Differential Diagnosis of - 新潟大学歯学部 · Differential Diagnosis of ... (indicating keratin pearls), (ii) large number ... available, for a total of 20 slides from the

ORIGINAL ARTICLE

Differential Diagnosis ofWell-Differentiated Squamous CellCarcinoma from Non-NeoplasticOral Mucosal Lesions: NewCytopathologic EvaluationMethod Dependent onKeratinization-Related Parametersbut Not Nuclear AtypismHitoshi Hara, PHD,1 Tsuneo Misawa, DDS, PHD,2 Eri Ishii, CT,1

Miki Nakagawa, CT,1 Saki Koshiishi, CT,1 Kenji Amemiya, CT,1

Toshio Oyama, MD,1 Kazuya Tominaga, DDS, PHD,3 Jun Cheng, MD, PHD,4

Akio Tanaka, DDS, PHD,3 and Takashi Saku, DDS, PHD3,4*

Background: The cytology of oral squamous cell carcinoma(SCC) is challenging because oral SCC cells tend to be well dif-ferentiated and lack nuclear atypia, often resulting in a falsenegative diagnosis. The purpose of this study was to establishpractical cytological parameters specific to oral SCCs.Methods: We reviewed 123 cases of malignancy and 53 of non-neoplastic lesions of the oral mucosa, which had been

diagnosed using both cytology and histopathology specimens.From those, we selected 12 SCC and 4 CIS cases that had ini-tially been categorized as NILM to ASC-H with the Bethesdasystem, as well as 4 non-neoplastic samples categorized as LSILor ASC-H as controls, and compared their characteristic find-ings. After careful examinations, we highlighted five cytologicalparameters, as described in Results. Those 20 cytology sampleswere then reevaluated by 4 independent examiners using theBethesda system as well as the 5 parameters.Results: Five cytological features, (i) concentric arrangement oforangeophilic cells (indicating keratin pearls), (ii) large numberof orangeophilic cells, (iii) bizarre-shaped orangeophilic cellswithout nuclear atypia, (iv) keratoglobules, and (v) uneven fila-mentous cytoplasm, were found to be significant parameters. Allmalignant cases contained at least one of those parameters, whilenone were observed in the four non-neoplastic cases with nuclearatypia. In reevaluations, the Bethesda system did not help thescreeners distinguish oral SCCs from non-neoplastic lesions, whileuse of the five parameters enabled them to make a diagnosis ofSCC.Conclusion: Recognition of the present five parameters is usefulfor oral SCC cytology. Diagn. Cytopathol. 2017;00:000–000. VC

2017 Wiley Periodicals, Inc.

Key Words: concentric cellular arrangement; filamentous cyto-plasm; keratinization; keratoglobules; oral squamous cellcarcinoma

1Pathology Division, Yamanashi Prefectural Central Hospital, Kofu,Japan

2Oral Surgery Division, Yamanashi Prefectural Central Hospital,Kofu, Japan

3Department of Oral Pathology, Osaka Dental University, Hirakata,Japan

4Division of Oral Pathology, Department of Tissue Regeneration andReconstruction, Niigata University Graduate School of Medical andDental Sciences, Niigata, Japan

*Correspondence to: Takashi Saku, DDS, PhD, Osaka Dental Univer-sity, 8-1 Kuzuha Hanazono-cho, Hirakata 573-1121, Japan.E-mail: [email protected] grant sponsor: JSPS KAKENHI; contract grant numbers:JP25305035; JP26305032, and JP15K15693.

Conflicts of Interest Disclosure: The authors have no conflicts of inter-est to declare.

Received 28 September 2016; Revised 21 January 2017; Accepted 24January 2017

DOI: 10.1002/dc.23685Published online 00 Month 2017 in Wiley Online Library (wileyonli-

nelibrary.com).

VC 2017 WILEY PERIODICALS, INC. Diagnostic Cytopathology, Vol. 00, No 00 1

http://orcid.org/0000-0002-7108-8391

Oral cancer is one of the most common types of human

cancers, ranking eighth among whole body cancers world-

wide,1 and its frequency of oral cancer is well known to

be higher in southern Asia, especially in developing coun-

tries,2 such as India3 and Pakistan.4 According to our own

studies, it was sixth most common in Myanmar5 and the

most prominent (19.6%) in Yemen.6 At the same time,

increasing oral cancer frequency has been reported in

European countries, including Denmark, France, Germa-

ny, and Scotland, with prevalence mainly in males.7 Since

the most common type of oral cancer is squamous cell

carcinoma (SCC), these statistics are considered to also

reflect those for SCC.3

Death and incidence rates of oral cancer according to

age tend to be opposite in the United States as compared

to Asian countries. In the United States, the age-adjusted

death and incidence rates for males were decreased by

0.6- and 0.8-folds, respectively, in the 37 years up to

2012.8 In contrast, those rates in Taiwan were increased

by 8.5- and 14.4-folds, respectively, in the 33-year period

up to 2012.9 As for Japan, the age-adjusted death rate

from oral and pharyngeal cancer was reported to have

increased by 2.1-folds, while the rough overall rate

increased by 8.5-folds in the past 57–62 years up to

2014.10,11 Such an obvious decreasing tendency of oral

cancer in the United States seems to be the result of an

early medical examination campaign for oral cancer pre-

vention that features early detection and early treatment,

for which cytology has been applied.12

When compared with cervical and esophageal SCCs,

oral SCCs show greater differentiation in histological

findings, a phenomenon that may be due to differences in

their embryonal origin, ectodermal or endodermal.13

Well-differentiated SCCs present one of the most chal-

lenging tasks for interpreting cytology findings of oral

mucosal lesions, since they lack nuclear atypia, as

described in the WHO Blue Book.14 However, cytological

diagnosis is basically performed by referring to nuclear

atypia appearing in any organ. Thus, when this cytologi-

cal concept is applied to oral mucosal lesions, false-

negative results are inevitable in significant numbers of

oral SCC cases, as the sensitivity (true positive/true posi-

tive plus false-negative) of cytological diagnosis for oral

cancer has been reported to vary at 87.4,15 96.5,16 and

85.4%.17 These findings may indicate some limits of

exfoliative cytology,18 although the reason for the higher

rate of false-negative results has not been investigated in

detail. Rather, in clinical practice, there is a general

impression among clinicians that cytological diagnosis is

not always reliable, especially in clinically gray-zone

cases. Thus, it is important to establish oral-SCC-specific

diagnostic criteria based on the concept that most oral

SCC cases can be well differentiated and occasionally

lack nuclear atypia, and that the Bethesda System widely

used for clinical cytological examinations is primarily for

cervical and not oral lesions.

