Prevalence and Distribution of Carcinoma in Patients
with Solitary and Multiple Thyroid Nodules on
Mary C. Frates, Carol B. Benson, Peter M. Doubilet, Elizabeth Kunreuther, Maricela Contreras,
Edmund S. Cibas, Joseph Orcutt, Francis D. Moore, Jr., P. Reed Larsen, Ellen Marqusee, and
Erik K. Alexander
Department of Radiology (M.C.F., C.B.B., P.M.D., M.C.); Thyroid Section, Division of Endocrinology, Hypertension and
Diabetes, Department of Medicine (E.K., J.O., P.R.L., E.M., E.K.A.); Department of Pathology (E.S.C.); and Department of
Surgery (F.D.M), Brigham & Women’s Hospital and Harvard Medical School, Boston, Massachusetts 02115
Context: Controversy remains as to the optimal management of
patients with multiple thyroid nodules.
Objective: The objective of this study was to determine the preva-
lence, distribution, and sonographic features of thyroid cancer in
patients with solitary and multiple thyroid nodules.
Design: We describe a retrospective observational cohort study that
was carried out from 1995 to 2003.
Setting: The study was conducted in a tertiary care hospital.
Patients: Patients with one or more thyroid nodules larger than 10
mm in diameter who had ultrasound-guided fine needle aspiration
(FNA) were included in the study.
Main Outcome Measures: The main outcome measures were prev-
alence and distribution of thyroid cancer and the predictive value of
demographic and sonographic features.
Results: A total of 1985 patients underwent FNA of 3483 nodules.
The prevalence of thyroid cancer was similar between patients with
a solitary nodule (175 of 1181 patients, 14.8%) and patients with
multiple nodules (120 of 804, 14.9%) (P ? 0.95, ?2). A solitary nodule
had a higher likelihood of malignancy than a nonsolitary nodule (P ?
0.01). In patients with multiple nodules larger than 10 mm, cancer
was multifocal in 46%, and 72% of cancers occurred in the largest
nodule. Multiple logistic regression analysis of statistically signifi-
cant features demonstrates that the combination of patient gender
nodule composition (P ? 0.01), and presence of calcifications (P ?
0.001) can be used to assign risk of cancer to each individual nodule.
Risk ranges from a 48% likelihood of malignancy in a solitary solid
nodule with punctate calcifications in a man to less than 3% in a
noncalcified predominantly cystic nodule in a woman.
Conclusions: In a patient with one or more thyroid nodules larger
than 10 mm in diameter, the likelihood of thyroid cancer per patient
is independent of the number of nodules, whereas the likelihood per
nodule decreases as the number of nodules increases. For exclusion
of cancer in a thyroid with multiple nodules larger than 10 mm, up
to four nodules should be considered for FNA. Sonographic charac-
teristics can be used to prioritize nodules for FNA based on their
individual risk of cancer. (J Clin Endocrinol Metab 91: 3411–3417,
lence of such nodules increases with age (5). As increasing
uations, more and more thyroid nodules are being detected.
Current diagnostic recommendations for patients with thyroid
nodules are based primarily on data obtained from the evalu-
ation of nodules before widespread use of ultrasonography.
Based on these studies, nodules have a 5–15% prevalence of
malignancy (5–9) and, thus, fine needle aspiration (FNA) of
solitary nodules larger than 10–15 mm in maximal diameter is
usually recommended if the patient is euthyroid (6, 10, 11).
On ultrasound examination, many patients thought to
have a solitary nodule by physical exam are found to have
additional nodules larger than 10 mm in diameter (4, 12–15).
