0021-972X/06/$15.

00/0 The Journal of Clinical Endocrinology & Metabolism 91(11):4295– 4301

Printed in U.S.A. Copyright © 2006 by The Endocrine Society
doi: 10.1210/jc.2006-0527

Serum Thyrotropin Concentration as a Novel Predictor

of Malignancy in Thyroid Nodules Investigated by Fine-
Needle Aspiration
K. Boelaert, J. Horacek, R. L. Holder, J. C. Watkinson, M. C. Sheppard, and J. A. Franklyn
Division of Medical Sciences (K.B., J.C.W., M.C.S., J.A.F.), The Medical School, Department of Primary Care and General
Practice (R.L.H.), University of Birmingham, Birmingham B15 2TT, United Kingdom; and Department of Medicine (J.H.),
Charles University (Prague), Faculty of Medicine and University Hospital, CZ-50005 Hradec Kralove, Czech Republic

Context: Thyroid nodules and goiter are common, and fine-needle
aspiration biopsy (FNAB) is the first investigation of choice in dis-
tinguishing benign from malignant disease.
Objective: The objective of the study was to assess whether simple
clinical and biochemical parameters can predict the likelihood of
thyroid malignancy in subjects undergoing FNAB.
Design: The design was a prospective cohort.
Setting: The study was conducted at a single secondary/tertiary care
clinic.
Participants: One thousand five hundred consecutive patients with-
out overt thyroid dysfunction (1304 females and 196 males, mean age
47.8 yr) presenting with palpable thyroid enlargement between 1984
and 2002 were evaluated by FNAB of the thyroid.
Intervention(s): There were no interventions.
Results: The overall sensitivity and specificity of FNAB in predicting
malignancy were 88 and 84%, respectively. The risk of diagnosis of
malignancy rose in parallel with the serum TSH at presentation, with
significant increases evident in patients with serum TSH greater than
0.9 mU/liter, compared with those with lower TSH. Binary logistic
regression analysis revealed significantly increased adjusted odds
ratios (AORs) for the diagnosis of malignancy in subjects with serum
TSH 1.0 –1.7 mU/liter, compared with TSH less than 0.4 mU/liter
[AOR 2.72, 95% confidence interval (CI) 1.02–7.27, P ⫽ 0.046], with
further increases evident in those with TSH 1.8 –5.5mU/liter (AOR
3.88, 95% CI 1.48 –10.19, P ⫽ 0.006, compared with TSH ⬍ 0.4 mU/
liter) and greater than 5.5 mU/liter (AOR 11.18, 95% CI 3.23– 8.63,
P ⬍ 0.001, compared with TSH ⬍ 0.4 mU/liter). Males (AOR 1.8, 95%
CI 1.04 –3.1, P ⫽ 0.04), younger patients (AOR 1.1, 95% CI 1.01–1.15,
P ⫽ 0.025), and those with clinically solitary nodules (AOR 2.53, 95%
CI 1.5– 4.28, P ⫽ 0.001) were also at increased risk. Based on these
findings, a formula to predict the risk of the diagnosis of thyroid
malignancy in individual patients, taking into account their gender,
age, goiter type determined clinically, and serum TSH, was calculated.

Main Outcome Measures: Goiter type was assessed clinically and
classified as diffuse in 183, multinodular in 456, or solitary nodule in
861 cases. Serum TSH concentration at presentation was measured
in a sensitive assay in patients presenting after 1988 (n ⫽ 1183). The
final cytological or histological diagnosis was determined after sur-
gery (n ⫽ 553) or a minimum 2-yr clinical follow-up period (mean 9.5
yr, range 2–18 yr).
Conclusions: The risk of malignancy in a thyroid nodule increases
with serum TSH concentrations within the normal range. In addition
to patient’s gender, age, and goiter type, the serum TSH concentration
at presentation is an independent predictor of the presence of thyroid
malignancy. We propose that these simple clinical and biochemical
factors can serve as an adjunct to FNAB in predicting risk of
malignancy. (J Clin Endocrinol Metab 91: 4295– 4301, 2006)

