Scientific Foundations for Direct-to-Consumer Genetic Testing
More than a century before the human genome sequence was fully revealed in 2003, biomedical researchers had begun to recognize that certain human diseases were hereditary and carried a very high probability of being transmitted to offspring; that is, such diseases behaved as Mendelian disorders. Over the decade preceding the completion of sequencing the human genome, researchers were beginning to identify specific genes for some of these Mendelian disorders: for example, cystic fibrosis in 1989, Huntington’s in 1993, and breast cancer in 1994 and 1995 (BRCA1 and BRCA2, respectively). And this was only the beginning.
GENOMIC ASSOCIATIONS1
By 2002, genes had been identified for roughly 1,700 of the 5,000-6,000 known Mendelian human disorders, most of them resulting each from a single erroneous allele, or version of a gene; in contrast, fewer than 10 genes associated with more common, complex human disorders had been discovered.2 It appeared that with then-current technologies, the cost to locate all the single-nucleotide polymorphisms (SNPs)—individual variations in the coding DNA sequence of the genome—associated with any one complex disorder would be about $10 billion: far too expensive to employ in a full-out assault on common diseases.
1 |
Unless otherwise noted, the section “Genomic Associations” reflects the remarks of Alan Guttmacher, former director of the National Human Genome Research Institute, National Institutes of Health. |
2 |
Glazier A, Nadeau J, Aitman T., Finding genes that underlie complex traits, Science, 2002: 298 (5602): 2345-2349. |
Research conducted during the early 2000s revealed that sets of SNPs within distinct regions of the chromosome tend to be inherited together. If specific SNPs were contributing to common disease progression, identifying variants between distinct populations could theoretically help construct a map of the genome that identifies areas of potential association with disease. These initial observations led to the formation of an international consortium called the Haplotype Map (HapMap) project, which resulted in the creation of a HapMap, a catalogue of SNPs commonly found in a consensus human genome and their locations within the genome. Using the HapMap, a researcher can conduct a genome-wide association study (GWAS)—a rapid scan of many individuals’ entire genomes to discover alleles associated with a particular disorder. DNA microarray chips—miniature glass chips encoded with thousands of short, synthetic, single-stranded DNA sequences representing identified SNPs—have been developed to facilitate this approach. The technology relies on the discriminatory nature of DNA sequences to bind to exactly complementary pieces of DNA with high affinity. Genomic DNA, for example, can be denatured into single strands, cut into smaller pieces, labeled fluorescently, and allowed to bind to the DNA encoded on the chip. Differences between the binding pattern of DNA from individuals with a particular disease and the binding pattern of DNA from a control group can lead to the identification of specific alleles which may contribute to disease progression. A GWAS is particularly useful for finding alleles that contribute to common diseases, and it enables researchers to do so far more easily and cheaply than before. However, as sequencing costs continue to fall, many predict that the “$1,000 genome” will soon become a reality. At this price, it is conceivable that whole genome sequencing will eventually replace genome wide association studies.
Although a GWAS can be a powerful source of information about genetic predisposition to disease, so far these studies explain only a very small fraction of heritability and fail to capture other contributing factors such as multiple common genetic variants acting together, copy number variants, or epigenetics—chemical changes in base pairs, or physical changes in chromosome structure—that may greatly modulate phenotypic expression but that are themselves typically not heritable.
Therefore, a GWAS is a weak forecaster of an individual’s risk for a genetic disorder. Obesity linkages, for instance, account for less than 2 percent of variance in heritable body mass index. Furthermore, a GWAS has a major drawback: it misses rare genetic variants that, when present,
may have major health effects. As a result, assessing the value of DTC genetic tests only begins with determining whether or not they are scientifically valid—a fairly straightforward task. Much more problematic is determining whether or not they are clinically useful.
ANALYTIC AND CLINICAL VALIDITY AND CLINICAL UTILITY3
As with any biological test, assessing the analytic validity of a particular genetic test is relatively straightforward: does the test correctly measure or detect what it is intended to measure or detect? Specifically:
-
Analytic sensitivity: How often a test result is positive when the genetic variant of interest is present in the tested sample.
-
Analytic specificity: How often a test result is negative when the tested sample does not contain the genetic variant of interest.
Documenting analytic validity is important even though it may not be publicly disclosed. Many of the tests currently offered by reputable DTC genetic testing companies have been assessed for analytic validity, sensitivity and specificity (although there is no current regulatory scheme that assesses or ensures this). A recent exception came to light just prior to the workshop: a software glitch had caused one DTC genetic testing company to confuse human and animal mitochondrial DNA.4
Assessing other factors in clinical validity, however—particularly a test’s value in predicting disease and calculating risk—presents more complex challenges, many of them in the context of GWAS SNPs:
-
Credible genetic associations: Whether or not a clear association has been established between a particular SNP and an increased likelihood of developing a specific disease.
-
Positive predictive value: The probability that—if a target SNP is present in a tested subject’s genome—the individual will eventually be affected by the associated disorder.
