Molecular analysis of pancreatic cyst fluid: a comparative analysis with current practice of...

Molecular Analysis of PancreaticCyst Fluid

A Comparative Analysis With Current Practice of Diagnosis

Jian Shen, MD, PhD1; William R. Brugge, MD2; Christopher J. DiMaio, MD2; and Martha B. Pitman, MD1

BACKGROUND: The management of patients with pancreatic cysts is based on the preoperative distinction

of nonmucinous and mucinous cysts in general and of benign and malignant cysts in particular. An accurate

diagnosis is challenging, because endoscopic ultrasound (EUS) and cyst fluid analysis for carcinoembryonic

antigen (CEA) and cytology have low sensitivity and specificity. Currently, molecular analysis is a commer-

cially available test that promises an accurate diagnosis. The objective of the current study was to correlate

a commercially provided molecular diagnosis (MDx) with a clinical consensus diagnosis (CCD) in the gen-

eral categories of malignant, benign mucinous, and benign nonmucinous pancreatic cysts. METHODS:

Pancreatic cysts that had aspirated fluid submitted for a commercially available molecular test (PathFin-

derTG) were reviewed. The CCD, defined by histology, malignant cytology, or 2 concordant tests (such as

EUS, cytology, or CEA �192 ng/mL for mucinous cysts), was categorized as malignant, benign mucinous, or

benign nonmucinous cyst in 35 patients. Their MDx, based on the PathFinderTG report, including analysis

of k-ras mutation, loss of heterozygosity, and quantity/quality of DNA, also was classified as malignant,

benign mucinous, or benign nonmucinous cyst. These 2 diagnoses were compared and correlated.

RESULTS: The concordance between CCD and MDx was 5 of 6 (83%), 13 of 15 (87%), and 13 of 14 (93%),

respectively, for malignant, benign mucinous, and benign nonmucinous cysts, with an overall Cohen kappa

statistic of 0.816. The sensitivity, specificity, and positive predictive value of the MDx were 83%, 100%, and

100%, respectively, for a malignant cyst and 86%, 93%, and 95%, respectively for a benign mucinous cyst.

CONCLUSIONS: Molecular analysis of pancreatic cyst fluid adds diagnostic value to the preoperative diag-

nosis with high sensitivity, specificity, and positive predictive value for the diagnosis of malignant and

benign mucinous pancreatic cysts. Cancer (Cancer Cytopathol) 2009;117:217–27. VC 2009 American Cancer

Society.

KEY WORDS: pancreatic neoplastic mucinous cysts, mucinous cystic neoplasm, intraductal papillary

mucinous neoplasm, k-ras gene, loss of heterozygosity, molecular diagnosis, clinical consensus diagnosis.

Received: November 7, 2008; Revised: January 26, 2009; Accepted: January 27, 2009

Published online: May 4, 2009 VC 2009 American Cancer Society

DOI: 10.1002/cncy.20027, www.interscience.wiley.com

Corresponding author: Martha B. Pitman, MD, Department of Pathology, Massachusetts General Hospital, 55 Fruit Street, Boston, MA 02114;

Fax: (617) 7260-7474; [email protected]

1James Homer Wright Pathology Laboratories, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts; 2Department of

Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts

We thank Dr. Jingyun Yang from the Biostatistics Center at Massachusetts General Hospital, Boston, Massachusetts for help with the calculation of

kappa statistics.

Cancer Cytopathology June 25, 2009 217

Original Article

The clinical management of patients with neoplasticcysts of the pancreas is evolving as we learn more aboutthe biologic behavior of these cysts. Management is basedon the preoperative distinction of nonmucinous andmucinous cysts in general and of benign and malignantcysts in particular. Treatment is increasingly nonsurgical,even when neoplasia is confirmed.1-6

Accurate preoperative diagnosis of these cysts is

therefore critical to proper patient management. Serous

cysts are benign and, unless they are symptomatic or large

(>4 cm in greatest dimension), do not necessarily need

resection.1 Algorithms have been developed for the man-

agement of mucinous cysts (Fig. 1).7 The basic premise of

the algorithm is that small cysts (<3 cm) in asymptomatic

patients can be followed safely without a specific diagnosis

to confirm the clinical suspicion of a mucinous cyst. The

proposed management algorithm recommends that

patients with cysts<1 cm should be re-evaluated at 1 year

with magnetic resonance imaging or thin-slice computed

tomography studies and should undergo resection if the

cyst increases in size. Diagnostic intervention is not rec-

ommended for cysts measuring <1 cm. Resection is rec-

ommended for any patient who has symptoms

attributable to the cyst, a cyst >3 cm, or a cyst measuring

between 1 cm and 3 cm in a patient with high-risk stig-

mata. High-risk stigmata include the presence of a dilated

main pancreatic duct (>6 mm), the presence of a mural

nodule in the cyst wall on radiologic studies, or ‘‘positive

cytology.’’ If none of the listed high-risk stigmata are pres-

ent in an asymptomatic patient who has a cyst between

1 cm and 3 cm, then repeat imaging studies are recom-

mended every 3 months to 12 months, depending on the

size of the cyst at presentation.

