Will Genetics Revolutionize Medicine?
On both sides of the Atlantic, revolutionary claims have been made about the ultimate impact of genetics on clinical medicine. John Bell at Oxford has asserted that "within the next decade genetic testing will be used widely for predictive testing in healthy people and for diagnosis and management of patients. . . . The excitement in the field has shifted to the elucidation of the genetic basis of the common diseases."1 And in the United States the director of the National Human Genome Research Institute, Francis Collins, has stated that the good that would come from mapping the human genetic terrain "would include a new understanding of genetic contributions to human disease and the development of rational strategies for minimizing or preventing disease phenotypes altogether."2
Statements like these clothe medicine in a genetic mantle. The result of efforts to identify genes that have a role in common diseases suggests a different picture: the genetic mantle may prove to be like the emperor's new clothes. In this article we argue that the new genetics will not revolutionize the way in which common diseases are identified or prevented. Mapping and sequencing the human genome will lead to the identification of more genes causing mendelian disorders and to the development of diagnostic and predictive tests for them. The development of safe and effective treatments, however, will usually lag behind,3 although occasionally a treatment does precede the discovery of the disease-causing allele, as was the case for hemochromatosis.4 Furthermore, only a small proportion of the population has mendelian disorders, and this will limit the ultimate impact of the Human Genome Project.
Our doubts stem from the incomplete penetrance of genotypes for common diseases, the limited ability to tailor treatment to genotypes, and the low magnitude of risks conferred by various genotypes for the population at large. Consequently, most people will have little interest in learning their genotypes. In the following sections, we use the term "genotype" to denote the alleles that a person possesses at a single gene locus on homologous chromosomes.
Penetrance
In contrast to genotypes for mendelian disorders such as Huntington's disease, which is due to a single, highly penetrant autosomal dominant gene, most genotypes for common, complex diseases are incompletely penetrant, and correlations between the genotype and the phenotype are therefore weak. Associations between a disease and a genetic marker can occur by chance,5,6 and some have proved to be spurious.7,8,9 Although many disease-related genes have been mapped to regions of specific chromosomes, highly penetrant susceptibility-conferring genotypes at loci related to asthma, hypertension, schizophrenia, bipolar disorder, and other disorders have not been found despite intensive efforts.
Searches for susceptibility-conferring genotypes for breast cancer,10 colon cancer,11 rare, early-onset forms of type 2 diabetes,12 and Alzheimer's disease13 have been successful, but in each case these genotypes account for less than 3 percent of all cases. One explanation is that the risk of disease conferred by alleles at one locus depends not only on alleles at other, independently segregating loci, which by themselves do not increase the risk,14,15 but also on environmental factors.16 The problem of identifying susceptibility-conferring genotypes is compounded when different combinations of gene loci are implicated in a disease, for it means that finding enough patients to serve as research subjects in a study will be extremely difficult.9,17
Frequently occurring genotypes, or polymorphisms (frequency of 1 percent or more), are unlikely to have a high penetrance for diseases that reduce reproductive fitness; such genotypes would be selected against except when the presence of a gene on only one chromosome (a single gene dose) confers a selective advantage that counterbalances the disadvantage of its presence on both chromosomes (a double, or homozygous, gene dose). Polymorphisms may confer higher risks for diseases that usually begin after the reproductive years end (e.g., Alzheimer's disease) or diseases for which selection pressures have not had a chance to reduce their frequency because the environment or lifestyles have changed only in recent generations. In populations that have been relatively isolated, genotypes that confer susceptibility to diseases with a recent increase in incidence may have frequencies in the polymorphic range.18
Tailoring Treatments to Genotypes
A recent article in the New York Times echoed the assertions of proponents of the genetic revolution: "Health care will shift from a focus on detection and treatment to a process of prediction and prevention."19 One researcher was quoted as saying, "You can imagine having an infant tested at birth . . . and a result that says you are susceptible to diseases A, B, and C." Physicians will, the argument goes, be able to tailor drugs to a patient's genetic profile.