The purpose of this study was to elucidate which cyto-

logical parameters other than nuclear atypia are effective

for differential diagnosis of well-differentiated oral SCCs

and non-neoplastic lesions with nuclear atypia. To this

end, cytological findings were carefully compared with

histological findings in oral carcinoma in situ (CIS) and

SCC cases, which were objectively diagnosed based on

immunohistochemistry findings.

Materials and Methods

Specimens

Cytology specimens were obtained from patients who vis-

ited the Oral Surgery Clinic of Yamanashi Prefectural

Central Hospital, Kofu, Japan, during the 96-month peri-

od from 2004 to 2011. We examined a total of 603 speci-

mens, which accounted for 0.82% of the 73,736 cytology

specimens collected at that hospital during the same peri-

od. They were composed of 424 (70.3%) curette-scraping

samples from oral mucosal surfaces and 179 (29.7%) fine

needle aspiration samples from deeper soft parts, includ-

ing lymph nodes and salivary glands. Initial screening of

those 424 oral mucosal specimens by curette scraping

was randomly and independently performed by 4 screen-

ers, each of whom had >10 years of experience in cyto-

pathology (1 expert cytologist, 3 pathologists), to select

specimens from non-neoplastic cases, which were deter-

mined by whether they contained >200 squamous epithe-

lial cells smeared on a slide glass. Malignant cases were

chosen when 30 or more diagnosable cells were present

in the smear. From those, we then selected 123 cases of

malignancy (119 SCCs, 4 CISs) and 53 cases with non-

neoplastic lesions, both of which also had histology speci-

mens in adequate condition available for re-evaluation

using immunohistochemical hallmarks for oral malignan-

cy, such as loss of keratin (K) 13 or K19, as well as

emergence of K10, K16, and K17,19–22 and podopla-

nin.23,24 The study protocol for analyzing biopsy and

cytology samples was reviewed and approved by the Ethi-

cal Board of Yamanashi Prefectural Central Hospital.

Patients were not asked to give informed consent, because

anonymous specimens and clinical data were used.

Categorization Using Bethesda System

The 2001 Bethesda System (TBS) for reporting cervical

or vaginal cytology was the initial method used for cyto-

logical evaluations in a hospital setting.25 Among the 123

malignant cases, 107 (87.0%) were initially diagnosed as

having malignant [9 with high-grade squamous intraepi-

thelial lesion (HSIL) (7.3%), and 98 with SCC (79.7%)].

Of the remaining 16 cases not correctly diagnosed as

malignant, 5 (4.1%) were initially diagnosed as negative

Diagnostic Cytopathology DOI 10.1002/dc

HARA ET AL.

2 Diagnostic Cytopathology, Vol. 00, No 00

for intraepithelial lesions or malignancy (NILM), 1

(0.8%) with atypical squamous cells of undetermined sig-

nificance (ASC-US), 3 (2.4%) with low-grade squamous

intraepithelial lesion (LSIL), and 7 (5.7%) with ASC,

though those do not exclude HSIL (ASC-H). These 16

malignant tumors occurred in the tongue, gingiva, buccal

mucosa, or lip, as shown in Table I. Thus, the 1 expert

cytologist and 3 pathologists who served as screeners

reexamined those 16 cases by referring to the 53 non-

neoplastic lesion cases, including 1 case of LSIL and 3 of

ASC-H, all of which occurred in the gingiva as well as

the 107 malignant cases above mentioned, to determine

the cytological characteristics of oral SCC and CIS cells

other than nuclear factors.

Selection of Five Parameters for Cytologic Features

While reviewing the 20 (16 malignant, 4 non-neoplastic)

cases, each of which had samples available on 2–10

slides (Table I), the screeners listed the varieties of cyto-

plasmic appearance as possible cytological parameters by

comparing histological features with special attention giv-

en to modes of keratinization. Furthermore, they again

reviewed the remaining 107 malignant and 49 non-

neoplastic cases in addition to these 20 cases for their

cytological findings. As described in the Results section,

we finally selected the following five parameters: (i) con-

centric arrangement of orangeophilic cells indicating kera-

tin pearls; (ii) large number of orangeophilic cells; (iii)

bizarre-shaped orangeophilic cells with a spindle- or

caudate-shape with deep orange-yellow color and distinct

cell borders, but without nuclear atypia; (iv) keratoglo-

bules, a term used to designate small round cells with

deep orangeophilic cytoplasm; and (v) uneven filamentous

cytoplasm. We termed this cytologic evaluation method

based on the presence or absence of these five parameters

as the five parameter system (FPS).

Reevaluation by TBS and FPS

Among the 16 malignant cases (12 SCC, 4 CIS), which

were selected for the reevaluation test, 2 with SCC and 3

with CIS were initially diagnosed as NILM, 1 with SCC

as ASC-US, 3 with SCC as LSIL, and 6 with SCC and 1

with CIS as ASC-H, while the 4 non-neoplastic cases

were diagnosed as LSIL or ASC-H (Table I). Each case

had 2–10 slide specimens available (Table I), thus only 1

from each was selected by choosing a slide that either

contained a reasonable amount of cells or the one with

the larger numbers of cells when only 2 or 3 slides were

available, for a total of 20 slides from the 20 cases exam-

ined. This selection of slides for the reevaluation test was

performed by the first author and corresponding authors.

To avoid bias in the selection, it was not reconfirmed

whether all five of the parameters were present. Reevalua-

tions were then performed in a blind fashion by 4 Tab

leI.