We demonstrated previously that not all small,

asymptomatic, branch duct intraductal papillary muci-

nous neoplasms (IPMNs) are benign8 and that ‘‘positive’’

cytology may not be identified in malignant IPMNs.8,9 In

addition, solitary branch duct-appearing cysts may

actually be secondarily cystic solid neoplasms that require

resection.10 The overall sensitivity and specificity of the

cytologic diagnosis of pancreatic cysts is rather low and

highly variable, depending on technical and interpretive

expertise.8,9,11-16 Preoperative diagnosis currently

depends on a multidisciplinary approach that incorpo-

rates radiologic analysis, cyst fluid analysis, and/or cytol-

ogy.17 Analysis of cyst fluid for carcinoembryonic antigen

(CEA) is considered the most accurate predictor of a

mucinous cyst.18 Using a cutoff CEA value of 192 ng/mL,

we have demonstrated an accuracy of 79.2% for mucinous

etiology.18 Although radiologic features can be rather spe-

cific for some pancreatic cysts, many incidental pancreatic

cysts are small and have nonspecific radiologic fea-

tures.6,19 Preoperative diagnosis is especially a challenge

in small cysts that produce scant fluid with low to no cel-

lular yield and insufficient quantity for CEA analysis.

The overall low specificity of the current diagnostic

methods have led to the development of a commercially

available molecular test (PathFinderTG; RedPath Inte-

grated Pathology, Pittsburgh, Pa) as an additional ancil-

lary test in this multimodal approach to diagnosis.

Previous studies have reported that an initial k-ras gene

mutation followed by allelic loss was predictive of malig-

nancy.20-22 Although molecular analysis of cyst fluid has

produced promising results, it has not been widely stud-

ied. The objective of the current study was not to analyze

or validate the scientific methods of the molecular analysis

but to correlate the molecular diagnosis (MDx) derived

from this commercially available test with a clinical con-

sensus diagnosis (CCD) in the general categories of malig-

nant, benign mucinous, and benign nonmucinous cysts.

FIGURE 1. International Consensus Algorithm for the manage-

ment of mucinous cysts. EUS indicates endoscopic ultrasound;

MRCP, magnetic resonance cholangiopancreatography;

ERCP; endoscopic retrograde cholangiopancreatography;

MR, magnetic resonance; CT, computed tomography. Repro-

duced with permission from Tanaka M, Chari S, Adsay V et al.

International consensus guidelines for management of intra-

ductal papillary mucinous neoplasms and mucinous cystic

neoplasms of the pancreas. Pancreatology. 2006;6:17-32.

Original Article

218 Cancer Cytopathology June 25, 2009

MATERIALS AND METHODS

Patient Cohort

Forty-five consecutive patients with pancreatic cysts who

underwent endoscopic ultrasound (EUS)-guided fine-

needle aspiration (FNA) biopsy between 2005 and 2007

at Massachusetts General Hospital (Boston, Mass) and

had aspirated fluid submitted for molecular analysis

(PathFinderTG) were evaluated. Patients’ electronic med-

ical records were reviewed for all follow-up information

and test values. Institutional Review Board approval was

obtained for this study.

Pancreatic cysts were identified and evaluated for

location, number, size, and characteristic features (wall

thickness, septations, associated mass lesions, calcifica-

tions, pancreatic duct communication, and/or dilation).

FNA was performed with a 22-gauge needle (Echotip;

Cook Medical, Winston-Salem, NC). The cysts were

evacuated until collapse was achieved. Fluid color, clarity,

and viscosity were recorded; and the samples were sent for

cytologic examination, CEA concentration, and molecu-

lar analysis (PathFinderTG). Of these 45 patients, 8 had

histologic follow-up, 29 had concurrent cytologic inter-

pretation, and 30 had CEA cyst fluid chemical analysis.

Clinical Consensus Diagnosis

All EUS findings and images were reviewed blindly by 2

separate endosonographers (W.R.B. and C.J.D.). The

cysts were categorized as malignant, mucinous (IPMN or

mucinous cystic neoplasm [MCN]) with or without ma-

lignant features, serous cystadenoma, pseudocyst, or

other, and the general features used for this classification

are listed in Table 1. Overall, the diagnosis was assigned

based primarily on cyst morphology and/or cyst fluid

characteristics; and cyst location, patient age, and patient

sex were used to differentiate competing diagnoses in

unclear cases (Table 1). Suspicion of malignancy was

determined by the presence of high-risk features, as listed

in Table 1. Large cyst size in the absence of any other

high-risk feature was not indicative of malignancy.

A CCD was established for each patient, and the

cyst was classified as malignant, benign mucinous, or

benign nonmucinous using either the histologic diagnosis

or a combination of 2 of 3 concordant characteristics:

1) EUS features, 2) cyst fluid CEA level, or 3) cytology

(Table 2). The following classification criteria were used.

1) Malignant cyst (invasive or in situ carcinoma):

required either histologic confirmation or positive

(malignant) cytology;

2) Benign mucinous cyst (nonmalignant): required histo-

logic confirmation of IPMN or MCN with low-grade

dysplasia (adenoma) or moderate dysplasia (borderline

malignancy) or 2 of the following: a) cyst fluid CEA

level �192 ng/mL; or b) cytology with thick, colloid-

like extracellular mucin with or without atypical epi-

thelial cells; or c) EUS features consistent with an

IPMN or MCN and the macroscopic appearance of

‘‘thick, viscous, mucoid cyst fluid’’; and

Table 1. Endoscopic Criteria for Pancreatic Cystic Lesions

CystType

Age,Decade

Sex Location Morphology FluidCharacteristics

MalignantFeatures

IPMN 7th Men ¼ women Head (or

multifocal,

diffuse)

Dilated pancreatic duct

(main duct or side

branch)

Viscous Dilated main pancreatic duct,

cyst size >3 cm, intramural

mass/nodules

MCN 5th-7th Women > men Body/tail Unilocular or multilocular,

thick wall, peripheral

calcifications

Viscous Cyst size >3 cm, thick wall,

intramural mass/nodules,

peripheral calcifications

Pseudocyst Any Men ¼ women Evenly

distributed

Unilocular, thick wall Low viscosity,

dark, opaque

SCA 7th Women > men Evenly

distributed

Microcystic/honeycomb

(less commonly

oligocystic, macrocystic);

central calcification

Low viscosity, clear

IPMN indicates intraductal papillary mucinous neoplasm; MCN, mucinous cystic neoplasm; SCA, serous cystadenoma.