Finding drugs to thwart a disease will depend on the complexity of the genetic contribution to the disease. If genotypes at only one locus markedly increase the risk of disease, drugs to compensate for the malfunction could be devised. Yet, over 40 years have passed since the molecular basis of sickle cell anemia was discovered,20 and no definitive treatment has emerged. If genotypes at more than one locus must be present simultaneously in order to increase the risk of disease, finding the loci will be difficult. Once they are found, a drug that blocks the effect of only one allele might interrupt the pathogenic process, but this remains to be proved.
Inherited differences in sensitivity to drugs may be more amenable to pharmacologic tailoring than differences in susceptibility to disease. Determining patients' genotypes before they are given certain drugs may lead physicians to avoid administering drugs that could be harmful or to lower the dosages in sensitive patients, but the overall risk of adverse reactions may not be very high because of the low penetrance or low frequency of the genotypes. Alternatively, patients could begin taking a drug, be carefully monitored, and undergo genotyping only after an adverse reaction has occurred. This approach has been recommended for women in whom deep-vein thrombosis develops while they are taking oral contraceptives and who may have a susceptibility-conferring genotype at the prothrombin gene locus.21
The Magnitude of Absolute, Relative, and Attributable Risks
The lifetime risk of breast cancer is 12.6 percent for women, the lifetime risk of prostate cancer is 15.9 percent for men, and the lifetime risk of colon cancer is 5.6 percent for men and women combined.22 The prevalence of asthma at all ages is 5.5 percent; at the age of 45 to 64 years, the prevalence of ischemic heart disease is 5.2 percent and that of diabetes is 5.8 percent.23 The lifetime risk of a major depressive episode is 17.1 percent, and the lifetime risk of nonaffective psychosis is 0.7 percent.24 We can use these data together with genotype frequencies and penetrance to calculate the relative risks of genotypes that confer susceptibility. Thus, susceptibility-conferring genotypes at the BRCA1 and BRCA2 gene loci confer a relative risk of breast cancer of about 5.22,25 Susceptibility-conferring genotypes at DNA-mismatch–repair gene loci confer a relative risk of colon cancer of about 9.3.22,26 Susceptibility-conferring genotypes with polymorphic frequencies would be expected to confer relative risks that are not much more than 2 for various diseases.27 Given the high prevalences of these disorders, even a relative risk of 2 could make the absolute risk conferred by susceptibility-conferring genotypes appreciable, and people might therefore flock to be tested for these genotypes.
Several other factors must be considered in the decision whether or not to be tested. First, the probability that the disease will develop in a person with a positive test result (the positive predictive value) is approximately equal to the penetrance of the disease and is usually low. As illustrated in Table 1, the positive predictive value is a function of the frequency of a susceptibility-conferring genotype, the relative risk of the disease, and the risk of disease in a given population.28 Only if the frequency of the susceptibility-conferring genotype is 1 percent or less and if the relative risk approaches 20 will the positive predictive value exceed 50 percent when the risk of disease in a given population is 5 percent. When the risk in a given population is lower, the positive predictive value will also be lower.
|
Second, the proportion of cases of a common disease that can be
attributed to susceptibility-conferring genotypes is small under the
most likely circumstances (Table
2).29
Other factors, such as the environment, can have a substantial role.
Consequently, healthy people will gain little reassurance that a
negative test result means they will remain free of a particular
disease. For instance, only about 0.25 percent of women carry BRCA1
or BRCA2 susceptibility-conferring genotypes,30
and only about 0.1 percent of people have susceptibility-conferring
genotypes at the DNA-mismatch–repair loci.31
Given the relative risks associated with these
susceptibility-conferring genotypes, people who have them will
account for fewer than 5 percent of all patients with breast or
colon cancer. Only in the case of polymorphisms that have
frequencies in the range of 10 to 30 percent and that increase
susceptibility to disease is the attributable risk appreciable.
