinical setting, Callier said. Some investigators say that ancestry tests misrepresent human genetic diversity, and they argue that these tests should not be used in the medical setting (Blell and Hunter, 2019). Others, including Fullerton, have recognized the potential that consumers will share and discuss genetic ancestry results with providers, for which providers must be prepared (Royal et al., 2010). There could be some potential value to DTC ancestry tests, Callier said, but it will be important as the field moves forward to bring clarity and consensus to the way that ancestry is discussed. A recent study of 10 clinical laboratories found that the labs were not providing the same descriptive categories to designate a group or population (Popejoy et al., 2018). Several groups have argued that it is time to rethink how ancestry is talked about and reported in the literature (Bonham et al., 2018; Cooper et al., 2018; Nelson et al., 2018).

As the cost of sequencing declines, improving genetic and genomic literacy, particularly among underserved populations, will be an important endeavor, Callier said. In closing, she asked, Are there missed opportunities for developing clear and concise language related to race, ethnicity, and ancestry and opportunities for building trust related to genetic and genomic testing services?

RURAL AND UNDERSERVED POPULATIONS AND ENGAGEMENT WITH GENOMICS SERVICES

“As more companies enter the direct-to-consumer retail market for genomics, the increase in numbers of genotyped consumers will allow for an exponential increase in innovation,” Sadie Hutson said, “but this is only going to happen if the reach expands beyond individuals of European descent.” The analysis of rare variations, she explained, becomes much more powerful when sample sizes increase and become more diverse.

Echoing previous speakers in this session, Hutson described the under-representation of non-European, rural, and underserved populations in genomic databases as problematic, as it creates bias in foundational databases that exacerbates disparities. The result is that clinical interventions, polygenic risk scores, and guidelines for risk reduction and the management of complex chronic diseases may be largely inaccurate, particularly for diverse populations (Wojcik et al., 2019).

To provide some background about rural and underserved populations and their engagement with genomic services, Hutson described a 2016 study of a rural community in West Virginia in which more than half of the participants reported a high interest in participating in genetics and genomics studies to improve health (Mallow et al., 2016). Many of the individuals, the study found, were concerned about the influence of environmental health, including how exposure to harmful substances may be changing their genomes. This concern, Hutson said, was frequently cited by patients in her own practice in eastern Kentucky, where coal mining remains a major industry.

The participants in Mallow's study were divided in their attitudes, with some reporting little fear of the risks of genetics studies, while others cited a fear of the unknown, which can be a barrier to pursuing genetic testing. Other barriers that the participants reported included their concerns that knowledge about disease risk would not actually translate into action to prevent disease occurrence and that they mistrusted or had low confidence in their health care providers' knowledge about genetics and genomics.

Hutson said that the factors affecting the uptake of consumer genomics services that she sees regularly in her practice include a lack of awareness of DTC genomics testing (Salloum et al., 2018; Sussner et al., 2009; Vadaparampil et al., 2006), a lack of awareness about genetic counseling, and a lack of access to genetic counseling (Fogleman et al., 2019) (see Box 3-1). It is important to recognize, she said, that the literature on factors affecting the use of consumer genomics is still evolving, with little knowledge at present that is specific to rural and underserved populations.

BOX 3-1

Factors Affecting the Use of Consumer Genomics Services (as presented by Sadie Hutson).

Cost is one of the most frequently cited barriers affecting the use of DTC genomics testing services. In a survey by the Personalized Medicine Coalition (Personalized Medicine Coalition, 2018) of more than 1,000 American adults over the age of 18, 25 percent of those surveyed said they would willingly pay $50 to $100 for such services, but 30 percent were unwilling to pay any amount out of pocket. The respondents also mentioned privacy concerns, and only 10 percent said their provider had ever talked to them about genetic testing to diagnose a disease or guide treatment. A 2012 literature review found that a majority of consumers preferred receiving guidance from a health care provider regarding the use of DTC genomic tests (Goldsmith et al., 2012).

