improving detection of precancerous … detection of precancerous and cancerous oral lesions ......

Cancer of the oropharynx is a major

cause of cancer-related death in the

United States, exceeding the annual

death rates for cervical cancer and

malignant melanoma.1 According to the American

Cancer Society’s Department of Epidemiology and

Surveillance,2 an estimated 30,750 new cases of

oropharyngeal cancer are expected to be diag-

nosed in the United States in 1999, a figure that

accounts for about 3 percent of all cancers diag-

nosed annually. Despite advances in surgery,

radiation and chemotherapy, the mortality rate

associated with oral cancer has not improved in

the last 40 years.3 Ultimately, 50 percent of people

who have oral cancer die as a result of the malig-

nancy2; 8,440 such deaths are predicted in the

United States this year alone.2 A recent report of

compiled databases from the World Health Orga-

nization4 suggests that there will be a continuing

increase worldwide in the absolute numbers of

patients with oral cancer to be treated in the

coming decades.

JADA, Vol. 130, October 1999 1445

IMPROVING DETECTION OF PRECANCEROUSAND CANCEROUS ORAL LESIONS Computer-Assisted Analysis of the Oral Brush BiopsyJAMES J. SCIUBBA, D.M.D., PH.D., FOR THE U.S. COLLABORATIVE ORALCDX STUDY GROUP*

* Thegroup islisted in itsentirety atthe conclu-sion of thearticle.

A B S T R A C T

Background.A study group composedof researchers from across the United Statesundertook a study to evaluate the sensitivityand specificity of OralCDx (OralScan Laborato-ries Inc.), a computer-assisted method of anal-ysis of the oral brush biopsy, in the detection ofprecancerous and cancerous lesions of the oralmucosa.Methods.The study group conducted amulticenter double-blind study comparingresults of OralCDx analysis with those of scalpelbiopsy of suspicious oral lesions, as well as usingOralCDx on oral lesions that appeared benignclinically.Results. In 945 patients, OralCDx inde-pendently detected every case of histologicallyconfirmed oral dysplasia and carcinoma (sensi-tivity = 100 percent, false-negative rate = 0 per-cent). Every OralCDx “positive” result was sub-sequently confirmed by histology as dysplasia orcarcinoma. The specificity for the OralCDx “posi-tive” result was 100 percent, while the specificityfor the OralCDx “atypical” result was 92.9 per-cent. In 4.5 percent of clinically benign-appearing lesions that would not have receivedadditional testing or attention other than clinicalfollow-up, OralCDx uncovered dysplasia or carci-noma (statistical sensitivity > 96 percent, P < .05, n = 131; statistical specificity for theOralCDx “positive” result > 97 percent and forthe “atypical” result > 90 percent, P < .05,n = 196).Conclusions.The authors proposethat this multicenter trial demonstrates thatOralCDx is a highly accurate method ofdetecting oral precancerous and cancerouslesions. OralCDx can aid in confirming thenature of apparently benign oral lesions and,more significantly, revealing those that are pre-cancerous and cancerous when they are not clin-ically suspected of being so. All OralCDx “atyp-ical” and “positive” results should be referred forscalpel biopsy and histology to completely char-acterize the lesion. Clinical Implications.Given the difficulty in clinically differentiating premalig-nant and malignant lesions from benign lesionswith a similar appearance, OralCDx appears todetermine the significance of an oral lesiondefinitively and detect innocuous-appearing oralcancers at early, curable stages.

JA D

A

CO

NT

I

NU

I N G E D UC

AT

IO

N

✷✷

ARTICLE 1

Copyright ©1998-2001 American Dental Association. All rights reserved.

COVER STORY

unreliability of the oral exami-nation as a method of detectingearly-stage oral cancer afterstudying 208 oral cancers andfinding that approximately 25percent of them appearedbenign, lacking any clinical fea-tures of malignancy. Further-more, the high prevalence oforal abnormalities discovered asa result of oral cancer screeningprograms, reported between5 percent and 15 percent,13-15

makes it impractical to subjectevery oral lesion to histologicevaluation.16 Scalpel biopsy is

an invasive procedure associ-ated with potential morbidity.Thus, many oral lesionsundergo biopsy only when theydisplay either symptoms or clin-ical features typical of malig-nancy, while many innocuous-appearing early-stage oralcancerous lesions are merelyobserved clinically and leftundiagnosed.17 This mayexplain, in part, why more than50 percent of oral cancers arediagnosed in the advancedstages.

Delays in biopsy and, thus, inrecognition of early-stage oral

DETECTING ORALCANCER

Detection of oral cancer in theearly asymptomatic stage dra-matically improves cure ratesand patients’ quality of life byminimizing extensive, debili-tating treatments. The five-yearsurvival rate for patients withearly, localized disease approxi-mates 80 percent; for those withdistant metastases, it is 19 per-cent.5 Unfortunately, more than50 percent of patients with oralcancer display evidence ofspread to regional lymph nodesand metastases at time ofdiagnosis, and approxi-mately two-thirds ofpatients have apparentsymptoms, a negativeprognostic indicator.6

Although screeninghas been emphasized as amethod of reducing themorbidity and mortalityassociated with oral can-cers, the visual detectionof oral cancer at an earlystage is significantly hin-dered by the difficulty inclinically differentiatingpremalignant and malig-nant lesions from similar-looking benign lesions.7,8

In contrast, visual inspectionof the skin by dermatologists isa reliable screening method fordetecting melanoma, with sensi-tivity and specificity rates ofapproximately 93 percent to 98percent.9,10 Early-stage oral can-cers are asymptomatic. Further-more, they often may appearinnocuous, since the classic clin-ical characteristics associatedwith advanced oral cancers—including ulceration, indura-tion, elevation, bleeding andcervical adenopathy—usuallyare absent in early-stagelesions.11

Sandler12 emphasized the

cancers are well-documentedand are common. One studydemonstrated that one-third ofpatients eventually diagnosedwith oral cancer received inap-propriate therapy for incor-rectly diagnosed conditions.18

Given the limitations of the oralcavity examination in identi-fying oral cancer and the signif-icant morbidity and mortalityassociated with advanced oralcancer and its treatment, theneed for early detection ofapparently innocuous oral can-cers is compelling.

