|
Journal Publication |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
On
the Use of Population-based Registries in the Clinical Validation of
Genetic Tests for Disease Susceptibility by Quanhe Yang PhD1, Muin J. Khoury MD PhD2, Steven S. Coughlin PhD3, Fengzhu Sun PhD4, W. Dana Flanders MD ScD5 1 Birth Defects and Pediatric Genetics Branch, Centers for Disease Control and Prevention (CDC), MS F-45, 4770 Buford Highway, Atlanta, GA 30341. Telephone: (770) 488-7186, fax: (770) 488-7197, email: qyang@cdc.gov 2 Office of Genomics and Disease Prevention, Centers for Disease Control and Prevention (CDC), MS K-89, 4770 Buford Highway, Atlanta GA 30341. Telephone: (770) 488-3238, fax: (770) 488-3236, email: muk1@cdc.gov 3 Division of Cancer Prevention and Control, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention (CDC), MS K-55, 4770 Buford Highway, Atlanta GA 30341. Telephone: (770) 488-4776, fax: (770) 488-4639, email: sic9@cdc.gov 4 Department of Mathematics, University of Southern California, Los Angeles CA 90089. Telephone: (213) 740 2413, fax: (213) 740 2424, email: fsun@hto.usc.edu 5 Department of Epidemiology, School of Public Health, Emory University, Atlanta GA 30322. Telephone: (404) 727-8716, fax: (404) 727-8737, email: wflande@sph.emory.edu Address reprint requests and correspondence: Quanhe Yang, Division of Birth Defects and Pediatric Genetics, National Center for Environmental Health, Centers for Disease Control and Prevention (CDC), Mailstop F-45, 4770 Buford Hwy, Atlanta, GA 30341. Phone: (770) 488-7186, Fax: (770) 488-7197, Email: qay0@cdc.gov Running head: Genetic test and case-control study Purpose: Many new genetic tests for susceptibility to adult-onset diseases are developed on the basis of selected and high-risk groups. Before such tests can be used in medical practice, however, epidemiologic studies must be conducted to evaluate their clinical sensitivity, specificity, and positive predictive value in the general population. For many common adult-onset diseases, this process may take decades of follow-up. Method: We illustrate how clinical validation of new predictive genetic tests can be done retrospectively using case-control studies that are derived from population-based registries of diseases. We use the examples of birth defects and cancer registries to illustrate a hypothetical process by which such tests can be clinically validated. Results: We demonstrate how such epidemiologic studies can be successfully used to derive measures of a test’s sensitivity, specificity, positive predictive value, negative predictive value and of the population attributable fraction of disease due to the disease-susceptibility genes. Under certain assumptions, data derived from population-based case-control studies provide adequate estimates of lifetime risks for disease (penetrance) among people with specified genotypes. Discussions: With adequate protections of human subjects, studies involving population-based registries of disease will increasingly become valuable in validating the numerous genetic tests that will emerge from advances in human genetic research and the Human Genome Project. Key words: Genetic testing, registries, case-control study, epidemiologic methods The rapid pace of genetic discoveries has resulted in the proliferation of genetic tests for a wide variety of diseases. Currently, DNA-based tests are available for more than 600 conditions,1 and more than 6,000 conditions have been mapped to specific chromosomes.2 Genetic tests of susceptibility to common adult-onset disorders in the general populations have become increasing available. Examples include tests for BRCA1/BRCA2 gene mutations in relation to breast and ovarian cancer,3 4 5 tests for apolipoprotein E-E4 allele in relation to Alzheimer disease,6 7 and tests for hereditary prostate cancer 1 (HPC1) in relation to prostate cancer.8 9 Before
such genetic tests can be used in medical practice, their clinical validity
needs to be established.10
Determining the clinical validity of the tests involves measuring their clinical
sensitivity, specificity, positive predictive value (PPV), and negative
predictive value (NPV) with respect to disease occurrence.11
Ideally, estimates of these clinical measurements of genetic testing would come
from large-scale population-based cohort studies; such studies, however, will
take considerable time and resources. In this paper, we illustrate how clinical
validation of new predictive genetic tests can be done retrospectively using
incident case-control studies that are derived from population-based registries
of diseases. We demonstrate how such epidemiologic studies can be successfully
used to measure the clinical sensitivity, specificity, PPV, and NPV of such
tests, as well as the population attributable fraction of disease due to the