Family communication is generally very important to individuals in rural areas, Hutson said, and it can have both a positive and negative influence on discussions about genetics and family history. “While the number one reason for participating in genetic testing is to learn information for family members, concerns about guilt or fear of passing on genetically linked health conditions can be a major deterrent for testing,” she said. While fear and concerns about discrimination can be a barrier, Hutson said she has seen this factor decrease in importance for some populations in her practice. Many of the barriers to genetic testing and counseling in rural populations have been changing over time, she said.

Supporters of DTC genomics argue that this approach to genetic testing can benefit underserved populations, citing patient empowerment as the primary reason for that optimism, Hutson said. Other positive factors include the relative affordability of DTC testing, non-invasive sample collection, and increased accessibility to genetic testing. Increased patient engagement can lead to overall improvements in genetics literacy, Hutson said, but challenges remain, including the need to have DTC genetic testing results confirmed by more expensive methods which many patients may not be able to afford. As a result, she said, important decisions about treatment or disease management may be based on incomplete, inaccurate, or misunderstood information on the part of both the patient and the health care provider.

The availability of DTC genomics tests does not necessarily improve access for those who have existing barriers to access for general health care needs, Hutson said. In addition, there are privacy concerns about the unauthorized use of an individual's genomic data and about how such data will affect relatives. Wrap-around contracts⁶ can result in consumers unknowingly consenting to terms and conditions regarding the privacy of their genetic data, Hutson said.

There are, however, opportunities to expand the reach of all types of genetic testing into underserved populations. In her own practice at the Pikeville Medical Center, Hutson said, she developed a cancer genetics program in which two genetics nurse navigators help patients complete family and personal health history information during a routine visit with the oncologist or over the phone. The navigators also provide an extensive overview of what will happen at their visit with Hutson, including providing background about her as someone who is not originally from the area, something that she found is important for improving trust with her patients. Patients are assisted in completing a three-generation family history questionnaire before meeting with Hutson. This pre-visit process with the navigators decreased the no-show rate for the cancer genetics service and increased patient encounters by 54 percent in 1 year, Hutson said. The workflow relies on the nurses, who come from the community, to provide background information on the patients before they are seen.

Hutson is also partnering with a nurse-led initiative called the Health Wagon, which provides high-quality care to medically underserved communities in southwest Virginia. Given the cancer disparities in this region, where 98 percent of the residents are uninsured and 70 percent live on less than $20,000 per year, many patients have a clinical indication for genetic testing, she said. To address the cost barrier, she partnered with Invitae, which provides free testing for all individuals referred by the Health Wagon, and she provides free genetic counseling. Telemedicine and faith-based initiatives can also serve as avenues for increasing the reach of and promoting accessibility to genetic testing, Hutson said.

Meeting people where they are and communicating with them in terms that are easily understood is important, Hutson shared. For example, Hutson does not use the term “genetic mutations,” but rather “genetic changes,” an easier and potentially less frightening concept to grasp. It is also critical to set aside enough time to truly work with individuals who come in with questions about genetics. Ideally, Hutson said, she likes to spend 45 minutes with all new patients and 30 minutes for appointments at which she discloses and discusses test results. Spending enough time with each patient is of critical importance, Hutson said, and it requires a thorough review of all patient and family history information before the visit, which assures efficiency and the capacity to build trust with the patient. That trust is critical, especially if other family members need to be tested, she said.

Genomics-based test results, including those from DTC tests, can be confusing, particularly since different test services and providers may be providing conflicting information regarding specific test results. “In my experience, some providers may order testing and not provide results or provide misinformation or incomplete information about results to patients,” Hutson said. Given that the landscape of genetic and genomic testing is evolving rapidly, efforts are needed to educate providers of all types and experience levels about these tests and the resources available for their practices and patients.

For DTC genomics tests to reduce disparities in underserved communities, there will need to be a balance between patient autonomy, clinical utility, ensuring patient safety, and technological innovation, Hutson said. Consumer genomics disintermediates the health care provider from the process of obtaining personal genetic information, but this approach may have broader implications related to medical management as well as psychosocial consequences for patients who decide to undergo DTC genetic testing, she said. There is a critical opportunity for multiple sectors to work together to ensure the proper inclusion of all individuals in genomic testing, she added.

DISCUSSION