Molecular Analysis of Pancreatic Cysts/Shen et al


3) Benign nonmucinous cyst: required histologic confir-

mation or 2 of the following: a) cyst fluid CEA <5

ng/mL; or b) nonmucinous cytology (no extracellu-

lar mucin or mucinous or atypical epithelium); or c)

EUS features typical of a serous cyst and/or thin,

watery, and nonmucoid macroscopic cyst fluid.

A CCD was reached in 35 of 45 patients according

to the above criteria, and those 35 patients established the

study cohort. Ten cysts with molecular analysis did not

fulfill the above criteria, because they lacked either cyst

fluid analysis or cytology for a concordant clinical cyst

classification and were excluded from the final analysis.

Molecular Diagnosis

The MDx was taken directly from the PathFinderTG

report. It was not the purpose of this study to evaluate the

scientific methods or validity of this commercially avail-

able test. The molecular analyses developed by RedPath

Integrated Pathology include 3 tests: k-ras gene point

mutation, loss of heterozygosity (LOH) analysis using a

preselected panel of 15 genomic loci associated with tu-

mor suppressor genes, and determination of DNA quan-

tity/quality in cyst fluid.20-22 Each of the 3 tests is defined

as ‘‘abnormal’’ when the following are identified: 1) k-ras

gene point mutation, 2) LOH mutations present in �2

genomic loci, and, 3) a high quantity/quality of DNA

content. The cyst is consistent with a benign nonmuci-

nous cyst when none of the above-defined abnormalities

are present and with a benign mucinous cyst when 1 of 3

abnormal findings described above is identified. In addi-

tion to the criteria for a mucinous MDx, if the k-ras or

LOH mutation is present at high amplitude (>75% of

total DNA), suggesting a significant clonal expansion,

then a malignant diagnosis is rendered (Table 3).

Statistical Analysis

The data were analyzed using Fisher exact tests, and the

differences were considered statistically significant when

P < .05. A kappa statistic was calculated using MatLab

(release 2008a; The MathWorks, Natick, Mass) for the

concordance between CCD andMDx.

RESULTS

The 35 patients included 27 women and 8 men who

ranged in age from 31 years to 85 years (mean age,

65 years). On the basis of our criteria for CCD, there were

6 malignant cysts, 15 benign mucinous cysts, and 14

benign nonmucinous cysts (Table 4). Malignant cysts

were located in the head (5 of 6 cysts) and body (1 of 6

cysts) and ranged in size from 1.6 cm to 5 cm (mean size,

3.1 cm). Benign mucinous cysts were located in the head

or uncinate process (8 of 15 cysts), body (5 of 16 cysts),

and tail (2 of 15 cysts) and ranged in size from 0.6 cm to

4.5 cm (mean size, 2.2 cm). Benign nonmucinous cysts

were located in the head or uncinate process (5 of 14 cysts),

body (7 of 14 cysts), and tail (2 of 14 cysts) and ranged in

size from 0.3 to 7.6 cm (mean size, 3.7 cm).

Comparison of Molecular Analysis

(PathFinderTG) Among Cyst Types

The cyst fluid contained high DNA quantity and quality

in 5 of 6 (83%) malignant cysts, which was significantly

different (P < .01) from benign mucinous cysts (2 of 15

cysts; 13%) and benign nonmucinous cysts (0 of 14 cysts;

Table 2. Criteria for Clinical Consensus Diagnosis

Clinical Consensus Diagnosis Criteria

Malignant (in situ or invasive carcinoma) Histology: invasive or in situ adenocarcinoma arising from IPMN or MCN OR cytology positive for

adenocarcinoma (in situ or invasive carcinoma) with abundant extracellular mucin and radiologic

evidence of cystic lesion

Benign mucinous Histology: IPMN or MCN with low-to-moderate grade dysplasia OR 2 of the following: 1) cyst fluid

CEA �192 ng/mL, 2) EUS consistent with IPMN or MCN, 3) cytology consistent with mucinous cyst

Benign nonmucinous Histology: nonmucinous cyst OR 2 of the following: 1) cyst fluid CEA <5ng/mL, 2) EUS consistent with

nonmucinous cyst, 3) cytology consistent with nonmucinous cyst

IPMN indicates intraductal papillary mucinous neoplasm; MCN, mucinous cystic neoplasm; CEA, carcinoembryonic antigen; EUS, endoscopic ultrasound.

Original Article


0%) (Table 5). There was no difference in DNA quantity/

quality between benign mucinous and nonmucinous

cysts. k-ras gene mutation was present in 5 of 6 (83%) ma-

lignant cysts compared with 7 of 15 (47%) benign muci-

nous cysts and 0 of 14 (0%) benign nonmucinous cysts

(Table 5), a finding that was identified more frequently in

both malignant cysts (P < .01) and mucinous cysts (P <

.05) compared with nonmucinous cysts, but there was no

significant difference between malignant and benign

mucinous cysts. LOHmutations in�2 genomic loci were

identified in 5 of 6 (83%) malignant cysts, 4 of 15 (27%)

benign mucinous cysts, and 1 of 14 (7%) benign nonmu-

cinous cysts (Table 5), an abnormality that occurred more

frequently in malignant cysts than in benign mucinous

and benign nonmucinous cysts (P < .05), but no signifi-

cant difference was observed between benign mucinous

and benign nonmucinous cysts.

