Perils, pitfalls, and promise of direct-to-consumer genetic testing

It was not difficult to foresee the appearance of direct-to-consumer (DTC) genetic testing as a logical or perhaps even inescapable outcome of the Human Genome Project. The Project was an international undertaking of massive scale, and part of the value of obtaining the first complete human genome sequences lay in its opportunity for scientific outreach. Capturing the public imagination about the potential of genetic testing, as well as educating the public and legislators about the powers and pitfalls of genomic technology, were (and remain) important goals. But it is often easier to stir the imagination than to educate, and this may be the case with regard to DTC genetic testing. It can be a challenge for physicians today when a patient comes in bearing a pile of DTC testing results but not fully understanding them. In this month’s Primer, we’ll review some DTC services that companies provide, and explore some of the potential pitfalls to meaningful interpretation of the data.

A number of DTC companies actively advertise services to the general public. Some are very widely known; others are less well known but equally willing, for a fee, to take some form of minimally invasive sample, such as a buccal swab, and report back on any number of results. In general the advertised testing appears to exist in two broad categories, one providing information on ancestry and one providing information on potential health susceptibilities (disease risk factors, pharmacokinetics, and the like). Both work on the same basic principle: they check for the presence or absence of particular genetic markers (mostly variable nucleotide tandem repeats [VNTRs] for ancestry and mostly single nucleotide polymorphisms [SNPs] for health issues). The two classes then diverge with regard to data analysis stages.

Ancestry results: mostly harmless

We’ll deal with the ancestry results class first, as it is probably the simpler one and the one with less potential for harm. Essentially, this testing looks at a customer’s set of markers and compares them against known frequencies of occurrence of those markers in reference populations. The most informative markers are those which are relatively rare in all but one population; if a sample has that marker, there’s a good chance that there’s a proportion of that ancestry in the customer’s background. Over a larger number of markers (10 to 20, and likely with multiple alleles possible for each), it’s possible to compare the raw data against publicly available marker ancestry databases to suggest an overall probability of someone’s ancestral genetics.

When a friend recently had one of these tests done and provided the raw data, your author had an opportunity to experiment first-hand with doing these sorts of analyses. What’s important to understand for this class of analysis is that statistical likelihoods are just that—likelihoods, not certainties. A 70 percent likelihood of having picked up a marker from a given population means there’s a 30 percent chance it didn’t come from that population. Also, a certain amount of data filtering or incorporation of additional evidence can be important; for the case the author looked at, despite one marker having a high statistical match to Australian aboriginal populations, based on prior extensive traditional genealogical work it was pretty safe to assume this marker was a statistical anomaly and no aboriginal ancestry was actually present. Where such alternate lines of evidence are not available or not utilized, examples such as this show that significant misinterpretation of results is possible in this application. However, if done purely out of curiosity, and if there’s an appreciation that results need to be taken with appropriate statistical weight, there seems to be minimal potential for harm in DTC ancestry testing.

Health-related testing: buyer beware

Unfortunately, the same cannot be said for the second class of genetic testing: testing for potential health susceptibilities. First, let’s be clear that the following discussion will assume there are no methodological shortcomings (unreliable test results) being considered; I shall assume that the DTC labs have this covered. Depending on jurisdiction, however, please note that DTC labs may not be required to meet accreditation standards equal to those of a clinical lab; and where such proof of competence and results uniformity is not required, there is at least potential for an additional range of problems.

If we stipulate that the test results are accurate, any concerns must be based on the interpretation of said results. Indeed, this is where significant issues may arise. While a number of possible scenarios could be envisioned, for brevity I’ll select just a few that demonstrate what some of these concerns could be.

Case 1: A DTC customer gets a result indicating that a particular SNP is linked to a “100 percent increased risk” of some given condition. For the sake of argument, let’s naively assume this is a completely accurate statement, not confounded by additional genetic or environmental factors. A critical second piece of information, however, which many customers may not stop to consider, is what the base risk is. If the base risk is one in a million for Condition X, it now rises to 1/500,000—100 percent more, but perhaps still basically irrelevant. In other words, knowing base risk is essential to understanding the real-life impact of the result–yet the media usually seem to focus only on increased risk values.