3 |
Unless otherwise noted, this section reflects the remarks of Muin J. Khoury, director of the Centers for Disease Control and Prevention’s (CDC) Office of Public Health Genomics and a National Cancer Institute Senior Consultant in Public Health Genomics. |
4 |
Aldhous, P., My “non-human” DNA: a cautionary tale, New Scientist, August 26, 2009. Available at: www.newscientist.com/article/dn17683-my-nonhuman-dna-a-cautionary-tale.html; Accessed: May 10, 2010. |
-
Negative predictive value: Similarly, the probability that—if a target SNP is not present in a tested subject’s genome—the individual can be confident of never being affected by the associated disease.
-
Additional predictive value: Whether or not a test provides predictive information beyond what could be readily learned by other means, such as a family history, clinical evaluation, or standard laboratory testing. Looking at heart disease, for example, it takes an estimated 20 or more SNPs to get a two-fold increased risk, whereas family history has much greater predictive value; it could be therefore, that awareness of SNPs contributes little or no additional value. In at least two cases, however—coronary heart disease (nine SNPs) and breast cancer (seven SNPs)—it appears that awareness of SNPs may indeed contribute appreciable value by stimulating earlier screening for these diseases.5 The utility of this information for predicting individual risk, however, remains to be determined.
-
The uncertainty of risk estimation. In addition to the problem of hidden heritability (the possibility of as-yet undiscovered risk alleles), there are factors other than one’s genome—one’s “raw DNA”—that can significantly influence the probability that an individual will eventually be affected by a genetic disorder. These factors include environmental and behavioral factors, epigenetics, the epidemiology of the disease, and the individual’s age at testing.
Even more problematic than assessing clinical validity is the question of clinical utility. Assessing clinical utility—the net balance of benefits and harms to the individual, the population at large, and the overall health care system—is a more complex, and often more subjective, task. Much of the discussion of clinical utility revolves around actionability. Regarding genetic disorders for which no prevention or intervention is currently known, the practical value of testing for increased risk is often unclear. For a few devastating diseases—Huntington’s and Alzheimer’s for example—it is understandable that an individual might want to know in order to make financial or other arrangements. Unlike Huntington’s disease, for which a positive test carries a virtually 100 percent probability of developing the disease, risk assessment for Alzheimer’s disease is more complex and problematic and far less certain.
In addition, preventive actions that might be taken in response to increased risk for common diseases such as diabetes, cancer and cardio-
vascular disease—smoking cessation, exercise, a healthier diet—are not unlike the advice most individuals already receive from their health care providers, especially in the presence of relevant family histories or clinical risk factors. Moreover, a negative result might cause an individual to forego prudent prevention and early-detection measures. Furthermore, it is at least plausible that the kinds and efficacy of patient/consumer interventions in response to this information are affected by the extent of counseling or other professional intermediation in providing the information.
In December 2008, the NIH and CDC jointly sponsored a multidisciplinary workshop, “Personal Genomics: Establishing the Scientific Foundation for Using Personal Genome Profiles for Risk Assessment, Health Promotion, and Disease Prevention” in Bethesda, Maryland. The workshop report6 made several recommendations:
-
Develop and implement industry-wide scientific standards.
-
Develop and apply a multidisciplinary research agenda, including observational studies and clinical trials, to fill knowledge gaps in the clinical validity and utility of personal genomics.
-
Enhance credible knowledge synthesis and dissemination of information to providers and consumers.
-
Link scientific research on validity and utility to evidence-based recommendations for use of personal genomic tests.
-
Consider the value of personal utility.
One final point, which is neither directly nor exclusively relevant to genetic testing, per se, is nevertheless sufficiently important to raise in this context. One long-lasting and far-reaching benefit of the Human Genome Project to scientific research—and one that is often overlooked in favor of headline-grabbing advances in genomics—is the landmark agreement among participating centers to post their sequencing data every 24 hours. As beneficial as this advance has been in many ways, it has highlighted the need to balance three imperatives: enabling rapid public access to all data generated from publicly funded research; encouraging research by offering researchers exclusive use of their data for a period of time; and protecting participants’ rights, including privacy. As a result, the National Institutes of Health is in the process of developing trans-NIH principles on data access.7
QUESTIONS RAISED FOR FURTHER DISCUSSION
-
What personal, familial, and societal issues are raised by DTC genetic testing and what issues will be raised upon the advent of affordable whole genome sequencing? How can individuals and society prepare for them?
-
In contrast with fairly objective criteria of analytic and clinical validity, what should the criteria be for clinical utility, and who should make that determination—researchers, physicians, individuals, governments, society?
-
Is personal utility a sufficient rationale for genetic testing?
-
With the cost of locating all the SNPs associated with genetic contributions to any complex disorder rapidly becoming affordable, what process and criteria will be used to decide which common diseases will be explored and in what order?
-
If DTC genetic testing companies are capable of confusing human and animal DNA, should they be providing genetic testing services to customers at all?