Correlation Between Clinical Consensus

Diagnosis and Molecular Diagnosis

(PathFinderTG)

Malignant cysts

Of 6 patients who had a malignant CCD diagnosis,

5 had positive cytology, 3 had histologic confirmation, 3

demonstrated clinically or pathologically proven meta-

static disease, and 2 died of disease within 2 years from the

initial diagnosis (Table 4). Within this group, 83% had

molecular changes consistent with malignancy, specifi-

cally, k-ras gene point mutation, a high quantity/quality

of DNA content in fluid samples, and LOH mutations in

�2 genomic loci (Table 5).

The concordance rate between the CCD and the

MDx was 83% in this group, and the overall Cohen kappa

statistic was 0.816 (95% confidence interval [95% CI],

0.6464-0.9857), indicating perfect agreement (Table 6).

The 1 false-negative MDx (Patient 6, who had an IPMN

with carcinoma in situ) (Table 4) was a 1.6-cm cyst

located in the head of the pancreas with indeterminate cy-

tology and EUS features of a low-grade IPMN. This

patient eventually underwent a Whipple resection after 1

year because of a strong family history of pancreatic cancer

and detection of BRCA2mutation.

Benign mucinous cysts

Fifteen patients had their cysts classified as benign

mucinous based on the CCD: Three patients had histo-

logic follow-up (an MCN with low-grade dysplasia and 2

IPMNs with moderate dysplasia), 4 patients had muci-

nous cysts according to cytology, and CEA levels were

available in 14 of 15 patients and were elevated (�192 ng/

mL) in 13 of 14 patients. The concordance rate between

the CCD and the MDx was 87%, and the overall Cohen

kappa statistic was 0.816 (95% CI, 0.6464-0.9857) (Ta-

ble 6). In this group, 47% (7 of 15 patients) had k-ras

gene point mutations, 27% (4 of 15 patients) had LOH

mutations present in �2 genomic loci, and 13% (2 of 15

patients) had high DNA content (Table 5). The 2 non-

concordant cases included the following.

1. Patient 21 (Table 4) had a 4.2-cm, thinly septated

cyst located in the tail of pancreas with 3 associated

mural nodules that ranged in size from 0.3 cm to

0.8 cm on EUS. Cytology demonstrated extracellular

mucin that was positive on alcian blue stain and

contained atypical epithelial cells consistent with at

least moderate dysplasia (Fig. 2). The cyst fluid was

macroscopically viscous, but the CEA level was only

71.6 ng/mL. The overall clinical impression was side

branch IPMN or MCN. Molecular analysis (Path-

FinderTG) revealed low quantity/quality DNA with

no k-ras point mutation or LOH identified, consist-

ent with benign nonmucinous cyst. The patient

remained alive and well at 18 months.

Table 3. Molecular Criteria (PathFinderTG) Developed by RedPath Integrated Pathology

Molecular Diagnosis Criteria

Benign nonmucinous 1) DNA quantity/quality low to moderate; AND 2) k-ras gene point mutation not present; AND 3) LOH,

<2 genomic loci present

Benign mucinous 1) DNA quantity/quality: high; OR 2) k-ras gene point mutation present; OR 3) LOH, �2 genomic loci present

Malignant (in situ or invasive carcinoma) 1) k-ras gene point mutation, high amplitude (>75%); OR 2) �2 more LOH, high amplitude (>75%)

LOH indicates loss of heterozygosity.



Table

4.PatientList

CCD

Patient

No.

Age,

ySex

Cyst(s)

Size,cm

Cyst(s)

Location

EUS

CEA,

ng/m

LCytology

Histology/

Clinical

follow-up

MDx

Malig

nant

171

Woman

5Head

ACA

NA

MA

NR

MA

260

Woman

2Body

MCN-m

alig

nant

4151.5

MA

IPMN-A

CA

MA

365

Woman

2Head

IPMN-m

alig

nant

NA

MA

IPMN-A

CA

MA

472

Woman

4Head

IPMN-m

alig

nant

320,099

MA

DOD

MA

585

Woman

4Head

MCN-m

alig

nant

NA

MA

DOD

MA

665

Woman

1.6

Head

IPMN-low

grade

NA

IND

IPMN-C

ISNM

Benign

mucinous

755

Man

1.8

Head

IPMN-low

grade

378.4

NA

AWD

at12mo

MU

831

Woman

4.5

Tail

MCN-m

oderate

927.4

IND

MCN-low

grade

MU

976

Man

3.7

Uncinate

IPMN-m

oderate

4602

IND

NA

MU

10

63

Man

2Body

IPMN-low

grade

213.7

NA

IPMN-m

oderate

MU

11

66

Woman

0.6

Head

IPMN-low

grade

NA

MU

NA

MU

12

69

Woman

1.8

Body

MCN-low

grade

6860

MU

NA

MU

13

61F

Woman

1.2

Head

IPMN-low

grade

�37,900

IND

AWD

at60mo

NM

14

81

Woman

2Head

IPMN-low

grade

200.3

IND

AWD

at15mo

MU

15

73

Woman

1.7

Body(m

ultiple)

IPMN-low

grade

604

IND

NA

MU

16

68

Woman

2.4

Head

IPMN-low

grade

4514

MU

IPMN-low

grade

MU

17

70

Man

2.5

Head

IPMN-low

grade

6827

NA

AWD

at17mo

MU

18

85

Woman

1Head

IPMN-m

alig

nant

392-528

IND

AWD

at36mo

MU

19

81

Woman

1.7

Body

IPMN-low

grade

>50,000

NA

NA

MU

20

68

Woman

1Body

MCN-low

grade

22,720

NA

AWD

at36mo

MU

21

75

Man

4.2

Tail

MCN-m

alig

nant

71.6

MU

AWD

at18mo.

