Epidemiologic
Approach to Genetic Tests: Population-Based Data for Preventive Medicine
Marta Gwinn, Muin J. Khoury
Tables | Appendix | References
Abbreviations
CRC – colorectal cancer
DNA – deoxyribonucleic acid
FAP – familial adenomatous polyposis
HNPCC – hereditary non-polyposis colorectal cancer
MMR – mismatch repair
Names of genes: APC, BRCA1, MLH1, MSH2
Sequencing the human genome ahead of schedule has raised expectations
for quick translation of the data into tools for medical practice. When
the initial sequence was published in February 2001, Francis Collins
and Victor McKusick wrote that “genetic prediction of individual
risks of disease and responsiveness to drugs will reach the medical
mainstream in the next decade or so.”(1) The
idea that genetic tests could offer patients personal estimates of risk
and interventions on the basis of their genotypes has captured the imagination
of scientists and the public.
Genetic Tests
Until now, genetic tests have been used mostly to aid the diagnosis
of rare hereditary disorders. In November 2000, when we reviewed the
list of tests in GeneTests, a Web-accessible database that serves as
the main directory of United States clinical and research laboratories
offering genetic testing,(2-3) we found that fewer
than 5 percent of tests available for clinical use applied to common,
adult-onset diseases.(4) Most of these were tests for
variants of single genes associated with disease susceptibility in high-risk
families (e.g., BRCA1 for breast cancer).
A recent review of entries in the online version of Mendelian Inheritance
in Man(5) suggests that this situation is unlikely
to change soon: although the discovery of disease-associated gene variants
is accelerating rapidly, the number of identified “susceptibility
genes” remains small.(6)
Genetic tests that predict future risk for disease in asymptomatic
people, thereby suggesting specific strategies for prevention or early
detection, are the starting point for models of individualized preventive
medicine. An example that helps illustrate the expectations, limitations,
and future potential of predictive genetic tests is Francis Collins’
“hypothetical case in 2010,”(7) in which
a 23-year-old man named John undergoes DNA testing for genes related
to several common chronic diseases. The genetic test report includes
relative risks (range: 0.3-6) as well as lifetime risks (range: 7%-30%)
for each of these diseases, predicted on the basis of John’s genotype
for one to three genes related to each condition. John’s physician
recommends that he stop smoking, undergo regular colonoscopy beginning
at age 45 years, and take lipid-lowering medications.
In this example, not only the patient but the relative and lifetime
risks are hypothetical. Where will we obtain the data needed to interpret
and act on the results of genetic tests? Despite the media’s tendency
to depict genetic tests as definitive, no test can predict with certainty
the behavior of a complex biologic system (in this case, John) over
a lifetime. Furthermore, risk cannot be predicted solely on the basis
of individual information; it must be estimated by analysis of the characteristics,
experience, and outcomes of a group of people “similar”
to the individual of interest. Information about the family is used
to assess risk of classic, Mendelian disorders, and the ability to predict
disease based on inheritance is the foundation for the clinical specialty
of genetic counseling. However, estimating the risk for complex disorders
(without a clear pattern of inheritance) requires genetic information
from larger population samples. Information on prevalence of gene variants,
genotype-phenotype correlations, and gene-gene and gene-environment
interactions must be collected systematically by epidemiologic studies
conducted in populations resembling those to which inferences will be
drawn.(8) These populations are likely to be much more
diverse than the genetically homogeneous groups in which susceptibility
genes are usually first identified.
Epidemiologic Approach
Epidemiologic studies that collect genetic information depend on access
to valid, reproducible, economical tests for the genetic variants of
interest. New technology has made such tests available for use in large-scale,
population-based studies.(9) Choosing among analytic
methods involves practical considerations, such as availability of collaborators,
as well as characteristics of the variant itself. Before a test can
be used in epidemiologic research, its analytic validity must be established.
Analytic validity refers to the sensitivity,
specificity, and predictive value of the test in relation to genotype;
these characteristics are measured by comparing the test result with
a gold standard in a set of well-described samples, only some of which
contain the genetic variant.
