Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Received: 25 September 2018 Accepted: 28 September 2018

DOI: 10.1002/dc.24097

ORIGINAL ARTICLE

Retrospective assessment of the effectiveness of the Milan


system for reporting salivary gland cytology: A systematic
review and meta-analysis of published literature
Sahar J Farahani MD, MPH | Zubair Baloch MD, PhD

Department of Pathology and Laboratory


Medicine, Hospital of the University of Abstract
Pennsylvania, University of Pennsylvania, Introduction: Fine needle aspiration (FNA) has been widely utilized in establishing the nature of
Philadelphia, Pennsylvania
salivary gland lesions and guiding the clinical management. This study aimed to determine the
Correspondence
accuracy of FNA in detecting salivary gland neoplasms and malignancies, employing the “Milan
Sahar J Farahani, Department of Pathology
and Laboratory Medicine, Hospital of the System for Reporting Salivary Gland Cytopathology” (MSRSGC).
University of Pennsylvania, University of Method: A systematic search was conducted. The data on FNA and histologic diagnosis were
Pennsylvania, Perelman School of Medicine, extracted and categorized based on the MSRSGC and risk of malignancy (ROM) was calculated.
3400 Spruce St, 6 Founders, Philadelphia, PA
The risk of publication bias and level of heterogeneity were evaluated. A mixed-effects model
19104
Email: sahar.jalali-farahani@uphs.upenn.edu was used to estimate FNA accuracy. Meta-regression was conducted to assess the potential
effect of different variables on FNA accuracy.
Results: Ninety-two studies with a total of 16 456 FNA with surgical follow-up were included.
ROM was estimated as 17%, 8%, 34%, 4%, 42%, 58%, and 91%, in nondiagnostic, nonneoplastic,
atypia of undetermined significance, benign neoplasm, salivary gland neoplasm of uncertain
malignant potential, suspicious for malignancy, and malignant groups, respectively. High level of
heterogeneity was detected (P-value <.001). Including cases with definite FNA diagnosis of neo-
plasm or malignancy, summary estimates of FNA sensitivity, specificity, diagnostic odds ratio,
and positive and negative likelihood ratio in detecting neoplasms and malignancies were 96.9%,
95.3%, 636.8, 20.5, and 0.03, and 80.5%, 97.9%, 189.5, 37.8, and 0.2, respectively. Meta-
regression showed several variables significantly impacting FNA accuracy; however, subgroup
analysis did not reduce the level of heterogeneity.
Conclusion: FNA can be used as a reliable diagnostic tool in the preoperative evaluation and man-
agement of salivary glands lesions. Concise of abstract is using Milan system for reporting salivary
gland FNA could increase FNA reliability, facilitate communication, and improve patient care.

KEYWORDS

diagnostic accuracy, FNA, meta-analysis, Milan system for reporting salivary gland cytology,
salivary gland malignancy, salivary gland neoplasm

1 | I N T RO D UC T I O N differentiation between a benign and malignant neoplasm is crucial in


determining the urgency, and extent of surgery.3,4 Moreover, in some
Salivary gland tumors account for less than 4-6% of head and neck cases distinguishing between different subtypes of primary and sec-
masses and comprise a heterogeneous group of lesions with diverse ondary malignancies of salivary glands is essential to predict the risk
etiologies and variable biological behavior. The mentioned variations of occult metastasis, local or distant recurrence and the need for addi-
warrant a specific treatment plan based on the underlying pathology. tional treatment modalities such as local irradiation or systemic che-
The nonneoplastic lesions of salivary glands comprising approximately motherapy in the biologically aggressive subtypes.5
20%-50% of all salivary gland lesions are managed conservatively with Fine needle aspiration (FNA) is a minimally invasive, inexpensive,
medical treatment and clinical follow-up.1,2 The salivary gland neo- and widely accessible diagnostic procedure. It has proven to be a use-
plasms often require surgical resection; nevertheless, the preoperative ful tool in obviating unnecessary surgical resection of nonneoplastic

Diagnostic Cytopathology. 2018;1–21. wileyonlinelibrary.com/journal/dc © 2018 Wiley Periodicals, Inc. 1


2 FARAHANI AND BALOCH

lesions and in management of neoplasms by differentiating between performance of FNA across the published literature, and determine
salivary gland nonneoplastic lesions, benign and malignant neoplasms the efficacy of MSRSGC in standardization and reduction of this het-
with high accuracy.6–9 However, in some cases, the FNA fails to fulfill erogeneity in reporting FNA diagnoses of salivary gland lesions.
this task due to the architectural and cellular complexity of salivary
gland neoplasms encountered among the same subtypes and even 2.1 | Literature search
within an individual tumor. Also, the subtle cytomorphological fea-
A comprehensive systematic search was conducted across PubMed/
tures of low-grade malignant neoplasms, and extensive cellular and
MEDLINE, Embase, and Scopus database, without restrictions on publi-
architectural overlap between tumor subtypes with entirely different
cation year. The appropriate keywords and subject headings were gener-
biological behaviors can be challenging in arriving at a definite diagno-
10,11 ated with a focus on three parameters: salivary gland pathology, FNA,
sis based on the limited FNA material. Moreover, the accuracy of
and diagnostic performance of FNA. For each parameter, an individual
FNA can also be significantly affected by variation in sample collec-
search was carried out for selected keywords and related medical subject
tion, preparation, evaluation and interpretation.
heading (MeSH) terms. This step was followed by a search combining
In response to the limitations of salivary gland FNA to increase its
reproducibility and promote its clinical relevance; a group of interna- keywords and MeSH terms using “OR” commands. Afterward, the search

tional pathologists proposed a classification system known as the results of the three parameters were connected with “AND” command

“Milan system for reporting salivary gland cytology report (MSRSGC)”. to shape the final results. The same strategy was followed searching

The MSRSGC classifies the results of salivary gland FNA into the fol- Embase database using the keywords and relevant subject heading

lowing diagnostic categories: “nondiagnostic”, “nonneoplastic”, “atypia (Emtree). Scopus database was searched using the same keywords as

of undetermined significance (AUS)”, “neoplasm-benign”, “neoplasm- PubMed and Embase. The reference lists of relevant identified studies

salivary gland neoplasm of uncertain malignant potential (SUMP)”, were manually searched to ensure the capture of all available literature.

“suspicious for malignancy”, and “malignant”. These are based on risk After filtering the duplicate articles using Endnote software (X8.2,

stratification rather than the specific histological subtypes and are Bld. 13302), title and abstract of the retrieved studies were reviewed

broad enough to accommodate the morphologic diversity of salivary to determine their eligibility to be included in the primary screening

gland tumors. It is expected that the MSRSGC classification will facili- process. The eligibility criteria for s included: studies written in English

tate the communication between pathologists and treating clinicians language and which evaluated the diagnostic performance of FNA in

and improve patient care. 12,13 detecting salivary gland tumors and differentiating malignant from
In this study, we conducted a systematic review and meta- benign neoplasms and/or neoplasms from nonneoplastic lesions.
analysis of all available literature on salivary gland FNA to determine The full text of eligible studies was retrieved and reviewed to
its accuracy in detecting salivary gland neoplasms, to report potential select only those which met the prespecified inclusion criteria. The
factors impacting its diagnostic performance, and to evaluate the effi- inclusion criteria were defined as original study, a total sample size of
ciency and clinical utility of the MSRSGC classification. 100 or more cases of salivary gland or head and neck tumors compris-
ing of at least 50 cases of salivary gland lesions in latter, and corre-
sponding histopathologic follow-up. The studies labeled a review, case
2 | MATERIALS AND METHODS reports, commentary, and editorials or abstracts were excluded.
Moreover, studies were also excluded if number of FNA true and false
Institutional Review Board (IRB) approval was not obtained for this positive, and true and false negative in comparison to histology cannot
study as it only involves a review of published literature. The be reproduced based on the reported data, or the risk of malignancy
Cochrane Guideline for meta-analysis of diagnostic studies and Pre- (ROM) could not be separately determined for the intermediate and
ferred Reporting Items for Systematic Review and Meta-Analysis definite diagnostic categories.
(PRISMA) were followed in developing the study protocol, in conduct-
ing the review and analysis, and reporting of the results.14,15 Two
2.2 | Data extraction
authors (ZB and SJF) formulated research questions, primary and sec-
ondary objectives, study protocol with regards to the search strate- A data extraction sheet was developed based on the PRISMA data
gies, eligibility criteria for primary screening, study selection, and final extraction template. After pilot-testing of the primary sheet on 10 ran-
inclusion, and application of statistical methods. domly-selected studies and making adjustments, following information
The research question was formulated according to PICO were extracted from each study: study population (inclusion and exclu-
(Patient, Intervention, Comparison, Outcome) format to investigate sion criteria, age, gender, co-morbidities, secondary tumor); setting and
the diagnostic performance of FNA in: differentiating between non- location of subject recruitment; study methods (design and random or
neoplastic, benign and malignant neoplasms in patients presenting consecutive enrollment); index test (description of FNA procedure in
with major and minor salivary gland masses, in comparison to surgical regards to the needle size, number of passes, conventional or liquid-
pathology follow-up, and in the light of the MSRSGC proposed diag- based preparation, stains, ancillary techniques, use of image guidance;
nostic categories. The primary objective of the study was defined as rapid on-site cytological examination [ROSE]; and interpretation of the
determining the accuracy of FNA in diagnosing salivary gland lesions. results); description of the reference test and its interpretation (histo-
The secondary objectives were to evaluate the level of, and identify pathologic follow-up); the interval between index (FNA) and reference
the potential factors were causing heterogeneity in diagnostic tests; and the number of FNA cases with histological follow-up. The
FARAHANI AND BALOCH 3

FNA diagnoses and their corresponding histopathologic follow-up were and I2 index of heterogeneity. The Spearman's Rho rank correlation test
used to tabulate 2*2 contingency table. was used to evaluate the presence of threshold effect by assessing the
correlation between the logit of true positive rate (sensitivity) and false

2.3 | Quality assessment of included studies positive rate (1-specificity). A P-value <.05 in spearman's test was con-
sidered as the presence of the threshold effect.
Two authors (ZB and SJF) assessed the quality of the selected studies
A bivariate generalized linear mixed-effects model with maximum
according to the Quality Assessment of Diagnostic Studies
likelihood estimation was used to calculate the summary points for the
(QUADAS)-2 questionnaire to evaluate the risk of bias in the domains
FNA sensitivity, specificity, DOR, PLR, and NLR in differentiating
concerning the patient selection; index and reference test execution,
between nonneoplastic, benign, and malignant neoplasms.17 A summary
evaluation, and interpretation; and study flow and timing.
receiver operating characteristics (SROC) curve with 95% confidence
and prediction regions was generated using the parameters estimated
2.4 | Descriptive statistics, analysis, and data from a hierarchical receiver operating characteristics (HSROC) model
synthesis and the area under the curve (AUC) was calculated accordingly.18 In a

2.4.1 | Milan system for reporting salivary gland cytology SROC curve, 95% confidence region around the summary estimates of
sensitivity and specificity is the area that real value of sensitivity and
The extracted data on cytology diagnosis of included cases were
specificity would be expected to lie in, based on the observed data. On
sorted and grouped into six diagnostic categories of the MSRSGC.
the other hand, 95% prediction region indicates the region that the
The corresponding histological diagnoses were categorized as non-
results of any future study would be expecting to be located. In other
neoplastic, benign neoplasm, and malignant neoplasm. The number
words, the confidence region depicts uncertainty in the overall average
and percentage of neoplastic, benign, and malignant neoplasms in
values caused by the intra-study variations, while the prediction region
each MSRSGC categories were determined.
demonstrates the variation from study heterogeneity. Where heteroge-
The prevalence of nonneoplastic, benign neoplasm, and malignant
neoplasm in each of the MSRSGC categories was estimated for the indi- neity is high, it is expected that 95% prediction region is much larger

vidual study. The I2 index of heterogeneity and Cochrane Q statistics than the 95% confidence region.14

test were used to assess the presence of potential variations in classify-


Likelihood ratios, pretest, and posttest probabilities
ing cytology and histology diagnosis across the selected studies. When
a moderate to high level of the heterogeneity was observed (indicated The clinical utility of FNA in detecting salivary gland neoplasms and