Early evaluation of oralprecancerous lesions canhave a dramatic impacton oral cancer mortalityrates.19 Erythroplakia,occurring as either an iso-lated lesion or as a com-ponent of leukoplakia(erythroleukoplakia), hasbeen emphasized repeat-edly as a marker ofsevere epithelial dys-plasia or carcinoma insitu.11,20 The significanceof the leukoplakic lesion,the most common pre-cursor of oral cancer (85percent of all precan-

cerous lesions are leuko-plakic), also has important

prognostic implications.19 Likeoral cancer, leukoplakia has avaried appearance, andalthough certain clinical fea-tures may indicate the lesionsthat have a greater risk ofbecoming malignant, leuko-plakias that histologically dis-play severe dysplasia, carci-noma in situ or frank carcinomaoften are asymptomatic andappear totally harmless.21 More-over, lesions that are large andominous-looking may prove tohave no significant histologicabnormalities.

The clinical evaluation of

1446 JADA, Vol. 130, October 1999

Early evaluation

of oral precancerous

lesions can have

a dramatic impact

on oral cancer

mortality rates.


leukoplakia is further compli-cated by the fact that theappearance of the lesionschanges over time.22 The rangeof malignant transformation ofleukoplakia varies consider-ably—from less than 4 per-cent19,21 to more than 40percent,21,23 depending on thespecific subtype studied. Sincethe malignant transformation ofleukoplakia cannot be accu-rately predicted solely on thebasis of clinical characteristics,histologic evaluation has beenrecommended for all suspiciouslesions.19 However, given thelarge number of patients withleukoplakia, estimated at 3 per-cent of the U.S. adult popula-tion,14,23 it is not surprising thatonly 25 percent of leukoplakiasever are evaluated histologi-cally.24 Consequently, signifi-cant numbers of apparentlyinnocuous premalignant lesionsremain undiagnosed and mayprogress to oral cancer. There-fore, all cases of leukoplakia,erythroplakia and erythroleuko-plakia require evaluation.

In light of the need for moreprecise methods of identifyingoral cancer in its early stages,the U.S. Collaborative OralCDxStudy Group undertook a studyto evaluate the sensitivity andspecificity of OralCDx(OralScan Laboratories, Inc.), acomputer-assisted method ofanalysis of the oral brushbiopsy, in the detection of pre-cancerous and cancerous lesionsof the oral mucosa.

MATERIALS ANDMETHODS

A prospective multicenter trialusing OralCDx testing was con-ducted at 35 U.S. academicdental sites. Dentists special-izing in oral and maxillofacialpathology, oral medicine and

oral surgeryobtained thespecimens inthe course oftheir routineclinical prac-tice. Duringthe studyinterval (1998-1999), allpatients olderthan 18 yearsof age who hadintraorallesions dis-playing anepithelial com-ponent wereeligible forenrollment. Atinvestigatorsites thatrequired con-sent approvedby an institu-tional reviewboard, patientssigned a consent form beforeparticipating. Oral lesions cov-ered with clinically intactnormal epithelium—such asmucoceles, fibromas and pig-mented lesions—were notincluded in the study. Lesionsof the vermilion border of thelips and cutaneous surfaceswere also excluded.

The investigators clinicallycharacterized all oral lesions inthe study either as innocuous oras causing suspicion of intraep-ithelial neoplasia. Suspiciouslesions (categorized as Class I)were analyzed by use of bothOralCDx and scalpel biopsy.Apparently innocuous lesions(categorized as Class II) that, inthe investigators’ opinion,required no further attentionother than clinical follow-upwere tested only by use ofOralCDx. Patients with appar-ently innocuous lesions that

produced abnormal OralCDxresults, as defined below, subse-quently were subjected toscalpel biopsy at the investiga-tors’ discretion.

OralCDx kits supplied toinvestigators consisted of anoral brush biopsy instrument, aprecoded glass slide andmatching coded test requisitionform, an alcohol/polyethyleneglycol fixative pouch and apreaddressed container inwhich to submit the contents.The test requisition formincluded demographic data suchas the patient’s age, sex andhistory of tobacco and alcoholuse, as well as the location, clin-ical description and category(Class I or Class II) of the orallesions.

All oral brush biopsies wereperformed using the suppliedsterile instrument (Figure 1)that was specially designed to


COVER STORY

Figure 1. The sterile brush biopsy instrument isspecifically designed to provide an appropriate full-thickness sample of the lesion in question.


obtain a complete transepithe-lial specimen. Patients whosesamples were considered inade-quate for laboratory interpreta-tion because they were incom-plete transepithelial biopsyspecimens—in other words,because they did not containadequate representation of cellsfrom all three epithelial layersof the oral mucosa (superficial,intermediate and basal)—wereexcluded from the study.

Depending on the lesion’sintraoral location and accessi-bility, either the flat surface orcircular border of the brush wasplaced against the surface of thelesion and, while firm pressurewas maintained, rotated five to10 times. Pinkness of tissue orpinpoint bleeding at the brushbiopsy site was evidence of

proper technique. Neither top-ical nor local anesthetic wasused. The cellular material col-lected on the brush then wastransferred to the bar-codedglass slide and rapidly floodedwith the fixative to avoid air-drying. After approximately 15minutes, the dry slide wasplaced in a plastic slide con-tainer and sent, with the bar-coded requisition form, in thepreaddressed mailing container.The great majority of investiga-tors had been trained in theoral brush biopsy and slidepreparation technique at aninvestigators’ meeting, and allwere provided with writteninstructions.