NM

Benign

nonmucinous

22

76

Woman

3Neck

SCA

3.3-4.6

NM

AWD

at31mo

NM

23

54

Woman

2.5

Tail

MCN-m

oderate

>5000

IND

Entericduplication

cyst

NM

24

80

Woman

3.5

Head

SCA

0.7

NM

NA

MU

25

44

Woman

5Head

Pseudocyst

176.5

NM

Pseudocyst

NM

26

49

Man

2.5

Body

SCA

0.2

IND

NA

NM

27

56

Woman

0.3

Head

SCA

0.2

NM

NA

NM

28

76

Woman

5Head

SCA

0.4

NA

AWD

at60mo

NM

29

70

Woman

2.5

Uncinate

SCA

0.2

IND

NA

NM

30

48

Woman

2-2.4

Body

SCA

0.6

IND

AWD

at14mo

NM

31

53

Woman

1.7

Body

SCA

1.4

IND

NA

NM

32

64

Man

5.7

Body

Pseudocyst

2.3

NM

NA

NM

33

50

Woman

7.6

Tail

Gastric

duplicationcyst

2.4

IND

NA

NM

34

36

Man

4.3

Body

Gastric

duplicationcyst

0.2

NM

NA

NM

35

71

Woman

3.7

Body

SCA

NA

NM

AWD

at13mo

NM

CCD

indicatesclinicalconsensusdiagnosis;EUS,endoscopic

ultrasound;CEA,carcinoembryonic

antigen;MDx,moleculardiagnosis;ACA,adenocarcinoma;NA,notavailable;MA,malig

nant;NR,notresect-

able;MCN,mucinouscysticneoplasm;IPMN,intraductalpapillary

mucinousneoplasm;DOD,diedofdisease;IND,indeterm

inate;CIS,carcinomain

situ;AWD,alivewithdisease;MU,benignmucinous;NM,

benignnonmucinous;SCA;seruscystadenoma.

Original Article


2. Patient 13 (Table 4) had a 2.5-cm cyst in the pan-

creatic head with indeterminate cytology but a CEA

value that ranged from 27.6 ng/mL, 962 ng/mL,

1184 ng/mL, and 37,900 ng/mL between 2001 and

2004. The patient underwent ethanol ablation,

which resulted in a reduction of the cyst size from

2.5 cm to 1.2 cm. The MDx obtained 2 years after

ethanol ablation was nonmucinous based on low

DNA quality and quantity and only 1 LOH muta-

tion. The patient remained alive and disease free >5

years.

Benign nonmucinous cysts

Fourteen patients had their cysts classified as benign

nonmucinous according to CCD criteria. Thirteen of 14

cyst fluid samples were tested for CEA, and all but 2 sam-

ples demonstrated levels <5 ng/mL. Only 1 sample dem-

onstrated a CEA level >192 ng/mL. Cytology was

deemed nonmucinous in 7 of 13 cysts and indeterminate

in 6 of 13 cysts. Histologic follow-up was available in 2

patients, including 1 patient who had an enteric duplica-

tion cyst and 1 patient who had a pseudocyst. In this

group, none of the cyst fluids had k-ras point mutations

or high DNA content, and 7% (1 of 14 cysts) had LOH

mutations detected in >2 genomic loci (Table 5). The

concordance rate between the CCD and the MDx (Path-

FinderTG) was 93%, and the overall Cohen kappa sta-

tistic was 0.816 (95% CI, 0.6464-0.9857). The 1

discordant case (Patient 24) (Table 4) had thin, clear cyst

fluid with a CEA level of 0.7 ng/mL, nonmucinous cytol-

ogy, and EUS features consistent with a serous cystade-

noma. Molecular analysis revealed 3 LOH at loci of 10q,

21q, and 17q, leading to a mucinous MDx. No follow-up

was available on this patient at the time of this writing.

Table 5. Molecular Data in Different Categories of Clinical Consensus Diagnosis

Clinical Consensus Diagnosis: No (%)

Molecular Data Status Malignant,n56

Benign Mucinous,n515

Benign Nonmucinous,n514

DNA quantity/quality High/good 5 (83) 2 (13) 0 (0)

k-ras point mutation Present 5 (83) 7 (47) 0 (0)

No. of LOH present �2 Genomic loci 5 (83) 4 (27) 1 (7)

LOH indicates loss of heterozygosity.

Table 6. Correlation Between Clinical Consensus Diagnosis and Molecular Diagnosis (PathFinderTG)

Clinical Consensus Diagnosis: No. (%)

BenignMolecular Diagnosis Malignant Mucinous Nonmucinous Sensitivity, % Specificity, % PPV, %

Malignant 5 (83)* 0 0 83 100 100

Benign mucinous 0 13 (87)* 1 86 93 95

Benign nonmucinous 1 2 13 (93%)* 93 86 81

PPV indicates positive predictive value.

* Agreement (kappa statistic, 0.816).

FIGURE 2. A single, atypical epithelial cell cluster with cyto-

plasmic mucin is noted in the mucinous cyst fluid of an intra-

ductal papillary mucinous neoplasm with moderate dysplasia

(Papanicolaou stain; high-power view).