by an I2 index of heterogeneity≥50 with a P-value<.05); a univariate malignancies was evaluated based on the PLRs and NLRs. The PLR
random-effects model was utilized to estimate the mean prevalence of and NLR can be used to update the pretest probability of a target con-
nonneoplastic, benign neoplasm, and malignant neoplasm in each of the dition, which is equal to its baseline prevalence in the studied popula-
MSRSGC categories to take into account the intra and inter-study varia- tion, in the form of posttest probability using the followed formulas:
16
tions. In estimating the mean prevalence of benign (nonneoplastic/
benign neoplasms) and malignant lesions in each of the MSRSGC cate- • Posttest Probability (Prob.) on Positive FNA results = (Pretest
gories, all the studies were included. However, only the studies which Prob.*PLR)/ (1+ {Pretest Prob.*[PLR-1]})
made the distinction between these two categories on histologic • Posttest Probability (Prob.) on Negative FNA results = (Pretest
follow-up were used to estimate the prevalence of nonneoplastic and Prob.*NLR)/ (1+ {Pretest Prob.*[NLR-1]})
benign neoplasms (to be representative of general population).
In general, a PLR > 10 and NLR < 0.1 indicate an informative
2.4.2 | Diagnostic performance of FNA in detecting diagnostic test in ruling in and ruling out the target condition (neo-
salivary gland neoplasms and malignancy plasms or malignancies in this instance), respectively.
Descriptive statistics
Data from 2*2 contingency table was used to calculate the FNA sensi- Investigating the potential sources of data heterogeneity, meta-
tivity and specificity, diagnostic odds ratio (DOR), and positive and neg- regression, and subgroup analysis
ative likelihood ratio (PLR and NLR) in differentiating nonneoplastic To further investigate the potential source of heterogeneity among
lesions versus neoplasms and benign versus malignant neoplasms of sal- the included studies, multiple univariate linear models were used to
ivary glands for each study. In order to avoid zero as the denominator evaluate the impact of different covariates on the summary measures
in calculating DOR, PLR, NLR, a fixed continuity correction factor of 0.5 of sensitivity and specificity. These included: variations in study design
was added to each of TP, FP, FN, TN cells when a zero in FP and FN and methodology, patient population and setting, execution, prepara-
cells encountered. The forest plots were generated to illustrate the tion, and evaluation of the index (FNA) and reference (histologic
FNA sensitivity and specificity in detecting salivary gland neoplasms follow-up) tests.
and malignancies with their corresponding 95% confidence interval (CI). An HSROC model was used to assess the potential effect of co-
variables on the shape and asymmetry of SROC. While the general-
Assessment of heterogeneity and data synthesis ized linear models account for intra and inter-study variations by
The potential presence of intra and inter-study variations were assessed modeling the mean logit of sensitivity and specificity against each
visually using the forest plots and statistically by the Cochrane Q test other, the HSROC models do it by modeling the accuracy and
4 FARAHANI AND BALOCH

threshold parameters.19 Therefore, in investigating the potential dependent on the observed (such as the results of FNA or other
source of heterogeneity, the effect of covariates which affect the sen- reported factors that is, patients age or ultrasound or palpation-guided
sitivity or specificity or both without affecting the shape of SROC FNA) or on unobserved data and most importantly on the histologic
could be assessed by a linear model. However, the co-variables that diagnosis (eg, the patients with malignant tumors had less chance of
increase or decrease the sensitivity and specificity with different mag- undergoing excision compared to benign cases due to their poor
nitude alter the shape and symmetry of an SROC. The latter effect health condition).25,26
18,19
could be evaluated using an HSROC model. The analyses were conducted in Stata software (Stata/SE 13.1,
In addition, in an attempt to generate a more homogeneous group College Station, TX) and Microsoft Excel using MetaXL extension.
of studies, the variables with a significant effect (P-value <.05) on the
mean sensitivity, specificity or the SROC shape were used to subdi-
vide the primary studies into the smaller subgroups, and the summary 3 | RE SU LT S
estimates of sensitivity, specificity, DOR, PLR, and NLR associated
with each subgroup were calculated. 3.1 | Literature search

Investigating and handling publication bias After removing the duplicate studies, the literature searches yielded
2209 articles (Figure 1). After reviewing the titles and abstracts, 2034
The presence of publication bias was assessed using the Deek's funnel
articles were excluded. The full text of the remaining 175 articles were
plot asymmetry test. Visual asymmetry in the funnel plot and P-value
reviewed, and 92 studies were identified as eligible for systematic
<.1 for the slope coefficient could imply the presence of publication
20 review and meta-analysis. Three of the 92 articles reported data sepa-
bias. When the Deek's test indicated the presence of publication bias,
the Trim and Filled method was used to generate a hypothetical cohort rately on two different cohorts of patients from two different centers.

of studies by adding the potentially missing studies due to the publica- These were treated as two separate studies, and the systematic

tion bias. The DOR and I2 index of heterogeneity were estimated for review and meta-analysis included 95 individual patient cohorts.

this hypothetical cohort. Reduction in the I2 index of heterogeneity


with a significant difference between the DORs evaluated from the 3.2 | Selected studies characteristics and quality
original data and the hypothetical group (P-value <.05) could indicate assessment
that the accuracy of the studied index test is overestimated due to the
The pooled data comprised of 31 852 cases of salivary gland tumors,
publication bias.21,22 However, if the I2 index of heterogeneity does not
which included 908 cases with only clinical follow-up, 3963 cases of
change or even increases in the hypothetical data set, the observed
surgically resected lesions without preoperative FNA, 10 525 cases of
asymmetry in the funnel plot of the original data could be attributable
FNA without histologic follow-up, and 16 456 cases of FNA with the
to the high level of heterogeneity among the included studies.23
histologic follow-up. On average, 44.2%  5.1% (mean  [Standard

Investigating and handling partial verification bias Error*1.96]) of FNA cases proceeded to surgical resection. In the stud-
ies which reported the demographics of enrolled patients, the mean
The majority of included studies were a retrospective review of the
age, and male to female ratio were 51.0  11.1 (ranged between
FNA diagnosis and histological follow-up. A majority of cases with a
nonneoplastic diagnosis on FNA did not undergo the surgical resec- <1 month to 100 years) and 1.04:1, respectively. Of the 16456 FNA

tion. As only studies with the surgical histological follow-up were con- cases with histologic follow-up, 6733 (41%) were located in parotid,

sidered eligible for this study; it was expected that only a smaller 575 in submandibular (3.5%), and 19 in sublingual or minor salivary

percentage of cases with the nonneoplastic diagnosis on cytology had glands (0.1%). The site of salivary gland mass was not specified in

the surgical follow-up as compared to the ones diagnosed as a neo- 9129 (55%) cases.
plasm. The same scenario could be expected for FNA cases diagnosed Of the 92 included studies, 5 were prospective,27–31 and the rest

as benign neoplasm, however, in a smaller percentage, since the treat- were retrospective analyses. In 14 studies the subject enrollment was

ment of choice for most cases will be surgical resection. carried out consecutively or randomly.27–40
This phenomenon is known as the partial verification bias, it is The FNA procedure was performed by cytologists in 11, by radiol-
common in the diagnostic studies and can result in the overestimation ogists in 2, by clinicians in 12, and by clinicians/cytologists for the
of the disease prevalence and sensitivity, and underestimation of the palpable masses and by radiologists for the impalpable tumors in
specificity.24 16 studies. In seven studies, all FNAs were performed under ultra-
An interactive web-based tool known as Global Sensitivity Analy- sound guidance,28,30,38,41–44 while eight studies explicitly mentioned
sis (uwmsk.org/gsa) was utilized to calculate the corrected sensitivity that no radiologic guidance was employed.45–52 The use of ancillary
and specificity based on the ratio of the FNA cases with the positive studies for the selected cases was reported in 8 studies.34,53–59 ROSE
and negative results for a neoplasm/malignancy with or without surgi- was available in 14 studies.29,32,37,39,47,51,59–66
cal follow-up. Besides, this tool conducted several sensitivity analyses The quality of included studies was assessed according to the
using all the possible results on histologic follow-up for the FNA cases QUADAS-2 questionnaire. None of the studies were considered as
which did not undergo surgical excision. This helped to determine low-bias-risk in all four domains. Figure 2 summarizes the overall
whether the chance of not receiving the histologic follow-up was results of the QUADAS-2 questionnaire for the included studies.
FARAHANI AND BALOCH 5

FIGURE 1 PRISMA flow chart of identified, eligible, included and excluded studies

3.3 | Analysis and data synthesis histopathologic diagnoses for the diagnostic categories of MSRSGC
classification, using a mixed-effects model.
3.3.1 | The Milan system for reporting salivary gland
cytopathology classification
Nondiagnostic
The 16 456 FNA cases with the histologic follow-up were sorted into
The diagnostic terms “nondiagnostic”, “inadequate”, “insufficient”,
the diagnostic categories of MSRSGC as follows: “nondiagnostic”,
“unsatisfactory”, “sampling error” were used in 67 studies. In 49 stud-
622; “nonneoplastic”, 1269; “AUS”, 110; “benign neoplasm”, 10 979;
ies, these terms were used to describe the samples containing only
“SUMP”, 397; “suspicious for malignancy”, 122, and “malignant”, 2957
blood, necrotic material, acellular cyst content, benign salivary gland
cases.
tissue in the presence of a mass; or when there was insufficient cellu-
In estimating the mean prevalence of histologic follow-up diagno-
larity or analyzable material to render a diagnosis based on the FNA
sis of nonneoplastic, benign neoplasm and malignant neoplasm for dif-
specimen. In 16 studies, the term “nondiagnostic” or “unsatisfactory”
ferent MSRSGC categories, I2 index of heterogeneity and Cochrane Q
was used without any further explanation. In three studies, the term
test indicated moderate to high level of heterogeneity in all categories “nondiagnostic” was also used to classify specimens with questionable
(70%-83%, P-value <.001), except for cases classified as AUS (45%, P- or indefinite diagnosis or with atypical features. The cases extracted
value: .06). Based on these findings, the pooled mean prevalence was from these three studies were not classified as a “nondiagnostic” in
estimated using mixed-effects generalized linear model. the present study, b MSRSGC criteria for the nondiagnostic category.
Table 1 summarizes the number and percentage of the FNA spec- The histological follow-up of “nondiagnostic” cases was reported in
imens with the histologic follow-up as well as the mean prevalence of 44 studies. Of these, in 25 studies the authors made a distinction

QUADAS-2 Questionnaire
Enrolled Patients Consecutively or Randomly 15 80

Avoided Case-Control Design 95

Avoided Inappropiate Exclusion 64 3 28

Cytology Was Interpreted Blindly to Histology Result 45 47 3

Hitology Classified SGT Correctly 95

Histology Was Interpreted Blindly to Cytology Result 7 85 3

Appropiate Interval Between Cytology and Histology 3 92

All Patients Received Histology/Clinical Follow-up 27 9 59

All Patients Included in the Analysis 95

0 10 20 30 40 50 60 70 80 90 100

Yes Unclear No

FIGURE 2 Overall summaries of QUADAS-2 checklists in patient selection, index and reference tests, and study flow and timing domains for the
included studies
6 FARAHANI AND BALOCH

TABLE 1 Number of specimens and pooled mean prevalence of histological diagnosis of non-neoplastic, benign neoplasms, and malignant
neoplasms across six categories of the MSRSGC classification
Neoplasm
Cytology Nondiagnostic Nonneoplastic AUS Benign neoplasm SUMP Suspicious for Malignant
Histology Number Number Number Number Number malignancy Number Number
Specimen Number (%) 622 (3.8) 1269 (7.7) 110 (1) 10 979 (66.8) 379 (2.3) 122 (0.1) 2957 (18.0)
Nonneoplastic
Specimen Number (%) 100 (16.1) 1006 (79.3) 23 (20.9) 49 (0.45) 30 (7.56) 4 (3.3) 53 (1.79)
Mean prevalencea (%) 20.0 82.2 36.7 0.7 7.9 5.0 2.1
(95% CI) (15.1-23.9) (79.1-83.3) (25.2-50.2) (0.5-0.9) (4.2-12.2) (1.1-11.1) (1.6-2.7)
Benign neoplastic
Specimen Number (%) 394 (63.4) 150 (11.8) 46 (41.8) 10 342 (94.2) 242 (61.0) 43 (35.3) 269 (9.10)
Mean prevalencea 62.7 9.4 29.5 95.8 50.6 36.6 6.8
(95% CI) (55.1-65.9) (7.8-11.0) (18.7-42.5) (95.2-96.3) (42.5-57.2) (26.2-47.4) (5.9-7.8)
Malignant
Specimen Number (%) 128 (20.6) 113 (8.9) 41 (37.3) 588 (5.4) 125 (31.5) 75 (61.5) 2635 (89.1)
Mean prevalencea 17.3 8.4 33.8 3.5 41.5 58.4 91.1
(95% CI) (12.7-21.0) (6.9-9.9) (22.6-47.1) (3.1-4.0) (33.8-48.2) (47.1-68.7) (90.4-92.5)

Abbreviations: AUS, atypia of undetermined significance; SUMP, salivary gland neoplasm of uncertain malignant potential; CI, confidence interval.
a
Pooled prevalence of nonneoplastic, benign neoplasms, and malignant neoplasms across the six categories of the MSRSGC in the studies which made a
distinction between nonneoplastic lesions and benign neoplasms, adjusted for inter and intra-study variations.