Cli

nic

alan

dC

yto

logic

alC

har

acte

rist

ics

of

14

Mal

ignan

tan

d4

Non-n

eopla

stic

Cas

es,

whic

hW

ere

Not

Corr

ectl

yD

iagnose

dby

the

Bet

hes

da

Syst

em(T

BS

)an

duse

das

Subje

cts

for

Ree

val

uat

ion

by

TB

San

dth

eF

ive

Par

amet

erS

yst

em(F

PS

)

Dis

ease

cate

gori

es

Mal

igna

ntN

on-n

eopl

asti

c

SCC

CIS

CS

PV

Cas

e#

12

34

56

78

910

11

12

1314

1516

1718

1920

Age

and

sex

94F

60M

60F

64M

64F

41F

81F

76M

49F

66M

56M

77M

52F

70M

71M

82M

29M

86F

77F

54F

Loca

tion

GL

GT

TT

BG

GG

TG

TT

TT

GG

GG

Sli

de

num

ber

sa8

410

22

23

46

22

22

210

22

22

2T

BS

bA

SC

HA

SC

HA

SC

HL

SIL

NIL

ML

SIL

AS

CU

SA

SC

HA

SC

HA

SC

HN

ILM

LS

ILN

ILM

NIL

MN

ILM

AS

CH

AS

CH

AS

CH

LS

ILA

SC

H

SC

C:

squam

ous

cell

carc

inom

a;C

IS,

carc

inom

ain

-sit

u;

CS

,ch

ronic

stom

atit

is;

PV

,pem

phig

us

vulg

aris

;G

,gin

giv

a;L

,li

p;

T,

tongue;

B,

bucc

alm

uco

sa.

aN

um

ber

sof

cyto

logy

slid

eson

whic

hori

gin

aldia

gnose

sw

ere

mad

eby

TB

S.

bT

BS

:ori

gin

aldia

gnose

sby

TB

S.


KERATINIZATION IN ORAL CANCER CYTOLOGY

Diagnostic Cytopathology, Vol. 00, No 00 3

examiners, 2 experts with >20 years of experience each

as a cytologist and 2 nonexperts with <2 years of experi-

ence each. One of the expert cytologists, the first author,

was involved in the screening of the 20 cases and fixed

the 5 parameters for oral SCC as described above. Since

the other 3 examiners were not familiar with the 5 cyto-

logical parameters noted above, they underwent compre-

hensive training, during which slides containing the 5

parameter features from the other 107 cases, that had cor-

rectly been diagnosed as malignant by TBS, of the 123

malignancies were presented to them, until they were

able to recognize the parameters. The 20 slides used for

reevaluation were not used for training the 3 examiners,

hence they examined those for the first time during the

reevaluation. A repeatability test was performed by asking

the examiners for their evaluations of 20 specimens by

use of TBS or by FPS, in which the presence (1) or

absence (2) of parameters (i) (ii), (iii), (iv), and (v) were

determined. Their final diagnoses were then compared

between the two systems.

Statistical Analysis

Cytological parameter data were statistically analyzed

using Fisher’s exact test with Graph-Pad Instat (version

3.06 for Windows; GraphPad Software, San Diego, CA).

P-values <0.05 were considered to be statistically

significant.

Results

TBS Evaluation

Of the 123 malignant cases, 107 (87.0%) had been cor-

rectly diagnosed as malignant by TBS, including 1

(0.8%) HSIL (severe dysplasia), 4 (3.3%) HSIL (severe

dysplasia/CIS), 1 (0.8%) HSIL (CIS), 3 (2.4%) HSIL

(CIS/SCC), and 98 (79.7%) SCC cases. However, the

remaining 16 (13.0%) had not been diagnosed as malig-

nant by TBS and were selected as the main materials for

the present study, as described in Materials and Methods.

Cytology specimens from the 16 target cases common-

ly contained keratinizing superficial layer type (ST) and

keratinizing intermediate layer type (IMT) cells, and rare-

ly non-keratinizing ST/IMT/parabasal layer type cells.

Interestingly, malignancy in these 16 cases was clinically

doubtful, though histopathological findings later con-

firmed them to be malignant. Among the 53 non-

neoplastic cases, 1 (1.9%) was categorized as LSIL, 3 as

ASC-H (5.7%) (Table I), and the remaining 49 (92.5%)

as NILM. Histopathologically, they were confirmed to be

mucositis, granulation tissue (fibroepithelial polyp or pyo-

genic granuloma), or fistulas. Thus, the 4 non-neoplastic

cases with nuclear atypia and the 16 malignant cases

were selected for a reevaluation study.

Extraction of Parameters Specific to Oral SCC

In our review of cytology specimens from the 123 malig-

nant cases including the 16 malignant cases selected for

reevaluation and 53 non-neoplastic cases, including 4 nor-

mal control samples, we gave attention to not only nucle-

ar but also cytoplasmic findings. As a result, 5 findings

were considered to be most conspicuous in the 16 malig-

nant but not in the non-neoplastic cases. In addition, these

5 findings were confirmed to be present in some of the

remaining 107 cases. For this portion of the study, three

of the four examiners who took the reevaluation testing

were not involved. Those five parameters are described in

detail in the following sections.

Concentric arrangement of orangeophilic cells indicat-ing keratin pearls. Small clusters of keratinized cells

were often found arranged in a concentric manner, which

might represent a minimal formation of keratin pearls

within oral SCC foci in HE-stained biopsy specimens

(Fig. 1A, arrows). In cytology specimens, clusters of

orangeophilic keratinized cells arranged in a concentric

manner were often isolated from aggregates of smeared

cells (Fig. 1B). In higher magnification observations, con-

centric arrangements of orangeophilic cells in cytology

findings varied in size and shape (Figs. 1C–E), although

they were mostly identical to each histological counter-

part (Figs. 1F–H). SCC cells surrounding the concentric

core of keratinized cells loosing nuclei had polygonal

(Fig. 1F) or flat (Fig. 1G) shapes. Among those with a

flat shape, surrounding flat cells were occasionally

arranged in a concentric manner as squamous eddies or

classic keratin pearls (Fig. 1H). Thus, these concentric

arrangements of orangeophilic cells were considered to

indicate an initial stage of keratin pearl development.

When keratinized cells began to form concentric struc-

tures, they tended to detach from each other (Figs. 1C, F,

and G), though they were densely packed in larger con-

centric structures (Figs. 1E and H). It was difficult to dis-

tinguish any nuclear atypia in those orangeophilic cells

that formed concentric structures. Such concentric

arrangement of keratinized cells was observed in 9 (75%)

of the 12 SCC cases, whereas that was not seen (0%) in

the 4 CIS and 53 non-neoplastic cases.