All OralCDx specimens wereanalyzed at OralScan Laborato-ries in Suffern, N.Y., whereas

oral and maxillofacial patholo-gists at the investigators’ dentalinstitutions histologically evalu-ated all scalpel biopsy speci-mens. Trial coordinators atOralScan Laboratories receivedall OralCDx slides and docu-ments and entered demographicand clinical data retrieved fromthe requisition forms. Thepathologist analyzing theOralCDx specimen was maskedfrom all of the clinical anddemographic data as well ashistologic results.

All OralCDx slides werestained in accordance with amodified Papanicolau method.Stained slides then werescanned by the OralCDx com-puter system, which consists ofa neural network-based image-processing system specificallydesigned to detect oral epithe-lial precancerous and cancerouscells. The OralCDx computersearches the brush biopsy spec-imen for a combination ofabnormal cellular morphologyand abnormal keratinization,which uniquely characterizesdysplasia and carcinoma of theoral epithelium. This imageanalysis process is performedusing a specially designed andtrained image processor thathas been demonstrated todetect as few as two abnormaloral epithelial cells scatteredamong thousands of normalcells distributed on an oralbrush biopsy specimen.

In addition, the OralCDxcomputer was adapted to com-plement existing oral cancerscreening modalities that usevital dyes. Specifically, the sig-nificance of oral lesions stainedwith toluidine blue, a metachro-matic vital dye that has beenshown to increase the visualdetection of oral cancers after anegative clinical examina-


COVER STORY

TABLE 1

Sex (n)

Male

Female

Age (Years)

Mean

Range

Cigarette Use

None

< 1 pack cigarettes/day

≥ 1 pack cigarettes/day

Other Tobacco Use (snuff, pipes,cigars)

Alcohol Use

None

Social use (less than 7 oz. perday)

Heavy use (7 oz. or more per day)

DEMOGRAPHIC INFORMATION

TOBACCO AND ALCOHOL USE

443

502

55

18-83

Percentage of Total

63

4

33

7

51

43

6

PROFILE OF STUDY PATIENTS (N = 945) AND CLINICALCHARACTERISTICS OF ORAL LESIONS.

Copyright ©1998-2001 Amer ican Dental Association. All rights reserved.

tion,25,26 can be determined withOralCDx testing.

Images of abnormal cellsidentified by the computersystem are individually dis-played on a high-resolutioncolor video monitor for reviewby a pathologist speciallytrained in computer-assistedanalysis of the oral brushbiopsy specimen. The com-puter video microscope outputis used by the pathologist inconjunction with a standardmicroscopic evaluation of eachoral brush biopsy specimen.The computer does not providea diagnosis of the brush biopsyspecimen; rather, it assists inthe search for and identifica-tion of abnormal cells, whichare then visually assessed andinterpreted by the pathologist,who renders a final diagnosis.The specimens were classifiedinto one of the following fourcategories:d“negative”: no epithelialabnormality;d“atypical”: abnormal epithe-lial changes of uncertain diag-nostic significance;d“positive”: definitive cel-lular evidence of epithelialdysplasia or carcinoma;d“inadequate”: incompletetransepithelial biopsy speci-mens (these specimens wereexcluded from the study).

In cases that the patholo-gist judged to be “atypical” or“positive” according toOralCDx, a summary screencontaining representative cel-lular abnormalities wasselected from the computer’svideo display, and these anno-tated images were printed andsupplied to the dentist whosubmitted the oral brushbiopsy specimen.

Statistical significance wasdetermined using the normal

approximation to the binomialdistribution with the continuitycorrection for the normal test.

RESULTS

A total of 945 patients were


COVER STORY

TABLE 2

Predominant Color

White

Red

Mixed

Mucosal-colored

Not specified

Morphology

Flat

Plaquelike

Verrucous

Not specified

Ulcerated

Yes

No

Not specified

Signs/Symptoms

None

Pain

Bleeding

Not specified

Location

Floor of mouth

Ventral tongue

Lateral tongue

Dorsal tongue

Attached gingiva

Buccal mucosa

Alveolar and labial mucosa

Hard palate

Soft palate

Oropharynx

Retromolar trigone

Not specified

CHARACTERISTIC INCIDENCE (n)

484

127

263

20

51

417

310

59

159

204

583

158

594

299

22

30

54

58

143

47

153

243

116

54

21

3

26

27

CHARACTERISTICS AND INCIDENCE OF LESIONS.


enrolled during the studyinterval; 502 (53 percent) werewomen and 443 (47 percent)were men. The patients’ agesranged from 18 to 83 years. Thedemographic features of studypatients are summarized inTable 1.

Brush biopsy specimens wereobtained from oral lesions withdiverse clinical features arisingon mucosa from all regions ofthe oral cavity (Table 2).

Of 945 lesions, 298 werejudged as clinically suspicious(Class I) and were evaluated byuse of OralCDx and scalpelbiopsy. The remaining 647lesions tested by OralCDx werecharacterized clinically as ClassII and, in the opinion of theinvestigators, did not requirehistologic evaluation. Theresults of all OralCDx andhistopathologic tests are sum-marized in Table 3. Theseinclude 29 Class II specimensthat had abnormal OralCDxresults and subsequently weretested by scalpel biopsy. Of the945 lesions, 131 revealedhistopathologic evidence of dys-plasia or carcinoma. OralCDxdetected every one of thesecases. Specifically, 78 orallesions were identified asOralCDx “positive” with defini-tive cellular evidence of epithe-lial dysplasia or carcinoma, and53 were reported as OralCDx“atypical,” with both types war-ranting histologic analysis. Thesensitivity rate, defined as ameasure of the likelihood that apatient with dysplasia or carci-noma will have an abnormalOralCDx result, is 100 percent(131/131). Of the 131 cases, 29initially were characterized asClass II lesions and ultimatelywere subjected to scalpel biopsyas a result of the OralCDxevaluation.