T HYROID ENLARGEMENT IS a common clinical prob-

lem. The Framingham study in the United States in-
dicated a 5–10% lifetime risk of developing a thyroid nodule
is reported to be rising (4). Most patients with thyroid en-
largement can be managed conservatively after malignancy
is ruled out, the challenge to the clinician being to identify the
(1), and the Whickham survey in the northeast of England minority of patients with thyroid cancer who therefore re-
reported a 15% prevalence of goiters or thyroid nodules (2). quire surgical intervention (5).
Additionally, high-resolution ultrasound can detect thyroid In most cases, thyroid glands harboring malignancy are
nodules in 19 – 67% of individuals, with higher frequencies in clinically indistinguishable from those that do not, and phys-
women and the elderly, even when the gland is normal to ical examination is therefore deemed largely unhelpful in
palpation (3). identifying those patients with thyroid cancer (5). A major
Thyroid cancer, in contrast, is rare, accounting for approx- aim of clinical evaluation of patients presenting with thyroid
imately 1% of all new malignant disease (0.5% of cancers in enlargement is to minimize the risk of overlooking thyroid
men and 1.5% in women) (4), although the annual incidence cancer. Recognized clinical parameters raising the suspicion
for malignancy include young (⬍20 yr) or old age (⬎70 yr),
First Published Online July 25, 2006 male gender, large (⬎4 cm) or rapidly growing nodules (es-
Abbreviations: ATA, American Thyroid Association; BTA, British pecially during thyroid hormone therapy), and radiation
Thyroid Association; FNAB, fine-needle aspiration biopsy; fT3, free T3; exposure history (5, 6). It has been widely perceived that rates
fT4, free T4.
of malignancy are higher in subjects with solitary nodules
JCEM is published monthly by The Endocrine Society (http://www.
endo-society.org), the foremost professional society serving the en- than in those with multinodular goiters (5, 7), although some
docrine community. have reported similar rates in these two groups (8). Our own