The final report of the National Institutes of Health-Department of
Energy Task Force on Genetic Testing(10) distinguished
analytic validity from clinical validity,
which they defined as the sensitivity, specificity, and predictive value
of a test in relation to a particular phenotype. In contrast to analytic
validity, which must be determined before a study begins, clinical validity
is defined by epidemiologic studies that measure gene-disease associations.
A third parameter defined by the Task Force, clinical
utility, refers to the net value of the information gained from
a genetic test in changing disease outcomes. Clinical utility is best
assessed in clinical trials or by synthesis of observational data (e.g.,
as in cost-effectiveness analysis). Table 11-1
summarizes these measures of validity and utility.
Classic epidemiologic study designs include cross-sectional, cohort,
and case control studies. Each design is useful for addressing particular
aspects of genetic variation or gene-environment interaction in relation
to disease outcomes.(11) Cross-sectional studies can
be used to estimate the prevalence of gene variants, although variants
associated with poor survival may be underrepresented. Cohort studies
are unique in providing direct estimates of absolute risk and relative
risk in people with different genotypes. “Experimental”
cohort studies (randomized, controlled trials) are ideal for evaluating
the effects of gene-environment interactions or specific interventions
(see chapter 15). Retrospective cohort studies are attractive because
genetic information is invariant and can be measured long after the
study has ended; however, they are subject to the usual biases of observational
studies.
Case-control studies can generally measure gene-disease associations
more quickly, efficiently, and at lower cost than cohort studies. Case-control
studies yield odds ratios, which approximate the relative risk of disease
as long as the disease is rare or controls are sampled randomly (independent
of disease status or genotype) from the source population.(12)
The defining characteristic of a population based case-control study
is the set of a priori criteria—applied to selection of controls
as well as cases—that specifies the study’s source population
(e.g., by geographic area and ethnicity). Studies comparing genotypes
of patients in a clinical case series with those of an undefined convenience
sample of control subjects (or worse, “control specimens”)
are numerous in the published literature but they provide little basis
for risk estimation. From a public health perspective, a population-based
estimate of relative risk is critical because it provides the basis
for estimating attributable fraction—the proportion of cases that
would not occur in the absence of a particular exposure (or genotype)
in the population. To examine the role of epidemiologic studies in eliciting
genetic factors in common diseases, we consider the example of colorectal
cancer (CRC).
Box 11.1
Example: Colorectal Cancer
Like other cancers, colorectal cancer (CRC) is a “genetic
disease” caused by mutations that disable normal regulation
of cell growth and differentiation. An estimated 945,000 new cases
of CRC occur annually worldwide.(13) Epidemiologic
studies have identified many exposures associated with increased
risk for CRC, including smoking, overweight, inactivity, and dietary
factors. Since familial clustering of CRC was first reported more
than 100 years ago, clinical research has identified several high-risk
syndromes, including familial adenomatous polyposis (FAP) and
hereditary nonpolyposis CRC (HNPCC).
Table 11-2 compares estimates of the
absolute (lifetime) risk and relative risk for CRC in people
who have FAP,(14) pedigrees consistent with
HNPCC,(15-16) or at least one first-degree
relative with CRC,(17) with risks in people
in none of these groups, who are considered at “average
risk.” Without intervention, people with FAP are virtually
certain to develop CRC (absolute risk approaching 1), usually
by their mid-30s; however, FAP is very rare (prevalence approximately
1/8000) and thus accounts for a very small share of CRC cases
in the U.S. population (attributable fraction <1 percent).(14)
FAP is diagnosed clinically on endoscopic and histologic criteria.