COVER STORY

TABLE 3

Positive

Atypical

Negative

Not Performed

TOTAL

BRUSH BIOPSYRESULTS

SCALPEL BIOPSY RESULTS

Malignantor Dys-plastic

Benign Not Per-formed

TOTAL

78

53

0

0

131

0

14

182

0

196

2

99

517

0

618

80

166

699

0

945

OVERVIEW OF ALL BRUSH AND SCALPEL BIOPSYRESULTS (N = 945).

TABLE 4

Positive

Atypical

Negative

Not Performed

TOTAL

BRUSH BIOPSYRESULTS




TOTAL

64

38

0

0

102

0

14

182

0

196

0

0

0

0

0

64

52

182

0

298

BRUSH AND SCALPEL BIOPSY RESULTS OF CLINICALLYSUSPICIOUS LESIONS: CLASS I (n = 298).

TABLE 5

Positive

Atypical

Negative

Not Performed

TOTAL

BRUSH BIOPSYRESULTS




TOTAL

14

15

0

0

29

0

0

0

0

0

2

99

517

0

618

16

114

517

0

647

BRUSH AND SCALPEL BIOPSY RESULTS OF CLINICALLYBENIGN LESIONS: CLASS II (n = 647).


No features of dysplasia orcarcinoma were evident in 196of the oral lesions that wereevaluated histologically (Table4). Of these, OralCDx reported182 as “negative,” with noepithelial abnormalities, and 14as “atypical.” The specificityrates—defined as a measure ofthe likelihood that a patientwith a lesion determined to bebenign by histology will nothave an abnormal OralCDxresult—are 100 percent(196/196) for “positive” OralCDxresults and 92.9 percent(182/196) for “atypical”OralCDx results.

Of the 647 apparently benignoral lesions characterized asClass II (innocuous-appearinglesions that in the investigators’opinion required no furtherattention other than clinicalfollow-up), 16 were reported byOralCDx as “positive” and 114as “atypical” (Table 5). Fourteenof the “positive” and 15 ofthe “atypical” oral lesionssubsequently were sub-jected to a scalpel biopsy.All 29 of the OralCDx“positive” and “atypical”Class II oral lesionsproved histologically to bedysplastic or cancerous.Some of the patients withan abnormal OralCDxresult who did not undergoa scalpel biopsy were lostto follow-up; in themajority of otherinstances, the investiga-tors determined clinicallythat the oral lesion wasbenign. For instance, inflam-matory conditions that weretested with OralCDx, such aspemphigus, lichen planus andgeographic tongue, often areattended by cellular atypia andmay result in OralCDx “atyp-ical” reports. Overall, OralCDx

uncovered 29 (4.5 percent) his-tologically confirmed, unsus-pected oral precancers and can-cers among the 647 apparently

benign Class II oral lesions.Additionally, 15.3 percent(99/647) of the Class II lesionsyielded an OralCDx “atypical”result but were not tested histo-logically.

The rate of OralCDx speci-

mens lacking adequate biopsyrepresentation of cells from allthree epithelial layers from allinvestigator sites was 7 percent.

Given the various levelsof experience among theinvestigators with thebrush biopsy technique,the median of 3.7 per-cent inadequate speci-mens (half the clinicianshad inadequate ratesthat were lower and halfhigher than this figure)and the mode of 0 per-cent (16 of the 35 siteshad no inadequate spec-imens) may be morerepresentative of the“inadequate specimen”rate to be expected

among experienced users.

DISCUSSION

Precancers and early-stage oralcancers cannot be adequatelyidentified by visual inspectionalone and easily may be over-looked and neglected, even byhighly trained professionals


COVER STORY

Figure 2. A photomicrograph of a specimen that underwent a scalpelbiopsy ordered on the basis of a prior abnormal OralCDx brush biopsyresult. Clearly, the defect created by the brush traverses the epithe-lium, terminating in an immediate subepithelial location.

Precancers andearly-stage oral

cancers cannot beadequately identifiedby visual inspectionalone and easily may

be overlooked andneglected.


with broad experience.7 Thus, amethod of detection at early,curable stages is crucial andmay lead to a reduction in thecurrently unacceptably highoral cancer morbidity and mor-tality rates. The results of thisstudy demonstrate thatOralCDx testing can be reliablyused on oral lesions with epithe-lial abnormalities as a methodof confirming their benignnature and, more importantly,revealing those that are precan-cerous and cancerous when theyare not clinically suspected ofbeing so.

Indeed, in this study, 4.5 per-cent of precancerous and can-cerous lesions were deemedclinically benign by academicclinicians and would haveremained undiagnosed at thattime had they not been detectedby OralCDx. All of these lesionsproved to be precancerous orcancerous when undergoingsubsequent histologic testing.An additional 15.3 percent ofclinically benign lesions werediagnosed as “atypical” by useof the OralCDx technique.Although these lesions were notsubjected to histologic evalua-tion, the results of this studysuggest that additional precan-cers and cancers could be antici-pated in this group.

On the basis of the results ofthis trial, it appears thatOralCDx could provide invalu-able assistance to clinicians indetermining the significance ofan oral lesion while examiningthe oral cavity. In this study,the brush biopsy was equivalentto a scalpel biopsy as a detec-tion tool, since all precancersand cancers detected by scalpelbiopsy also were detected byOralCDx. However, it should beemphasized that OralCDx doesnot substitute for a scalpel


COVER STORY

Figure 3. A. OralCDx biopsy provides a transepithelial specimen. B.Fine-needle–aspiration biopsy provides a transtumor specimen. C.Cytology provides information limited to the exfoliated superficial layer.

A

B

C


biopsy; rather, it identifies orallesions that require histologicevaluation. When this tech-nique detects cellular morpho-logic abnormalities, histology isnecessary to further assess thearchitecture of the lesion.Therefore, all OralCDx “atyp-ical” and “positive” resultsshould immediately indicate theneed for a scalpel biopsy andhistologic evaluation, to com-pletely characterize (that is, toassign a stage and grade to) thelesion. Oral lesions with “nega-tive” OralCDx results requirethe same careful clinical follow-up as negative histologicallysampled lesions. Any patientswhose samples are inadequateshould have samples takenagain to provide an optimalspecimen for analysis.