between the nonneoplastic and benign neoplasms on histologic between nonneoplastic and neoplasm could not be made. Using a
follow-up. random-effects model, the summary prevalence of histologic follow-up
A total of 622 cases with the FNA diagnosis of “nondiagnostic” diagnosis of benign (nonneoplastic/benign neoplasms) and malignant
were extracted from the 44 studies mentioned above. The rate of non- tumors was 63% (54%-71%) and 37% (29%-46%), respectively. In
diagnostic FNA ranged from <1% to 44%; 75% had a nondiagnostic rate 11 studies in which the histologic follow-up diagnoses of the nonneo-
of ≤10%. The rate of nondiagnostic salivary gland FNA in the studies plastic and benign neoplasms were reported separately, the mean prev-
conducted at general hospitals was 21.28%  17.47% as compared to alence of nonneoplastic, benign neoplasms, and malignant neoplasms
4.38%  2.32% performed at the tertiary or referral cancer centers. were 37% (25%-50%), 30% (19%-43%), 34% (23%-47%), respectively.
This difference was statistically significant (P-value: .002). The rate of
nondiagnostic FNA was significantly lower in studies in which the cytol- Neoplasm
ogists performed FNA procedures, when compared to the rest of the Benign. A total of 5737 FNA specimens were interpreted as “benign
studies (3.63%  6.2% vs 9.78%  2.55%, P-value: .004). neoplasm” in 56 studies. This diagnosis was confirmed on the surgical
On histologic follow-up, the mean prevalence of benign (nonneo- follow-up in 5464 cases; 46 cases were diagnosed as nonneoplastic
plastic/benign neoplasm) and malignant neoplasms in 44 studies, esti- and 227 as malignant neoplasms. The mean prevalence of nonneo-
mated by a random-effects model, was 81% (77%-83%) and 19% plastic, benign neoplasm, and malignant neoplasm, estimated by
(17%-23%), respectively. The pooled mean prevalence of nonneoplas- random-effects model was 1% (0.5%-0.9%), 96% (95%-96%), and 4%
tic, benign and malignant neoplasms, derived from 25 studies, which (3%-4%), respectively.
distinguished between nonneoplastic and benign neoplasms on histo- In the remaining studies from the pooled cohort, 5242 cases were
pathologic follow-up was 20% (15%-24%), 63% (55%-66%), and 17% classified as “benign” on FNA without distinguishing between the
(13%-21%), respectively. nonneoplastic lesion and benign neoplasm. On histologic follow-up,
the mean prevalence of benign and malignant lesions was 5% (5%-6%)
Nonneoplastic and 95% (94%-96%), respectively.
A total of 1269 specimens could be retrospectively classified as non-
neoplastic in 45 studies. On the histopathologic follow-up, 1006 cases SUMP. In 29 studies, the diagnostic terms “indeterminate”, “neo-
were diagnosed as nonneoplastic, 150 cases as benign and 113 cases plasm”, “neoplasm not otherwise specified”, “tumor”,” atypia”, or
as the malignant neoplasms. The summary estimates of nonneoplastic, “basal cell or basaloid neoplasm” were used for a total of 397 cases to
benign and malignant neoplasms on histologic follow-up were 82% describe a specimen with sufficient cellularity and definite features of
(79%-83%), 9% (8%-11%), and 8% (7%-10%), respectively. a neoplasm but lacking of required elements to render a definite diag-
nosis of benign versus malignant neoplasm. The summary estimates of
Atypia of undetermined significance benign (nonneoplastic/benign neoplasm) and malignant neoplasm
The 16 studies employed the terms “atypia”, “non-contributory”, “non- diagnosis on histology was 66% (62%-71%) and 34% (29%-38%)
diagnostic”, or “indeterminate” in a total of 110 cases to report the FNA respectively. In 20 studies, the distinction between the nonneoplastic
samples with sufficient cellularity and containing atypical cells. This lesion and benign neoplasm were made on the histologic follow-up.
term was employed for FNA cases in, which the differentiation The summary estimate of nonneoplastic, benign neoplasm, and
FARAHANI AND BALOCH 7

malignant neoplasm in these studies was 8% (4%-12%), 51% (43%- included studies. The Spearman rank-order correlation test showed a
57%), and 42% (34%-48%), respectively. significant, although a positive association between the mean sensitiv-
ity and specificity (coefficient: 0.34, P-value: .02).
Suspicious for malignancy Table 3 summarizes the estimated FNA summary sensitivity, speci-
A total of 122 cases extracted from 17 studies were considered as “sus- ficity, DOR, PLR, and NLR in distinguishing neoplasms and nonneoplas-
picious for malignancy” on FNA. On histologic follow-up, 4 cases were tic lesions of the salivary gland in five scenarios mentioned above.
diagnosed as nonneoplastic, 43 as benign and 75 as malignant Figure 4 demonstrates the SROC of FNA when only the definite cytol-
neoplasms. The mean prevalence of benign (nonneoplastic/benign ogy diagnosis of neoplasm (scenario 3) was included in the analysis.
neoplasms) and malignant neoplasm were 37% (29%-46%) and 63% In the current series, the prevalence of neoplasm was 82%, which
(54%-71%). Twelve studies reported the histologic follow-up diagnosis is expected to be an overestimation of the salivary gland neoplasm
of the nonneoplastic lesion and benign neoplasms separately, the mean prevalence in the general population due to the partial verification
prevalence of nonneoplastic, benign neoplasm and malignant neoplasm bias. The previous studies reported a prevalence of 50%-70% for the
were 5% (1%-11%), 37% (26%-47%), 58% (47%-69%), respectively. neoplastic lesions.1,2 A Fagan nomogram was drawn to demonstrate
the posttest probability of neoplasm with positive and negative FNA
Malignant results for neoplasm in four scenarios mentioned earlier for a pretest
A total of 2957 cases were diagnosed as either a primary or secondary probability of 60% (Figure 5).
malignant tumor of the salivary gland on FNA. The malignant diagnosis In the Deek's funnel plot asymmetry test; the regression coefficient
was confirmed on histology in 91% (90%-92%) cases. The remaining between the DOR and the inverse of the root of the SE of DOR was
2% (2%-3%) and 7% (6%-8%) cases were diagnosed as nonneoplastic −11.7 with a P-value: .001, suggesting the presence of significant publi-
and benign neoplasms on the histopathologic follow-up. cation bias. Using the Trim and Fill test, a hypothetical database was
generated by adding 27 studies in place of the “missing” studies
3.3.2 | Analysis and data synthesis assumed to be responsible for the publication bias. Subsequently, the
Differentiating non-neoplastic from neoplasm DOR of this hypothetical database was estimated using an HSROC
Of the 92 included studies, 45 had sufficient information to evaluate model. The analyses showed that the addition of the “missing” studies
the diagnostic accuracy of salivary gland FNA in distinguishing resulted in an increased variance between the studies, and the evidence
between a nonneoplastic lesion and a neoplasm. The number of true of heterogeneity in the dataset remained unchanged at a P-value <.01.
and false positive and negative cases, sensitivity, specificity, DOR,
PLR, and NLR in the individual primary studies are summarized in Differentiating malignant from benign lesions
Table 2. All 92 studies were included in calculating the accuracy of FNA in dis-
In estimating the accuracy of salivary gland FNA four scenarios tinguishing between the benign and malignant salivary gland neo-
were considered: (1) recognizing the FNA cases interpreted as “non- plasms. The benign lesions were considered as nonneoplastic or
neoplastic” as the negative index test result for neoplasm, and cases benign neoplasms. The number of true and false positive and negative
classified as “nondiagnostic”, “AUS”, “benign neoplasm”, “SUMP”, “sus- FNA cases with the measures of sensitivity, specificity, DOR, PLR, and
picious for malignancy”, and “malignant” as the positive index test NLR in each of the included study are summarized in Table 4.
result for neoplasm; (2) excluding “nondiagnostic” cases from the anal- The following five scenarios were considered in estimating the FNA
ysis, considering FNA cases in the diagnostic category of “nonneoplas- accuracy: (1) identifying the FNA cases classified as “nonneoplastic” and
tic” as the negative index test result for neoplasm, and the ones in “benign neoplasm” as the negative index test result for malignancy, and
“AUS”, “benign neoplasm”, “SUMP”, “suspicious for malignancy”, and “nondiagnostic”, “AUS”, “SUMP”, “suspicious for malignancy”, and
“malignant” categories as the positive index test result for neoplasm; “malignant” as the positive index test result for malignancy; (2) excluding
(3) including the cases with definite diagnosis for neoplasm in the “nondiagnostic” FNA cases from the analysis and recognizing the cases
analysis; considering the FNA cases in diagnostic category of “non- with the diagnosis of “nonneoplastic” and “benign neoplasm” as the neg-
neoplastic” as the negative index test result for neoplasm, and” benign ative index test result for malignancy, and “AUS”, “SUMP”, “suspicious
neoplasm”, “SUMP”, “suspicious for malignancy”, and “malignant” as for malignancy”, and “malignant” as the positive index test result for
the positive index test result for neoplasm; (4) recognizing FNA cases malignancy; (3) excluding “nondiagnostic” and “AUS” cases from the
classified as “nondiagnostic”, “nonneoplastic”, and “AUS” as the nega- analysis and considering the cases with diagnosis of “nonneoplastic”
tive index test result for neoplasm, and cases classified as “benign and “benign neoplasm” as the negative index test result for malignancy,
neoplasm”, “SUMP”, “suspicious for malignancy”, and “malignant” as and “SUMP”, “suspicious for malignancy”, and “malignant” as the posi-
the positive index test result for neoplasm. tive index test result for malignancy; (4) including FNA cases with the
The Cochrane Q test showed significant heterogeneity in FNA definite diagnosis for malignancy in the analysis; identifying the cases
sensitivity, specificity, and DOR (P-value <0.001). The I2 index of FNA with the diagnosis of “nonneoplastic” and “benign neoplasm” as the neg-
sensitivity, specificity, and DOR were 70% (95% CI: 65%-76%), 74% ative index test result for malignancy, and “suspicious for malignancy”
(95% CI: 68%-82%), and 74% (66%-82%), respectively. The forest and “malignant” as the positive index test result for malignancy;
plots (Figure 3A,B) depict the individual study sensitivity and specific- (5) including FNA cases with the diagnosis of “nonneoplastic”, “benign
ity and their corresponding 95% CI and the variations across the neoplasm”, “AUS”, and “SUMP” as the negative index test result for
8 FARAHANI AND BALOCH

TABLE 2 Diagnostic performance of salivary gland FNA in detecting neoplasm in individual primary studies

Author year TP FP FN TN Sensitivity (%) Specificity (%) DOR PLR NLR


27
Rajdeo RN 2017 93 0 3 4 96.9 100.0 240.4 9.6 0.04
Rohilla M 201767 69 0 6 18 92.0 100.0 395.6 34.8 0.09
Mohammed Nur 201668 124 3 9 7 93.2 70.0 32.2 3.1 0.10
Ameli F 201569 72 6 4 19 94.7 76.0 57.0 4.0 0.07
Arul P 201570 112 4 3 21 97.4 84.0 196.0 6.1 0.03
Mairembam P 201671 149 1 0 21 100.0 95.5 4285.7 15.3 0.00
Novoa E 201637 88 2 6 4 93.6 66.7 29.3 2.8 0.10
Omhare A 201472 47 0 0 39 100.0 100.0 7505.0 79.2 0.01
Pastore A 201373 283 4 23 24 92.5 85.7 73.8 6.5 0.09
Kechagias N 201258 72 1 1 4 98.6 80.0 288.0 4.9 0.02
Nanda KDS 201274 39 0 0 17 100.0 100.0 2765.0 35.6 0.01
Nguansangiam S 201255 62 1 2 68 96.9 98.6 2108.0 66.8 0.03
49
Ali NS 2011 113 2 4 5 96.6 71.4 70.6 3.4 0.05
Cho HW 201143 203 1 4 7 98.1 87.5 355.3 7.9 0.02
Ashraf A 201031 86 0 0 14 100.0 100.0 5017.0 29.8 0.01
Christensen RK 201075 294 2 23 63 92.7 96.9 402.7 30.1 0.07
76
Stramandinoli RT 2010 71 0 4 3 94.7 100.0 111.2 7.5 0.07
Daneshbod Y 200977 355 1 3 17 99.2 94.4 2011.7 17.9 0.01
Zhang S 200947 56 4 6 12 90.3 75.0 28.0 3.6 0.13
de Ru JA 200778 83 0 0 5 100.0 100.0 1837.0 11.9 0.01
64
Elagoz S 2007 29 1 11 15 72.5 93.8 39.6 11.6 0.29
Lim CM 200779 72 2 3 4 96.0 66.7 48.0 2.9 0.06
Tan LGL 200680 37 0 7 12 84.1 100.0 125.0 21.7 0.17
Bandyopadhyay A 200581 46 0 1 12 97.9 100.0 775.0 25.2 0.03
82
Riley N 2005 89 2 4 2 95.7 50.0 22.3 1.9 0.09
Seethala RR 200561 158 4 6 18 96.3 81.8 118.5 5.3 0.04
Postema RJ 200467 295 3 21 61 93.4 95.3 285.6 19.9 0.07
Contucci AM 200362 118 0 12 9 90.8 100.0 180.1 18.1 0.10
83
Sengupta S 2002 131 0 4 73 97.0 100.0 4295.7 143.1 0.03
Jayaram G 200150 48 0 2 3 96.0 100.0 135.8 7.6 0.06
Costas A 200054 73 3 2 22 97.3 88.0 267.7 8.1 0.03
Al-Khafaji BM 199832 127 3 9 9 93.4 75.0 42.3 3.7 0.09
68
Boccato P 1998 528 0 1 35 99.8 100.0 25 015.7 71.8 0.00
Filopoulos E 199859 109 0 1 10 99.1 100.0 1533.0 21.7 0.01
Cajulis RS 199784 67 0 4 59 94.4 100.0 1785.0 112.5 0.06
Christallini EG 199751 50 0 2 11 96.2 100.0 464.6 22.9 0.05
85
Atula T 1996 135 8 35 26 79.4 76.5 12.5 3.4 0.27
Atula T 199586 15 11 9 43 62.5 79.6 6.5 3.1 0.47
Orell SR 199587 290 3 2 23 99.3 88.5 1111.7 8.6 0.01
Zurida S 199388 204 0 0 19 100.0 100.0 15 951.0 39.9 0.00
89
Chan MKM 1992 78 1 6 25 92.9 96.2 325.0 24.1 0.07
Nettle WJS 198945 97 0 1 4 99.0 100.0 585.0 9.9 0.02
Layfield LJ 198790 111 1 10 44 91.7 97.8 488.4 41.3 0.08
Persson PS 197391 187 1 1 13 99.5 92.9 2431.0 13.9 0.01
92
Eneroth CM 1967 461 0 8 74 98.3 100.0 8089.8 147.3 0.02