disease-susceptibility genes (i.e., the proportion of cases in the population
attributable to a positive genetic test - an adjusted measure of clinical
sensitivity). Under certain assumptions, data derived from population-based
case-control studies provide adequate estimates of lifetime risk that people
with specific genotypes have for common diseases (penetrance). By
genetic testing, we mean the analysis of a specific gene, its product, or
function, or other DNA and chromosome analysis that is done to detect or exclude
an alteration likely to be associated with a genetic disorder.12
Others have provided a more detailed definition of genetic testing.13
Here, we focus on genetic tests for susceptibility to common complex diseases in
humans: the term ‘complex’ disease implies that the inheritance of an
allelic variant or even combination of alleles at multiple loci may increase
risk but does not always result in the disease (incomplete penetrance). The
primary objective of such testing is to determine the probability of developing
a disease given a positive genetic test result. Such genetic tests for
susceptibility differ from diagnostic genetic tests or population screening
program for certain genetic diseases. The measurements used in the clinical
validation of genetic testing can be applied to tests for various diseases with
different ages of onset, such as birth defects, cancers, and cardiovascular
disorders. In estimating lifetime risk for disease among carriers of
disease-susceptibility genes, we focus on common diseases with varying age of
onset. We use the examples of population-based birth-defect and cancer registries to illustrate a hypothetical process of clinically validating maternal MTHFR C677T mutation testing as a means of estimating child’s risk for neural tube defects (NTDs) and BRCA1 testing as a means of estimating a woman’s risk for breast cancer. The
clinical validity of a genetic test reflects its ability to correctly classify
people who have or will develop disease as test-positive and those who will not
develop disease as test-negative. Measurements of validity include sensitivity,
specificity, PPV, and NPV. Here, sensitivity is defined as the probability of a
positive genetic test results among people who will develop disease, specificity
is the probability of a negative genetic test results among people without the
disease, PPV is the probability of developing disease among people with a
positive test result, and NPV is the probability of not developing disease among
people with a negative test result.14 The parameters needed to estimate these measurements can be obtained from a population-based registry and a case-control study derived from such a registry. Population-based registries involve a continuous and systematic process of collection, analysis, interpretation, and dissemination of descriptive information for the disease concerned as illustrated by birth defects and cancer registries.15 16 17 18 For most population-based registries, the population usually represents residents within a specified geographic area, which may be a city, region, or nation. The primary objective of a population-based registry is usually to collect information on all new cases of the disease over time within the defined geographic region and monitor changes over time. The data collected from these registries provide basic information about the descriptive epidemiology of the health problems, such as the overall rate of the disease in the population and its change over time. Such information can also be used to evaluate intervention programs and to help formulate policy decision regarding health care and resource allocation.19 For the clinical validation of genetic testing, population-based registries can provide information about the overall disease rate in the population as well as serve as a framework for population-based case-control studies. Population-based case-control study Although population-based registries collect all cases of a disease with basic demographic information, they may not contain sufficient detailed information for in-depth epidemiologic studies. However, population-based registries can serve as a framework for investigators designing detailed epidemiologic research such as a population-based case-control study. In a population-based case-control study, investigators ascertain all new
cases of diseases diagnosed in a certain time interval in a defined population,
and the control subjects are randomly chosen from the same underlying
population.20
The advantages of population-based case-control studies include the ability to
quantify the magnitude of disease risks in the underlying population, to
estimate the population attributable risk for the disease due to specified risk
factors, and to estimate the absolute risk for disease.21
