/  7
 
CLINICIAN’S CORNER
COMMENTARY
The Incidentalome
A Threat to Genomic Medicine
Isaac S. Kohane, MD, PhDDaniel R. Masys, MDRuss B. Altman, MD, PhD
G
ENOMIC MEDICINE IS POISED TO OFFER A BROAD AR
-ray of new genome-scale screening tests. How-ever, these tests may lead to a phenomenon inwhich multiple abnormal genomic findings arediscovered, analogous to the “incidentalomas” that are of-ten discovered in radiological studies. If practitioners pur-sue these unexpected genomic findings without thought,theremaybedisastrousconsequences.First,physicianswillbe overwhelmed by the complexity of pursuing unex-pectedgenomicmeasurements.Second,patientswillbesub- jectedtounnecessaryfollow-uptests,causingadditionalmor-bidity. Third, the cost of genomic medicine will increasesubstantiallywithlittlebenefittopatientsorphysicians(butwith great financial benefits to the genomic testing indus-try), thus throwing the overall societal benefit of genome-basedmedicineintoquestion.Inthisarticle,wediscussthebasis for these concerns and suggest several steps that canbe taken to help avoid these substantive risks to the prac-tice of genomically personalized medicine.
Diagnostic Testing and Incidental Findings
Physiciansaregenerallytrainedtoordertestscarefullyandonlyifsuchtestswillresultinachangeinmanagement.Forthisreason,muchtimeisspentdecidingifarenalpanelwith7 blood measurements should be expanded to a compre-hensive panel with 20 or more measurements. Physiciansknow that as the number of tests increases, the chance thataspuriousabnormaltestresultwillarisealsoincreases.Theyalso know that it is difficult to ignore abnormal findings,and they often must embark on a sequence of more expen-sive tests to investigate the findings. Furthermore, the sig-nificance of an abnormal finding is related to the preva-lence of disease in the population from which the testedpatient is drawn. Therefore, if the risk associated with thefinding was established in a population with a high preva-lence of disease, the rate of false-positive results when test-inginapopulationwithalowerrateofdiseasewillbemuchhigher.There is a rich literature in radiology on the “incidenta-loma,” which is a finding (most commonly a mass) foundon computed tomography or magnetic resonance imagingstudies ordered for symptoms or concerns totally unre-lated to the gland in which the mass is found. The workupof an incidentaloma is complicated by concerns that it maybe associated with malignant disease and, at least initially,the lack of good data on the prevalence of malignant dis-ease in the general population. Incidentalomas occur be-cause imaging modes do not only report on the areas of di-rect clinical concern but, incidentally, report on all organsin the field of view.
1
This phenomenon of possible incidental genomic find-ings—the
incidentalome
—threatenstounderminetheprom-ise of molecular medicine. In particular, the application of comprehensivegenotypeandfunctionalgenomicmeasure-ments across the general population is likely to yield un-expectedincidentalfindingsfornearlyeveryone.Ofcourse,there are important differences in the interpretation of ge-nomic data and radiological data (eg, discovering inciden-talomas may be lifesaving), but the potential similarity isthattheclinicianandpatientareconfrontedwithresultsthatthey did not anticipate when the test was ordered.The sequencing of the human genome has brought in-creasinginterestintheuseofgenome-scaletechnologiestomeasure individual variation in the human genome. A va-riety of technologies have emerged that make it economi-cally attractive to assess the structure and function of hun-dreds of thousands of genes simultaneously. Although allhumans share more than 99.8% of their genome DNA se-quence,theremaining0.2%(alongwithenvironmentalex-posures) is responsible for much of the variation in risk of disease and response to therapies. Recent reports indicatethat more than 300000 single-nucleotide polymorphismscan be measured on an individual genome for a few hun-dreddollars.
2
Clinicalgenomicsstudieshaveshownthattheexpression pattern of thousands of genes can differentiatecancer cells from normal cells and can distinguish sub-
CME available online at www.jama.com
Author Affiliations:
Center of Biomedical Informatics and Division of Health Sci-ences Technology, Harvard Medical School, Boston, Mass (Dr Kohane); Depart-ment of Biomedical Informatics, Vanderbilt University School of Medicine, Nash-ville,Tenn(DrMasys);andDepartmentofGenetics,StanfordUniversity,Stanford,Calif (Dr Altman).
Corresponding Author:
Isaac S. Kohane, MD, PhD, Center of Biomedical Infor-matics and Division of Health Sciences Technology, Harvard Medical School, 10Shattuck St, Boston, MA 02115 (isaac_kohane@harvard.edu).
212
JAMA,
July 12, 2006—Vol 296, No. 2
(Reprinted)
©2006 American Medical Association. All rights reserved.
 at STANFORD Univ Med Center on May 2, 2008www.jama.comDownloaded from 
 