4295

The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
4296 J Clin Endocrinol Metab, November 2006, 91(11):4295– 4301
previous studies have also reported the presence of malig-
nancy in subjects with clinically diffuse as well as those with
multinodular goiters (9, 10). Although virtually all patients
with thyroid carcinoma are euthyroid, the presence of a
suppressed serum TSH concentration (generally indicative
of subclinical or overt thyrotoxicosis) does not rule out the
presence of malignancy (5).
Although several imaging modalities are available, fine-
needle aspiration biopsy (FNAB) remains the gold standard
in the evaluation of patients presenting with thyroid en-
largement as stated in recent guidelines published by the
American Thyroid Association (ATA) (11) and other author-
ities (12–14). FNAB has a reported diagnostic sensitivity and
specificity ranging between 65 and 98% and 72 and 100%,
respectively (4, 7, 15). Diagnostic FNAB results are obtained
in approximately 80% of cases, and repeat aspiration can
augment the accuracy of the procedure (15, 16). Several stud-
ies have demonstrated that ultrasound guidance, compared
with palpation-guided FNAB, reduces the number of non-
diagnostic aspirates, and some, but not all, studies suggest
that it increases diagnostic sensitivity and specificity (16 –18).
We set out to explore the hypothesis that simple clinical or
biochemical criteria might predict the likelihood of thyroid
malignancy in patients presenting with thyroid enlargement,
thereby identifying those at greatest risk of harboring thyroid
malignancy. Having previously undertaken a similar anal-
ysis in a cohort of 1005 patients (10), we reexamined this
hypothesis in a larger prospectively collected cohort of 1500
patients who underwent FNAB between 1984 and 2002 in our
clinic. We performed a detailed investigation of which pa-
rameters of clinical and laboratory assessment might predict
the diagnosis of thyroid cancer. We also used these param-
eters to derive a formula to predict the risk of diagnosis of
malignancy in individual subjects undergoing FNAB.
Subjects and Methods
Subjects and data recorded
We prospectively collected data on 1500 patients presenting consec-
utively to the Multidisciplinary Thyroid Clinic at the University Hos-
pital Birmingham National Health Service Trust, Birmingham, UK, be-
tween 1984 and 2002. This study was performed with the approval of
the Trust Research and Development Directorate. The cohort included
1304 females and 196 males with a mean age of 47.8 yr (range 6 – 88 yr).
All patients included were clinically and biochemically euthyroid [de-
fined as normal serum concentrations of free T4 (fT4) and free T3 (fT3)].
Patients were followed up in the clinic for a minimum duration of 2 yr
(mean duration of follow-up since presentation 9.5 yr, range 2–19 yr). A
final histological diagnosis was made in 553 patients who underwent
surgery, and a final cytological diagnosis was made after a minimum
2-yr follow-up period in the remainder. The goiter type was assessed by
physical examination of the neck by one of four senior clinicians and
classified as diffuse in 183, multinodular in 456, and solitary nodule in
861 subjects as we have previously described (10).
All 1500 patients had FNAB of their thyroid performed at initial
presentation, the clinic protocol being to perform this investigation at
first visit in all subjects with goiter or nodules and without overt thyroid
dysfunction, in keeping with guidelines from the ATA and British Thy-
roid Association (BTA) (11, 19). The aspirate was performed on the
dominant nodule in subjects with multinodular goiters, whereas those
with large diffuse goiters had multiple aspirates from different sites at
a single sitting. In patients with nondiagnostic aspirates as well as those
with changing clinical symptoms and signs, FNAB was repeated at
subsequent visits, typically after 3– 6 months. A minimum follow-up
period of 2 yr was designated to allow clinical reevaluation of the neck
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
Boelaert et al. • Serum TSH and Prediction of Thyroid Malignancy
over an extended period. Any change in symptoms or signs at 3- to
6-month review prompted repeat FNAB, a policy in agreement with the
recently published ATA guidelines, which state that easily palpable
nodules do not require sonographic monitoring but that patients should
be followed up clinically at 6- to 18-month intervals (11).
Cytological findings were classified according to accepted guidelines
(11, 19) as: nondiagnostic (Thy1 category according to BTA guidelines);
benign (BTA Thy2, nonneoplastic category); indeterminate (including
BTA categories Thy3 (follicular lesions) and Thy4 (suspicious of malig-
nancy)]; and diagnostic of malignancy (BTA Thy5 category). Patients
were assigned to a final diagnostic category for cytology results (true
negative, true positive, false negative, or false positive for malignancy)
after the minimum 2-yr follow-up period, except for those patients who
underwent surgery before the end of this period (and in whom a formal
histological diagnosis was obtained earlier). The cytological diagnosis
was considered true negative if, after a follow-up period or surgery, a
diagnosis of thyroid malignancy had not been made. Cytology indicat-
ing indeterminate or frankly malignant findings was considered true
positive if malignancy was confirmed histologically after surgery or
open biopsy. Cytological results were defined as false negative in sub-
jects with one or more diagnostic aspirates without suspicious or ma-
lignant features in whom later cytology and/or surgery provided evi-
dence for malignancy (including those patients harboring a
microcarcinoma).
A biochemical evaluation of all patients attending a morning clinic
was performed at presentation through measurement of fT4, fT3, and
TSH concentrations, in keeping with ATA and BTA guidance (11, 19).
Furthermore, 697 patients were investigated by respiratory flow-loop
examination to identify those with evidence of upper airway obstruction
as described by ourselves previously (20). Serum fT4, fT3, and TSH were
measured by automated luminescent immunoassays (ACS 180; Chiron
Diagnostics, Halstead, UK) with interassay coefficients of variation of
less than 12% over the ranges 6.2– 81 pmol/liter, 2.5–20 pmol/liter, and
0.3–39 mU/liter, respectively. A measurement of TSH in this sensitive
assay was obtained for all patients who presented after 1988 (n ⫽ 1183).
A total of 1229 subjects had antibodies to thyroid peroxidase measured
by gelatin particle agglutination (Fujirebo Inc., Tokyo, Japan), and a titer
of 1:100 or greater was considered positive.
Statistical analysis
The final diagnostic outcome was defined as the presence or absence
of malignancy (thyroid cancer). The influence of factors including age,
gender, thyroid characteristics on physical examination, serum TSH
concentration, and presence of antimicrosomal antibodies at presenta-
tion on the final diagnostic outcome was investigated statistically. We
used binary logistic regression analyses, using Minitab (version 14.0;
Minitab Ltd., Coventry, UK) as the statistical package. The serum TSH
concentration was recorded as a categorical variable in this analysis;
analysis of the influence of serum TSH measurement was confined to the
subgroup (n ⫽ 1183) in whom TSH had been measured in a sensitive
assay. Examination of the prevalence of malignancy in the different
serum TSH categories enabled a judgment to be made on the validity of
including TSH as a continuous variable. With a satisfactory outcome to
this investigation, a formula could then be proposed to predict the
probability of malignancy in an individual patient through binary re-
gression analysis using the serum TSH concentration as a continuous
variable.
Results
The goiter type determined by palpation was classified as
diffuse in 183, multinodular in 456, and solitary nodule in 861
subjects. Twelve subjects in the series of 1500 had clinical
features suggestive of malignancy (21) (six cervical lymph-
adenopathy, two hoarseness of voice, and four rapidly en-
larging or fixed goiters). Nine of these patients were found
to harbor malignancy.
A total of 697 patients had a respiratory flow loop exam-
ination performed, and evidence of significant airways ob-
Boelaert et al. • Serum TSH and Prediction of Thyroid Malignancy J Clin Endocrinol Metab, November 2006, 91(11):4295– 4301 4297