For the clinical validation of genetic testing, such studies can be used to
derive the sensitivity, specificity, PPV, and NPV of such tests. Under certain
assumptions, one can also use data from case-control studies to estimate the
life time risk (penetrance) of the common diseases with varying age of onset. Estimations of sensitivity, specificity,
PPV, and NPV For simplicity, we assume a single disease-susceptibility gene with two alleles. The estimations also apply to multiple alleles at multiple loci. If one views the result of genetic testing of disease-susceptibility genotype as a risk factor (positive or negative on the test), then estimating the probability of developing disease among people with disease-susceptibility genotype (PPV) can be done by estimating exposure-specific incidence rates using data from a case-control study.22 23 Table 1 shows how to estimate sensitivity, specificity, PPV, and NPV. For conditional probability, Bayes’ theorem states that:
where Pk(D/G) = the annual risk of disease among people with a positive genetic test result for the disease-susceptibility gene in the kth age interval, which is, in other words, PPV; Pk(G/D) = the probability of a positive test result for the disease-susceptibility gene among case subjects in the kth age interval; Pk(G/ö) = the probability of a positive test result among control subjects in the kth age category; and Pk(D) = the prior probability of disease (overall rate of disease) among people in the kth age interval. To simplify, we assume Pk(G/ö) is constant across age. For most diseases, the annual risk and the incidence rate per person year will be nearly the same numerically. Similarly, one can estimate
NPV as follows:
where
= the annual risk of disease among people with a negative genetic test result
for the disease-susceptibility gene in the kth
age interval (NPV); and
= probability of a negative test
result for the disease-susceptibility gene among case subjects in the kth age interval. It is important to note that the probability of people with the specified
genotype developing disease is age dependent. For diseases expressed at birth,
such as external structure malformations, k=0;
for all other diseases, k
represents
age at disease onset. If there is no matching or other important biases in
subjects selection, one can estimate the test (genotype) probability among case
and control subjects from the case-control study as Pk(G/D) = a/N1,
and Pk(G/ö)
= b/N2
(Table 1). Thus, Pk(G/D) approximates the
sensitivity of a genetic test, and
approximates the specificity
of a genetic test in a follow-up cohort study. In this sense, we define Pk(G/D)
and
as ‘sensitivity’ and
‘specificity’ of the test. For overall rate of disease in the population, Pk(D),
the numerator represents the number of incidence cases in a specific time
interval, and the denominator is the size of the population at risk. One may get
these overall disease rates from population-based registries, for example, using
data from birth defects surveillance programs or from Surveillance,
Epidemiology, and End Results Program (SEER) for cancer incidence rates in the
general population. If estimates of
risk ratio for disease among carriers of the disease-susceptibility gene and the
prevalence of disease-susceptibility gene in the population are available, the
estimate of PPV can also be obtained by:
where Rk = risk
ratio for disease among carriers of the disease-susceptibility gene in the kth age interval; and Pk(G) = the prevalence of the
disease-susceptibility gene in the kth
age interval in a population. The estimations of ‘sensitivity’ and
‘specificity’ can also be expressed in terms of Rk,
Pk(G) and Pk(D).14 In addition, using data from a population-based case-control study, one
can estimate the population attributable fraction, which is the proportion of
all new cases in a given period that were identified by a positive genetic test
result. Assuming no confounding of genetic test-disease association, one can
estimate attributable fraction using the following formula:24
where AFk is the
population attributable fraction for age group k, fk is the
proportion of people in age group k with the disease-susceptibility gene in the
population, and yk
is risk of having the disease in question among those who have the
disease-susceptibility gene compared with those who have no
disease-susceptibility gene. Other formulas are also available for estimating
the population attributable fraction.25
26 Estimation of lifetime risk (penetrance) of common diseases
Formulas (1) or (3) provide the point estimates of the genotype-specific
incidence rate by age; these estimates can be used to approximate the penetrance
of the disease-susceptibility for diseases expressed at birth, such as birth
defects, where k=0. For common complex