DISCUSSION

We evaluated the diagnostic concordance of a commer-

cially available molecular test for the diagnosis of pancre-

atic cysts (PathFinderTG) with the current method of

diagnosing pancreatic cysts, either by histology after resec-

tion or preoperatively using a combination of cytology,

cyst fluid analysis, and radiologic features. Our data pro-

duced 83% concordance for malignant cysts, 87% con-

cordance for benign mucinous cysts, 93% concordance

for benign nonmucinous cysts, and an overall Cohen

kappa statistic of 0.816 (95% CI, 0.6464-0.9857), indi-

cating perfect agreement (Table 6). The sensitivity, speci-

ficity, and positive predictive value for a malignant MDx

with PathFinderTG were 83%, 100%, and 100%, res-

pectively; and the same rates for a mucinous MDx with

PathFinderTG were 86%, 93%, and 95%, respectively

(Table 6). There were no false-positive results for a

malignant MDx.

The accuracy and validity of the commercially avail-

able molecular test PathFinderTG are controversial,

because to our knowledge there have been few published

reports20-22 and no validation studies outside of studies

that were sponsored by RedPath Integrated Pathology. It

was not the purpose of this study to delve into the accu-

racy of the scientific or proprietary techniques offered in

their molecular analysis of pancreatic cyst fluid. The basic

principals outlined above highlight the MDx criteria. Ma-

lignant cysts require either k-ras gene point mutation,

LOH mutations present in �2 genomic loci, or a high

quantity/quality of DNA with k-ras or LOH mutations

present at high amplitude (>75% of total DNA), suggest-

ing significant clonal expansion. Benign mucinous cysts

require either k-ras gene point mutation, LOH present in

�2 genomic loci, or a high quantity/quality of DNA; and

benign nonmucinous cysts do not have any of these mo-

lecular changes and have little or poor-quality DNA. We

abstracted the MDx directly from the PathFinderTG

report on each patient in this study and categorized the

cyst as either malignant, benign mucinous, or benign

nonmucinous.

On the basis of our CCD criteria, k-ras gene muta-

tion appears to be the best discriminator between muci-

nous cysts and nonmucinous cysts (P < .05). High DNA

quantity/quality and accumulation of LOH occurred more

frequently in malignant cysts but did not differ signifi-

cantly between benign mucinous cysts and nonmucinous

cysts (P > .05) (Table 5). These observations suggest that

k-ras gene mutation is a rather early event for a mucinous

cyst, whereas the accumulation of LOH and increased

DNA quantity/quality occur more commonly later during

malignant transformation. This is consistent with the pre-

vious findings, which suggested that an initial hit of k-ras

mutation followed by allelic loss is most predictive of the

presence of malignancy in pancreatic cyst fluid.20

In the current study, we compared the MDx (Path-

FinderTG) with our CCD, a diagnosis that was deter-

mined from either the histology from resection or from 2

of 3 concordant preoperative tests (EUS features, CEA

cyst fluid analysis, and cytology) that supported the same

clinical cyst classification. The weakness of the study rests

here, but the weakness in our ability to accurately diag-

nose pancreatic cysts in current practice also rests here,

hence the need for additional ancillary tests that will help

classify these often small cysts with nonspecific and non-

diagnostic features.

Five of the 6 malignant cysts in this study were sus-

pected of being malignant by EUS because of complex

features and/or direct invasion of surrounding structures,

and all 5 of those cysts were confirmed as malignant by cy-

tology and MDx. CEA analysis was performed in 2 cysts,

and both had markedly elevated CEA levels (Table 4).

There was no false-positive cytology result, but there was

1 false-negative cytology and MDx result. One small cyst

(1.6 cm) located in the pancreatic head in a woman aged

65 years (a low-grade IPMN by EUS without CEA analy-

sis and with indeterminate cytology) proved to be a malig-

nant IPMN on histology (moderate dysplasia with focal

carcinoma in situ). The MDx on this cyst fluid was non-

mucinous. It is highly likely that there was a complete

sampling error in this case, which is not surprising given

the small size of the cyst.

All remaining 29 cysts except 1 (Patient 18) (Table

4) did not demonstrate malignant EUS features and ini-

tially were categorized as mucinous (IPMN or MCN) or

nonmucinous by EUS (outlined in Table 1), and EUS-

FNA was performed either to support the EUS diagnosis

or to refine it. The triage of cyst fluid for analysis from our

institution generally is based on cyst fluid volume. Cyst

fluid with a volume �1 mL is submitted for cytology and

cyst fluid analysis. Molecular analysis is obtained at the

discretion of the gastroenterologist and usually is

Original Article


requested because of previous nondiagnostic cytology or

because of an amount of fluid that is too scant for both cy-

tology and/or cyst fluid analysis. In the current study, 6

cyst fluid samples were not tested for CEA or cytologic

analysis because of insufficient fluid sample. Of the 29

specimens that were examined cytologically, 13 of 29

samples (45%) were indeterminate or nondiagnostic.