In order to avoid zero as the denominator in calculating DOR, PLR, NLR, a fixed continuity correction factor of 0.5 was added to each of TP, FP, FN, TN
cells when a zero in FP and FN cells encountered.
Abbreviations: TP, true positive; FP, false positive; FN, false negative; TN, true negative; DOR, diagnostic odds ratio; PLR, positive likelihood ratio; NLR,
negative likelihood ratio.

malignancy, and “suspicious for malignancy” and “malignant” as the posi- 79% (95% CI: 75%-83%) with a P-value <.001 by the Cochrane Q test.
tive index test result for malignancy. The forest plots (Figure 6A,B) depict the individual study sensitivity
The I2 index of heterogeneity for the FNA sensitivity, specificity, and specificity and their corresponding 95% CI and the variations
and DOR was 68% (95% CI: 65%-72%), 76% (95% CI: 71%-80%), and across the included studies. The Spearman correlation test did not
FARAHANI AND BALOCH 9

show any significant association between the mean sensitivity and The malignancy prevalence in the current series was 22%, which
specificity. was in concordance with the previous studies.1,2 Considering a pretest
The results of the FNA summary sensitivity, specificity, DOR, probability of 22%, Figure 8 depicts the posttest probability of malig-
PLR, and NLR in the five scenarios above were calculated using a nancy with FNA positive and negative results for malignancy in a
bivariate generalized linear mixed-effects model (Table 5). Figure 7 Fagan nomogram for the five mentioned scenarios.
demonstrates the SROC of FNA when only the definite cytology diag- The Deek's funnel plot asymmetry test with a regression coeffi-
nosis for the malignancy (scenario 4) was included in the analysis. cient of −11.2 and a P-value: .06 indicated the presence of significant

FIGURE 3 (Continued on next page)


10 FARAHANI AND BALOCH

FIGURE 3 (A,B) FNA sensitivity and specificity forest plots in differentiating salivary glands neoplasms from nonneoplastic lesions. The forest
plot demonstrates a high level of intra and inter-study variations

publication bias. However, in the Trim and Fill test, adding the poten- can be extracted from 18 and 24 studies, respectively. The ratios of
tially missed studies and generating a new hypothetical database did the total number of cases with the cytology diagnosis of nonneoplas-
not result in the resolution of heterogeneity among the studies. tic, neoplasm, benign, and malignant, with to the ones, without histo-
logic follow-up, was 5.59, 1.44, 2.44, and 1.48, respectively.
Verification bias By employing the above ratios, a corrected estimate of salivary
The data on the number of FNA cases diagnosed as nonneoplastic, gland FNA sensitivity and specificity in detecting salivary gland neo-
and neoplasm, and benign and malignant, without histologic follow-up plasm and malignancy was calculated using the interactive web-based
FARAHANI AND BALOCH 11

TABLE 3 Summary diagnostic performance of salivary gland FNA in detecting neoplasm in different scenarios

Sensitivity (%) Specificity (%) DOR PLR NLR


(95% CI) (95% CI) (95% CI) (95% CI) (95% CI)
Scenario 1 97.0 90.2 302.3 9.9 0.03
(95.6-98.0) (84.6-93.9) (142.7-640.5) (6.2-15.9) (0.02-0.05)
Scenario 2 96.9 93.7 470.0 15.3 0.03
(95.5-98.0) (89.6-96.2) (217.8-1013.8) (9.2-25.3) (0.02-0.05)
Scenario 3 96.9 95.3 636.8 20.5 0.03
(95.5-97.9) (91.8-97.3) (284.2-1426.5) (11.8-35.6) (0.02-0.05)
Scenario 4 94.4 95.6 368.3 21.7 0.06
(90.6-97.2) (90.9-98.9) (220.1-1561.4) (14.3-34.9) (0.04-0.08)
I2 index of heterogeneity (%) 70.7 74.3 78.2
(65.1-78.4) (69.4-78.8) (74.3-82.5)

Abbreviations: CI, confidence interval; DOR, diagnostic odds ratio; PLR, positive likelihood ratio; NLR, negative likelihood ratio.
Scenario 1, FNA “nonneoplastic” as a negative index test result for neoplasm, and “nondiagnostic”, “AUS”, benign neoplasm”, “SUMP”, “suspicious for malig-
nancy”, and “malignant” as a positive index test result for neoplasm;
Scenario 2, FNA “nonneoplastic” as negative index test result for neoplasm, and “AUS”, “benign neoplasm”, “SUMP”, “suspicious for malignancy”, and
“malignant” as a positive index test result for neoplasm;
Scenario 3, FNA “nonneoplastic” as negative index test result for neoplasm, and “benign neoplasm”, “SUMP”, “suspicious for malignancy”, and “malignant”
as a positive index test result for neoplasm;
Scenario 4, FNA “nondiagnostic”, “nonneoplastic”, and “AUS” as negative index test result for neoplasm, and cases classified as “benign neoplasm”, “SUMP”,
“suspicious for malignancy”, and “malignant” as a positive index test result for neoplasm.

tool, Global Sensitivity Analysis. The corrected sensitivity and specific- 4 | DI SCU SSION
ity of FNA in differentiating a neoplasm from a nonneoplastic lesion
and a malignant from a benign neoplasm were estimated as 84.37% This comprehensive review and meta-analysis of the literature on sali-
(95% CI: 81.87%-87.39%) and 98.27% (95% CI: 97.51%-98.94%), and vary gland FNA is an attempt to; assess the accuracy of FNA in the
72.10% (95% CI: 67.65%-79.67%) and 98.62% (95% CI: 98.33%- diagnosis of salivary gland lesions; to investigate the potential factors
98.98%), respectively. Sensitivity analysis showed that the chance of impacting its accuracy and to retrospectively evaluate the feasibility
not receiving the histologic follow-up was not dependent on the his- and effectiveness of the MSRSGC (Table 1).
tologic diagnosis of neoplastic or malignant diagnosis of the examined
lesions.
4.1 | Diagnostic accuracy, the posttest probabilities,
and ROM
Investigating the possible causes of heterogeneity, meta-
regression and sub-group analysis The systematic review and meta-analysis of available literature showed

Multiple meta-regression analyses showed that the FNA sensitivity that the salivary gland FNA rendered a definite diagnosis of nonneoplas-

and specificity in differentiating a neoplasm from a nonneoplastic tic lesion, benign, and malignant neoplasm in 94% cases. FNA was indi-

lesion and a benign from malignant lesion were significantly higher in cated to be an informative test in both rulings in and ruling out the

the studies in which FNAs were performed in the cancer referral presence of neoplasm with a PLR of 20.5 and NLR of 0.03. Interestingly,
centers. by considering the cases with an indefinite diagnosis of neoplasm (AUS)
The FNA sensitivity and specificity in distinguishing benign from as a positive for neoplasm result, FNA specificity and PLR decreased
malignant neoplasm were also significantly higher in the studies, without any significant improvement in FNA sensitivity and NLR. A defi-
which employed ancillary techniques in selected cases, the FNA pro- nite diagnosis for neoplasm on FNA increased the odds of having a neo-
cedures were performed by cytologist/s, ROSE was available, and in plasm by 20.5 times, and a definite diagnosis of a nonneoplastic lesion
studies with higher prevalence of malignancy (> 25%). The salivary on FNA decreased the odds of having a neoplasm by 0.03 comparing to
gland FNA sensitivity and specificity were also noted to be higher its baseline odds. The odds of having a neoplasm on histologic follow-
after the release of the first revision of the WHO classification in up with “nondiagnostic” and “AUS” diagnosis on cytology was 2.5 and
1991 as compared to ones before. The availability of ROSE availability 1.3 times of having a nonneoplastic lesion, respectively. In addition, by
increased the FNA sensitivity with larger magnitude comparing to the considering “nondiagnostic” and “AUS” diagnoses in addition to “benign
specificity and affected the shape of the SROC curve. neoplasm”, “SUMP”, “suspicious for malignancy”, and “malignant” as a
Although the above variables were demonstrated to significantly positive result for neoplasm, the odds of having a neoplasm comparing
affect the FNA sensitivity, or specificity, or SROC curve or all, subdi- to the baseline odds decreased to 9.1 and therefore a positive result on
viding the included studies into the smaller subgroups based on these FNA could no longer be considered as informative in ruling in the diag-
variables did not result in a more homogeneous cohort of studies. nosis of neoplasm.
Table 6 summarized the summary estimates of FNA diagnostic In distinguishing between benign and malignant salivary gland
performance measures in distinguishing the salivary gland malignancy tumors, FNA had high specificity, yet moderate sensitivity even when
in different subgroups. it was able to deliver a definite diagnosis. With a PLR of 37.7, a
12 FARAHANI AND BALOCH

FIGURE 4 HSROC curve of FNA in detecting salivary gland neoplasm


when only including the FNA cases with a definite diagnosis for
neoplasm (the MSRSGC categories of “nonneoplastic”, “neoplasm-
benign/SUMP”, “suspicious for malignancy”, “malignant”) in the
analysis. Circles indicate primary studies sensitivity and specificity.
The size of circles corresponds with study sample size. The solid green
line represents FNA summary sensitivity, and specificity line with 95%
prediction region (dotted green line) and the red square indicates the
FNA summary points of sensitivity and specificity with 95%
confidence region (dotted orange line). The 95% confidence region FIGURE 5 Fagan nomogram for detecting neoplasm: scenario1,
shows the region the real value of FNA sensitivity and specificity recognizing the FNA cases interpreted as “nonneoplastic” as the
would expect to lie, based on the observed data. The 95% prediction negative index test result for neoplasm, and cases classified as
region depicts the region the results of a future study would expect to “nondiagnostic”, “AUS”, “benign neoplasm”, “SUMP”, “suspicious for
be located. As it could be expected due to the high level of malignancy”, and “malignant” as the positive index test result for
heterogeneity between the studies, the 95% prediction region is neoplasm; scenario 2, excluding “nondiagnostic” cases from the
much larger than the 95% confidence region analysis, considering FNA cases in the diagnostic category of
“nonneoplastic” as the negative index test result for neoplasm, and
“suspicious for malignancy” or “malignant” diagnosis on FNA could be the ones in “AUS”, “benign neoplasm”, “SUMP”, “suspicious for
considered as informative in ruling in the malignancy diagnosis. How- malignancy”, and “malignant” categories as the positive index test
result for neoplasm; scenario 3, including the cases with definite
ever, the NRL of 0.19 indicated that a diagnosis of a nonneoplastic
diagnosis for neoplasm in the analysis; considering the FNA cases in
lesion or benign neoplasm or generally benign on FNA could not be
diagnostic category of “nonneoplastic” as the negative index test
considered informative in ruling out the risk of malignancy in the result for neoplasm, and “benign neoplasm”, “SUMP”, “suspicious for
examined sample. By considering “nondiagnostic”, “AUS”, and “SUMP” malignancy”, and “malignant” as the positive index test result for
in addition to “suspicious for malignancy”, and “malignant” as a posi- neoplasm; scenario 4, recognizing FNA cases classified as
tive result, NLR of FNA did not improve to the point so it could be “nondiagnostic”, “nonneoplastic”, and “AUS” as the negative index test
result for neoplasm, and cases classified as “benign neoplasm”,
considered informative in ruling out the malignancy diagnosis.
“SUMP”, “suspicious for malignancy”, and “malignant” as the positive
By stratifying the cases from 92 eligible studies according to index test result for neoplasm. Scenario 1 (blue arrow); scenario 2 (red
MSRSGC classification and comparing the cytology diagnosis with the arrow); scenario 4 (black arrow); scenario 1-3 (dark green arrow);
corresponding histologic follow-up diagnosis, the ROM in “nondiag- scenario 3,4 (Purple arrow). Fagan nomogram depicts the posttest
nostic”, “nonneoplastic”, “AUS”, “neoplasm-benign”, “neoplasm- probability of having a neoplasm after cytology diagnosis positive and
negative for neoplasm, for a pretest probability of 60%. Including
SUMP”, “suspicious for malignancy”, and “malignant” was 17%, 8%,
cytology, indefinite diagnosis for neoplasm (MSRSGC categories of
34%, 4%, 42%, 58%, and 91%, respectively.
“nondiagnostic” and “AUS”) resulted in decreasing FNA PLR and
In this series, the mean prevalence of malignancy was 22%. A therefore the posttest probability of neoplasm, while the not reducing
diagnosis of “suspicious for malignancy” or “malignant” increased the the likelihood of having a neoplasm after FNA negative result for
probability of having a malignancy to 91%, while a diagnosis of neoplasm
FARAHANI AND BALOCH 13

TABLE 4 Diagnostic performance of salivary gland FNA in detecting malignancy in individual primary studies