Although it is an autosomal dominant disorder, about one third
of cases result from new mutations; thus genetic testing is
useful mainly for counseling an affected person’s family
members, offering increased surveillance if positive and reassurance
otherwise.(18)
HNPCC is a diagnosis based on a pedigree consistent with autosomal
dominant inheritance of CRC (as well as cancers at other sites
in some high-risk families).(16) Unlike FAP,
HNPCC cannot be diagnosed on the basis of clinical characteristics,
and thus far, genotypic information has not become part of the
case definition. Since 1994, mutations in several DNA mismatch-repair
(MMR) genes have been found in HNPCC families, with MSH2 and
MLH1 mutations by far most often implicated.(16)
Identifying a mutation within an HNPCC family can be useful
for testing and counseling family members. Although absolute
risk for CRC is approximately 80 percent in HNPCC family members
with inherited susceptibility,(16) data are
insufficient to estimate the absolute and relative risks for
CRC based on MMR genotype. The population prevalence of genetic
variants associated with HNPCC is unknown, although one study
analyzing data from Scotland, Finland, and the United States
arrived at an estimate of 1/3000 for MSH2 and MLH1 variants
combined.(15) Population-based data on frequency
of HNPCC among CRC cases is also scarce, although recent studies
suggest that it may be lower than previously estimated from
case series, perhaps as low as 1 percent.(19)
Family history can capture shared, unmeasured genetic risk,
as well as the potential influences of shared diet, behavior,
and other non-genetic factors.(20) Accumulated
evidence from epidemiologic studies of CRC suggests that having
at least one first-degree relative with CRC increases the relative
risk for colon cancer approximately twofold. However, because
the average risk for CRC is only about 4 percent, the absolute
risk of CRC in this group remains <10 percent;(17)
thus, family history offers poor predictive value as a “genetic
screening test” for CRC.
Genetic information obtained thus far from population-based,
epidemiologic studies is relevant to only about 10 percent of
CRC cases in the population. Although epidemiologic studies
consistently identify CRC in a first-degree relative as one
of the strongest risk factors for CRC, other factors (e.g.,
dietary habits, physical activity) that are less strongly associated
but more prevalent in the population have higher attributable
fractions.(21) |
Estimating Individual Risk from Population-Based
Data
Epidemiologic studies of gene-disease associations—particularly
those reporting results in terms of predicted risk—have lately
come under fire as an unwarranted extension of the individual risk paradigm,(22)
in which relations among disease risk factors are teased out at the
individual level. The exclusive focus on individuals can be criticized
on both practical and philosophic grounds, for failing to prompt effective
public health interventions while “blaming the victim.”(22-23)
The “privatization of risk”(24) also appears
to defy the concept of risk as an aggregate measure and to ignore the
reality that individual risk factors generally make poor screening tools.(25)
A recent commentary on the potential impact of genetics on preventive
medicine echoes these concerns,(26) arguing that because
most genetic tests have low predictive value, low clinical sensitivity,
and little potential for stimulating tailored intervention, they will
have limited value in preventing common complex diseases.
Viewing the potential contribution of a single test in isolation reflects
a time-honored perspective in public health screening, as well as the
traditional use of clinical genetic tests. Mass screening programs are
generally delivered to whole populations without regard to prior information
(such as family history or race/ethnicity), both to maximize sensitivity
and to achieve social goals, such as fairness and program efficiency.
On the other hand, clinical geneticists have used tests mostly for diagnosis
of hereditary disorders resulting from single gene variants with very
high penetrance. In this setting, a single genetic test may be definitive,
although DNA sequencing is revealing increasingly diverse genotype-phenotype
relations, even in classic “single gene disorders” like
cystic fibrosis.(27)
Most common chronic diseases arise from interactions among multiple
genes, environmental exposures, and behaviors; thus, genetic tests are
most likely to be useful when combined with results of other information
to uncover interactions associated with markedly elevated risks. This
concept can be demonstrated using basic principles from either epidemiology(28-29)
or genetics.(30) Real examples are still scarce, however,
partly because they are likely to be complicated, involving multiple
genes and multiple environmental exposures.
This state of affairs is familiar to most medical practitioners, who
are used to considering the results of multiple clinical tests in the
context of other, often incomplete, information about an individual
patient, including family history, lifestyle, and physical examination.
The basis for integrating and interpreting this information is experience—whether
clinical experience of an individual physician, expert consensus, or
data gathered systematically from scientific studies, such as clinical
trials. During the last decade, the methods developed by clinical epidemiologists
for critical analysis, synthesis, and application of accumulated experience
have become the foundation for “evidence-based medicine.”(31)
In this medical paradigm, diagnosis is a Bayesian process: as each test
result is added to the body of evidence, some possible diagnoses become
more likely, while others are less likely or altogether ruled out.(31,
pp.121-40)
Results of genetic tests can be integrated into the same framework,
as long as the association of genotype with disease outcome (genotype-phenotype
correlation) has been well described.(32) When the
goal is to predict future disease, rather than to make a diagnosis,
a genotype becomes part of the evidence that can be used to make a probabilistic
estimate of risk.(Yang Q, personal communication)(32)
The underlying relationship between a susceptibility genotype (defined
by variant alleles at one or more genetic loci) and disease outcome,
and its reflection in measures of test performance and risk, can be
summarized in the familiar framework of a 2 x 2 table (Appendix).