Brush biopsy vs. exfolia-tive cytology. The accuracy ofcomputer-assisted analysis ofthe oral brush biopsy as deter-mined in this multicentertrial sharply contrastswith the unreliable sensi-tivity of oral exfoliativecytology. A large numberof studies were conductedin the mid-1960s exam-ining exfoliative cytologyas a method of potentiallyidentifying precancerousand early cancerous orallesions. The results ofthose studies were notencouraging. Oral exfolia-tive cytology was found toyield unreliable results, asevidenced by the 31 per-cent false-negative rate in148 oral cancers in the studyby Folsom and colleagues27 andsimilarly high false negative-rates in other studies.28,29

Exfoliative oral cytology wasunsuccessful because of itsinherent limitations. The sensi-tivity of any cytologic evalua-

tion depends on a tedious visualsearch for potentially rareabnormalities on the micro-scopic slide. Microscopic

screening of a cytologic smearinvolves examination of hun-dreds of thousands of normalcells to identify abnormal cellsthat are sometimes few innumber and small in size.30

Normal cells exfoliating in enor-mous numbers as a result ofepithelial turnover outnumberabnormal cells exfoliating from

a dysplastic or cancerouslesion and, therefore,impede recognition ofcellular abnormalities.When exfoliativecytology is adapted tooral-mucosal abnormali-ties, the limitations areeven greater and areexacerbated by addi-tional factors. The totalnumber of abnormalcells available for cyto-logic sampling isreduced by a keratinlayer, and the high rateof epithelial turnover in

the oral cavity results ingreater exfoliation of normal

cells, further diluting thenumber of abnormal cells on thesmear.27,31,32

Brush vs. aspirationbiopsy. In the past twodecades, the field of aspirationbiopsy has developed consider-


COVER STORY

Figure 4. Color images of cellular abnormalities from OralCDx “atypical”and “positive” results with an accompanying explanation are mailed tothe dentist. These images enable the dentist to demonstrate to thepatient the abnormal test results.

The high rates ofsensitivity and speci-ficity achieved withOralCDx reflect its

close relationship tofine-needle–aspiration

biopsy.


ably as a natural outgrowth ofexfoliative cytology.33,34 Fine-needle–aspiration biopsies fromvirtually all body sites providecellular material from all layersof the lesion being analyzed.For example, fine-needle–aspi-ration biopsy of an enlargedlymph node that frequentlyaccompanies tumors of the headand neck often provides the ini-tial evidence of malignancy.35

Aspiration biopsy specimens arecomposed of individual cells aswell as of tissue fragments. Thepreparations obtained may becompared to a jumbled puzzle inwhich the components are fittedtogether by the pathologist toform a recognizable picture.36

Histology offers intact architec-ture, which is crucial for tumorclassification. However, as adisease detection tool, the aspi-ration biopsy offers similar evi-dence.37,38 In fact, it has theadvantage of offering signifi-cantly improved cytologic detail.

The high rates of sensitivityand specificity achieved withOralCDx reflect its close rela-tionship to fine-needle–aspira-tion biopsy. The fine needle isused for sampling deep-seatedlesions such as submucosal oralmasses and enlarged lymphnodes.39 The oral brush biopsyinstrument is preferred forlesions of the exposed oralmucosal surface. OralCDx

testing uses a brush biopsyinstrument with a bristle shapeand tangent modulus (that is,the bending force resulting fromthe bristle material and therelationship or angle of the indi-vidual bristles to the wire coreof the OralCDx biopsy instru-ment) that is optimized toobtain a full transepithelialbiopsy specimen (Figure 2) withminimal or no discomfort. Asdysplastic and cancerous orallesions frequently have an over-lying keratin layer, cellularabnormalities in the deep basallayer of the epithelium are bestsampled with this instrument.

Although not previously usedin the oral cavity, the brushbiopsy is a commonly usedcancer detection technique inother body sites. Studies of thebrush biopsy have validated itsuse as a diagnostic tool for avariety of upper gastroin-testinal, endobronchial lung,biliary, pancreatic, rectal andother cancers.40-44 The similari-ties between the two biopsytechniques, OralCDx and fine-needle aspiration, as well as thedifferences between these diag-nostic tools and exfoliativecytology are schematically illus-trated in Figure 3.

Advantages of the brushbiopsy. The OralCDx oralbrush biopsy is a rapidly con-ducted chairside procedure thatresults in minimal or nobleeding and requires no topicalor local anesthetic. A transep-ithelial brush biopsy is not adifficult or demanding proce-dure to master, as shown by therelatively low number of inade-quate specimens obtained inthis study by clinicians experi-enced in this technique. Thegreat majority of inadequatesamples were obtained at theonset of the trial, as investiga-


COVER STORY

Testing with OralCDx(OralScan Laboratories Inc.)is a potentially life-saving,reimbursable chairside ser-vice that can be integratedinto any dental practice.

Patients undergoingOralCDx testing incur twofees.dThe dentist performingthe brush biopsy charges forthe procedure using the ADAor CPT billing codes suppliedwith the test kit acquiredfrom the examining specialtylaboratory. OralCDx test kitsare provided to the dentist atno charge.dThe specialty laboratorybills the patient’s insuranceprovider directly for a sepa-rate analysis fee comparableto charges for other routineanatomic pathology services.

Most dental and medicalinsurance plans, including

Medicare, routinely reim-burse for both of these fees.

Although the brush biopsyis not a difficult procedure, itis recommended that generaldentists attend a briefinstructional seminar tomaximize proper use andapplication of the procedure.

These seminars are offeredacross the country at local,state, regional and nationaldental meetings, including anumber of continuing educa-tion seminars scheduled forthe fall and winter months of1999. Still more of these sem-inars were being arranged atpress time.