OPULATION STUDIES SUGGEST that 3–8% of asymp-
tomatic adults have thyroid nodules (1–4). The preva-
The recommended diagnostic approach for patients with
multiple nodules is variable. Some advocate routine FNA of
all nodules larger than 10 mm (16, 17), whereas others rec-
ommend FNA of only the largest nodule (18). Still others
advocate follow-up alone based on the belief that cancer is
rare when multiple nodules are present (19), or they suggest
thyroid scintigraphy as an initial diagnostic test even when
the patient is biochemically euthyroid (20). This lack of a
consistent recommendation stems in part from the absence
of studies investigating the prevalence and location of thy-
roid cancer in patients with multiple thyroid nodules.
FNA of a thyroid nodule is the method of choice for de-
termining the risk that a given nodule is malignant (6, 7, 21,
22). A number of studies have also assessed various sono-
graphic characteristics as predictors of thyroid cancer. Sono-
graphic features reported to be associated with an increased
risk of cancer include nodule size, presence of microcalcifi-
cations, hypoechogenicity, solid composition, absence of a
First Published Online July 11, 2006
Abbreviation: FNA, Fine needle aspiration.
JCEM is published monthly by The Endocrine Society (http://www.
endo-society.org), the foremost professional society serving the en-
Printed in U.S.A.
The Journal of Clinical Endocrinology & Metabolism 91(9):3411–3417
Copyright © 2006 by The Endocrine Society
none has systematically compared multiple nodules in the
same gland with respect to predicting the risk of thyroid
cancer based on combined sonographic criteria.
In our facility we perform a large number of ultrasound-
guided FNAs and have routinely recommended FNA for all
nodules larger than 10 mm regardless of the sonographic
appearance or number of nodules. Thus, this patient popu-
lation provides a large, unbiased sample to assess the risk of
cancer in patients with thyroid nodules with various com-
binations of sonographic findings.
In this study, we retrospectively reviewed the records of
all patients with one or more thyroid nodules larger than 10
mm in maximum diameter who had ultrasound-guided
FNA. Our goal was to compare the risk of thyroid cancer in
patients with solitary nodules to that in patients with mul-
tiple nodules. We also wished to determine whether or not
sonographic features of thyroid nodules would be useful in
predicting the risk of malignancy for a given nodule.
Patients and Methods
Approximately 3200 patients without prior thyroid surgery or ra-
dioiodine exposure were referred to the multidisciplinary Thyroid Nod-
ule Clinic at the Brigham and Women’s Hospital (Boston, MA) between
1995 and 2003 for evaluation of suspected thyroid nodular disease. The
nodules by physical exam or the presence of an “incidental” nodule
discovered by an imaging technique such as magnetic resonance im-
aging, computed tomography, or carotid ultrasound. All patients un-
derwent thyroid sonography as part of their evaluation, and those pa-
tients with one or more thyroid nodules larger than 10 mm in diameter
who had ultrasound-guided FNA are the subject of this report. Patients
whose serum TSH was normal or elevated were advised to have ultra-
sound-guided FNA of all nodules larger than 10 mm in maximal di-
ameter. If the serum TSH was less than 0.5 ?U/ml, patients had thyroid
not aspirated. Calcitonin measurements were not routinely performed.