diseases with varying age at onset, we are interested in the cumulative risk
over certain time intervals of the disease-susceptibility gene among people with
positive test results, for example, the cumulative risk of BRCA1-gene mutation
carriers for breast cancer by the different age group. To calculate cumulative
risk, one can apply either formula 1 or 3, depending on the availability of
data, to each age group to get separate estimates of the age-genotype-specific
incidence rate, and calculate the cumulative incidence (risk):
(5) where Rj
is the cumulative risk for disease (penetrance) among
carriers of the disease-susceptibility genotype by the Kth age group. IDkj
is the age-genotype-specific incidence rate in the kth age group (say j=0 for carriers of the nonsusceptibility
genotype and j=1 for carriers of the disease-susceptibility genotype), and ) tk
is the width of age interval. We estimate
IDkj
as Pk(D/G) for j=1 and
for j=0, assuming one year
risk, Pk(D/G), is rare. If we do not take into account other risk
factors or competing risks, the cumulative risk for disease (Rj)
is equivalent to the penetrance of the disease-susceptibility genotype for
diseases with varying ages of onset. When the study is stratified by one or more risk factors such as smoking
status, the above formulas may be applied to different strata to get
stratum-genotype-specific estimates. The difference of estimates for different
strata may provide some evidence of gene-environment interaction. Although Bayes’
theorem is used mainly in prevalence and cumulative incidence studies, it also
provides a good approximation in incidence density studies.20 We present two examples of how data from population-based case-control
studies can be used to estimate the ‘sensitivity’, ‘specificity’, PPV,
and NPV of genetic tests, and to estimate the lifetime risk for disease
(penetrance) among people with disease-susceptibility gene. In one example, we
use data derived from a population-based case-control study of the association
between homozygousity for 5,10-methylenetetrahydrofolate reductase (MTHFR) gene
C677T mutation and risk for neural tube defects (NTDs). In the other example, we
created a case-control data set to estimate, by age group, the penetrance of
BRCA1-gene for breast cancer among carriers. Several recent studies reported an association between homozygousity for
the MTHFR gene C677T mutation and risk for NTDs.27
28
29
30.
In a meta-analysis of that association, Botto and Yang31
found a pooled odds ratio of 1.73 (95% CI 1.39-2.16) among infants who were
homozygous for MTHFR C677T variation and an odds ratio of 1.15 (95% CI
0.98-1.34) among those who were heterozygous for the mutation. Shaw
et al.30 investigated the joint effects of having MTHFR C677T
genotype and maternal use of supplements containing folic acid on an infant’s
risk for NTDs in a population-based case-control study conducted in California
(1987-1991 birth cohorts). The study genotyped the allelic variants of MTHFR in
214 liveborn case infants with spina bifida and 503 control infants without
birth defects. Of the 214 case infants, 41, had two alleles for the mutation,
100 had one, and 73 had none, of the 503 control infants, the corresponding
figures were 72, 213, and 218 with a estimated allele frequency of 35.5% (95% CI
32.5% - 38.5%) in the population. Table 2 shows our estimates of the
‘sensitivity’, ‘specificity’, PPV, NPV and population attributable
fraction of infants for MTHFR gene C677T mutation for risk of NTDs. The
estimated prevalence of NTDs is about 1/1,000 in the general population. As
shown in Table 2, if homozygosity for the C677T variant is used to measure risk
for NTDs, the ‘sensitivity’ and PPV of such test are low, 19.2% and 0.14%
respectively, and the NPV is high (99.9%). The estimated population attributable
fraction of NTDs due to homozygosity for the MTHFR gene C677T mutation is about
7.9%. To ensure sufficient numbers of case and control subjects for the second
example, we used the results of Whittemore et al.32
to calculate the parameters necessary to estimate PPV, NPV, and lifetime risk
for disease (penetrance). In that study, Whittemore et al., using data from
three U.S. population-based case-control studies, estimated that the percentages
of breast cancer cases attributable to BRCA1/BRCA2 mutations were 11.2, 10.7,
8.6, and 5.8 percent for women aged 15-29, 30-39, 40-49, and 50-59 years,
respectively, and the estimated carrier prevalence of BRCA1 mutations in the
U.S. population is about 0.0029 (P(G/ö)).
Another study suggested that about 60% of BRCA1/BRCA2 mutation carriers were
BRCA1 carriers.33
Therefore, we multiplied Whitemore et al.’s estimates of the percentages of
breast cancer attributable to BRCA1/BRCA2 mutations by 0.6 to estimate the
percentages attributable to BRCA1 only (Pk(G/D)).