As observed by critics of individualized preventive medicine, few risk
factors for common chronic diseases have sufficient predictive ability
to serve as screening tools;(25) in this respect,
common polymorphisms associated with disease susceptibility are unlikely
to be different. Most risk estimates useful to individuals will be obtained
only by considering the joint effects of many factors; however, although
technical advances have made large-scale genotyping feasible in epidemiologic
studies, the ability to assimilate, synthesize, and interpret the data
has not yet fully caught up. The sheer number of variables potentially
available for analysis tests the limits of conventional methods.
Challenges for “Genomic” Epidemiology
Despite their independent origins, genetic and epidemiologic methods
for investigating causes and predicting risks for human diseases share
many concerns common to observational sciences. During the last 50 years,
the synthesis of genetic and epidemiologic methods has been accelerated
by growth in statistical and computing techniques and by development
of molecular methods for measuring environmental exposures, biological
processes, and genetic traits.(9,33)
Much recent development in genetic epidemiology and statistical genetics
has focused on methods for identifying disease susceptibility genes
in families; however, describing the distribution of genetic traits
in populations and evaluating the role of genetic factors in disease
occurrence requires larger studies of unrelated people. Epidemiologic
studies of genotype prevalence, gene disease association, and gene-environment
interaction are subject to the usual sources of bias, including confounding
and misclassification. Confounding is analogous to “population
stratification” in studies of gene-disease association;(34)
misclassification of genotype occurs as a function of analytic validity.
Also, type I errors are of concern when multiple gene-disease associations
are tested, and type II errors are likely when results are analyzed
for small subgroups defined by genotype or gene-environment interactions.
The nature of genomic data poses additional challenges for epidemiologic
analysis. For example, genetic variants at different loci cannot be
assumed to occur independently, even when they are found on different
chromosomes.(35) Furthermore, although the genetic
sequence of an individual remains the same, gene expression naturally
varies tremendously among tissues, in response to environmental stimuli,
and with age, reflecting cumulative effects over the course of a lifetime.
Thus, the interactions of gene products with each other and with other
factors in their milieu reflect an underlying complex “genetic
architecture,”(36) in which health and disease
are “defined by the same continuum of biological traits.”(36,
p. 217) New models are needed for analyzing these relations and using
them for prediction and intervention.
Most common chronic diseases result from multiple gene-environment
interactions over a long period of time, involving invariant features
(e.g., genotype), “context-dependent features” (e.g., diet),
and chance processes.(36) Prediction from “first
principles” (genotype) is thus an unrealistic goal. One strategy
for capturing the effects of multiple factors pursues data more proximate
to the disease outcome, such as acquired (somatic) mutations, gene expression,
protein markers, or intermediate states or conditions that recapitulate
prior gene-environment interactions. Predictions based on observations
made closer to the outcome are likely to be more accurate. To examine
the potential of this approach, we revisit the example of CRC.