each with special expertise in thyroid sonography, using a 5- to 15-MHz
in depth in a subset of the patient population that underwent ultra-
sound-guided FNA between June 1995 and October 2000. For each
nodule, sonographic images were reviewed by at least one of three
radiologist participants in the study, and sonographic characteristics
were recorded. Nodules were excluded from this part of the study if
images were not available for review. For each nodule, the following
sonographic characteristics were recorded: size, parenchymal compo-
sition, echogenicity, presence or absence of a halo, margin appearance,
presence or absence of calcifications, type of calcifications, and presence
or absence of other nodules larger than 10 mm in the gland. Size was
recorded as three orthogonal dimensions. If a halo was present, mea-
surements included the halo as part of the nodule. Parenchymal com-
position was classified based on subjective assessment of the approxi-
mate portion of the nodule that was cystic, as follows: completely solid,
predominantly solid (1–24% cystic), mixed solid/cystic (25–74% cystic),
of each nodule that was more than 50% solid was determined by com-
paring the solid portion of the nodule to surrounding thyroid paren-
chyma and was reported as hyperechoic when more echogenic, iso-
echoic when similar, or hypoechoic when less echogenic than thyroid
at least 50% of the nodule, present around less than 50% of the nodule,
or absent. Margin appearance was categorized as well defined or poorly
defined. Presence or absence of calcifications was noted for each nodule,
and calcifications were classified as punctate, coarse, or isolated to the
rim of the nodule. A nodule was considered solitary if there were no
other nodules in the gland measuring larger than 10 mm in maximum
diameter and nonsolitary if the thyroid contained at least one other
nodule larger than 10 mm in maximum diameter. If a particular nodule
only once, and the images acquired at the time of the diagnostic aspirate
or at the last ultrasound before surgery were used for the analysis. For
a small number of nodules, not all sonographic characteristics were
FNA was performed by one of four thyroidologists under ultrasound
guidance as previously described (15, 31). Three to four aspirates were
performed per nodule using a 25-gauge needle. A maximum of three
nodules was aspirated during a single visit. Most patients requiring
more than one visit for complete evaluation of multiple nodules larger
than 10 mm in diameter returned within 6 months of their initial eval-
was performed one or more times or the patient elected surgical
All FNA specimens were collected in CytoLyt (Cytyc Corp., Marl-
borough, MA), and two slides were prepared using Thin-Prep 2000
(Cytyc Corp.). Thin-Prep slides were stained with a modified Papani-
colaou procedure. In cases with residual cell sediment, a cell block was
prepared by sedimentation, and two cell block sections were stained
with hematoxylin and eosin. Specimens were considered nondiagnostic
if insufficient cellular material (fewer than six groups of cells containing
?10 cells each) was present and no evidence of cellular atypia was
found. Diagnostic aspiration was classified as follows: benign, atypical
cells of undetermined significance, suggestive of follicular neoplasm,
suspicious for papillary carcinoma, or positive for papillary carcinoma.
We recommended repeat FNA of nodules with a diagnosis of atypical
cells of undetermined significance (“atypical” cytology). If “atypical” on
repeat FNA, surgery was recommended. Nodules with FNA cytology
suspicious or positive for papillary carcinoma or suggestive of a follic-
ular neoplasm were referred for surgical resection. In all cases in which
surgery was performed, the final diagnosis for each nodule larger than
10 mm in the gland was based on histopathological examination of the
entire gland (32, 33).
Nodules larger than 10 mm were classified as benign if the FNA
diagnosis was benign on an adequate FNA specimen, if no evidence of
cancer was found on histological evaluation of a resected nodule, if
thyroid scintigraphy indicated that the nodule functioned autono-
nodule diameter on follow-up ultrasound examination. In patients with
multiple nodules, each nodule larger than 10 mm was individually
classified as benign or malignant based on the above-described criteria.
Thyroid cancers 10 mm or less in maximal diameter incidentally dis-
covered on histopathological examination were excluded from analysis.
Permission from the Investigational Review Board at the Brigham and
Women’s Hospital was granted to perform this review and analysis.
To determine which sonographic features were associated with thyroid
cancer, ?2analyses were used for categorical variables and the Student’s
t test for continuous variables. P ? 0.05 was accepted as significant.
Statistically significant individual variables were then evaluated using
multiple logistic regression analysis to generate a table that specifies the
risk of malignancy based on a combination of nodule characteristics.
Prevalence of thyroid cancer in the study population
Approximately 3200 patients were referred to the Nodule
Clinic for diagnostic thyroid ultrasound and possible FNA
from 1995–2003. About two thirds (2208) had at least one
thyroid nodule larger than 10 mm in maximal diameter and
underwent FNA. Of these, 223 patients were excluded from
analysis because of the absence of a histopathological diag-
nosis. This resulted in a study population of 1985 patients
with a total of 3483 nodules larger than 10 mm in largest
dimension (Fig. 1). Two hundred forty-three (12.2%) of the
1985 patients were male, and 1742 (87.8%) were female.