We used equation 4b to estimate age-specific risk ratios (Rk)
and assumed that the proportion of people with the disease-susceptibility gene
in the population was constant (fk=0.0029 for all k). For overall
disease rates Pk(D), we used the SEER 1988 breast cancer incidence
rate.34
With these estimated parameters (Pk(G/D), Rk, and
Pk(D)), we then apply equation 3 to estimate age-specific exposure
rates (Pk(D/G)). Finally, we used equation 5 (cumulative risk for
disease (penetrance)) to estimate penetrance of breast cancer among BRCA1
mutation carriers (Table 3). Figure 1 presents our estimates of BRCA1-gene’s
penetrance among BRCA1 carriers and the estimates of two other studies.35
36
36
36
These estimates were comparable, though we used estimates of the percentages of
BRCA1 mutation carriers among case and control subjects from a published study.32
This example illustrates how the population-based registries provide the
estimates of absolute risk and the population-based case-control study provides
estimates of relative risk, and how one can then use these values to estimate
lifetime risk (penetrance). It should be pointed out that the main purpose of
our estimations is to illustrate how the clinical validity of genetic testing
can be calculated using data from population-based case-control studies, not the
precision of such estimates. We illustrated how to use data derived from population-based case-control studies in the clinical validation of genetic testing. We presented a simple yet valid method of estimating lifetime risks for disease among carriers of disease-susceptibility genes for common diseases with varying ages at onset. The rapid development of molecular genetics and the Human Genome Project will make genetic testing for various diseases increasingly feasible. Clinical validation of genetic testing is an important component to consider before such tests can be used in medical practice. Ideally, the clinical validation of such tests and the lifetime risk (penetrance) estimates of disease susceptibility will come from large-scale population-based cohort studies. However, in the absence of such studies, the simple approaches presented here may prove to be helpful in clinical validation of genetic tests and in determining the lifetime risk for disease (penetrance) among carriers of disease-susceptibility genes. Other researchers who examined the relationship between genetic
testing parameters (sensitivity, specificity, and PPV) and the frequency of
disease-susceptibility genes, the frequency of the disease concerned, and the
relative risk for the disease among carriers of disease-susceptibility genes
concluded that the suitability of testing for a particular
disease-susceptibility gene depends on the objectives of the testing.10,13
For most of the common complex diseases, any single disease-susceptibility gene
may increase a carrier’s risk for disease only moderately, and the genotype
frequency in the population may be relatively high.37
Under these circumstances, the sensitivity and PPV of genetic tests will tend to
be low for rare diseases such as birth defects and cancers.
We
presented our examples using data derived from population-based birth defects
and cancer registries, which are well known disease registries. Other
established population-based registries include registries for diabetes, stroke,
and communicable diseases.
38
39
40
41
42
Population-based registries for other common complex diseases, such as
cardiovascular and mental disorders, are still relatively rare which limits the
application of the present method to other nonregistered diseases. In addition,
some genes may increase carriers’ risks for multiple health problems; for
example, BRCA1 mutations are associated with breast and ovarian cancers, and p53
gene variants are associated with multiple primary cancers. Population-based
registries may be available for some, but not all of these diseases. As
with any case-control design, the definition and selection of case and control
subjects are crucially important to the validity of the study findings.