Sensitivity Specificity
Author year TP FP FN TN (%) (%) DOR PLR NLR
Correia SI 201793 15 4 0 36 100.0 90.0 251.4 8.8 0.04
Eytan DF 201739 123 29 26 273 82.6 90.4 44.5 8.6 0.19
Mikaszewski B 201794 9 11 10 70 47.4 86.4 5.7 3.5 0.61
95
Ramirez-Perez F 2017 21 25 10 116 67.7 82.3 9.7 3.8 0.39
Rajdeo RN 201727 19 0 3 78 86.4 100.0 874.7 134.0 0.15
Rohilla M 201767 25 1 7 58 78.1 98.3 207.1 46.1 0.22
Edizer DT 201644 15 10 9 211 62.5 95.5 35.2 13.8 0.39
96
Feinstein A 2016 75 10 17 189 81.5 95.0 83.4 16.2 0.19
Gudmudsson JK 201697 8 3 3 100 72.7 97.1 88.9 25.0 0.28
Mohammed Nur 201668 25 2 5 111 83.3 98.2 277.5 47.1 0.17
Rossi ED (AGH) 201639 47 3 16 341 74.6 99.1 333.9 85.5 0.26
39
Rossi ED (HUP) 2016 48 3 30 177 61.5 98.3 94.4 36.9 0.39
Ameli F 201569 14 1 3 81 82.4 98.8 378.0 67.5 0.18
Arul P 201570 25 5 6 104 80.7 95.4 86.7 17.6 0.20
Halder S 201552 7 4 3 34 70.0 89.5 19.8 6.7 0.34
98
Hipp J (quick-diff prep) 2015 39 3 4 89 90.7 96.7 289.3 27.8 0.10
Hipp J (ThinPrep) 201598 26 5 6 46 81.3 90.2 39.9 8.3 0.21
Mairembam P 201671 33 3 4 124 89.2 97.6 341.0 37.8 0.11
Naz S 201556 7 3 2 19 77.8 86.4 22.2 5.7 0.26
37
Novoa E 2016 13 3 7 74 65.0 96.1 45.8 16.7 0.36
Song HI 201599 32 4 23 289 58.2 98.6 100.5 42.6 0.42
Diaz KP 2014100 16 0 1 156 94.1 100.0 3443.0 287.8 0.08
Omhare A 201472 15 2 2 67 88.2 97.1 251.3 30.4 0.12
101
Zerpa VZ 2014 4 4 3 79 57.1 95.2 26.3 11.9 0.45
Fassnatch W 201334 4 8 5 62 44.4 88.6 6.2 3.9 0.63
Jeong WJ 201336 9 1 2 90 81.8 98.9 405.0 74.5 0.18
Kim BY 2013102 61 7 34 419 64.2 98.4 107.4 39.1 0.36
74
Pastore A 2013 26 0 3 299 89.6 100.0 4535.3 530.0 0.12
Tryggvason G 2013103 138 2 23 380 85.7 99.5 1140.0 163.7 0.14
Fakhry N 2012104 43 16 11 132 79.6 89.2 32.3 7.4 0.23
Haung YC 201242 10 2 4 44 71.4 95.7 55.0 16.4 0.30
105
Inanch HM 2012 21 4 5 77 80.8 95.1 80.9 16.4 0.20
Javadi M 2012106 11 1 8 45 57.9 97.8 61.88 26.6 0.43
Kechagias N 201258 27 1 1 49 96.4 98.0 1323.0 48.2 0.04
Nanda KDS 201274 12 2 3 39 80.0 95.1 78.0 16.4 0.21
55
Nguansangiam S 2012 13 1 3 116 81.3 99.2 502.7 95.1 0.19
Ali NS 201149 26 2 4 92 86.7 97.9 299.0 40.7 0.14
Cho HW 201143 28 0 9 178 75.7 100.0 1071.0 268.5 0.25
Piccioni LO 201138 13 1 3 123 81.3 99.2 533.0 100.8 0.19
31
Ashraf A 2010 14 1 4 81 77.8 98.8 283.5 63.8 0.22
Christensen RK 201075 44 1 9 322 83.0 99.7 1574.2 268.2 0.17
Gobic MB 201041 13 3 3 147 81.3 98.0 212.3 40.6 0.19
Stramandinoli RT 201076 16 7 7 47 69.6 87.0 15.4 5.4 0.35
29
Carrillo JF 2009 60 1 5 69 92.3 98.6 828.0 64.6 0.08
Daneshbod Y 200977 128 10 13 225 90.8 95.7 221.5 21.3 0.10
Jafari A 2009107 12 3 5 81 70.6 96.4 64.8 19.8 0.31
Zhang S 200947 11 1 5 54 68.8 98.2 118.8 37.8 0.32
63
Inohara H 2008 19 3 2 57 90.5 95.0 180.5 18.1 0.10
Jan IS 2008108 19 9 6 97 76.0 91.5 34.1 9.0 0.26
Mohammad F 2008109 21 6 14 148 60.0 96.1 37.0 15.4 0.42
Zbaren P 2008110 50 5 18 37 73.5 88.1 20.6 6.2 0.30
78
de Ru JA 2007 22 0 0 66 100.0 100.0 5985.0 131.1 0.02

(Continues)
14 FARAHANI AND BALOCH

TABLE 4 (Continued)

Sensitivity Specificity
Author year TP FP FN TN (%) (%) DOR PLR NLR
Elagoz S 200764 7 0 9 40 43.8 100.0 64.0 36.2 0.57
Lim CM 200779 8 0 2 71 80.0 100.0 486.2 37.8 0.11
Van Lierop AC 2007111 8 1 3 55 72.7 98.2 146.7 8.3 0.21
112
Upton DC 2007 20 2 1 29 95.2 93.6 290.0 5.7 0.26
Aversa S 200660 34 0 7 269 82.9 100.0 2479.4 16.7 0.36
Tan LGL 200680 3 0 0 50 100.0 100.0 707.0 42.6 0.42
Balakrishnan K (MH) 2005113 5 3 7 54 41.7 94.7 12.9 287.8 0.08
113
Balakrishnan K (RHH) 2005 15 7 4 37 79.0 84.1 19.8 30.4 0.12
Bandyopadhyay A 200581 13 1 2 43 86.7 97.7 279.5 11.9 0.45
Paris J 200548 25 5 6 97 80.7 95.1 80.8 3.9 0.63
Riley N 200582 31 3 2 61 93.9 95.3 315.2 74.5 0.18
61
Seethala RR 2005 44 2 7 122 86.3 98.4 383.4 39.1 0.36
Cohen EG 2004114 51 10 20 70 71.8 87.5 17.9 530.0 0.12
Postema RJ 200465 73 4 10 293 88.0 98.7 534.7 163.7 0.14
Sergi B 2004115 12 0 9 118 57.1 100.0 311.8 7.4 0.23
116
Stow N 2004 29 1 2 62 93.6 98.4 899.0 16.4 0.30
Contucci AM 200362 14 0 7 118 66.7 100.0 458.2 16.4 0.20
Gooden E 200235 16 8 2 61 88.9 88.4 61.0 26.6 0.43
Sengupta S 200283 28 4 4 172 87.5 97.7 301.0 48.2 0.04
117
Tasi SCS 2002 3 1 2 34 60.0 97.1 51.0 16.4 0.21
Jayaram G 200150 11 1 6 32 64.7 97.0 58.7 95.1 0.19
Costas A 200054 26 5 5 64 83.9 92.8 66.6 40.7 0.14
Santamaria J 2000118 18 6 4 161 81.8 96.4 120.8 22.8 0.19
32
Al-Khafaji BM 1998 58 11 13 64 81.7 85.3 26.0 5.6 0.21
Boccato P 199866 104 0 1 449 99.1 100.0 62 630.3 887.3 0.01
Filopoulos E 199859 37 1 1 81 97.4 98.8 2997.0 79.8 0.03
Cajulis RS 199784 23 1 2 102 92.0 99.0 1173.0 94.8 0.08
51
Christallini EG 1997 4 0 3 56 57.1 100.0 145.3 64.1 0.44
Tew S 1997119 18 0 2 109 90.0 100.0 1620.6 193.8 0.12
Atula T 199685 12 0 21 151 36.4 100.0 176.2 111.8 0.63
Atula T 199586 2 1 10 60 16.7 98.4 12.0 10.2 0.85
87
Orell SR 1995 71 1 8 228 89.9 99.6 2023.5 205.8 0.10
Jayaram G 1994120 37 2 2 99 94.9 98.0 915.8 47.9 0.05
Zurida S 199388 31 0 14 178 68.9 100.0 775.6 245.2 0.32
Chan MKM 199289 31 1 5 72 86.1 98.6 446.4 62.9 0.14
121
Frable MA 1991 51 1 2 131 96.2 99.2 3340.5 127.0 0.04
Nettle WJS 198945 20 1 5 73 80.0 98.7 292.0 59.2 0.20
Layfield LJ 198790 44 6 8 108 84.6 94.7 99.0 16.1 0.16
Lau T 1986122 8 0 7 86 53.3 100.0 196.1 92.4 0.47
123
Qizibash AH 1985 21 0 3 77 87.5 100.0 952.1 134.2 0.14
Lindberg LG 1976124 38 20 20 250 65.5 92.6 23.8 8.8 0.37
Persson PS 197391 30 2 4 166 88.2 98.8 622.5 74.1 0.12
Eneroth CM 196792 44 21 40 396 52.4 95.0 20.7 10.4 0.50

In order to avoid zero as the denominator in calculating DOR, PLR, NLR, a fixed continuity correction factor of 0.5 was added to each of TP, FP, FN, TN
cells when a zero in FP and FN cells encountered.
Abbreviations: TP, true positive; FP, false positive; FN, false negative; TN, true negative; DOR, diagnostic odds ratio; PLR, positive likelihood ratio; NLR,
negative likelihood ratio; AGH, Agostino Gemelli Hospital; HUP, Hospital of the University of Pennsylvania; MH, Monklands Hospital; RHH, Royal Hallam-
shire Hospital.

“nonneoplastic”, “neoplasm-benign”, or benign reduced the probability 4.2 | Investigating heterogeneity


of malignancy to 5%. While the posttest probability of malignancy The results of the meta-analysis of FNA accuracy were highly hetero-
with a diagnosis of “nondiagnostic”, “AUS”, “neoplasm-SUMP”, “suspi- geneous across the different studies, and various variables including
cious for malignancy”, and “malignant” was 86%. population characteristics, the setting of the study, and technical
FARAHANI AND BALOCH 15

TABLE 5 Summary diagnostic performance of salivary gland FNA in detecting malignancy in different scenarios

Sensitivity (%) Specificity (%) DOR PLR NLR


(95% CI) (95% CI) (95% CI) (95% CI) (95% CI)
Scenario 1 82.3 93.8 70.8 13.3 0.19
(79.4-84.9) (92.2-95.1) (50.7-98.9) (10.5-16.9) (0.16-0.22)
Scenario 2 81.6 96.4 116.9 22.3 0.19
(78.6-84.2) (95.2-97.2) (82.9-164.7) 17.2-29.0) (0.16–0.22)
Scenario 3 81.4 96.9 136.3 26.2 0.19
(78.4-84.1) (95.9-97.6) (95.5-194.4) (19.9-34.4) (0.16–0.22)
Scenario 4 80.5 97.9 189.5 37.8 0.20
(77.3-83.3) (97.1-98.4) (130.0-276.2) 28.1-50.6) (0.17-0.23)
Scenario 5 74.0 98.0 138.8 36.9 0.27
(71.2-77.8) (97.2-98.3) (110-7-256.9) (26.9-49.4) (0.24-0.29)
I2 index of heterogeneity (%) 68.2 75.3 78.6
(64.7-72.5) (71.3-79.4) (74.1-81.9)

Abbreviations: CI, confidence interval; DOR, diagnostic odds ratio; PLR, positive likelihood ratio; NLR, negative likelihood ratio.
Scenario 1, FNA “nonneoplastic” and “benign neoplasm” as a negative index test result for malignancy, and “nondiagnostic”, “AUS”, “SUMP”, “suspicious for
malignancy”, and “malignant” as a positive index test result for malignancy;
Scenario 2, FNA “nonneoplastic” and “benign neoplasm” as a negative index test result for malignancy, and “AUS”, “SUMP”, “suspicious for malignancy”,
and “malignant” as a positive index test result for malignancy;
Scenario 3, FNA “nonneoplastic” and “benign neoplasm” as a negative index test result for malignancy, and “SUMP”, “suspicious for malignancy”, and
“malignant” as a positive index test result for malignancy;
Scenario 4, FNA “nonneoplastic” and “benign neoplasm” as a negative index test result for malignancy, and “suspicious for malignancy” and “malignant” as
a positive index test result for malignancy;
Scenario 5, FNA “nonneoplastic”, “benign neoplasm”, “AUS”, and “SUMP” as a negative index test result for malignancy, and “suspicious for malignancy”
and “malignant” as a positive index test result for malignancy.