Large numbers of orangeophilic cells. Most of the

cytology specimen smears from malignant cases con-

tained cells with a monotone appearance and orangeo-

philic cytoplasm containing nuclei (Fig. 2A). Those

orangeophilic cell clusters contained <10% nonorangeo-

philic cells, among which there were no green-stained

basal or parabasal cells. With higher magnification, those

keratinized cells consistently showed a low nuclear/cyto-

plasm (N/C) ratio (Fig. 2B). Thus, it was difficult to

determine the malignant or neoplastic characteristics of

these orangeophilic cells, though there was an


HARA ET AL.


extraordinarily large number, which was never seen in

the non-neoplastic lesion samples. In the corresponding

tissue specimens, keratinized SCC cells were thickly

stacked on the surface and formed keratin pearls in the

center of the SCC cell foci. These keratinized SCC cells

were present in the smears used for cytology (Fig. 2C).

Thus, in such oral SCC cases with extremely thick kera-

tin pearls, high cell density and monotone orangeophilic

Fig. 1. Concentric arrangement of orangeophilic cells in oral SCC. A–D, F, G: case 10. E, H: case 2. A, F–H: Hematoxylin and eosin (HE) stain;B–E: Papanicolaou stain. A, B: 3200, C–H: 3400. Small clusters of keratinized cells were found arranged in a concentric manner within oral SCCfoci in HE-stained biopsy specimens, which might represent the initial stage of keratin pearl formation (A). In cytology specimens, concentric-arranged orangeophilic keratinized cells formed clusters that were isolated from singular cells (B). Higher magnification revealed a concentricarrangement of orangeophilic cells in cytology findings (C–E) that were varied in size and shape, though mostly identical to their histological coun-terparts (F–H). There was no discernible nuclear atypia in orangeophilic cells that formed concentric structures. [Color figure can be viewed atwileyonlinelibrary.com]




http://wileyonlinelibrary.com

cells with nuclei should be regarded as malignant hall-

marks, even though their nuclei were small. Such large

numbers of orangeophilic cells were observed in 12

(100%) of the 12 SCC cases and 3 (75%) of the 4 CIS

cases, whereas that was not seen (0%) in the 53 non-

neoplastic lesions.

Bizarre-shaped orangeophilic cells without nuclearatypia. In addition to the non-atypical orangeophilic

cells noted above, orangeophilic cells with bizarre shapes,

from spindle to caudate, were also seen in cytology speci-

mens from oral SCC cases (Fig. 3A). Among the cells of

the spindle, those with extremely long and condensed

(deeply stained with orange) cytoplasm were conspicuous.

With higher magnification, those cells had elongated

nuclei without distinctive atypia, while some lacked

nuclei (Fig. 3B). In HE-stained biopsy specimens from

the corresponding cases, keratinized SCC cells with long

spindle shapes were singularly detached from the surface

of the SCC foci (Fig. 3C). These spindle-shaped surface

cells sometimes appeared as characteristic bizarre-shaped

SCC cells in cytology smears (Figs. 3A and B). Bizarre-

shaped orangeophilic cells were seen in 10 (83.3%) of the

12 SCC cases and 3 (75%) of the four CIS cases, whereas

none (0%) were observed in the 53 non-neoplastic cases.

Keratoglobules (small round-shaped cells with deeplystained nuclei and cytoplasm). Small round-shaped and

orangeophilic cells in the smears were individually

detached from green- or pink-stained squamous cells (Fig.

4A). Their cytoplasm was more deeply stained with

orange G than the bizarre-shaped orangeophilic cells

mentioned above, and atrophic nuclei were also deeply

stained with hematoxylin (Figs. 4C–H). We designated

these characteristic keratinized cells as “keratoglobules.”

There were some variations in the shapes of these kerato-

globules within the range of round to ovoid, while their

nuclei also showed variations from karyorrhexis (Figs. 4C

and D) to pyknosis (Figs. 4E and F) and karyolysis (Figs.

4G and H). Their shapes and sizes resembled those of

basal/parabasal cells, while their nuclei resembled those

of superficial cells. Such inconsistencies in cellular differ-

entiation between nuclei and cytoplasm were considered

to be malignant hallmarks. Histologically, keratoglobules

were located in the superficial zone of oral SCC nests

(Fig. 4B) and isolated from surrounding SCC cells, which

were in tight contact with each other (Fig. 4B). Their

shapes were obviously different from the flat- or spindle-

shaped keratinized cells that formed keratin pearls, though

their staining characteristics resembled those of keratin

pearls (Figs. 1–3). Keratoglobules were seen in 12

Fig. 2. Predominant population of orangeophilic keratinized cell in oral SCC (case 2). A, B: Papanicolaou stain; C: HE stain. A: 3100, B: 3400, C:3100. Most of the cells in the smears had a monotone appearance with orangeophilic cytoplasm containing nuclei (A). Higher magnification revealedno cellular atypia in these keratinized cells with a monotone appearance and a lower nuclear/cytoplasm (N/C) ratio (B). In surgical specimens, kerati-nized SCC cells were thickly stacked to form keratin pearls in the center of SCC cell foci (C). Thus, this level of cell density among orangeophiliccells with nuclei should be regarded as indicating malignancy in oral SCC cases. [Color figure can be viewed at wileyonlinelibrary.com]


HARA ET AL.



(100%) of the 12 SCC cases and 4 (100%) of the 4 CIS

cases, whereas there were none (0%) in the 53 non-

neoplastic lesions.

Uneven filamentous cytoplasm. From well-differentiated

SCC foci, orangeophilic SCC cells containing distinct

nuclei without atypia formed aggregates in the smears

(Fig. 5A). Histologically, the SCC foci were mainly com-

posed of densely packickle cell-like cells with a wide but

uneven eosinophilic cytoplasm (Fig. 5B). With higher

magnification, the orangeophilic cells were found to con-

tain wavy filamentous structures, which were irregularly

distributed with either a thick (Fig. 5C), granular and fine

wavy distribution in the periphery (Fig. 5D), along with

oval condensation around the nuclei (Figs. 5C–F). Higher

magnification of histological specimens showed that oral

SCC cells were densely packed and each had intercellular

bridges, to which thick cytoskeletal filamentous bundles

entered from the cytoplasmic side, as if painted with a

brush, and were condensed around the nuclei (Fig. 5G),

which was consistent with the cytological features (Figs.