However, the risk of disease conferred by polymorphic genotypes is
usually low, as we have already discussed (Table
1).
|
The I1307K allele at the adenomatous polyposis coli (APC ) gene locus,
which is found in about 6 percent of Ashkenazi Jews,32
and the apolipoprotein E 
4
allele, which is found in about 20 percent of a predominantly white
population,33
confer relative risks of approximately 2 for colon cancer and Alzheimer's
disease, respectively. According to the formula given in Table
2, these alleles will account for 5.7 percent of cases of colon
cancer and 16.7 percent of cases of Alzheimer's disease,
respectively, in the general population. When
susceptibility-conferring genotypes at two or more independent gene
loci must be present simultaneously for a disease to occur, the
attributable risk will be much smaller.
Third, only in the case of very few diseases are interventions available that could prevent the disease in healthy people with positive test results or that could improve their survival or quality of life if the disease eventually developed. No interventions based on the identification of disease-related genes have yet proved safe and effective. Approaches already in use, such as prophylactic surgery and monitoring for incipient disease, may prolong life in people with an inherited susceptibility to breast and colon cancer, but the extent of improvement has not been established and the effects on the quality of life have not been studied.
The Degree of Public Interest in Learning about Disease Risks
Given the uncertainties surrounding test results and the questionable effectiveness of interventions in persons with positive results, how much interest will people have in being tested or in making lifestyle changes or undergoing medical or surgical interventions that might reduce their risk of future disease? With respect to predictive genetic testing, people want to know the probability of their getting a disease if the test result is positive (i.e., the positive predictive value) or negative (i.e., the false negative rate, calculated as 1 – sensitivity).34 When the test result is positive, they also want to know what can be done to prevent the disease or improve its outcome. Many people will decide not to be tested if the positive predictive value and sensitivity of a test are low35 and when no treatment is available.36 Even when the positive predictive value of a test is high and interventions are available, as is the case for hereditary nonpolyposis colorectal cancer, interest in testing has been lower than anticipated.37 Interest is also influenced by how convenient it is to be tested,35,38 raising questions about the extent to which enthusiastic suppliers can manipulate demand.
Some evidence suggests that when risks have been determined by genetic testing, persons perceive the risks as less amenable to change,39 suggesting that the likelihood that genetically based risk assessment will result in behavioral changes is even lower than the likelihood of this outcome after traditional assessments of health risks. When medical or surgical interventions are available, healthy people might not want to undergo them. Most women with a family history of breast cancer say they would not undergo prophylactic mastectomy if they were found to have a susceptibility-conferring genotype.40 When treatment is not available for a specific disease, a few people may want the information on risk just for the sake of knowing and, possibly, planning. Finding out that a test for highly penetrant genotypes is negative may also reduce a person's anxiety and the need for other tests. For incompletely penetrant susceptibility-conferring genotypes, negative results may provide those tested with a false sense of security.
Conclusions
We do not want to downplay the importance of highly penetrant susceptibility-conferring genotypes or inherited drug sensitivity. Nonetheless, neither category represents a large enough proportion of the population to warrant widespread screening.41 Testing in families with a history of the disease would be a more efficient approach but does not a revolution make.
It would be revolutionary if we could determine the genotypes of the majority of people who will get common diseases. The complexity of the genetics of common diseases casts doubt on whether accurate prediction will ever be possible. Alleles at many different gene loci will increase the risk of certain diseases only when they are inherited with alleles at other loci, and only in the presence of specific environmental or behavioral factors. Moreover, many combinations of predisposing alleles, environmental factors, and behavior could all lead to the same pathogenic effect.
In our rush to fit medicine with the genetic mantle, we are losing sight of other possibilities for improving the public health. Differences in social structure, lifestyle, and environment account for much larger proportions of disease42,43 than genetic differences. Although we do not contend that the genetic mantle is as imperceptible as the emperor's new clothes were, it is not made of the silks and ermines that some claim it to be. Those who make medical and science policies in the next decade would do well to see beyond the hype.
Neil A. Holtzman, M.D., M.P.H.
Johns Hopkins Medical Institutions
Baltimore, MD 21205-2004
Theresa M. Marteau, Ph.D.
Guy's, King's and St. Thomas' Medical School
London SE1 9RT, United Kingdom
Supported by funds from the Wellcome Trust (to Dr. Marteau