Particularly, the effects of population stratification
may invalidate attempts to clinically validate a genetic test on the basis of
data from a case-control study. For example, if a disease-susceptibility gene
occurs more often in a particular ethnic population, and this population also
has relatively high risk for the disease, the association of disease with
genetic markers may be overestimated. To reduce such a spurious association
between disease and genetic factors in a case-control study, the investigators
should select the appropriate control subjects, possibly by matching control
subjects with case subjects within major racial and ethnic subgroups; they
should collect other cultural, and anthropologic information and environmental
factors for refined matching of case and control subjects; they should consider
using DNA markers that characterize the genetic differences in subgroups of
populations as the biologic markers for population admixture.21
In
estimating lifetime disease risk (penetrance), our approach did not take into
account other risk factors or competing risks. Feuer et al. presented an
extended version of equation 5 that accounts for competing risks.43
Although model-based estimates of exposure-specific incidence rate and
cumulative risk estimates developed for other studies,44
45
46
47
45
46
47
45
46
47
can be used to make inferences about the general population, they are most
useful when applied to assess individualized risk.43 44
A number of
genetic analysis models have also been used to estimate the penetrance of
disease-susceptibility genes.34 35
The calculations used in these
approaches, however, are more complicated and require specialized genetic
analysis programs. Clinical
validity is only one of the criteria used to develop safe and effective genetic
testing. Others include analytical validity, which measures the ability of a
test to predict correctly the underlying genotype, and clinical utility, which
measures the benefits and risks to people with positive and negative testing
results.11
Other important issues in the development of safe and
effective genetic testing include the quality of laboratories performing genetic
tests; patients psychological responses to genetic testing; and the ethical,
legal, and social implications of genetic testing, which are discussed by many
other studies.10-13 48
49
50
51
52
53
54
55 The rapid advances in human molecular genetics and
then Human Genome Project seen in the past few years indicate that within the
next decade genetic testing will be increasingly used widely in population
screening for susceptibility of various common diseases, and in the diagnosis of
diseases, and management of patients. Data derived from population-based
case-control studies can help researchers assess the clinical validity of such
genetic testing and to estimate the lifetime risk for disease (penetrance) among
carriers of disease-susceptibility genotypes. Acknowledgements We thank Drs J. David Erickson and Adolfo Correa for their helpful comments. Estimation of a genetic test’s ‘sensitivity’,
‘specificity’, positive predictive value (PPV) and negative predictive value
(NPV) on the basis of data derived from a population-based case-control study
a = the number of people who test positive and have
the disease. b = the number of
people who test positive but do not have the disease. c = the number of
people who test negative but have the disease. d = the number of
people who test negative and do not have the disease. ‘Sensitivity’ = P(G/D) = a/N1: the probability that people with disease test positive. ‘Specificity’
=
= d/N2: the probability that people
without disease test negative.
PPV is the
probability of developing disease in the kth
age interval given a positive genetic test result (see text for estimation of
PPV).
NPV is the
probability of not developing disease in the kth
age interval given a negative genetic test result (see text for estimation of
NPV). Rk= the risk ratio (odds ratio) of having disease-susceptibility gene in the kth age interval. Estimates of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and population attributable fraction of infant for MTHFR gene C677T mutation and risk for NTDs using data from a population-based case-control study, California USA
[i]
McKusick VA. Mendelian inheritance in man. Catalogs of human genes and
genetic disorders (12th edition). Baltimore: Johns Hopkins University Press,
1998. [ii]
Online Mendelian Inheritance in Man, OMIM (TM). Center for Medical Genetics,
Johns Hopkins University (Baltimore, [iii] Ponder B. Genetic testing for
cancer risk. Science 1997;278:1050-1054. [iv]
Shattuck-Eidens D, McClure M, Simard J, et al. A collaborative survey of 80
mutations in the BRCA1/BRCA2 breast and ovarian cancer susceptibility gene:
implications for presymptomatic testing and screening. JAMA 1995;273:535-41. [v] Hearly B. BRCA genes-booking,
fortune telling and medical care. N Engl J Med 1997;336:1448-9. [vi] Post SG, Whitehouse PJ,
Binstock RH, Bird TD, Eckert SK, Farrer LA, Fleck LM, et al. The clinical
introduction of genetic testing for Alzheimer diseases: an ethical
perspective. JAMA 1997;277:832-36. [vii] Morrion-Bogorad M, Phelps C,
Buckholtz N. Alzheimer disease research comes of age. The pace accelerates.
JAMA 1997;277:837-40. [viii] Smith JR, Freije D, Carpten
JD, et al.. Major susceptibility locus for prostate cancer on chromosome 1
suggested by a genome-wide search. Science 1996;274:1371-4. [ix]
Gronberg H, Isaacs S, Smith
JR, et al.. Characteristics of prostate cancer in families potentially
linked to the hereditary prostate cancer 1 (HPC1) locus. JAMA
1997;278:1251-1255. [x]
Holtzman NA, Murphy PD,
Watson MS, Barr PA. Predictive genetic testing: from basic research to
clinical practice. Science 1997;278:602-605. [xi] Task Force on Genetic
Testing. Promoting safe and effective genetic testing in the United States.