5C–F). The nuclei of these SCC cells with irregularly

thick filamentous cytoplasm were larger and appeared to

be viable (Fig. 5), as compared with those of the cells

with keratin pearls/keratoglobules (Figs. 1 and 4). Such

uneven filamentous cytoplasm was recognized in 12

(100%) of the SCC and 4 (100%) of the CIS cases, while

that was seen in none (0%) of the non-neoplastic lesions.

Reevaluation by TBS and FPS

According to the TBS findings, all of the 12 well-

differentiated SCC cases with no obvious nuclear atypia

were determined by the 2 nonexperts to be either NILM

(33.3%, 58.3%, respectively) or ASC-US (66.7%, 41.7%),

and by the 2 experts to be NILM (0%, 16.7%), ASC-US

(0%, 16.7%), LSIL (58.3%, 33.3%) or ASC-H (41.7%,

33.3%). Four CIS cases were diagnosed as NILM (50%,

100%) or ASC-US (50%, 0%) by the nonexperts, and as

NILM (0%, 75%), LISL (25%, 0%), or ASC-H (75%,

25%) by the experts. In contrast, the 4 non-neoplastic

cases with nuclear atypia were judged as ASC-H (100%,

75%) or SCC (0%, 25%) by the nonexperts, and as ASC-

US (0%, 25%), LSIL (25%, 0%), or ASC-H (75%, 75%)

by the experts. The highest grade of ASC-H was given to

both malignant and non-neoplastic cases, though more

frequently in non-neoplastic cases. These results indicated

that diagnosis of the 16 cases without obvious nuclear

Fig. 3. Orangeophilic bizarre and spindle-shaped cells without nuclear atypia in oral SCC (case 2). A, B: Papanicolaou stain; C: HE stain. A: 3100,B: 3400, C: 3200. In addition to concentric cells, keratoglobules, and the monotone orangeophilic cell population, bizarre spindle-shaped cells wereoccasionally seen in cytology specimens from oral SCC cases (A). In higher magnification, those cells showed extremely long cell shapes with or with-out nuclei, and had no distinctive nuclear atypia (B). In an HE-stained biopsy specimen from the same case, keratinized SCC cells on the surface weresingularly detached from each other and showed long spindle shapes (C). Those spindle-shaped cells might have been seen as characteristic orangeo-philic spindle-shaped SCC cells in the smears (A, B). [Color figure can be viewed at wileyonlinelibrary.com]





atypia was extremely challenging when using TBS

(Table II).

In contrast, when using FPS, most of the malignant

cases with at least 1 of the 5 parameters confirmed were

not overlooked by any of the 4 examiner cytologists. As

shown in Table II, detection rates were quite similar

between the second expert and 2 nonexperts. All of the

parameters were noted at higher rates in SCC (60–96%)

and CIS (63–81%), except for parameter (i), which was

not always present in malignant cases but never detected

in reactive lesions by the 4 cytologists (Table II, right-

most column). The total numbers of the 5 parameters

confirmed in each case (Table II, bottom of each lesion

category) indicated that all parameters were not necessari-

ly included in each malignant specimen. However, there

were no differences in malignant judgment between the 2

experts and 1 of the nonexperts (12/12 SCC cases; 4/4

CIS), and nearly no differences with the other nonexpert

(11/12 SCC; 4/4 CIS), although the nonexperts tended to

identify fewer of the 5 parameters. In spite of those

reduced numbers, the nonexperts gave a malignant judg-

ment even when only a single parameter was noted, indi-

cating that any of the 5 parameters is a helpful indicator

of malignancy. In the CIS cases, at most 3 or 4 parame-

ters were included, and the nonexperts tended to overlook

them more frequently (3/4 cases). Nevertheless, as shown

in Table II, none of the 5 parameters were identified in

non-neoplastic cases, even those with nuclear atypism.

Discussion

The goal of this study was to determine cytological crite-

ria specific to well-differentiated oral SCC and CIS,

which lack discernible nuclear atypia, because their char-

acteristic of no or few atypical features tends to lead to

false-negative diagnoses.18 To this end, we newly deter-

mined 5 parameters for cytoplasmic features of well-

differentiated (keratinized) oral SCC cells after compara-

tive examinations of cytological and histological findings.

The present trial was successful for discriminating SCC

cells in cytology specimens based on their cytoplasmic

characteristics. An advantage of FPS is that it can widen

the value of cytological examinations for oral SCC cases,

most of which are well differentiated and difficult to pre-

cisely diagnose based only on nuclear features shown by

TBS.

This study showed that TBS functioned well for oral

cases (119 SCC, 4 CIS) that had nuclear atypia, while it

did not function well for 16 (13%) cases with cytology

specimens showing no obvious nuclear atypia. Similar

results were previously reported by Koch et al.,26 who

were able to cytologically detect 92 (88.5%) of 104

malignant (SCC and CIS) cases that were histologically

Fig. 4. Keratoglobules in oral SCC (case 10). A, C–H: Papanicolaou stain; B: HE stain. A: 3400, B: 3200, C–H: 3400. Singular small round-shapedand orangeophilic cells were found detached from green- or pink-stained squamous cells in the smears (A). The cytoplasm of these cells was deeplystained with orange G, while atrophic nuclei were also deeply stained with hematoxylin (C–H). These characteristic keratinized cells were designatedin this study as “keratoglobules.” They showed some variations in shape from round to ovoid, while their nuclei also showed variations from karyor-rhexis (C, D) to pyknosis (E, F), and karyolysis (G, H). Histologically, keratoglobules were located in the superficial zone of oral SCC nests and isolat-ed from surrounding SCC cells, which were in tight contact with each other (B). These were obviously different from flat or spindle-shapedkeratinized cells that formed keratin pearls (shown in Figs. 1–3). [Color figure can be viewed at wileyonlinelibrary.com]


HARA ET AL.



confirmed. Even after a second review, they did not find

any evidence of malignancy in 8 of the remaining 12

cases.26 Thus, it is possible to conclude that at least 10%

of oral SCC/CIS cases may be overlooked in cytological

evaluations that depend on TBS findings, because >10%

of oral SCC/CIS cells lack nuclear atypia.