Fewer than 1% of patients reported exposure to childhood
head or neck irradiation, and no patients had a familial
history of medullary carcinoma of the thyroid.
Among the 1985 study patients, 295 (14.9%) had thyroid
J Clin Endocrinol Metab, September 2006, 91(9):3411–3417 Frates et al. • Thyroid Cancer Risk in Multinodular Glands
cancer, including 261 cases of papillary carcinoma, 27 of
follicular carcinoma, three of medullary carcinoma, two of
anaplastic carcinoma, and two of metastatic disease to the
thyroid. One hundred eighty-eight of the 227 patients with
thyroid cancer (82.8%) had American Joint Committee on
Cancer (6th edition) stage I or II disease, and the remainder
had stage III or IV disease. The mean age of patients with
thyroid cancer was 46.2 ? 0.9 yr (mean ? sd) compared with
40.1 ? 0.3 yr in those without cancer (P ? 0.01, t test). The
rate of cancer in a man with nodules (49 of 243, or 20.2%) was
higher than in a woman with nodules (246 of 1742, or 14.1%)
(P ? 0.03, ?2).
The 1985 study patients included 1181 (59.5%) with a sol-
itary nodule larger than 10 mm in largest dimension and 804
(40.5%) who had two or more such nodules. There was no
significant difference in sex distribution between the groups
[females constituted 1035 (87.6%) of 1182 patients with a
solitary nodule vs. 708 (88.1%) of the 804 patients with mul-
tiple nodules; P ? 0.74, ?2]. Patients with solitary nodules
were younger (48 ? 15 yr) than those with more than one
nodule larger than 10 mm (53 ? 14 yr) (P ? 0.01, t test).
The prevalence of thyroid cancer (Table 1) did not differ
between patients with a solitary thyroid nodule (175 of 1181
patients, 14.8%) and patients with multiple nodules (120 of
were also similar in the two groups: among the 175 solitary
nodules that were cancers, 151 (86.3%) were papillary, 21
(12.0%) were follicular, and three (1.7%) were other types of
cancer, whereas the corresponding numbers for the 120 can-
cers in glands with multiple nodules were 110 (91.7%) pap-
illary, six (5.0%) follicular, and four (3.3%) other (P ? 0.09,
?2comparison of papillary vs. nonpapillary cancers in the
TABLE 1. Prevalence of thyroid cancer per patient and per
nodule according to the number of nodules larger than 10 mm in
No. of thyroid
nodules ?10 mm
nodules (n ? 175 patients) to those with multiple nodules (n ? 120
bP ? 0.30, ?2comparing cancer rate per patient in those with one
nodule (n ? 175 patients), two nodules (n ? 73 patients), three nod-
ules (n ? 27 patients), and four or more nodules (n ? 20 patients).
cP ? 0.001, ?2comparing cancer rate per nodule in those with
single nodules (n ? 175 nodules) to those with multiple nodules (n ?
dP ? 0.001, ?2comparing cancer rate per nodule in those with one
nodule (n ? 175 nodules), two nodules (n ? 107 nodules), three
nodules (n ? 47 nodules), and four or more nodules (n ? 33).
FIG. 1. Study population.
Frates et al. • Thyroid Cancer Risk in Multinodular Glands J Clin Endocrinol Metab, September 2006, 91(9):3411–3417
Unlike the per-patient comparison, a nodule that is one of
several had a lower likelihood of being malignant than did
a solitary nodule: 175 of 1181 solitary nodules (14.8%) were
malignant, as opposed to 187 of 2302 nonsolitary nodules
(8.1%) (P ? 0.001, ?2). The per-nodule likelihood of cancer
decreased progressively as the number of nodules larger
than 10 mm increased (Table 1).