aspects involved in cytology material collection and preparation were the cytological evaluation of salivary gland neoplasms. The primary
shown to have a measurable effect on FNA accuracy. applications of these techniques are in differentiating between the
Presence of ROSE was associated with increased sensitivity and benign neoplasms such as basal cell adenoma and pleomorphic ade-
specificity of salivary gland FNA in differentiating neoplasm from a noma and pleomorphic adenoma and the malignant tumors such as
nonneoplastic lesion and a malignant from benign. Moreover, the rate adenoid cystic carcinoma and mucoepidermoid carcinoma; or in distin-
of nondiagnostic FNA was significantly lower in the studies in which guishing between the lymphoid hyperplasia and low-grade lymphoma
ROSE was available compared to the rest of the studies (4% vs 10%, in lymphocyte-rich specimens.129
P-value: .004) it. Previous studies have demonstrated that ROSE can The sensitivity and specificity of salivary gland FNA in differenti-
reduce the rate of nondiagnostic samples by 50% by evaluating the ating malignant and benign lesions were significantly higher in the
sample adequacy, especially in cases where a less experienced pathol- studies in which FNA was performed and evaluated in a cancer refer-
ogist or clinician are performing the FNA.125,126 ral center. On the other hand, the prevalence of malignancy was
Of the 92 included studies, eight only included FNA cases per-
higher in these studies compared to the rest of the studies, and this
formed by palpation and seven solely enrolled US-guided FNA cases.
appeared to be associated with higher sensitivity. This phenomenon
The sensitivity, specificity, and accuracy of FNA in differentiating
could be attributable to partial verification bias. Only a fraction of
malignant and benign tumors were significantly lower in both sub-
FNA cases with a diagnosis of a nonneoplastic lesion (benign) received
groups compared to the rest of the studies. This finding was in con-
the reference test (surgical resection and histologic diagnosis); this
trast with the studies which argued that performing FNA under
results in an underestimation of the false negative rate and overesti-
ultrasound-guidance increases the FNA yield by ensuring the place-
mation of sensitivity and malignancy prevalence. The higher FNA
ment of biopsy needle within the viable and cellular components of a
accuracy in this subgroup could also be attributable to the fact that a
salivary gland lesion leading to a reduction in the rate of nondiagnostic
large number of cases of salivary gland neoplasms are encountered in
sample and improve the overall accuracy of FNA.127,128 The lower
cancer referral centers. This most likely is also associated with a higher
accuracy of ultrasound-guided FNA comparing to palpation-guided
level of operator experience in performing and interpretation of sali-
FNA could be attributable to the difference in referral patterns, study
vary gland FNA leading to overall improvement of FNA accuracy.
population, malignancy prevalence, the histological complexity of
studied lesions, and ROSE. Further investigation of the concurrent
impact of mentioned factors on the accuracy of FNA was not possible
4.3 | Limitations
due to the limited numbers of studies in this subgroup.
Utilizing ancillary studies in the selective FNA cases showed to This study has many limitations. First, the majority of the included
increase the sensitivity and specificity of FNA in detecting neoplasms studies were retrospective studies, and in general, these are prone to
of the salivary gland. It has been demonstrated that various ancillary various biases especially regarding consecutive enrollment of subjects,
techniques including immunohistochemistry, fluorescence in-situ variability in execution and evaluation, and blindness to the results of
hybridization (FISH), reverse transcription-polymerase chain reaction the index test when evaluating the reference test and vice versa. The
(RT-PCR), and next-generation sequencing can prove to be useful in second limitation was the presence of partial verification bias in
16 FARAHANI AND BALOCH

FIGURE 6 (A, B) FNA sensitivity and specificity forest plots in distinguishing between malignant and benign lesion of salivary glands. The forest
plot demonstrates a high level of intra and inter-study variations

almost all selected studies, which can result in overestimation of dis- neoplasm/malignancy with histologic follow-up showed that FNA sen-
ease prevalence and sensitivity, and underestimation of specificity. sitivity was overestimated by 8%-10%. However, this method is
By using an interactive web-based tool, first, we demonstrated designed initially for the tests with a binary outcome and classifying
that the chance of not having the histologic follow-up was not depen- FNA cases as indeterminate (eg, “nondiagnostic”, “AUS”, or “neo-
dent on the neoplastic or malignant nature of the examined cases. plasm-SUMP”) reduces the chance of having false negative FNA cases
Second, the result of corrected sensitivity and specificity calculated with a negative result for neoplasm/malignancy on histologic follow-
using the ratio of FNA cases with a positive and negative result for up. Therefore, it could be concluded that although the partial
FARAHANI AND BALOCH 17

FIGURE 7 HSROC curve of FNA in detecting salivary gland


malignancy when only including the FNA cases with a definite
diagnosis for malignancy (the MSRSGC categories of “nonneoplastic”,
“neoplasm-benign”, “suspicious for malignancy”, “malignant”) in the
analysis. Circles indicate primary studies sensitivity and specificity.
The size of circles corresponds with study sample size. The solid green
line represents FNA summary sensitivity, and specificity line with 95%
prediction region (dotted green line) and the red square indicates the FIGURE 8 Fagan nomogram for detecting malignancy: scenario
FNA summary points of sensitivity and specificity with 95% 1, identifying the FNA cases classified as “nonneoplastic” and “benign
confidence region (dotted orange line). The 95% confidence region neoplasm” as the negative index test result for malignancy, and
shows the region the real value of FNA sensitivity and specificity “nondiagnostic”, “AUS”, “SUMP”, “suspicious for malignancy”, and
would expect to lie, based on the observed data. The 95% prediction “malignant” as the positive index test result for malignancy; scenario
region depicts the region the results of a future study would expect to 2, excluding “nondiagnostic” FNA cases from the analysis and recognizing
be located. As it could be expected due to the high level of the cases with the diagnosis of “nonneoplastic” and “benign neoplasm” as
heterogeneity between the studies, the 95% prediction region is the negative index test result for malignancy, and “AUS”, “SUMP”,
much larger than the 95% confidence region “suspicious for malignancy”, and “malignant” as the positive index test result
for malignancy; scenario 3, excluding “nondiagnostic” and “AUS” cases from
verification bias could overestimate the sensitivity, the true value of the analysis and considering the cases with diagnosis of “nonneoplastic”
sensitivity would lie somewhere between the value corrected and and “benign neoplasm” as the negative index test result for malignancy, and
“SUMP”, “suspicious for malignancy”, and “malignant” as the positive index
uncorrected sensitivity.
test result for malignancy; scenario 4, including FNA cases with the definite
The third limitation was the high level of heterogeneity across the
diagnosis for malignancy in the analysis; identifying the cases with the
included studies. Although various variables were demonstrated to diagnosis of “nonneoplastic” and “benign neoplasm” as the negative index
have a significant effect on FNA sensitivity and specificity in meta- test result for malignancy, and “suspicious for malignancy” and “malignant”
regression analysis; subdividing the studies into smaller subgroups as the positive index test result for malignancy; scenario 5, including FNA
based on these variables did not result in generating a more homoge- cases with the diagnosis of “nonneoplastic”, “benign neoplasm”, “AUS”, and
“SUMP” as the negative index test result for malignancy, and “suspicious for
neous cohort. According to Cochrane guideline for meta-analysis, the
malignancy” and “malignant” as the positive index test result for malignancy.
available statistical tests which are used to evaluate the level of het- Scenario 1 (blue arrow); scenario 2 (red arrow); scenario 3 (light green
erogeneity are designed for meta-analysis of treatment effect and arrow); scenario 5 (black arrow); scenario 1-4 (dark green arrow); scenario
they generally overestimate the level of heterogeneity in meta- 4,5 (purple arrow). Fagan nomogram depicts the posttest probability of
analysis of diagnostic studies since they cannot take into account the having a malignancy after cytology diagnosis positive and negative for
malignancy, for a pretest probability of salivary gland malignancy of 22%.
correlation between sensitivity and specificity.
Including cytology, indefinite diagnosis for malignancy (MSRSGC categories
In conclusion, this meta-analysis confirms that FNA has proven to
of “nondiagnostic”, “AUS”, “neoplasm-SUMP”) resulted in decreasing FNA
be a useful tool in the diagnosis of salivary gland lesions with appre- PLR and therefore the posttest probability of malignancy without any
ciable sensitivity, specificity, and diagnostic accuracy. The retrospec- significant increase in NLR or reduction in posttest probability of
tive application of diagnostic categories of MSRSGC to the existing malignancy after a negative FNA result for malignancy
18 FARAHANI AND BALOCH

TABLE 6 Diagnostic performance of salivary gland FNA in detecting malignancy in different subgroups

Sensitivity (%) Specificity (%) DOR PLR NLR


Variable (95% CI) (95% CI) (95% CI) (95% CI) (95% CI)
Malignancy prevalence
≤0.25 77.1 98.5 223.7 52.0 0.23
(72.5-81.1) (97.8-99.0) (128.6-389.1) (33.5-80.8) (0.2-0.3)
>0.25 84.7 96.3 144.7 23.0 0.16
(80.8-87.9) (94.8-97.4) (87.8-238.4) (16.1-32.9) (0.13-0.20)
Setting
Tertiary cancer centers 86.4 97.8 282.7 39.2 0.14
(75.5-93.0) (96.3-98.9) (98.2-813.9) (21.4-69.9) (0.07-0.26)
Private clinics and general hospitals 79.8 97.9 180.8 37.5 0.21
(76.4-82.6) (97.1-98.5) (121.0-270.2) (27.2-51.8) (0.17-0.24)
Ancillary studies
Yes 86.4 97.4 238.2 33.4 0.14
(74.4-93.2) (93.0-99.1) (46.0-1232.9) (11.3-98.4) (0.07-0.28)
No or not reported 80.1 98.0 194.4 39.5 0.20
(76.8-83.1) (97.2-98.5) (131.1-288.2) (28.8-54.1) (0.17-0.24)
ROSE
Yes 83.2 99.0 496.1 84.4 0.17
(72.5-90.2) (97.4-99.6) (144.3-1705.8) (31.5-226.5) (0.10-0.29)
No or not reported 80.1 97.7 167.9 34.2 0.20
(76.7-83.1) (96.8-98.3) (113.3-248.8) (25.1-46.6) (0.17-0.23)

Abbreviations: CI, confidence interval; DOR, diagnostic odds ratio; PLR, positive likelihood ratio; NLR, negative likelihood ratio.

literature proves that this classification scheme will be helpful in stan- 7. Layfield LJ. Fine-needle aspiration in the diagnosis of head and neck
dardization and unification of salivary gland FNA results to increase lesions: a review and discussion of problems in differential diagnosis.
Diagn Cytopathol. 2007;35(12):798-805.
the FNA reliability and reproducibility. The standardization in report- 8. Layfield LJ, Gopez E, Hirschowitz S. Cost efficiency analysis for fine-
ing the results of the salivary glands cytology will result in improving needle aspiration in the workup of parotid and submandibular gland
communication between pathologists and treating clinicians and facili- nodules. Diagn Cytopathol. 2006;34(11):734-738.
9. Liu CC, Jethwa AR, Khariwala SS, Johnson J, Sensitivity SJJ. Specific-
tate the intra and inter-institution data collection and quality improve-
ity, and posttest probability of parotid fine-needle aspiration: a sys-
ment measures. tematic review and meta-analysis. Otolaryngol Head Neck Surg. 2016;
154(1):9-23.
10. Colella G, Cannavale R, Flamminio F, Foschini MP. Fine-needle
aspiration cytology of salivary gland lesions: a systematic review.
DIS CLOSUR E OF INTERESTS
J Oral Maxillofac Surg Med Pathol. 2010;68(9):2146-2153.
Authors have nothing to disclose. 11. Griffith CC, Schmitt AC, Pantanowitz L, Monaco SE. A pattern-based
risk-stratification scheme for salivary gland cytology: a multi-institu-
tional, interobserver variability study to determine applicability. Can-
ORCID
cer Cytopathol. 2017;125(10):776-785.
Sahar J Farahani http://orcid.org/0000-0003-0845-4659 12. Faquin WC, Rossi ED, Baloch ZW, et al. The Milan System For Report-
ing Salivary Gland Cytopathology. Cham: Springer Nature; 2018.
Zubair Baloch http://orcid.org/0000-0001-9342-3069
13. Rossi ED, Baloch ZW, Pusztaszeri M, Faquin WC. The Milan system
for reporting salivary gland cytopathology (MSRSGC): an ASC-IAC-
sponsored system for reporting salivary gland fine-needle aspiration.
RE FE R ENC E S
Acta Cytol. 2018;62:157-165.
1. Mohan H, Tahlan A, Mundi I, Punia RPS, Dass A. Non-neoplastic sali- 14. Deeks JJ, Wisniewski S, Davenport C. Chapter 4: guide to the
vary gland lesions: a 15-year study. Eur Arch Otorhinolaryngol. 2011; contents of a Cochrane diagnostic test accuracy protocol. In:
268(8):1187-1190. Deeks JJ, Bossuyt PM, Gatsonis C, eds. Cochrane Handbook for Sys-
2. Sandhu VK, Sharma U, Singh N, Puri A. Cytological spectrum of tematic Reviews of Diagnostic Test Accuracy Version 1.0.0. The Cochrane
salivary gland lesions and their correlation with epidemiological Collaboration; 2013. Available from http://handbook.cochrane.org.
parameters. J oral maxillofacl pathol. 2017;21(2):203-210. 15. McInnes MDF, Moher D, Thombs BD, et al. Preferred reporting items
3. Bussu F, Parrilla C, Rizzo D, Almadori G, Paludetti G, Galli J. Clinical for a systematic review and meta-analysis of diagnostic test accuracy
approach and treatment of benign and malignant parotid masses, per- studies: the PRISMA-DTA statement. JAMA. 2018 Jan 23;319(4):
sonal experience. Acta Otorhinolaryngol Ital. 2011;31(3):135-143. 388-396.
4. Sood S, McGurk M, Vaz F. Management of salivary gland tumours: 16. Barendregt JJ, Doi SA, Lee YY, Norman RE, Vos T. Meta-analysis of
United Kingdom national multidisciplinary guidelines. J Laryngol Otol. prevalence. J Epidemiol Community Health. 2013;67(11):974-978.
2016;130(S2):S142-S149. 17. Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM,
5. Wang X, Luo Y, Li M, Yan H, Sun M, Fan T. Management of salivary Zwinderman AH. Bivariate analysis of sensitivity and specificity
gland carcinomas - a review. Oncotarget. 2017;8(3):3946-3956. produces informative summary measures in diagnostic reviews. J Clin
6. Schmidt RL, Hall BJ, Wilson AR, Layfield LJ. A systematic review and Epidemiol. 2005;58(10):982-990.
meta-analysis of the diagnostic accuracy of fine-needle aspiration 18. Rutter CM, Gatsonis CA. A hierarchical regression approach to meta-
cytology for parotid gland lesions. Am J Clin Pathol. 2011;136(1): analysis of diagnostic test accuracy evaluations. Stat Med. 2001;
45-59. 20(19):2865-2884.
FARAHANI AND BALOCH 19