TBS was mainly devised for gynecological cytology, in

which ASC-US and LSIL are not objects for a biopsy or

close examination. As confirmed in the present re-

evaluation test results, cytologists not trained for oral can-

cer have difficulty at times detecting well-differentiated

SCC/CIS cells. In cases of cervical cancer, HPV tests are

performed when a diagnosis of ASC-US or LSIL is made,

while HPV-positive (1) cases are further examined using

biopsy findings and HPV-negative (2) cases are cytologi-

cally reexamined after 1 year.27 However, oral surgeons

must maintain suspicion regarding oral cancer even after

repeated cytology examinations for cases with clinically

suspected malignancy when they receive a diagnosis of

NILM, ASC-US, or LSIL, because HPV test results are

not informative for oral malignancy cases, as they are not

always infected by HPV. Thus, it is important for oral

surgeons and cytologists/pathologists to communicate

regarding patients with clinically suspected malignancy,

otherwise those would not be subjected to follow-up

examinations until lesions developed in a later stage. This

is because HPV-positive head and neck SCCs are thought

to be classified as nonkeratinized or less-differentiated

histological types, and arise in the oropharynx.28 We con-

sider that obtaining cytology results from well-

differentiated SCC is challenging, thus HPV tests are not

always required in cases of oral SCC, as most of those

Fig. 5. Uneven filamentous cytoplasm of orangeophilic cells in oral SCC (case 9). A, C–F: Papanicolaou stain; B, G: HE stain. A: 3100, B: 34, C–G: 3400. Some orangeophilic SCC cells were characterized by a filamentous cytoplasm appearance in both (A) cytology and (B, G) histology findings.Cytoskeletal filaments were unevenly distributed in the cytoplasm, with a granular or fine wavy appearance in the periphery, and condensed around thenucleus (C–F). In the most extreme instances, oral SCC cells with a prickle-cell-like appearance were compactly packed and connected with intercellu-lar bridges, with thick cytoskeletal filamentous bundles entering from the cytoplasm, as if painted with a brush (G). Those thick filaments were morecondensed in perinuclear space, which was consistent with the cytological features (C–F). [Color figure can be viewed at wileyonlinelibrary.com]





are well differentiated. Indeed, the 20 samples we investi-

gated in this study were obtained from the tongue, gingi-

va, buccal mucosa, and lip, while none came from the

oropharynx (Table I).

Previous diagnostic cytology methods have not focused

on keratin pearls even in examinations of the oral cavi-

ty.29 Among the 5 cytological features proposed for FPS,

the characteristic concentric arrangement of orangeophilic

cells (parameter i) indicating keratin pearls was observed

in 50% of the SCC cases, while that was not found in

any CIS or reactive lesion cases. This is because cancer

pearls are located in the deeper portion of CIS

lesions,13,21,30 and may not be easily included in cytology

specimens. Recently, Watanabe et al. also reported that

Table II. Comparative Evaluation of 16 Malignant and 4 Reactive Lesions of the Oral Mucosa between TBS and FPS

Lesions System

Diagnosticcategories (TBS)/criteria used fordiagnoses (FPS)

Diagnosed case numbers by reviewers

Expert 1(first author)

Expert2

Nonexpert1

Nonexpert2

Total(%)

Squamous cellcarcinoma (SCC):12 cases

Bethesda (TBS) NILM 0 2 4 7 13 (27)ASC-US 0 2 8 5 15 (31)

LSIL 7 4 0 0 11 (23)ASC-H 5 4 0 0 9 (19)HSIL 0 0 0 0 0 (0)SCC 0 0 0 0 0 (0)

Five Parameter (FPS) i 9 7 7 6 29 (60)ii 12 10 8 12 42 (88)iii 10 8 9 10 37 (77)iv 12 11 8 9 40 (83)v 12 12 11 11 46 (96)

Cases diagnosed as SCC 12 12 11 11 46 (96)Total number of five

parameters recognizedin each case

0 0 0 1 0 1 (2)1 0 1 0 0 1 (2)2 0 1 1 1 3 (6)3 2 2 3 4 11 (23)4 1 8 3 7 19 (40)5 9 0 4 0 13 (27)

Carcinoma in-situ(CIS): 4 cases




Cases diagnosed as HSIL 4 4 3 3 14 (88)Total number of five


0 0 0 0 0 0 (0)1 0 0 1 1 2 (13)2 0 1 2 2 5 (31)3 2 0 0 0 2 (13)4 2 3 1 1 7 (44)5 0 0 0 0 0 (0)

Reactive lesions:4 cases




Cases diagnosed as NILM 4 4 4 4 16 (100)Total number of five


0 0 0 0 0 0 (0)1 0 0 0 0 0 (0)2 0 0 0 0 0 (0)3 0 0 0 0 0 (0)4 0 0 0 0 0 (0)5 0 0 0 0 0 (0)


HARA ET AL.


they failed to find keratin pearls in oral CIS specimens.31

In endocervical (EC) smears, epithelial cells without

nuclear atypia contained in keratin pearls have been sim-

ply reported as “parakeratotic cells,” the same as those in

non-neoplastic lesions, without being distinguished from

each other.32 This fact clearly indicates that differential

diagnosis between malignant and benign EC cases is

dependent on nuclear atypia.32 However, such a concen-

tric arrangement of orangeophilic cells does exist in EC

smears, as shown in Fig. 1.39 in the chapter by Abdul-

Karim et al.18 and in Fig. 2.15a (squamous pearl, example

of typical parakeratosis) in the Bethesda 3rd edition,33

though that has been routinely ignored or judged as ASC-

US.18,33 As revealed in this study, detection of keratin

pearls is a reliable method for definitive diagnosis of oral

SCC.