Although thick and viscous extracellular mucin is a

helpful clue in the diagnosis on cytology, its sensitivity

remains suboptimal.9 In fact, extracellular mucin is not

always observed in neoplastic mucinous cysts, and MCN

and IPMN can produce thin, clear fluid that resembles se-

rous cysts as well as thin, brown fluid that mimics a pseu-

docyst.9 CEA cyst fluid analysis has provided an objective

test to determine whether the cyst fluid is mucinous or

nonmucinous and is considered the most accurate predic-

tor for the diagnosis of a mucinous cyst with 79.2% accu-

racy using a threshold CEA level of 192 ng/mL.18 Various

studies have produced different CEA cutoff values for a

mucinous cyst.23 Increasing the required value for CEA to

support the diagnosis of a mucinous cyst increases the

specificity but decreases the sensitivity of the test.18,23 It is

preferable to have a more sensitive and less specific value

for CEA given the malignant potential of all mucinous

cysts and the tendency for cytology to underestimate the

grade of the neoplasm because of the heterogeneity of the

cyst lining epithelium.9 Of the 14 cysts in the nonmuci-

nous category, only 1 case of an enteric duplication cyst

produced a CEA value >192 ng/mL (>5000 ng/mL); 1

other cyst, a pseudocyst, produced a CEA value >5 ng/mL

(176 ng/mL) but less than our 192 ng/mL threshold for a

mucinous cyst. The etiology for both of those cysts was

confirmed as benign nonmucinous on MDx. The median

value for CEA in the benign nonmucinous category was

0.7 ng/mL. Of the 14 benign mucinous cysts that had

CEA analysis, all but 1 produced a CEA value >192 ng/

mL. The range of CEA values for benign mucinous cysts

was from 71.6 ng/mL to 50,000 ng/mL (median, 4514

ng/mL). Only 2 malignant cysts were tested for CEA: One

had a CEA level >4000 ng/mL, and the other had a CEA

level >320,000 ng/mL.

Discordant findings in the benign mucinous cate-

gory included Patients 13 and 21 (Table 4). Patient 13, a

woman aged 61 years with a 1.2-cm cyst in the pancre-

atic head that had EUS features consistent with a branch

duct IPMN, produced cyst fluid with nondiagnostic cy-

tology, but her previous CEA level was elevated up to

37,900 ng/mL. Although the MDx was nonmucinous

based on no k-ras mutation and only 1 LOH mutation,

it was made 2 years after she underwent ethanol abla-

tion/lavage of the cyst, a process that denudes the cyst

lining from cell necrosis.24,25 No doubt such instrumen-

tation and treatment had a directly impact on the cytol-

ogy results, CEA analysis, and molecular analysis in this

patient. Patient 21 (Table 4) was a man aged 75 years

who had a 4.2-cm, malignant-appearing cyst located in

the pancreatic tail by EUS that had benign mucinous cy-

tology but a CEA level of only 71.6 ng/mL, which was

elevated but well below our cutoff level of 192 ng/mL for

supporting a diagnosis of mucinous cyst. The MDx in

this patient also was nonmucinous. Although no histol-

ogy was available in this patient, he remained alive and

well at 18 months.

In the benign nonmucinous group of cysts, accord-

ing to the CCD, there was 1 false mucinous MDx. Patient

24 (Table 4), who had a CCD of a serous cystadenoma

because of EUS features, a 0.7 ng/mL CEA analysis, and

nonmucinous cytology, had molecular analysis indicating

3 LOH mutations, including loci of 10q, 21q, and 17q.

No k-ras mutation was identified in this patient. The rea-

son for this discrepancy between the CCD and the MDx

is unclear. Although the presence of >2 LOH mutations

was used as 1 of the criteria for mucinous etiology by Red-

Path Integrated Pathology, as discussed above, this is less

specific for mucinous etiology than for k-ras mutation,

which was absent in this case.

Accurate diagnosis is essential for proper patient

management, which is evolving as we learn more about

the biologic behavior of neoplastic cysts in the pancreas.1-

6 Historically, treatment has been surgical excision of all

pancreatic cysts suspected of being neoplastic, but studies

of both serous cysts1 and mucinous cysts indicate that

not all cysts require resection,7,26 especially small (�3

cm) branch duct IPMNs,7,8 given the typical older age of

this often asymptomatic patient population, which is

fraught with comorbid conditions that increase surgical

risk, in addition to the likelihood that the patient with

not outlive the time to malignant progression of the cyst.

Nonsurgical therapies, such as ethanol ablation, also are

under investigation.24,25,27

International consensus guidelines recently were

proposed to address the diagnosis and treatment of



neoplastic mucinous cysts of the pancreas in general and

branch duct IPMNs in particular.7 An algorithm for the

preoperative evaluation of branch duct IPMNs to assist in

determining which patients should be offered surgical

resection was proposed (Fig. 1). We have demonstrated

that not all small cysts in asymptomatic patients are be-

nign and that not all cysts>3 cm are malignant.8 We also

demonstrated that ‘‘positive cytology’’ may not be

observed in IPMNs with malignancy.8,9 Therefore, add-

ing molecular analysis as an additional ancillary test to the

current multimodal approach offers improved sensitivity

and specificity for the major cyst classifications of malig-

nant, benign mucinous, and benign nonmucinous. A

combination of different ancillary tests is critical for a

more accurate diagnosis.

In summary, the results of the current study demon-

strate that there is significant agreement between the cur-

rent multimodal approach to pancreatic cyst diagnosis

and MDx of cyst fluid provided by PathFinderTG. The

current multimodal approach is achieved by the combina-

tion of routine clinical tests, such as EUS features, CEA

levels, and (most important) cytologic examination. How-

ever, a significant proportion of cases fall into the clinical

gray zone between benign mucinous and nonmucinous

cysts. Molecular analysis of k-ras gene mutation and LOH

can provide diagnostic information with minimal sample

volume, which may be the only information available

for clinical management when aspirated cyst fluid is

insufficient for chemical and cytologic analyses. The

current study provides valuable outcome data on molec-

ular testing of pancreatic cyst fluid. A large study with

validation of PathFinderTG molecular testing of pan-

creatic cyst fluid will be needed before we can draw a

firm conclusion.

Conflict of Interest Disclosures

This article was not supported financially by RedPath IP or anyother entity.

References

1. Tseng JF, Warshaw AL, Sahani DV, et al. Serous cystade-noma of the pancreas: tumor growth rates and recommenda-tions for treatment. Ann Surg. 2005;242:413-419; discussion419-421.

2. Matthes K, Mino–Kenudson M, Sahani DV, Holalkere N,Brugge WR. Concentration-dependent ablation of pancre-

atic tissue by EUS-guided ethanol injection. GastrointestEndosc. 2007;65:272-277.