19. Harbord RM, Deeks JJ, Egger M, Whiting P, Sterne JAC. A unification 41. Gobic MB, Pedisic D, Bekafigo IS, et al. Fine needle aspiration cytol-
of models for meta-analysis of diagnostic accuracy studies. Biostatis- ogy in the evaluation of parotid gland tumors. Coll Antropol. 2010;
tics. 2007;8(2):239-251. 34(2):345-348.
20. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publica- 42. Huang YC, Wu CT, Lin G, Chuang WY, Yeow KM, Wan YL. Compari-
tion bias and other sample size effects in systematic reviews of diag- son of ultrasonographically guided fine-needle aspiration and core
nostic test accuracy was assessed. J Clin Epidemiol. 2005;58(9): needle biopsy in the diagnosis of parotid masses. J Clin Ultrasound.
882-893. 2012;40(4):189-194.
21. Bürkner PC, Doebler P. Testing for publication bias in diagnostic 43. Cho HW, Kim J, Choi J, et al. Sonographically guided fine-needle
meta-analysis: a simulation study. Stat Med. 2014;33(18):3061-3077. aspiration biopsy of major salivary gland masses: a review of
22. Van Enst WA, Ochodo E, Scholten RJ, Hooft L, Leeflang MM. Investi- 245 cases. AJR Am J Roentgenol. 2011;196(5):1160-1163.
gation of publication bias in meta-analyses of diagnostic test accu- 44. Edizer DT. Role of Fine-Needle Aspiration Biopsy in the Manage-
racy: a meta-epidemiological study. BMC Med Res Methodol. 2014; ment of Salivary Gland Masses. Turk Arch Otorhinolaryngol. 2016;
14:70. 54(3):105-111.
23. McInnes MD, Bossuyt PM. Pitfalls of systematic reviews and meta- 45. Nettle WJ, Orell SR. Fine needle aspiration in the diagnosis of sali-
analyses in imaging research. Radiology. 2015;277(1):13-21. vary gland lesions. Aust N Z J Surg. 1989;59(1):47-51.
24. Kohn MA, Carpenter CR, Newman TB. Understanding the direction 46. Song IH, Song JS, Sung CO, et al. Accuracy of Core needle biopsy
of bias in studies of diagnostic test accuracy. Acad Emerg Med. 2013; versus fine needle aspiration cytology for diagnosing salivary gland
20(11):1194-1206. tumors. J Pathol Transl Med. 2015;49(2):136-143.
25. Kosinski AS, Barnhart HX. A global sensitivity analysis of perfor- 47. Zhang S, Bao R, Bagby J, Abreo F. Fine needle aspiration of salivary
mance of a medical diagnostic test when verification bias is present. glands: 5-year experience from a single academic center. Acta Cytol.
Stat Med. 2003;22(17):2711-2721. 2009;53(4):375-382.
26. Petscavage JM, Richardson ML, Carr RB. Verification bias an under- 48. Paris J, Facon F, Pascal T, Chrestian MA, Moulin G, Zanaret M. Pre-
recognized source of error in assessing the efficacy of medical imag- operative diagnostic values of fine-needle cytology and MRI in
ing. Acad Radiol. 2011;18(3):343-346. parotid gland tumors. Eur Arch Otorhinolaryngol. 2005;262(1):27-31.
27. Rajdeo RN, Shrivastava AC, Bajaj J, Shrikhande AV, Rajdeo RN. Clini- 49. Ali NS, Akhtar S, Junaid M, Awan S, Aftab K. Diagnostic accuracy of
copathological study of salivary gland tumors: an observation in ter- fine needle aspiration cytology in parotid lesions. ISRN Surg. 2011;
tiary hospital of Central India. Int J Res Med Sci. 2017;3(7):6. 2011:721525.
28. Novoa E, Gürtler N, Arnoux A, Kraft M. Diagnostic value of core nee- 50. Jayaram G, Dashini M. Evaluation of fine needle aspiration cytology
dle biopsy and fine-needle aspiration in salivary gland lesions. Head of salivary glands: an analysis of 141 cases. Malays J Pathol. 2001;
Neck. 2016;38:E346-E352. 23(2):93-100.
29. Carrillo JF, Ramirez R, Flores L, et al. Diagnostic accuracy of fine nee- 51. Cristallini EG, Ascani S, Farabi R, et al. Fine needle aspiration biopsy
dle aspiration biopsy in preoperative diagnosis of patients with of salivary gland, 1985-1995. Acta Cytol. 1997;41(5):1421-1425.
parotid gland masses. J Surg Oncol. 2009;100(2):133-138. 52. Haldar S, Mandalia U, Skelton E, et al. Diagnostic investigation of
30. Mikaszewski B, Markiet K, Smugała A, Stodulski D, Szurowska E, parotid neoplasms: a 16-year experience of freehand fine needle
Stankiewicz C. Diffusion- and perfusion-weighted magnetic reso- aspiration cytology and ultrasound-guided core needle biopsy. Int J
nance imaging—an alternative to fine needle biopsy or only an Oral Maxillofac Surg. 2015;44(2):151-157.
adjunct test in preoperative differential diagnostics of malignant and 53. Orell SR, Nettle WJS. Fine needle aspiration biopsy of salivary gland
benign parotid tumors? J Oral Maxillofac Surg. 2017;75(10):2248- tumours. Problems and pitfalls. Pathology. 1988;20(4):332-337.
2253. 54. Costas A, Castro P, Martín-Granizo R, Monje F, Marrón C, Amigo A.
31. Ashraf A, Shaikh AS, Kamal F, Sarfraz R, Bukhari MH. Diagnostic reli- Fine needle aspiration biopsy (FNAB) for lesions of the salivary
ability of FNAC for salivary gland swellings: a comparative study. glands. Br J Oral Maxillofac Surg. 2000;38(5):539-542.
Diagn Cytopathol. 2010;38(7):499-504. 55. Nguansangiam S, Jesdapatarakul S, Dhanarak N, Sosrisakorn K.
32. Al-Khafaji BM, Nestok BR, Katz RL. Fine-needle aspiration of Accuracy of fine needle aspiration cytology of salivary gland lesions:
154 parotid masses with histologic correlation: ten-year experience routine diagnostic experience in Bangkok, Thailand. Asia Pacific J
at the University of Texas M. D. Anderson cancer center. Cancer. Cancer Preven. 2012;13(4):1583-1588.
1998;84(3):153-159. 56. Naz S, Hashmi AA, Khurshid A, et al. Diagnostic role of fine needle
33. Eom HJ, Lee JH, Ko MS, et al. Comparison of fine-needle aspiration aspiration cytology (FNAC) in the evaluation of salivary gland
and core needle biopsy under ultrasonographic guidance for detect- swelling: an institutional experience. BMC Res Notes. 2015;8:101.
ing malignancy and for the tissue-specific diagnosis of salivary gland 57. Mamikunian C, Gatti WM, Reyes CV. Subcutaneous blastomycosis:
tumors. Am J Neuroradiol. 2015;36(6):1188-1193. diagnosis by fine-needle aspiration cytology. Otolaryngol Head Neck
34. Fassnacht W, Schmitz S, Weynand B, Marbaix E, Duprez T, Surg. 1989;101(5):607-610.
Hamoir M. Pitfalls in preoperative work-up of parotid gland tumours: 58. Kechagias N, Ntomouchtsis A, Valeri R, et al. Fine-needle aspiration
10-year series. B-ENT. 2013;9(2):83-88. cytology of salivary gland tumours: a 10-year retrospective analysis.
35. Gooden E, Witterick IJ, Hacker D, Rosen IB, Freeman JL. Parotid Oral Maxillofac Surg. 2012;16(1):35-40.
gland tumours in 255 consecutive patients: Mount Sinai Hospital's 59. Filopoulos E, Angeli S, Daskalopoulou D, Kelessis N, Vassilopoulos P.
quality assurance review. J Otolaryngol. 2002;31(6):351-354. Pre-operative evaluation of parotid tumours by fine needle biopsy.
36. Jeong WJ, Park SJ, Cha W, Sung MW, Kim KH, Ahn SH. Fine needle Eur J Surg Oncol. 1998;24(3):180-183.
aspiration of parotid tumors: diagnostic utility from a clinical perspec- 60. Aversa S, Ondolo C, Bollito E, Fadda G, Conticello S. Preoperative
tive. J Oral Maxillofac Surg Med Pathol. 2013;71(7):1278-1282. cytology in the management of parotid neoplasms. Am J Otolaryngol.
37. Novoa E, Gurtler N, Arnoux A, Kraft M. Diagnostic value of core nee- 2006;27(2):96-100.
dle biopsy and fine-needle aspiration in salivary gland lesions. Head 61. Seethala RR, LiVolsi VA, Baloch ZW. Relative accuracy of fine-needle
Neck. 2016;38(Suppl 1):E346-E352. aspiration and frozen section in the diagnosis of lesions of the
38. Piccioni LO, Fabiano B, Gemma M, Sarandria D, Bussi M. Fine-needle parotid gland. Head Neck. 2005;27(3):217-223.
aspiration cytology in the diagnosis of parotid lesions. Acta Otorhino- 62. Contucci AM, Corina L, Sergi B, Fadda G, Paludetti G. Correlation
laryngol Ital. 2011;31(1):1-4. between fine needle aspiration biopsy and histologic findings in
39. Rossi ED, Wong LQ, Bizzarro T, et al. The impact of FNAC in the parotid masses. Personal experience. Acta Otorhinolaryngol Ital.
management of salivary gland lesions: institutional experiences lead- 2003;23(4):314-318.
ing to a risk-based classification scheme. Cancer Cytopathol. 2016; 63. Inohara H, Akahani S, Yamamoto Y, et al. The role of fine-needle
124(6):388-396. aspiration cytology and magnetic resonance imaging in the manage-
40. Eytan DF, Yin LX, Maleki Z, et al. Utility of preoperative fine needle ment of parotid mass lesions. Acta Otolaryngol. 2008;128(10):1152-
aspiration in parotid lesions. Laryngoscope. 2018;128(2):398-402. 1158.
20 FARAHANI AND BALOCH