The second feature of FPS is a predominant popula-

tion of keratinized cells (parameter ii). Nuclear atypia

should be minimized in oral SCC cells during the kerati-

nization process, with keratinized cells without nuclear

atypia eventually becoming predominant. Such a tenden-

cy towards keratinization is a very strong characteristic

of oral SCC.13,20 Bizarre-shaped orangeophilic cells

without nuclear atypia (parameter iii) were also found to

be characteristic of oral SCC and CIS, while they were

never found in reactive lesion samples. In EC smears,

these orangeophilic cells may only be recognized as ker-

atotic changes (parakeratosis or atypical parakeratosis).

According to the 3rd edition of TBS 2015, they must be

diagnosed in a range that includes NILM, ASC-US, and

LSIL, but not as HSIL.33 However, when nuclear atypia

is evident, they are regarded as keratinizing SCC.18

Meanwhile, in oral cytology findings, such bizarre-

shaped orangeophilic cells without nuclear atypia have

not been documented. We consider that these cell types

can be a very specific criterion of oral malignancy and

likely originate from stacks of keratinized SCC foci on

the surface.

Keratoglobules (parameter iv), isolated and small

round-shaped keratinized cells containing pyknotic

nuclei, were observed in all of the present SCC and CIS

cases, but never in reactive lesion samples. In addition

to pyknosis, nuclear shapes were interpreted as karyoly-

sis or karyorrhexis. Any type of nuclear morphology

may indicate the process of cell death in association

with keratinization,34 though it remains unknown how

nuclei are cleared in those cells. Nevertheless, these

nuclear shape varieties are definitely different from apo-

ptosis or any other already-known cell death machinery.

Although there are no known reports of keratoglobules

or their analogues in previous studies, the cytoplasm of

keratoglobules may resemble that of keratinized cells

forming keratin pearls. Previously, round-shaped dysker-

atotic foci were shown to be generated in the basal layer

due to hemorrhage resulting from collapsed intraepithe-

lial blood vessels in oral CIS and SCC cases,30 because

hemophagocytosis induces keratinization of epithelial

cells facing connective tissues.21 In gynecological cytol-

ogy, keratinizing basaloid cells are often seen in non-

neoplastic lesions that originate from senile atrophy,18

and miniature squamous cells with small bland and

pyknotic nuclei are considered to represent typical para-

keratosis (Fig. 2.15b in the Bethesda 3rd edition).33

Uneven filamentous cytoplasm (parameter v) was iden-

tified in 100% of the 16 malignant cases, whereas that

was not seen in any of the reactive lesions. This cytologi-

cal feature may represent dyskeratotic changes among

IMT cells in the lower prickle cell layers or singular cel-

lular keratinization among those in the basal half of CIS/

SCC foci,13 in which keratin filaments (tonofilaments) are

not evenly distributed. Such dyskeratotic cells were found

to contain definite and large nuclei with clear nucleo-

plasm and intercellular bridges in the present study,

which is in sharp contrast to keratoglobules with charac-

teristic nuclear collapse shapes of pyknosis, karyolysis, or

karyorrhexis.34 Both keratoglobules and uneven filamen-

tous cytoplasm were simultaneously identified in 100% of

the 16 SCC and CIS cases, while keratin pearls were

found in 60 and 0%, respectively. Uneven filamentous

cytoplasm is considered to be an initial sign of abnormal

keratinization among activated keratinocytes, which

may secondarily accelerate keratinization with nuclear

collapse.

In conclusion, FPS is a powerful tool for cytological

diagnosis of well-differentiated oral SCC. When some of

the five parameters are detected, even though all may not

be simultaneously identified, it is possible to make a diag-

nosis of malignancy. We consider that FPS is much more

effective than TBS for diagnosis of oral SCC and CIS

specimens lacking nuclear atypia. However, the present

series of specimens submitted to reexamination included

only 20 cases, thus it will be necessary for us to perform

further studies to test the diagnostic value of FPS. In

addition, we also anticipate that FPS will be widely

investigated by other groups to confirm its usefulness for

oral SCC cytology examinations.

References

1. de Camargo Cancela M, Voti L, Guerra-Yi M, Chapuis F, MazuirM, Curado MP. Oral cavity cancer in developed and in developingcountries: Population-based incidence. Head Neck 2010;32:357–367.

2. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics,2002. CA Cancer J Clin 2005;55:74–108.

3. Sherin N, Simi T, Shameena PM, Sudha S. Changing trends in oralcancer. Indian J Cancer 2008;45:93–96.

4. Bhurgri Y, Bhurgri A, Hassan SH, et al. Cancer incidence in Kara-chi, Pakistan: First results from Karachi Cancer Registry. Int J Can-cer 2000;85:325–329.




5. Oo HN, Myint YY, Maung CN, et al. Oral cancer in Myanmar: apreliminary survey based on hospital-based cancer registries. J OralPathol Med 2011;40:20–26.

6. Sawair FA, Al-Mutwakel A, Al-Eryani K, et al. High relative fre-quency of oral squamous cell carcinoma in Yemen: Qat and tobac-co chewing as its aetiological background. Int J Environ Health Res2007;17:185–195.

7. Petersen PE, Bourgeois D, Bratthall D, Ogawa H. Oral health infor-mation systems – towards measuring progress in oral health promo-tion and disease prevention. Bull World Health Organ 2005;83:686–693.

8. Howlader N, Noone AM, Krapcho M, editors. SEER Cancer Statis-tics Review, 1975-2012, National Cancer Institute, Cancer Statics,CSR Sections, 20. Oral Cavity and Pharynx, Website title: AnnualIncidence Rates, Annual Death Rates. Available at: http://seer.can-cer.gov/csr/1975_2012/results_merged/sect_20_oral_cavity_pharynx.pdf. Accessed on 5/21/2016.

9. Taiwan Cancer Registry. Cancer Incidence Rate in Taiwan, 1998-2002 & 2003-2007. Available at: http://tcr.cph.ntu.edu.tw/main.php?Page5N2. Accessed on 5/21/2016.

10. Portal Site of Official Statistics of Japan, Statistics Bureau, TheMinistry of Internal Affairs and Communications. Trends in age-adjusted death rates (per 100,000 population) from malignant neo-plasms by sex and site: Japan (1950-2014) (Table 5-26). Availableat: http://www.e-stat.go.jp/SG1/estat/eStatTopPortalE.do. Accessedon 5/21/2016.