Of the 120 patients with thyroid cancer in a gland with
more than one nodule, 87 (72.5%) had cancer in the largest
thyroid nodule (Table 2). When more than two nodules
larger than 10 mm were present, the malignancy was in the
largest nodule about half the time. As the number of nodules
increased, the frequency of cancer in the largest nodule de-
of the largest nodule. Sixty-five of the 120 patients (54.2%)
with cancer in a gland with more than one nodule had a
unifocal cancer, whereas the cancer was multifocal in 55
A strategy of biopsying the largest nodule would have
detected only 86% of patients with two nodules who had
nodules who had cancer (Table 3).
Prediction of cancer based on sonographic characteristics
we reviewed a subset of the nodules in the total population.
Of the 1201 nodules that underwent FNA between June 1995
and October 2000, 201 were excluded because images could
not be retrieved for review. Another 135 were excluded be-
cause pathological diagnosis was not available. The remain-
comprised the study set for evaluation of sonographic fea-
tures. Of these 865 nodules, 771 (89.1%) were benign and 94
(10.9%) were malignant, including 80 papillary cancers, 12
follicular cancers, and two anaplastic cancers.
In this study set, 780 of the nodules (90.2%) occurred in
women and 85 (9.8%) in men. As in the larger group, the
cancer was more likely in a nodule in a man (16 of 85, or
18.8%) than in a woman (78 of 780, or 10%) (P ? 0.02, ?2).
To determine whether the likelihood of cancer was asso-
ciated with nodule size (measured as maximum diameter),
we divided nodules into size categories by 5-mm increments
no significant relationship between nodule size and likeli-
hood of thyroid cancer (P ? 0.48, ?2).
Those individual sonographic characteristics that had a
statistically significant association with thyroid cancer in-
cluded nodule composition (cystic vs. solid component),
echogenicity (hypoechoic), presence and type of calcifica-
tions, and whether or not the nodule was solitary (Table 4).
In particular, the more cystic a nodule was, the lower the
likelihood of cancer. Hypoechoic nodules had a higher rate
of malignancy than nodules that were isoechoic or hypere-
choic. When compared with the malignancy rate for nodules
without calcifications, the presence of coarse or rim calcifi-
cations increased the likelihood of cancer almost 2-fold, and
As in the entire study population, the rate of cancer per
nonsolitary nodules. The presence and extent of a halo
around a nodule and the appearance of a nodule’s margin
(well defined or poorly defined) were not significantly as-
sociated with presence or absence of thyroid cancer.
The results listed above pertain to the relationship be-
tween an individual nodule characteristic and thyroid can-
TABLE 2. Location of cancer in 120 patients with multiple
nodules and thyroid cancer
No. of thyroid
nodules ?10 mm
Cancer in two or
TABLE 3. Diagnostic yield of sequential aspiration strategies in
120 patients with multiple nodules and cancer
FNA performed on
No. of nodules ?10 mm
2 (n ? 73)
3 (n ? 27)
?4 (n ? 20)
Largest 2 nodules
Largest 3 nodules
Largest 4 nodules
Results are percentages.
TABLE 4. Relationship between sonographic characteristics of a
thyroid nodule and its likelihood of malignancy
Mixed solid and cystic
Solitary vs. multiple
Present around ?50%
Present around ?50%
do not always add up to the total number of benign and malignant
nodules in the study sample, because information was not available
for all characteristics of all nodules.
bP value via ?2test.
J Clin Endocrinol Metab, September 2006, 91(9):3411–3417 Frates et al. • Thyroid Cancer Risk in Multinodular Glands
cer. To assess how the likelihood of malignancy can be de-
termined based on a nodule’s characteristics in combination,
we applied multiple logistic regression analysis to those
characteristics with the strongest association with malig-
nancy, including female vs. male gender, solitary vs. non-
type of calcifications. This analysis yielded a table that spec-
ifies the risk of malignancy based on nodule characteristics
in solitary and nonsolitary (Table 5) nodules. For any set of
sonographic characteristics, the likelihood of a nodule being
malignant is approximately twice as high in a solitary com-
pared with a nonsolitary nodule and more than 1.5 times as
high in a man compared with a woman.