Final report. Baltimore: Johns Hopkins University Press (in press). World
Wide Web URL: http://www.nhgri.nih.gov/ELSI/TFGT_final/ [xii] Harper PS. What do we mean by
genetic testing? J Med Genet 1997;34:749-52. [xiii] Holtzman NA, Shapiro D. The
new genetics: genetic testing and public health. BMJ 1998;316:852-56. [xiv]
Khoury MJ, Newill CA, Chase
GA. Epidemiologic evaluation of screening for risk factors: application to
genetic screening. Am J Pub Health 1985;75:1204-1208. [xv] Edmonts LD, Layde PM, James
LM, Flynt JW Jr., Erickson JD, Oakley GP Jr. Congenital malformation
surveillance: two American systems. Int J Epidemiol 1981;10:247-252. [xvii]
Horm
JW, Asire AJ, Young JLJ.
Surveillance, Epidemiology, and End Results Program: cancer incidence and
mortality in the United States, 1973-91. Bethesda, MD: National Cancer
Institute, 1994. [xix] Buehler JW. Surveillance. In Rothman KJ, Greenland S, editors. Modern Epidemiology. Philadelphia, PA: Lippincott-Raven Publishers 1998. [xx]
Rothman
KL, Greenland S.
Modern Epidemiology. Philadelphia, PA: Lippincott-Raven Publishers. 1998. [xxi] Khoury MJ, Yang QH. The
future of genetic studies of complex human diseases: an epidemiologic
perspective. Epidemiology 1998;9:350-54. [xxii]
Neutra RR, Drolette ME. Estimating exposure-specific disease rates from
case-control studies using Bayes theorem. Am J Epidemiol 1978;108:214-222. [xxiii]
Greenland S. Multivariate estimation of exposure-specific incidence from
case-control studies. J Chron Dis 1981;34:445-453. [xxiv] Levin ML. The occurrence of
lung cancer in man. Acta Un Intern Cancer 1953;9:531-41. [xxv] Khoury MJ, Beaty TH, Cohen BH.
Fundamentals of Genetic Epidemiology. Oxford: Oxford University Press 1993. [xxvi]
Rockhill B, Newman B,
Weinberg C. Use and misuse of population attributable fraction. Am J Pub
Health 1998;88:15-19. [xxvii]
van der Put NM, Steegers‑Theunissen
RP, Frosst P, Trijbels FJ, Eskes TK, van den Heuvel LP, Mariman EC, den
Heyer M, Rozen R, Blom HJ. Mutated methylenetetrahydrofolate reductase as a
risk factor for spina bifida. Lancet 1995;346:1070-71. [xxviii]
Whitehead AS. Gallagher P.
Mills JL. Kirke PN. Burke H. Molloy AM. Weir DG. Shields DC. Scott JM. A
genetic defect in 5,10 methylenetetrahydrofolate reductase in neural tube
defects. QJM 1995;88:763-6. [xxix]
Ou CY. Stevenson RE. Brown VK.
Schwartz CE. Allen WP. Khoury MJ. Rozen R. Oakley GP Jr. Adams MJ Jr. 5,10
Methylenetetrahydrofolate reductase genetic polymorphism as a risk factor
for neural tube defects. Am J Med Genet 1996;63:610-4. [xxx]
Shaw GM. Rozen R. Finnell RH.
Wasserman CR. Lammer EJ. Maternal vitamin use, genetic variation of infant
methylenetetrahydrofolate reductase, and risk for spina bifida. Am J
Epidemiol 1998;148:30-7. [xxxi]
Botto LD, Yang QH. MTHFR and
birth defects. Am J Epidemiol 2000;151:862-77. [xxxii]
Whittemore AS, Gong G, Itnyre
J. Prevalence and contribution of BRCA1/BRCA2 mutation in breast cancer and
ovarian cancer: results from three U.S. population-based case-control
studies of ovarian cancer. Am J Hum Genet 1997;60:496-504. [xxxiii]
Breast Cancer Linkage Consortium. Pathology of familial breast cancer:
differences between breast cancer in carriers of BRCA1 or BRCA2 mutations
and sporadic cases. Lancet 1997;349:1505-1510. [xxxiv]
Kosary CL, Ries LAG, Miller BA, Hankey BF, Harras A, Edwards (eds). SEER
Cancer Statistics Review, 1973-1992: tables and graphs. NIH Pub. No.