types of cancer,
3-5
inflammatory bowel disease,
6,7
neurode-generative disease,
8,9
and many others.
10
Many companies are competing to create high-reliability, high-throughput assays for measuring thou-sandsofgenotypesandcellularphenotypesinordertocre-atetheinfrastructureformolecularmedicine.Inmanycases,comprehensive panels are being created that provide mea-surementsoflargenumbersofgenomicvariables.Thesepan-els raise the possibility that clinical use of genomics maynot be entirely focused on specific gene variants but mayalsoincludeanumberofrelatedvariants.Forexample,onepanelaccuratelymeasuresmorethan30polymorphismsin just 2 genes. These genes come from a group of more than200 that may be important for drug dosing, so future gen-erationsofthispanelmaytestfor200
15=3000polymor-phisms.Inthebestcase,thisinfrastructurewillprovidethebasisforgenome-informedmedicaldecisionmakingthatwillleadtodiagnosesandtherapiesthataremoretargeted,havereducedvariabilityinoutcome,maximizeefficacy,andmini-mize adverse effects.
Potential Implications of Genomic Testing
Thefollowing3-parthypotheticalscenarioillustratessomeof the implications of application of genomic testing:Thereexistsasinglegenomictestthathas99.9%sensitiv-ity (true-positive rate) and a false-positive rate of 0.1% (ie,specificity or a true-negative rate of 99.9%) for a rare treat-able disease, “X.” For comparison, cystic fibrosis tests havebeenreportedwith99%sensitivity
11
and
BRCA1
testingshowsasensitivityof81%whenthefalse-positiverateis42%.
12
Thehypotheticaltestinthiscasewasdevelopedinastudyoffami-lieswithdiseaseX,anditworkswellonthispopulation,whichhas a disease prevalence of 1 in 1000, much greater than inthegeneralpopulation.Specifically,if1000individualsfromthispopulationaretested,thentherewillbe1true-positive,1false-positive,and998true-negativeresults.Twoindividu-als(thosewiththetrue-positiveandfalse-positiveresults)aretested further at some expense; the one with the true-positiveresultisgivenadiagnosisandistreated.Thepersonwiththefalse-positiveresultistestedanddiseaseisruledout.TheconclusioninthiscaseisthatgenomemeasurementsareusefulfordiagnosingdiseaseX.Thisoccursbecausethepre-testprobabilityofdiseaseof0.001yieldsaposttestprobabil-ityof0.5ifthetestresultispositive,thesensitivityofthetestis0.999,andthefalse-positiverateofthetestis0.001,usingthe Bayes theorem.
13
However, if this same genomic test is applied to the gen-eral population (with no such occurrence of the disease intheirkinship),theoveralldiseaseprevalenceis1in100000,orapretestdiseaseprobabilityof0.00001.Ifageneralpopu-lation of 10000000 individuals is tested, 100 will have thedisease,10000peoplewilltestpositivewithnodisease,100people with disease will be missed, and 9989900 peoplewillhaveanegativetestandnodisease.Thus,10100peopletest positive and require follow-up. One hundred are accu-ratelyidentifiedashavingthedisease,butthecostofdoingsoisveryhighbecause10000haveaworkupandarefoundto not have the disorder. Thus, in this population a posi-tive test result raises the posttest probability that an indi-vidualhasdiseaseonlyfrom0.00001to0.0099,orlessthan1in1000.Now,theconclusionisthatthisisapoortestforscreening and leads to too many false-positive results. Thisproblemwillbereplayedwithindividualscomingfromdif-ferent ethnic and geographic backgrounds. As demon-stratedbytheHapMapproject,
14
thesepopulationscandif-fer in the frequencies of several genomic markers.The first example illustrates the use of a single genomictest. What if the general population is screened for severalgenetic variants at once? Suppose there is a panel of ge-nomic tests, each with superb testing performance: a sen-sitivity of 100% and a false-positive rate of 0.01%. That is,of100000individuals,eachtestwillonlyproduce10false-positive results. Assuming a disease prevalence of 1 in100000, in a population of 100000 the number of true-positive results will increase by 1 with each additional test.The increase in the number of false-positive results will be10 with each independent test, but some individuals willbe subject to multiple false-positive results; therefore, theincrease of the number of individuals with a false-positiveresultwillbeslightlylessthan10pertest.The
F
IGURE
showsthe increase in the proportion of individuals with a false-positivetestresultundertheseassumptions.Asillustrated,with 10000 independent tests, more than 60% of the en-tire population tested would have false test results.
Figure.
Percentage of Total Population With a False-Positive TestResult
7030605040201000 2000 4000 6000 8000 10
 
000
No. of Independent Tests
   P  e  r  c  e  n   t  a  g  e  o   f   T  o   t  a   l   P  o  p  u   l  a   t   i  o  n  w   i   t   h  a   F  a   l  s  e  -   P  o  s   i   t   i  v  e   T  e  s   t   R  e  s  u   l   t
Asthenumberoftestsincreasesto10000,thefractionofthepopulationthathasa false-positive test result increases to more than 60%. Any large-scale genomicpanel is therefore likely to routinely report false-positive results. The data for thisfigure were generated by running a simulation in which a population of 100000was tested with 1 through 10000 tests, each with a sensitivity of 100% and afalse-positive rate of 0.01%. That is, 10 individuals with false-positive tests wererandomly selected from the population for each test. Because some individualscould be selected more than once with a larger panel of tests, the increase in thenumber of individuals with false-positive test results is less than linear.
COMMENTARY
©2006 American Medical Association. All rights reserved.
(Reprinted) JAMA,
July 12, 2006—Vol 296, No. 2
213
 at STANFORD Univ Med Center on May 2, 2008www.jama.comDownloaded from 

212

Download this Document for FreePrint
This is a private document.

Info and Rating

personalgenome

Share & Embed

More from this user

Add a Comment

Characters: ...