Box
11.2 Example
Revisited: Colorectal Cancer
The ability to examine DNA sequence information in clinical and
epidemiologic studies of CRC reveals that traditional categories
for classifying CRC cases and their causes are not as distinct
as they once seemed. On the other hand, insights at the molecular
level may suggest more useful models for sorting out pathogenetic
mechanisms of CRC, along with more specific targets for prevention,
diagnosis, and treatment. For example, we now know that inherited
variation in the APC gene gives rise to several different cancer
syndromes with involvement at various extracolonic sites (e.g.,
Gardner syndrome, Turcot syndrome); a form of attenuated FAP in
which far fewer polyps are found; and a susceptibility polymorphism
in Ashkenazi Jews associated with only modestly increased risk
for CRC.(37)
HNPCC for now remains a diagnosis based on pedigree because
genotype-phenotype correlation is not well enough understood
to establish sensitive and specific diagnostic criteria on the
basis of genotype. Predictive tests for HNPCC (other than family
history) would thus currently seem to be out of reach. Even
DNA-based diagnosis remains problematic. Most tumors in affected
persons exhibit microsatellite instability (MSI), which has
been proposed as an initial screening test before sequencing
MSH2 and MLH1; however, MSI testing itself is a costly and complex
procedure that is not entirely sensitive or specific for cancer
in HNPCC families.(16)
The molecular pathways that give rise to FAP and HNPCC are
also important in sporadic CRC; thus events leading to loss
of functional APC and MMR gene products can begin either with
inherited or acquired mutations. In 1999, John Potter reviewed
the evidence implicating these and other pathways in the pathogenesis
of CRC, along with recognized or postulated gene-environment
and gene-gene interactions.(37) He pointed
out that epidemiologic studies that examine agents affecting
only one or some of the pathways could be expected to find weak
or inconsistent associations. However, he also predicted that
as these pathways became better understood, population subgroups
of similar susceptibility could be recognized for more specific
preventive interventions, and that early molecular changes could
serve as screening markers.
A recent study reported the feasibility of examining fecal
DNA for APC mutations that occur early in the pathogenesis of
CRC, suggesting future potential for new, noninvasive approaches
to screening.(38) Although highly specific,
the fecal DNA analysis was only 57 percent sensitive, positive
in 26 of 46 patients with neoplasia (9/18 adenomas, 11/28 carcinomas).
A commentary accompanying this article(39)
suggested that while this study had “drawn back a curtain
to reveal a tantalizing possibility, …there are other
curtains and other possibilities. The basis of the next molecular
screening test for CRC may not be a mutant gene but an abnormal
protein that the new science of proteomics may find.”(39,
p.304) |
New Opportunities and Challenges for Public Health
Pathogenesis at the molecular level is far better understood for CRC
than for other chronic diseases, including most other cancers. Ready
access to a premalignant lesion—the adenomatous polyp—has
afforded researchers a rare window for dissecting the process of tumorigenesis
and public health a ready opportunity for prevention. However, new technology
coupled with new analytic methods may be opening other windows onto
diseases for which preventive medicine and public health have had little
to offer until now. For example, collaborating researchers from several
federal agencies, academic medical centers, and industry recently reported
a method for using proteomic patterns in serum to identify ovarian cancer.(40)
They developed and tested a new algorithm for discriminating serum protein
profiles in patients with ovarian cancer from those in controls. The
algorithm used data-driven methods (cluster analysis and “genetic
algorithms,” in which principles of natural selection were used
to select key measurements for analysis) to distinguish patterns generated
by mass spectrometry. The algorithm successfully classified all 50 cancer
and 50 non-cancer serum samples in a “training set,” and
all 50 cancer samples in a “masked set;” however, 3 of 66
non-cancer samples were incorrectly classified (specificity 95 percent).
The authors concluded, “These findings justify a prospective population-based
assessment of proteomic pattern technology as a screening tool for all
stages of ovarian cancer in high-risk and general populations.”(40,
p.572)
New technologies are likely to continue to improve the prospects for
early intervention in disease processes and prevention of morbidity,
but epidemiologic studies are the key to their potential, beginning
at the population level for translation into tools for individualized
preventive medicine. However, evaluating the contribution of genomic
data to the evidence base for prevention only begins to address the
issue of clinical utility, which presents a larger array of complex
issues. These include the probable timing and severity of disease outcomes,
availability and effectiveness of interventions, and costs of alternative
tests, interventions, and treatments.(41) Furthermore,
the needs and preferences of tested individuals and their family members
must be taken into account.
Even if we recognize the potential of genomics for preventive medicine,
what does it mean for public health? Developments in the science and
technology of genomics have prompted widespread use of the term “paradigm
shift”(42) to describe the future of biological
research and clinical medicine. Epidemiology, the basic science of public
health, also faces challenges. Future studies of the distribution and
causes of disease in human populations will be incomplete if they do
not consider the potential contribution of genetic variation. Amid this
sea of changes, the mission of public health remains the same: to prevent
morbidity and mortality using science-based approaches that serve the
interests of the total population, with special responsibility for underserved
communities. However, new understanding of gene-environment interactions
in disease etiology and progression may suggest interventions that require
rethinking the “one size fits all” paradigm for public health
interventions.