64. Elagoz S, Gulluoglu M, Yilmazbayhan D, Ozer H, Arslan I. The value 85. Atula T, Greenman R, Laippala P, Klemi PJ. Fine-needle aspiration
of fine-needle aspiration cytology in salivary gland lesions, biopsy in the diagnosis of parotid gland lesions: evaluation of
1994-2004. ORL. 2007;69(1):51-56. 438 biopsies. Diagn Cytopathol. 1996;15(3):185-190.
65. Postema RJ, van Velthuysen ML, van den Brekel MW, Balm AJ, 86. Atula T, Grenman R, Laippala P, Klemi PJ. Fine-needle aspiration
Peterse JL. Accuracy of fine-needle aspiration cytology of salivary cytology of submandibular gland lesions. J Laryngol Otol. 1995;
gland lesions in the Netherlands cancer institute. Head Neck. 2004; 109(9):853-858.
26(5):418-424. 87. Orell SR. Diagnostic difficulties in the interpretation of fine needle
66. Boccato P, Altavilla G, Blandamura S. Fine needle aspiration biopsy aspirates of salivary gland lesions: the problem revisited. Cytopathol-
of salivary gland lesions: a reappraisal of pitfalls and problems. Acta ogy. 1995;6(5):285-300.
Cytol. 1998;42(4):888-898. 88. Zurrida S, Alasio L, Tradati N, Bartoli C, Chiesa F, Pilotti S. Fine-
67. Rohilla M, Singh P, Rajwanshi A, et al. Three-year cytohistological needle aspiration of parotid masses. Cancer. 1993;72(8):2306-2311.
correlation of salivary gland FNA cytology at a tertiary center with 89. Chan MKM, McGuire LJ, King W, Li AKC, Lee JCK. Cytodiagnosis of
the application of the Milan system for risk stratification. Cancer 112 salivary gland-lesions: correlation with histologic and frozen
Cytopathol. 2017;125(10):767-775. section diagnosis. Acta Cytol. 1992;36(3):353-363.
68. Mohammed Nur M, Murphy M. Adequacy and accuracy of salivary 90. Layfield LJ, Tan P, Glasgow BJ. Fine-needle aspiration of salivary
gland fine needle aspiration cytology. Ir J Med Sci. 2016;185(3): gland lesions. Comparison with frozen sections and histologic find-
711-716. ings. Arch Pathol Lab Med. 1987;111(4):346-353.
69. Ameli F, Baharoom A, Md Isa N, Noor Akmal S. Diagnostic challenges 91. Persson PS, Zettergren L. Cytologic diagnosis of salivary gland
in fine needle aspiration cytology of salivary gland lesions. Malays J tumors by aspiration biopsy. Acta Cytol. 1973;17(4):351-354.
Pathol. 2015;37(1):11-18. 92. Eneroth CM, Franzén S, Zajicek J. Cytologic diagnosis on aspirate
70. Arul P, Akshatha C, Masilamani S, Jonathan S. Diagnosis of salivary from 1000 salivary-gland Tumours. Acta Otolaryngol. 1967;
gland lesions by fine needle aspiration cytology and its Histopatho- 63(sup224):168-172.
93. Correia-Sa I, Correia-Sa M, Costa-Ferreira P, Silva A, Marques M.
logical correlation in a tertiary Care Center of Southern India. J Clin
Fine-needle aspiration cytology (FNAC): is it useful in preoperative
Diagn Res. 2015;9(6):EC07-EC10.
diagnosis of parotid gland lesions? Acta Chir Belg. 2017;117(2):
71. Mairembam P, Jay A, Beale T, et al. Salivary gland FNA cytology: role
110-114.
as a triage tool and an approach to pitfalls in cytomorphology. Cyto-
94. Mikaszewski B, Markiet K, Smugała A, Stodulski D, Szurowska E,
pathology. 2016;27(2):91-96.
Stankiewicz C. Diffusion- and perfusion-weighted magnetic reso-
72. Omhare A, Singh SK, Nigam JS, Sharma A. Cytohistopathological
nance imaging—an alternative to fine needle biopsy or only an
study of salivary gland lesions in bundelkhand region, Uttar Pradesh,
adjunct test in preoperative differential diagnostics of malignant and
India. Pathol Res Int. 2014;2014:804265.
benign parotid tumors? J Oral Maxillofac Surg. 2017;75(10):2248-
73. Pastore A, Borin M, Malagutti N, et al. Preoperative assessment of
2253.
salivary gland neoplasms with fine needle aspiration cytology and
95. Ramirez-Perez F, Gonzalez-Garcia R, Hernandez-Vila C, Monje-Gil F,
echography: a retrospective analysis of 357 cases. Int J Immunopathol
Ruiz-Laza L. Is fine-needle aspiration a reliable tool in the diagnosis
Pharmacol. 2013;26(4):965-971.
of malignant salivary gland tumors? J Craniomaxillofac Surg. 2017;
74. Singh Nanda KD, Mehta A, Nanda J. Fine-needle aspiration cytology:
45(7):1074-1077.
a reliable tool in the diagnosis of salivary gland lesions. J Oral Pathol
96. Feinstein AJ, Alonso J, Yang SE, John MS. Diagnostic accuracy of
Med. 2012;41(1):106-112.
fine-needle aspiration for parotid and submandibular gland lesions.
75. Christensen RK, Bjorndal K, Godballe C, Krogdahl A. Value of fine-
Otolaryngol Head Neck Surg. 2016;155(3):431-436.
needle aspiration biopsy of salivary gland lesions. Head Neck. 2010;
97. Gudmundsson JK, Ajan A, Abtahi J. The accuracy of fine-needle aspi-
32(1):104-108.
ration cytology for diagnosis of parotid gland masses: a clinicopatho-
76. Stramandinoli RT, Sassi LM, Pedruzzi PA, et al. Accuracy, sensitivity
logical study of 114 patients. J Appl Oral Sci: revista FOB. 2016;24(6):
and specificity of fine needle aspiration biopsy in salivary gland
561-567.
tumours: a retrospective study. Med Oral Patol Oral Cir Bucal. 2010; 98. Hipp J, Lee B, Spector ME, Jing X. Diagnostic yield of ThinPrep prep-
15(1):e32-e37. aration in the assessment of fine-needle aspiration biopsy of salivary
77. Daneshbod Y, Daneshbod K, Khademi B. Diagnostic difficulties in gland neoplasms. Diagn Cytopathol. 2015;43(2):98-104.
the interpretation of fine needle aspirate samples in salivary lesions: 99. Song IH, Song JS, Sung CO, et al. Accuracy of Core needle biopsy
diagnostic pitfalls revisited. Acta Cytol. 2009;53(1):53-70. versus fine needle aspiration cytology for diagnosing salivary gland
78. de Ru JA, van Leeuwen MS, van Benthem PP, Velthuis BK, Sie- tumors. J pathol Transl Med. 2015;49(2):136-143.
Go DM, Hordijk GJ. Do magnetic resonance imaging and ultrasound 100. Diaz KP, Gerhard R, Domingues RB, et al. High diagnostic accuracy
add anything to the preoperative workup of parotid gland tumors? and reproducibility of fine-needle aspiration cytology for diagnosing
J Oral Maxillofac Surg Med Pathol. 2007;65(5):945-952. salivary gland tumors: cytohistologic correlation in 182 cases. Oral
79. Lim CM, They J, Loh KS, Chao SS, Lim LH, Tan LK. Role of fine- Surg Oral Med Oral Pathol Oral Radiol. 2014;118(2):226-235.
needle aspiration cytology in the evaluation of parotid tumours. ANZ 101. Zerpa Zerpa V, Cuesta Gonzáles MT, Agostini Porras G, Marcano
J Surg. 2007;77(9):742-744. Acuña M, Estellés Ferriol E, Dalmau Galofre J. Diagnostic accuracy of
80. Tan LGL, Khoo MLC. Accuracy of fine needle aspiration cytology and fine needle aspiration cytology in parotid Tumours. Acta Otorrinolar-
frozen section histopathology for lesions of the major salivary glands. ingol Esp. 2014;65(3):157-161.
Annal Academy Med Singapore. 2006;35(4):242-248. 102. Kim BY, Hyeon J, Ryu G, et al. Diagnostic accuracy of fine needle
81. Bandyopadhyay A, Das TK, Raha K, Hati GC, Mitra PK, Dasgupta A. aspiration cytology for high-grade salivary gland tumors. Ann Surg
A study of fine needle aspiration cytology of salivary gland lesions Oncol. 2013;20(7):2380-2387.
with histopathological corroboration. J Indian Med Assoc. 2005; 103. Tryggvason G, Gailey MP, Hulstein SL, et al. Accuracy of fine-needle
103(6):312-316. aspiration and imaging in the preoperative workup of salivary gland
82. Riley N, Allison R, Stevenson S. Fine-needle aspiration cytology in mass lesions treated surgically. Laryngoscope. 2013;123(1):158-163.
parotid masses: our experience in Canterbury, New Zealand. ANZ J 104. Fakhry N, Antonini F, Michel J, et al. Fine-needle aspiration cytology
Surg. 2005;75(3):144-146. in the management of parotid masses: evaluation of 249 patients.
83. Sengupta S, Roy A, Mallick MG, et al. FNAC of salivary glands. Indian Eur Ann Otorhinolaryngol Head Neck Dis. 2012;129(3):131-135.
J Otolaryngol Head Neck Surg. 2002;54(3):184-188. 105. İnançlı HM, Kanmaz MA, Ural A, Dilek GB. Fine needle aspiration
84. Cajulis RS, Gokaslan ST, Yu GH, Frias-Hidvegi D. Fine needle aspira- biopsy: in the diagnosis of salivary gland neoplasms compared with
tion biopsy of the salivary glands: a five-year experience with empha- histopathology. Indian J Otolaryngol Head Neck Surg. 2013;65(Suppl
sis on diagnostic pitfalls. Acta Cytol. 1997;41(5):1412-1420. 1):121-125.
FARAHANI AND BALOCH 21

106. Javadi M, Asghari A, Hassannia F. Value of fine-needle aspiration 120. Jayaram G, Verma AK, Sood N, Khurana N. Fine needle aspiration
cytology in the evaluation of parotid tumors. Indian J Otolaryngol cytology of salivary gland lesions. J Oral Pathol Med. 1994;23(6):
Head Neck Surg. 2012;64(3):257-260. 256-261.
107. Jafari A, Royer B, Lefevre M, Corlieu P, Perie S, St Guily JL. Value of 121. Frable MAS, Frable WJ. Fine-needle aspiration biopsy of salivary
the cytological diagnosis in the treatment of parotid tumors. Otolar- glands. Laryngoscope. 1991;101(3):245-249.
yngol Head Neck Surg. 2009;140(3):381-385. 122. Lau T, Balle VHVH, Bretau P. Fine needle aspiration biopsy in sali-
108. Jan IS, Chung PF, Weng MH, et al. Analysis of fine-needle aspiration vary gland tumours. Clin Otolaryngol Allied Sci. 1986;11(2):75-77.
cytology of the salivary gland. J Formos Med Assoc. 2008;107(5): 123. Qizilbash AH, Sianos J, Young JE, Archibald SD. Fine needle aspira-
364-370. tion biopsy cytology of major salivary glands. Acta Cytol. 1985;29(4):
109. Mohammed F, Asaria J, Payne RJ, Freeman JL. Retrospective review 503-512.
of 242 consecutive patients treated surgically for parotid gland 124. Lindberg LG, Akerman M. Aspiration cytology of salivary gland
tumours. Otolaryngol Head Neck Surg. 2008;37(3):340-346. tumors: diagnostic experience from six years of routine laboratory
110. Zbaren P, Guelat D, Loosli H, Stauffer E. Parotid tumors: fine-needle work. Laryngoscope. 1976;86(4):584-594.
aspiration and/or frozen section. Otolaryngol Head Neck Surg. 2008; 125. Witt BL, Schmidt RL. Rapid onsite evaluation improves the adequacy
139(6):811-815. of fine-needle aspiration for thyroid lesions: a systematic review and
111. Van Lierop AC, Fagan JJ. Parotidectomy in Cape Town--a review of meta-analysis. Thyroid. 2013;23(4):428-435.
pathology and management. S Afr J Surg. 2007;45(3):96-98. 126. Ghofrani M, Beckman D, Rimm DL. The value of onsite adequacy
100, 102-103. assessment of thyroid fine-needle aspirations is a function of opera-
112. Upton DC, McNamar JP, Connor NP, Harari PM, Hartig GK. Paroti- tor experience. Cancer. 2006;108(2):110-113.
dectomy: ten-year review of 237 cases at a single institution. Otolar- 127. Jung AS, Sharma G, Maceri D, et al. Ultrasound-guided fine needle
yngol Head Neck Surg. 2007;136(5):788-792. aspiration of major salivary gland masses and adjacent lymph nodes.
113. Balakrishnan K, Castling B, McMahon J, et al. Fine needle aspiration Ultrasound Q. 2011;27(2):105-113.
cytology in the management of a parotid mass: a two-Centre retro- 128. Khan N, Afroz N, Agarwal S, et al. Comparison of the efficacy of the
spective study. Surgeon. 2005;3(2):67-72. palpation versus ultrasonography-guided fine-needle aspiration
114. Cohen EG, Patel SG, Lin O, et al. Fine-needle aspiration biopsy of sal- cytology in the diagnosis of salivary gland lesions. Clinl Cancer Investi-
ivary gland lesions in a selected patient population. Otolaryngol Head gat J. 2015;4(2):134-139.
Neck Surg. 2004;130(6):773-778. 129. Pusztaszeri M, Reis-Filho JS, de Lander Schmitt FC, Edelweiss M.
115. Sergi B, Contucci AM, Corina L, Paludetti G. Value of fine-needle Ancillary studies for salivary gland cytology. In: Faquin WC,
aspiration cytology of parotid gland masses. Laryngoscope. 2004; Rossi ED, Baloch ZW, et al., eds. The Milan System for Reporting Sali-
114(4):789. vary Gland Cytopathology. Cham: Springer Nature; 2018:139-155.
116. Stow N, Veivers D, Poole A. Fine-needle aspiration cytology in the
management of salivary gland tumors: an Australian experience. Ear
Nose Throat J. 2004;83(2):128-131.
117. Tsai SC, Hsu HT. Parotid neoplasms: diagnosis, treatment, and intra-
How to cite this article: Farahani SJ, Baloch Z. Retrospective
parotid facial nerve anatomy. J Laryngol Otol. 2002;116(5):359-362.
118. Santamaria J, Irazu L, Santamaria G, et al. The efficacy of diagnostic assessment of the effectiveness of the Milan system for
procedures in medical-surgical pathology of the salivary glands. Med reporting salivary gland cytology: A systematic review and
Oral. 2000;5(3):198-207. meta-analysis of published literature. Diagn Cytopathol. 2018;
119. Tew S, Poole AG, Philips J. Fine-needle aspiration biopsy of parotid
1–21. https://doi.org/10.1002/dc.24097
lesions: comparison with frozen section. Aust N Z J Surg. 1997;67(7):
438-441.

You might also like