96-2789. Bethesda, MD: National Cancer Institute, 1995. [xxxv] Claus EB, Risch N, Thompson
WD. Genetic analysis of breast cancer in the Cancer and Steroid Hormone
Study. Am J Hum Genet 1991;48:232-242. [xxxvi] Ford D, Easton DF, Peto J.
Estimates of the gene frequency of BRCA1/BRCA2 and its contribution to
breast and ovarian cancer incidence. Am J Hum Genet 1995;57:1457-1462. [xxxvii] Risch N, Merikangas K. The
future of genetic studies of complex human diseases. Science
1996;273:1516-1517. [xxxviii] Diabetes Epidemiology
Research International Group. Geographic patterns of childhood
insulin-dependent diabetes mellitus. Diabetes 1988;37:1113-9. [xxxix] LaPorte RE, Tajima N,
Akerblom HK, Berlin N, Brosseau J, Christy M, Drash AL, Fishbein H, Green A,
Hamman R, et al. Geographic differences in the risk of insulin-dependent
diabetes mellitus: the importance of registries. Diabetes Care 1985;8(Suppl
1):101-7. [xl] Mayo NE, Chockalingam A,
Reeder BA, Phillips S. Surveillance for stroke in Canada. Health Reports
1994;6:62-72. [xli] Declich S, Carter AO. Public
health surveillance: historical origins, methods and evaluation. Bull World
Health Organ 1994;72:285-304. [xlii] Halperin W, Baker EL (eds).
Public Health Surveillance. New York: Van Nostrand Reinhold 1992. [xliii]
Feuer EJ, Wun LM, Boring CC, Flanders WD, Timmel MJ, Tong T. The lifetime
risk of developing breast cancer. J Natl Can Inst 1993;85:892-7. [xliv] Greenland S. Multivariate
estimation of exposure-specific incidence from case-control studies. J Chron
Dis 1981;34:445-453. [xlv]
Gail MH, Brinton LA, Byar DP, et al. Projecting individualized probabilities
of developing breast cancer for white females who are being examined
annually. J Natl Cancer Inst 1989;81:1879-1886. [xlvi].
Benichou J, Wacholder S. A comparison of three approaches to estimate
exposure-specific incidence rates from population-based case-control data.
Stat Med 1994;13:651-661. [xlvii].
McTiernan A, Gilligan MA, Redmond C. Assessing individual risk for breast
cancer: risky business. J Clin Epidemiol 1997;50:547-56. [xlviii] Kinmonth AL, Reinhard J,
Bobrow M, Pauker S. The new genetics: implications for clinical services in
Britain and the United States. BMJ 1998;316:767-770. [xlix] Marteau TM, Croyle RT. The
new genetics: psychological responses to genetic testing. BMJ
1998;316:693-696. [l] Harper PS. Genetic testing,
common diseases, and health service provision. Lancet 1995;346:1645-1646. [li] Bell J. The new genetics: the
new genetics in clinical practice. BMJ 1998;316:618-620. [lii] Ponder B. Genetic testing for
cancer risk. Science 1997;278:1050-1054. [liii] Grady C. Ethics and genetic
testing. Adv Intern Med 1999;44:389-411. [liv]
Barber
JCK. Code of practice
and guidance on human genetic testing services supplied direct to the
public. Advisory Committee on Genetic Testing. J Med Genet 1998;35:443-445. [lv] King PA. The past as
prologue: race, class, and gene discrimination. In Annas GJ, Sherman E (eds).
Gene mapping : using law and ethics as guides. New York: Oxford University
Press, 1992:pp.94-111. Address correspondence to Dr Khoury at |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||