To learn about brushbiopsy seminars offered inyour area, call 1-800-560-4467. For information by e-mail, contact the company at“[email protected]”.

ORALCDx: REIMBURSEMENTAND FURTHER INFORMATION


tors were becoming familiarwith using the brush.

In addition to precancer andcancer detection, OralCDx canprovide morphologic evidence ofa variety of benign oral pro-cesses. In this study, OralCDxuncovered epithelial abnormali-ties consistent with candidiasis,herpes simplex virus infection,human papillomavirus infec-tion, pernicious anemia, radia-tion effects and pemphigus. Thecharacteristic morphologic fea-tures of these diseases havebeen described else-where.32,45

Computer-assistedanalysis. A critical compo-nent of OralCDx is the useof image analysis of theoral brush biopsy sample.Although automatedcytology had been pro-posed in the late 1950s asa method of reducing false-negative findings, earlyattempts that relied onanalysis by algorithmiccomputers were not suc-cessful. This limitationwas finally overcome bythe application of new,nonalgorithmic, neural net-work computers that weredeveloped in the late 1980s formissile defense. In recent years,neural networks have been suc-cessfully applied to severalmedical diagnostic procedures,including cervical smearscreening and interpretation ofdigital radiologic images suchas chest radiographs and mam-mograms.46-48 However, thediagnosis of oral squamous cellcarcinomas using a neural net-work screening system devel-oped to recognize abnormal cellsin cervical smears resulted inan unacceptably high false-neg-ative rate of 39 percent.49

By contrast, OralCDx uses

an image analysis system thatis adapted and optimized todetect epithelial abnormalitiesunique to oral brush biopsysamples, thereby enhancing itsaccuracy. In the current study,the results of OralCDx tests ofClass I and Class II lesionsshowed excellent correlationwith those obtained by scalpelbiopsy. Therefore, although amajority of oral lesions in thisstudy were categorized as ClassII and not subjected to scalpelbiopsy, it is likely that the cor-

relation of OralCDx resultswith scalpel biopsy results forthis group would be the same asthat obtained from lesionstested by both methods.

The OralCDx neural networkassists in the search of oralbrush biopsy samples for poten-tially abnormal cells, whichthen are interpreted by thepathologist. The identificationof these abnormal cells is labor-intensive, fatiguing and time-consuming; more importantly,however, abnormalities areeasily overlooked. The OralCDximages of the neural network–selected cells presented to the

pathologist for review identifycellular abnormalities thatmight otherwise have beenmissed with manual micro-scopic screening, optimizing thecombination of human and com-puter capabilities.

Interpretation of thebrush biopsy samples. Thebrush biopsy and use of neuralnetwork technology are interre-lated with a third component ofOralCDx, the specialized patho-logic interpretation of the oralbrush biopsy sample. The litera-

ture is replete withreports of failure ofcytology laboratories todiagnose invasivecancer. The interpreta-tion of oral brush biopsysamples is more prob-lematic, given thatthere are few patholo-gists with expertise inthis field and that thereis not one U.S. under-graduate dental cur-riculum requirementfor oral cytology.50

Moreover, subtlechanges of cytologicabnormalities in oral

mucosa are more difficult toappreciate than those of

abnormalities in other sites.51

As a means of minimizing diag-nostic error, OralCDx labora-tory pathologists undergo spe-cialty training in oral cytology.Furthermore, the laboratorythat analyzes these specimensfunctions exclusively in theinterpretation of oral brushbiopsy samples. As emphasizedby Hayes and colleagues,32

Allegra and colleagues51 andMorrison and Wu,52 specializedtraining for pathologists inter-preting oral cytology is manda-tory.

OralCDx is easily integratedinto conventional dental prac-


COVER STORY

Subtle changes ofcytologic abnormali-ties in oral mucosaare more difficult to

appreciate than thoseof abnormalities in

other sites.


tice. The den-tist submit-ting the spec-imen receivesa faxed report,usually withinthree daysafter the spec-imen arrivesat the spe-cialty labora-tory. In addi-tion, colorimages of cel-lular abnor-malities fromOralCDx spec-imens with“atypical” and“positive”results, withan accompan-ying explana-

tion, are mailed to the dentist(Figure 4). These images enablethe dentist to demonstrate tothe patient the abnormal testresults.

CONCLUSION

The most effective method ofcombating oral cancer is earlydetection, diagnosis and eradi-cation of early-stage lesions andtheir precursors.16 The results ofthis multicenter clinical trialsuggest that by bridging thegap between clinical inspectionand histologic evaluation of orallesions with epithelial abnor-malities, OralCDx could becomeinstrumental in achieving thisgoal.

Since the oral cavity is theonly region of the aerodigestivetract that can be effectivelyscreened, dentists should con-tinue to be encouraged to per-form oral cancer examinationsof all patients.53 Public educa-tion that stresses the impor-tance of yearly oral cancerexaminations, identification of

the warning signs of oralcancer, and recognition of thehazards associated with tobaccoand alcohol use is necessary toreverse the high morbidity andmortality rates associated withthis disease. The results of thismulticenter trial demonstratethe potential value of OralCDxas an adjunct to the oral cavityexamination in identifying pre-cancerous and cancerous lesionsat early stages, when curativetherapies are most effective. ■