As more genetic tests are developed and marketed for use in public
health and health-care settings, it will be important to evaluate the
value they add to existing interventions. Public health policies, backed
by strong epidemiologic research, must provide a balance to intense
commercial pressures, which have identified high-technology screening
tests, including those based on genomics, as a new opportunity for direct
marketing to individuals.(43-44) Ultimately, the public
will benefit only when genetic tests are used appropriately, interventions
are tailored to those at risk, and access is assured. Thus public health
institutions clearly have a role in helping realize the potential of
genetic information to prevent disease and improve health, by developing
appropriate research and policies, and by helping educate health-care
providers and the public.
- Collins FS, McKusick VA. Implications of the Human
Genome Project for medical science. JAMA 2001;285:540-544.
- Pagon RA, Tarczy-Hornoch P, Baskin PK, et al. GeneTests-GeneClinics:
Genetic testing information for a growing audience. Hum Mutat 2002;19:501-509.
- GeneTests. Web site. Available at http://www.genetests.org/
. Accessed April 26, 2002.
- Yoon PW, Chen B, Faucett A, et al. Public
health impact of genetic tests at the end of the 20th century. Genet
Med 2001;3:405-410.
- Online Mendelian Inheritance in Man. Web
site. Available at http://www3.ncbi.nlm.nih.gov/Omim/.
Accessed April 26, 2002.
- Peltonen L, McCusick VA. Genomics and medicine:
dissecting human disease in the postgenomic era. Science 2001;291:1224-1229.
- Collins FS. Shattuck lecture--medical and societal
consequences of the Human Genome Project. N Engl J Med 1999;341:28-37.
- Khoury MJ, Little J. Human genome epidemiologic
reviews: the beginning of something HuGE. Am J Epidemiol 2000;151:2-3.
- Ellsworth DL, Manolio TA. The emerging importance
of genetics in epidemiologic research. I. Basic concepts in human
genetics and laboratory technology. Ann Epidemiol 1999;9:1-16.
- Task Force on Genetic Testing. Ensuring the safety
and effectiveness of genetic tests. In Holtzmann NA, Watson MS (eds)
(1997). Promoting safe and effective genetic testing in the United
States: final report of the Task Force on Genetic Testing. Web site.
Available at http://www.nhgri.nih.gov/ELSI/TFGT_final/.
Accessed April 26, 2002.
- Whittemore AS, Nelson LM. Study design in genetic
epidemiology: theoretical and practical considerations. Monogr Natl
Cancer Inst 1999;26:61-69.
- Rothman KJ, Greenland S. Case-control studies.
In Modern Epidemiology, 2nd ed, Rothman KJ (ed). Philadelphia: Lippincott
Williams & Wilkins. 1998. pp.93-114.
- Ferlay J, Bray F, Pisani P, Parkin DM. Globocan
2000: Cancer Incidence, Mortality and Prevalence Worldwide. International
Agency for Research on Cancer, World Health Organization. Lyon, France:
IARC Press, 2001.
- Bodmer W. Familial adenomatous polyposis (FAP)
and its gene, APC. Cytogenet Cell Genet 1999;86:99-104.
- Dunlop MG, Farrington SM, Nicholl I, et al. Population
carrier frequency of hMSH2 and hMLH1 mutations. Brit J Cancer 2000;83:1643-1645.
- Lynch HT, Lynch JF. Hereditary nonpolyposis colorectal
cancer. Semin Surg Oncol 2000;18:305-313.
- Fuchs CS, Giovannucci EL, Colditz GA, et al. A
prospective study of family history and the risk of colorectal cancer.
New Engl J Med 1994;331:1669-1674.
- Cole TRP, Sleightholme HV. ABC of colorectal cancer:
the role of clinical genetics in management. BMJ 2000;943-946.
- Samowitz WS, Curtin K, Lin HH, et al. The colon
cancer burden of genetically defined hereditary nonpolyposis colon
cancer. Gastroenterology 2001;121:830-836.