This analysis of 1985 patients with one or more thyroid
nodules larger than 10 mm evaluated by ultrasound and
ultrasound-guided FNA demonstrates that the likelihood of
thyroid cancer in a patient with one or more nodules larger
sonographic characteristics are unable to accurately distin-
guish benign from malignant disease. In our cohort, 15% of
patients had thyroid cancer, a prevalence that applied
whether the patient had a solitary nodule or multiple nod-
ules. Thus, the presence of multiple nodules was not more
associated with a lower risk of thyroid cancer than was the
presence of a solitary thyroid nodule. In patients with two or
more nodules, aspiration of only the largest nodule would
have missed almost one third of the malignancies.
In studies performed before ultrasound or without ultra-
sound-guided FNA, thyroid cancer was detected in approx-
ultrasound-guided FNA report a 7–14% prevalence of thy-
roid cancer, but none has evaluated all nodules larger than
10 mm in a large cohort of patients including those with
multiple nodules (36–38). The prevalence of thyroid cancer
in thyroidectomy specimens from patients with one or more
14% of patients in that study had thyroid cancer, and no
nodules. However, only 132 nodules were analyzed in 132
patients, suggesting that only one nodule was assessed per
patient, even though multiple nodules were reported to be
present in 47%. To the best of our knowledge, no other
investigations have provided a complete malignancy assess-
ment of all thyroid nodules larger than 10 mm.
Numerous studies have examined the relationship be-
tween sonographic features of thyroid cancer and malig-
nancy. Although certain sonographic criteria are associated
with increasing cancer risk, the predictive value of these
criteria is not sufficiently high or low to preclude the neces-
sity of FNA. Most studies of sonographic features have been
based on small patient populations (27, 29, 30, 39) or were
subject to ascertainment bias, in that definitive pathology
was available for only a select subset of patients that was
skewed toward those with the most suspicious nodules (23,
28, 30, 40, 41). Other studies have been limited to univariate
analyses that evaluate one sonographic feature at a time (25,
42). This last issue is especially limiting, because in clinical
practice each individual nodule has a constellation of sono-
graphic features (e.g. solitary, predominately solid, with
punctate calcifications), and its likelihood of malignancy de-
pends on all of its characteristics. Our study of sonographic
characteristics addresses, and largely overcomes, these lim-
itations. Our sample is large, and we avoided ascertainment
bias by performing FNA on all nodules, with further patho-
logical information derived from the surgical specimen if the
patient underwent thyroidectomy. Although sonographic
criteria were analyzed in only about one third of our cohort,
In our study, patients were referred for ultrasound by
primary care physicians because of an abnormal physical
examination or an incidental imaging finding of a possible
nodule. Only two thirds of the referred patients were found
to have one or more nodules larger than 1 cm in diameter.
We analyzed data from 90% of these patients. Therefore, we
know of no selection or sample bias. Among those excluded
from the final analysis, the proportion of patients with sol-
itary and multiple nodules was nearly equal. Furthermore,
all thyroid nodules were classified based on cytological or
histopathological analysis. The rate of false-negative results
making it unlikely many cancers were missed.
There is considerable variation among published recom-
mendations and guidelines for appropriate evaluation of
patients with more than one thyroid nodule. Some advocate
no FNA unless worrisome ultrasound features are detected
(19), whereas others suggest routine aspiration of only the
TABLE 5. Likelihood of malignancy in a thyroid nodule: results of multiple logistic regression of sonographic characteristics
Women Men (%)
Coarse or rim
Coarse or rim
Mixed solid and cystic
One of multiple nodules
Mixed solid and cystic
Frates et al. • Thyroid Cancer Risk in Multinodular GlandsJ Clin Endocrinol Metab, September 2006, 91(9):3411–3417