The other participants in The U.S. Collabo-rative OralCDx Study Group are Allen CM,The Ohio State University College of Den-tistry, Columbus; Arm RN, Medical Center ofDelaware, Wilmington; Braun RJ, TempleUniversity School of Dentistry, Philadelphia;Cade JE, Louisiana State University Schoolof Dentistry, New Orleans; Carpenter WM,University of the Pacific School of Dentistry,San Francisco; Cohen DM and BhattacharyyaI, University of Nebraska Medical Center Col-lege of Dentistry, Lincoln; Damm DD, Univer-sity of Kentucky College of Dentistry, Lex-ington; Drinnan AJ, University of BuffaloSchool of Dental Medicine, Buffalo, N.Y.;Eisenberg E, University of ConnecticutHealth Center School of Dental Medicine,Farmington; Flaitz CM, University of Texas-Houston Dental Branch; Gordon S, Universityof Detroit Mercy School of Dentistry; Green-berg MS and Hoffman K, University of Penn-sylvania School of Dental Medicine, Philadel-phia; Greer RO and McDowell J, University ofColorado School of Dentistry, Denver;Houston GD, The University of OklahomaHealth Sciences Center, Oklahoma City;Howell RM, West Virginia University Collegeof Dentistry, Morgantown; Jacobson JJ,Helman JI and Ship JA, University ofMichigan School of Dentistry, Ann Arbor;Kaugars GE (now deceased) and Svirsky JA,Medical College of Virginia School of Den-tistry, Richmond; Kelsch RD, University ofIllinois at Chicago College of Dentistry; LillyGE, The University of Iowa College of Den-tistry, Iowa City; Lynch DP, University ofTennessee College of Dentistry, Memphis;Murrah VA, University of North CarolinaSchool of Dentistry, Chapel Hill; Neville BW,Medical University of South Carolina Collegeof Dental Medicine, Charleston; Oda D, Uni-versity of Washington School of Dentistry,Seattle; Phelan JA, Department of VeteransAffairs Medical Center, Northport, N.Y.; RoduB, University of Alabama School of Dentistry,Birmingham; Sauk JJ, University of Mary-land at Baltimore Dental School; Silverman SJr. and Lozada-Nur F, University of Cali-fornia School of Dentistry, San Francisco;Sirois DA, University of Medicine and Den-tistry of New Jersey-New Jersey DentalSchool, Newark; Stern D, Plantation, Fla.;Woo SB, Harvard School of Dental Medicine,Boston; Zegarelli DJ, Columbia PresbyterianMedical Center School of Dental & OralSurgery, New York City; Zunt SL, Indiana

University School of Dentistry, Indianapolis;Frist S, OralScan Laboratories, Suffern, N.Y.;Eisen D, OralScan Laboratories, Cincinnati.

1. Wingo PA, Ries LA, Rosenberg HM,Miller DS, Edwards BK. Cancer incidenceand mortality, 1973-1995: a report card forthe U.S. Cancer 1998;82:1197-207.

2. Landis SH, Murray MT, Bolden S, WingoPA. Cancer statistics. CA Cancer J Clin1999;49:8-31.

3. Murphy GP LWJ, Lawrence W, LenhhardRE Jr, eds. American Cancer Society textbookof clinical oncology. 2nd ed. Atlanta: Amer-ican Cancer Society; 1995.

4. Macfarlane GJ, Boyle P, Evstifeeva TV,Robertson C, Scully C. Rising trends of oralcancer mortality among males worldwide: thereturn of an old public health problem.Cancer Causes Control 1994;5:259-65.

5. National Cancer Institute. Cancer statis-tics review, 1973-1990. Bethesda, Md.: U.S.Department of Health and Human Services,Public Health Service, National Institutes ofHealth; 1993. DHHS publication (NIH) 93-2789.

6. Silverman S. American Cancer Society.Oral cancer. Hamilton, Ontario, Canada: B.C.Decker; 1998:xvi, 174.

7. Silverman S Jr. Early diagnosis of oralcancer. Cancer 1988;62:1796-9.

8. Shugars DC, Patton LL. Detecting, diag-nosing, and preventing oral cancer. NursePract 1997;22:105,109-10,113-5.

9. Whited JD, Grichnik JM. Does thispatient have a mole or a melanoma? JAMA1998;279:696-701.

10. Rampen FH, Casparie-van Velsen JI,van Huystee BE, Kiemeney LA, Schouten LJ.False-negative findings in skin cancer andmelanoma screening. J Am Acad Dermatol1995;33:59-63.

11. Mashberg A, Feldman LJ. Clinical cri-teria for identifying early oral and oropharyn-geal carcinoma: erythroplasia revisited. Am JSurg 1988;156:273-5.

12. Sandler H. Cytological screening forearly mouth cancer. Cancer 1962;15:1119-24.

13. Burzynski NJ, Firriolo FJ, Butters JM,Sorrell CL. Evaluation of oral cancerscreening. J Cancer Educ 1997;12:95-9.

14. Bouquot JE. Common oral lesions foundduring a mass screening examination. JADA1986;112:50-7.

15. Malaovalla AM, Silverman S, Mani NJ,Bilimoria KF, Smith LW. Oral cancer in57,518 industrial workers of Gujarat, India: aprevalence and followup study. Cancer1976;37:1882-6.

16. Silverman S. Oral cancer. Semin Der-matol 1994;13:132-7.

17. Schnetler JF. Oral cancer diagnosis anddelays in referral. Br J Oral Maxillofac Surg1992;30:210-3.

18. Dimitroulis G, Reade P, Wiesenfeld D.Referral patterns of patients with oral squa-mous cell carcinoma, Australia. Eur J CancerB Oral Oncol 1992;28B:23-7, Part B.

19. Sciubba JJ. Oral leukoplakia. Crit RevOral Biol Med 1995;6:147-60.

20. Silverman S Jr., Gorsky M, Lozada F.Oral leukoplakia and malignant transforma-


COVER STORY

Dr. Sciubba is

chairman, Depart-

ment of Dental

Medicine, Long

Island Jewish Med-

ical Center, New

Hyde Park, N.Y. He

also is a professor,

Oral and Maxillofa-

cial Pathology,

State University of

New York at Stony

Brook. Address

reprint requests to

Dr. Sciubba at

Johns Hopkins Med-

ical Center, Division

of Dental and Oral

Medicine, 600 N.

Wolfe St., Brady

202, Baltimore, Md.

21287.

The institutions at which Dr. Sciubba andthe other participants in the multicenterclinical trials are employed received reim-bursement for their participation fromOralScan Laboratories Inc.


tion: a follow-up study of 257 patients. Cancer1984;53:563-8.