Case 2: This is the same as Case 1, except now let’s be less naive and ask whether reported risk odds associated with a particular marker are actually well founded and whether they fully apply in the case in question. The correct answer to these questions in most cases is that we don’t yet know. We don’t have enough data from enough people tied to the right metadata on their other genetic markers and environmental influences to make this assessment. Of course, for some markers (notably those which are directly causal of monogenic traits, such as, say, a mutation for sickle cell anemia), as opposed to ones which are just linked and/or are part of complex polygenic traits, this isn’t the case; a definite phenotype or condition can be confidently reported. Most SNPs, however, don’t fall in this nicely defined direct-causal monogenic setting. (On the bright side, with more data and retrospective analysis it should become increasingly possible to improve on our understanding of complex polygenic traits and environmental influences. Current interpretational risks here may well decrease over time.)

Case 3: A result may indicate an unsuspected yet highly significant negative finding, such as a late-onset degenerative disease. The ethical implications of providing this information, whether correct or incorrect, are very serious and are only beginning to be considered.

Case 4: Rather than the results suggesting an immediate effect, they suggest carriership for some undesirable condition or trait. As we’ll see below, there are differing opinions on how dangerous this information could be. A case could be made, however, that lacking perceived immediacy of impact, the client receiving these results may not choose to discuss them with medical professionals, yet may make significant and possibly detrimental or misinformed reproductive choices based on the data.

All of these cases, and most directly the last two, share a common underlying theme best summed up by the longstanding adage, “a little knowledge is a dangerous thing.” In a clinical setting, genotypic testing would generally be coupled with genetic counselling sessions precisely to ensure that the client understands the implications of any particular result. Providing this level of expertise and personalized interpretation, in the context of having well documented test validation, would appear to be a minimal requirement for avoiding potential harm to customers of DTC testing.

Regulation and ethical concerns

Regulatory bodies have not been blind to these risks. In the United States, the Food and Drug Administration (FDA) sent warning letters in November 2015 to three different companies offering DTC genetic testing with purported clinical relevance, but without having cleared assays (that is, ones with formally demonstrated accuracy and significance of interpretation).1 In another, non-U.S. jurisdiction case, the appropriate accreditation body (after being made aware of the matter) issued a letter to a company that had publicly misstated what accreditations were held, warning that the provision of advertised testing was not allowed; simultaneously, the accrediting bodies behind the claimed but nonexistent accreditation required the offending company to remove the spurious claims.

On the whole, we seem to be in a challenging situation brought about by technological capacity advancing faster than society’s ability to responsibly handle it. This is hardly a novel occurrence, and with regard to genetic information it has been postulated and discussed for some decades now. The technology needed to do a full-exome sequence or an SNP profile of a person is rapidly becoming very cheap and widely available, and the general public has been at least partly educated to the terms and some concepts of genomics. With so many DTC testing options available, the ethical conundrums related to genetic testing that were postulated twenty years ago by bioethicists are no longer hypothetical. And understandable curiosity increasingly creates customers willing to pay for personalized genotyping, at the same time that companies which offer it see a market that can be addressed with decreasing financial hurdles to entry.

One approach to this situation has been increased regulation. In the United States, one DTC genetics company has been approved to provide test results, but interestingly is allowed to report only heterozygous (carrier) states; it must report homozygous (disease present) results as “no call.”2 While this is an attempt at harm reduction, one has to wonder if it might not backfire; will non-specialist customers, aware of this rule and getting a “no call” result, take this as a de facto disease diagnosis? Is even carriership information truly harmless in the absence of appropriate counselling? The proposed answers tend to raise more questions.

So where does this leave today’s laboratorian if a physician calls the lab and wants to discuss DTC results a patient has brought in and wishes to act on? Since the DTC results in most regulatory settings can’t be used to diagnose or treat any condition, that part is pretty simple: they have no standing. However, just because the test results are DTC results, that does not necessarily mean they’re wrong; so if the attending physician has suspicions that a DTC test result is both accurate and relevant, it could form at least part of the basis for making a decision to order an appropriate controlled clinical grade genomics test. Used in that context, there may yet be a useful role for DTC testing to play as a form of voluntary pre-screen.


  1. Mezher M. FDA warns three companies over DTC genetic tests. Regulatory Affairs Professionals Society.
  2. Cope TA. FDA challenges direct-to-consumer genetic tests. Law 360.

John Brunstein, PhD, is a member of the MLO Editorial Advisory Board. He serves as President and Chief Science Officer for British Columbia-based PathoID, Inc., which provides consulting for development and validation of molecular assays.