- Yoon P, Peterson-Oehlke KL, Scheuner MT, et al.
Can family history be used as a tool for public health and preventive
medicine? Genet Med 2002;4:304-310.
- Tomeo CA, Colditz GA, Willett WC, et al.
Harvard report on cancer prevention. Volume 3: Prevention of colon
cancer in the United States. Cancer Causes and Control 1999;10:167-180.
- Rockhill B, Kawachi I, Colditz GA. Individual
risk prediction and population-wide disease prevention. Epidemiol
Rev 2000;22:176-180.
- Pearce N. Traditional epidemiology, modern
epidemiology, and public health.
Am J Public Health. 1996;86:678-683.
- Rockhill B. The privatization of risk.
Am J Public Health 2001;91:365-368.
- Wald NJ, Hackshaw AK, Frost CD. When can
a risk factor be used as a worthwhile screening test? BMJ 1999;319:1562-1565.
- Holtzman NA, Marteau TM. Will genetics
revolutionize medicine? New Engl J Med 2000;343:141-144.
- Wang X, Moylan B, Leopold DA, et al. Mutation
in the gene responsible for cystic fibrosis and predisposition to
chronic rhinosinusitis in the general population. JAMA 2000;284:1814-1819.
- Khoury MJ. Will genetics revolutionize
medicine? (letter) New Engl J Med 2000;343:1497.
- Khoury MJ, Wagener DK. An epidemiological
evaluation of the use of genetics to improve the predictive value
of disease risk factors. Am J Hum Genet 1995;56:835-844. [Erratum,
Am J Hum Genet 1996;58:253.]
- Pharoah PDP, Antoniou A, Bobrow M, Zimmern
RL, Easton DF, Ponder BAJ. Polygenic susceptibility to breast cancer
and implications for prevention. Nat Genet 2002;31:33-36.
- The Evidence-Based Medicine Working Group.
Users’ guides to the medical literature: a manual for evidence-based
clinical practice. Guyatt G, Rennie D (eds.) Chicago: American Medical
Association, 2002.
- a. Sotos JG,
Rienhoff Y Jr. Will genetics revolutionize medicine? (letter). New
Engl J Med 2000;343:1496.
b. Yang Q, Khoury MJ, Botto L, Friedman JM, Flanders W. Improving
the predicition of complex diseases by testing for multiple disease-susceptibility
genes. Am J Hum Genet 2003; 72:636-649
- Beaty TH, Khoury MJ. Interface of genetics
and epidemiology. Epidemiol Rev 2000;22:120-125.
- Satten GA, Flanders WD, Yang Q. Accounting
for unmeasured population substructure in case-control studies of
genetic association using a novel latent-class model. Am J Hum Genet
2001;68:466-477.
- Zerba KE, Ferrell RE, Sing CF. Genetic
structure of five susceptibility gene regions for coronary artery
disease: disequilibria within and among regions. Hum Genet 1998;103:346-354.
- Sing CF, Haviland MB, Reilly SL. Genetic
architecture of common multifactorial diseases. Ciba Foundation Symposium
1996;197:211-229.
- Potter JD. Colorectal cancer: molecules
and populations. J Natl Cancer Inst 1999;91:916-932.
- Traverso G, Shuber A, Levin B, et al. Detection
of APC mutations in fecal DNA from patients with colorectal tumors.
N Engl J Med 2002;346:311-20.
- Schwartz RS. A needle in a haystack of
genes. N Engl J Med 2002;346:302-304.
- Petricoin EF, Ardekani AM, Hitt BA, et
al. Use of proteomic patterns in serum to identify ovarian cancer.
Lancet 2002;359:572-7.
- Evans JP, Skrzynia C, Burke W. The complexities
of predictive genetic testing. BMJ 2001;322:1052-1056.
- Subramanian G, Adams MD, Venter JC, Broder
S. Implications of the human genome for understanding human biology
and medicine. JAMA 2001;286:2296-307.
- Lee TH, Brennan TA. Direct-to-consumer
marketing of high-technology screening tests. N Engl J Med 2002;346:529-531.
- Personalized medicine: opportunities and
challenges for commercialization. Web site accessed April 26, 2002.
http://www.lifesciencesinfo.com/email/2707/.
|