21. Waldron CA, Shafer WG. Leukoplakiarevisited: a clinicopathologic study 3256 oralleukoplakias. Cancer 1975;36:1386-92.

22. Bouquot JE, Whitaker SB. Oral leuko-plakia—rationale for diagnosis and prognosisof its clinical subtypes or “phases.”Quintessence Int 1994;25:133-40.

23. Axell T. A prevalence study of oralmucosal lesions in an adult Swedish popula-tion. Odontol Revy 1976;27(suppl):1-103.

24. Bouquot JE, Gorlin RJ. Leukoplakia,lichen planus, and other oral keratoses in23,616 white Americans over the age of 35years. Oral Surg Oral Med Oral Pathol1986;61:373-81.

25. Ephros H, Mashberg A. Toluidine blue—viewpoints (letter). Oral Surg Oral Med OralPathol Oral Radiol Endod 1999;87:526-7; dis-cussion 528-9.

26. Mashberg A. Tolonium (toluidine blue)rinse—a screening method for recognition ofsquamous carcinoma. Continuing study oforal cancer IV. JAMA 1981;245:2408-10.

27. Folsom TC, White CP, Bromer L, CanbyHF, Garrington GE. Oral exfoliative study:review of the literature and report of a three-year study. Oral Surg Oral Med Oral Pathol1972;33:61-74.

28. Shklar G, Cataldo E, Meyer I. Relia-bility of cytologic smear in diagnosis of oralcancer: a controlled study. Arch Otolaryngol1970;91:158-60.

29. Rovin S. An assessment of the negativeoral cytologic diagnosis. JADA 1967;74:759-62.

30. Koss LG. The Papanicolaou test for cer-vical cancer detection: a triumph and atragedy. JAMA 1989;261:737-43.

31. Koss LG. Cytologic diagnosis of oral,esophageal, and peripheral lung cancer. JCell Biochem Suppl 1993:66-81.

32. Hayes RL, Berg GW, Ross WL. Oralcytology: its value and its limitations. JADA1969;79:649-57.

33. Frable WJ. Fine-needle aspirationbiopsy: a review. Hum Pathol 1983;14:9-28.

34. Lieu D. Fine-needle aspiration: tech-nique and smear preparation. Am Fam Physi-cian 1997;55:839-46, 853-4.

35. Koss LG. Diagnostic cytology and itshistopathologic bases. Philadelphia: Lippin-cott; 1961:xiii, 380.

36. Koss LG, Woyke SA, Olszewski WO.Aspiration biopsy: cytologic interpretationand histologic bases. New York: Igaku-Shoin;1992:xvi,742.

37. Muslumanoglu M, Dolay K, Ozmen V,Igci A, Bozfakioglu Y. Comparison of fineneedle aspiration cytology and excisionalbiopsy in palpable breast cancers. Radiol Med(Torino) 1995;89:225-8.

38. Koss LG. Quantitative and analyticalcytology in historical perspective. J CellBiochem Suppl 1994;19:23-7.

39. Layfield LJ. Fine-needle aspiration ofthe head and neck. Pathology 1996;4:409-38.

40. Hardwick RH, Morgan RJ, Warren BF,Lott M, Alderson D. Brush cytology in thediagnosis of neoplasia in Barrett’s esophagus.Dis Esophagus 1997;10:233-7.

41. Bilaceroglu S, Gunel O, Cagirici U,Perim K. Comparison of endobronchial needleaspiration with forceps and brush biopsies inthe diagnosis of endobronchial lung cancer.Monaldi Arch Chest Dis 1997;52:13-7.

42. Bardales RH, Stanley MW, SimpsonDD, et al. Diagnostic value of brush cytologyin the diagnosis of duodenal, biliary, andampullary neoplasms. Am J Clin Pathol1998;109:540-8.

43. Farouk R, Dodds J, MacDonald AW, et al.Feasibility study for use of brush cytology as acomplementary method for diagnosis of rectalcancer. Dis Colon Rectum 1997;40:609-13.

44. Pedersen U, Balle VH, Greisen O. Diag-nostic value of brush biopsy in suspectedbronchial carcinoma with the use of the flex-ible fibre bronchoscope. Clin Otolaryngol1981;6:329-33.

45. Silverman SJ. The cytology of benignoral lesions. Acta Cytol 1965;9:287-95.

46. Ishida T, Katsuragawa S, Ashizawa K,MacMahon H, Doi K. Application of artificialneural networks for quantitative analysis ofimage data in chest radiographs for detectionof interstitial lung disease. J Digit Imaging1998;11:182-92.

47. Mango LJ. Computer-assisted cervicalcancer screening using neural networks.Cancer Lett 1994;77:155-62.

48. Wu Y, Giger ML, Doi K, Vyborny CJ,Schmidt RA, Metz CE. Artificial neural net-works in mammography: application to deci-sion making in the diagnosis of breast cancer.Radiology 1993;187:81-7.

49. Levine TS, Njemenze V, Cowpe JG,Coleman DV. The use of the PAPNET auto-mated cytological screening system for thediagnosis of oral squamous carcinoma.Cytopathology 1998;9:398-405.

50. Kaugars GE, Silverman S Jr., Ray AK,et al. The use of exfoliative cytology for theearly diagnosis of oral cancers: is there a rolefor it in education and private practice? JCancer Educ 1998;13:85-9.

51. Allegra SR, Broderick PA, Corvese N.Oral cytology. Seven year oral cytologyscreening program in the State of RhodeIsland: analysis of 6448 cases. Acta Cytol1973;17:42-8.

52. Morrison LF HE, Wu ES. Diagnosis ofmalignancy of the nasopharynx: cytologicalstudies by the smear technique. Ann Otol1949;58:18-32.

53. Meskin LH. Do it or lose it (editorial).JADA 1997;128:1058-60.


COVER STORY


improving detection of precancerous … detection of precancerous and cancerous oral lesions ......

Documents