3. Garcea G, Ong SL, Rajesh A, et al. Cystic lesions of thepancreas. A diagnostic and management dilemma. Pancrea-tology. 2008;8:236-251.

4. Allen PJ, Brennan MF. The management of cystic lesionsof the pancreas. Adv Surg. 2007;41:211-228.

5. Farnell MB. Surgical management of intraductal papillarymucinous neoplasm (IPMN) of the pancreas. J GastrointestSurg. 2008;12:414-416.

6. Lahav M, Maor Y, Avidan B, Novis B, Bar-Meir S. Non-surgical management of asymptomatic incidental pancreaticcysts. Clin Gastroenterol Hepatol. 2007;5:813-817.

7. Tanaka M, Chari S, Adsay V, et al. International consensusguidelines for management of intraductal papillary muci-nous neoplasms and mucinous cystic neoplasms of the pan-creas. Pancreatology. 2006;6:17-32.

8. Pitman MB, Michaels PJ, Deshpande V, Brugge WR,Bounds BC. Cytological and cyst fluid analysis of small(�3 cm) branch duct intraductal papillary mucinous neo-plasms adds value to patient management decisions. Pan-creatology. 2008;8:277-284.

9. Michaels PJ, Brachtel EF, Bounds BC, Brugge WR, PitmanMB. Intraductal papillary mucinous neoplasm (IPMN) ofthe pancreas: cytohistologic analysis and correlation withhistologic grade. Cancer (Cancer Cytopathol). 2006;108:163-173.

10. Deshpande V, Lauwers GY. Cystic pancreatic endocrine tu-mor: a variant commonly confused with cystic adenocarci-noma. Cancer (Cancer Cytopathol). 2007;111:47-53.

11. Vignesh S, Brugge WR. Endoscopic diagnosis and treat-ment of pancreatic cysts. J Clin Gastroenterol. 2008;42:493-506.

12. Jhala NC, Jhala D, Eltoum I, et al. Endoscopic ultrasound-guided fine needle aspiration biopsy: a powerful tool toobtain samples from small lesions. Cancer (Cancer Cytopa-thol). 2004;102:239-246.

13. Layfield LJ, Cramer H. Fine-needle aspiration cytology ofintraductal papillary-mucinous tumors: a retrospective anal-ysis. Diagn Cytopathol. 2005;32:16-20.

14. Recine M, Kaw M, Evans DB, Krishnamurthy S. Fine-nee-dle aspiration cytology of mucinous tumors of the pancreas.Cancer (Cancer Cytopathol). 2004;102:92-99.

15. Stelow EB, Stanley MW, Bardales RH, et al. Intraductalpapillary-mucinous neoplasm of the pancreas. The findingsand limitations of cytologic samples obtained by endoscopicultrasound-guided fine needle aspiration. Am J Clin Pathol.2003;120:398-404.

16. Shin HJ, Lahoti S, Sneige N. Endoscopic ultrasound-guided fine-needle aspiration in 179 cases: the M. D.Anderson Cancer Center experience. Cancer (Cancer Cyto-pathol). 2002;96:174-180.

17. Pitman MB, Deshpande V. Endoscopic ultrasound-guidedfine needle aspiration cytology of the pancreas: a

Original Article


morphological and multimodal approach to the diagnosisof solid and cystic mass lesions. Cytopathology. 2007;18:331-347.

18. Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al.Diagnosis of pancreatic cystic neoplasms: a report of theCooperative Pancreatic Cyst Study. Gastroenterology. 2004;126:1330-1336.

19. Brugge WR. Cystic pancreatic lesions: can we diagnosethem accurately what to look for? FNA marker molecularanalysis resection, surveillance, or endoscopic treatment?Endoscopy. 2006;38(suppl 1):S40-S47.

20. Khalid A, McGrath KM, Zahid M, et al. The role of pan-creatic cyst fluid molecular analysis in predicting cyst pa-thology. Clin Gastroenterol Hepatol. 2005;3:967-973.

21. Khalid A, Finkelstein S, McGrath K. Molecular diagnosisof solid and cystic lesions of the pancreas. GastroenterolClin North Am. 2004;33:891-906.

22. Schoedel KE, Finkelstein SD, Ohori NP. K-Ras and micro-satellite marker analysis of fine-needle aspirates from intra-ductal papillary mucinous neoplasms of the pancreas. DiagnCytopathol. 2006;34:605-608.

23. van der Waaij LA, van Dullemen HM, Porte RJ. Cyst fluidanalysis in the differential diagnosis of pancreatic cystic lesions:a pooled analysis. Gastrointest Endosc. 2005;62:383-389.

24. Gan SI, Thompson CC, Lauwers GY, Bounds BC, BruggeWR. Ethanol lavage of pancreatic cystic lesions: initial pilotstudy. Gastrointest Endosc. 2005;61:746-752.

25. Oh HC, Seo DW, Lee TY, et al. New treatment for cystictumors of the pancreas: EUS-guided ethanol lavage withpaclitaxel injection. Gastrointest Endosc. 2008;67:636-642.

26. Walsh RM, Vogt DP, Henderson JM, et al. Managementof suspected pancreatic cystic neoplasms based on cyst size.Surgery. 20008;144:677-684; discussion 684-5.

27. Brugge WR. Management and outcomes of pancreaticcystic lesions. Dig Liver Dis. 2008;40:854-859.



Molecular analysis of pancreatic cyst fluid: a comparative analysis with current practice of...

Documents

Transcript of Molecular analysis of pancreatic cyst fluid: a comparative analysis with current practice of...