11. Cancer Registry and Statistics. Cancer Information Service, Nation-al Cancer Center, Japan. Cancer mortality (1958-2014). Availableat: http://ganjoho.jp/reg_stat/statistics/dl/index.html. Accessed on 5/21/2016.

12. Lingen MW, Kalmar JR, Karrison T, Speight PM. Critical evalua-tion of diagnostic aids for the detection of oral cancer. Oral Oncol2008;44:10–22.

13. Kobayashi T, Maruyama S, Cheng J, et al. Histopathological varie-ties of oral carcinoma in situ: Diagnosis aided by immunohisto-chemistry dealing with the second basal cell layer as theproliferating center of oral mucosal epithelia. Pathol Int 2010;60:156–166.

14. Pindborg JJ, Reichart PA, Smith CJ, van der Waal I. Histologicaltyping of cancer and precancer of the oral mucosa. WHO Classifi-cation of Tumours. 2nd ed. Berlin: Springer; 1997. p 24–26.

15. Kaugars GE, Silverman S, Ray AK, et al. The use of exfoliativecytology for the early diagnosis of oral cancers: is there a role forit in education and private practice? J Cancer Educ 1998;13:85–89.

16. Potter TJ, Summerlin DJ, Campbell JH. Oral malignancies associat-ed with negative transepithelial brush biopsy. J Oral MaxillofacSurg 2003;61:674–677.

17. Shklar G, Meyer I, Cataldo E, Taylor R. Correlated study of oralcytology and histopathology. Report on 2,052 oral lesions. OralSurg Oral Med Oral Pathol 1960;13:897–907.

18. Abdul-Karim FW, Brainard J, Michael CW. Cytopathology. In:Chang L, Bostwick DG, editors. Essentials of Anatomic Pathology.3rd ed. New York: Springer; 2011. p 17–23.

19. Mikami T, Cheng J, Maruyama S, et al. Emergence of keratin 17vs. loss of keratin13: their reciprocal immunohistochemical profilesin oral carcinoma in situ. Oral Oncol 2011;47:497–503.

20. Kobayashi T, Maruyama S, Ab�e T, et al. Keratin 10-positive ortho-keratotic dysplasia: A new leukoplakia-type precancerous entity ofthe oral mucosa. Histopathology 2012;61:910–920.

21. Al-Eryani K, Cheng J, Ab�e T, et al. Hemophagocytosis-mediatedkeratinization in oral carcinoma in-situ and squamous cell carcino-ma: A possible histopathogenesis of keratin pearls. J Cell Physiol2013;228:1977–1988.

22. Ida-Yonemochi H, Maruyama S, Kobayashi T, Yamazaki M, ChengJ, Saku T. Loss of keratin 13 in oral carcinoma in-situ: A compara-tive study of protein and gene expression levels using paraffin sec-tions. Mod Pathol 2012;25:784–794.

23. Funayama A, Cheng J, Maruyama S, et al. Enhanced expression ofpodoplanin in oral carcinomas in situ and squamous cell carcino-mas. Pathobiology 2011;78:171–180.

24. Tsuneki M, Yamazaki M, Maruyama S, Cheng J, Saku T. Podopla-nin-mediated cell adhesion through extracellular matrix in oralsquamous cell carcinoma. Lab Invest 2013;93:921–932.

25. Solomon D, Nayar R, editors. The Bethesda System for ReportingCervical Cytology: Definitions, Criteria, and Explanatory Notes.2nd ed. New York: Springer, 2004. p 21–56, 67-122.

26. Koch FP, Kunkel M, Biesterfeld S, Wagner W. Diagnostic efficien-cy of differentiating small cancerous and precancerous lesions usingmucosal brush smears of the oral cavity—a prospective and blindedstudy. Clin Oral Invest 2011;15:763–769.

27. Safaeian M, Solomon D, Wacholder S, et al. Risk of precancer andfollow-up management strategies for women with human papillo-mavirus–negative atypical squamous cells of undetermined signifi-cance. Obstet Gynecol 2007;109:1325–1331.

28. Marur S, D’Souza G, Westra WH, Forastiere AA. HPV-associatedhead and neck cancer: A virus-related cancer epidemic. LancetOncol 2010;11:781–789.

29. Divani S, Exarhou M, Theodorou LN, Georgantzis D, Skoulakis H.Advantages and difficulties of brush cytology in the identificationof early oral cancer. Arch Oncol 2009;17:11–12.

30. Funayama A, Maruyama S, Yamazaki M, et al. Intraepitheliallyentrapped blood vessels in oral carcinoma in-situ. Virchows Arch2012;460:473–480.

31. Watanabe N, Ohkubo T, Shimizu M, Tanaka T. Preneoplasiaand carcinogenesis of the oral cavity. Oncol Discov 2015;3:1.Available at: http://www.hoajonline.com/journals/pdf/2052-6199-3-1.pdf. Accessed on 9/9/2016.

32. Nauth HF. Benign changes of the female genital tract. Gynecologi-cal Cytology. New York: Thieme Press; 2007. p 127–146.

33. Nayar R, Wilbur DC, editors. The Bethesda System for ReportingCervical Cytology. Definitions, Criteria, and Explanatory Notes. 3rded. New York: Springer; 2015. p 29–89, 103-192.

34. Koss LG. Light microscopic estimation of cellular activity, injury,and death. In: Koss LG. editor. Diagnostic Cytology and its histo-pathologic basis. 3rd ed. vol.1 Philadelphia. Toronto: JB LippincottCompany; 1979. p 53–60.


HARA ET AL.


http://seer.cancer.gov/csr/1975_2012/results_merged/sect_20_oral_cavity_pharynx.pdf



http://tcr.cph.ntu.edu.tw/main.php?Page=N2



http://www.e-stat.go.jp/SG1/estat/eStatTopPortalE.do

http://ganjoho.jp/reg_stat/statistics/dl/index.html

http://www.hoajonline.com/journals/pdf/2052-6199-3-1.pdf

http://www.hoajonline.com/journals/pdf/2052-6199-3-1.pdf

Documents

Differential Diagnosis of - 新潟大学歯学部 · Differential Diagnosis of ... (indicating keratin pearls), (ii) large number ... available, for a total of 20 slides from the