struction was detected in 85 patients, all of whom subse- malignancy in our series was 88 and 84%, respectively. When
quently underwent surgery. analyzing the performance of FNAB for the different goiter
types defined clinically, we obtained sensitivities of 80, 77,
Accuracy of FNAB and 90% and specificities of 90, 84, and 83% for clinically
Of the cohort of 1500, FNAB was repeated on one occasion defined diffuse goiters, multinodular goiters, and solitary
in 479 patients and more than once in 177. The final cyto- nodules, respectively.
logical diagnosis reached after a minimum of 2 yr clinical Male gender, extremes of age, and solitary nodules are
follow-up was benign cytology in 1086 (72.4%), indetermi- associated with increased risk
nate cytology in 291 (19.4%), and malignant cytology in 30
(2%) subjects. Nondiagnostic aspirates were initially ob- The predictive value of the patient’s age and gender, as
tained in 257 subjects (17.1%), but after repeated FNAB, only well as the goiter type as defined by clinical examination, on
93 (6.2%) had cytological features that remained unclassified. final diagnostic outcome of malignancy in the total cohort
Thirty-three of these patients proceeded to surgery, and in (n ⫽ 1500) was investigated. Male subjects (n ⫽ 196) pre-
the remainder, the thyroid enlargement was no longer evi- senting with thyroid enlargement had significantly higher
dent at follow-up. rates of malignancy (12.2%) when compared with female
A total of 553 patients proceeded to open biopsy or sur- patients (n ⫽ 1304, 7.4%, P ⫽ 0.02). Significant increases in
gery. This group included all those with malignant (n ⫽ 30) the prevalence of malignancy (P ⫽ 0.005) were detected in
or indeterminate cytology (n ⫽ 291), 85 subjects with upper patients who were aged younger than 30 yr and in those
airways obstruction defined by respiratory flow-loop exam- older than 80 yr at presentation (Fig. 1A). The change in
ination, 31 with recurrent accumulation of cyst fluid, 33 with prevalence with age was not linear as evidenced by high
persistent nondiagnostic cytology, and 83 who requested prevalences at both extremes of age. We therefore incorpo-
surgery for cosmetic reasons. The histological diagnoses rated an age squared term in the binary logistic regression
were compared with the cytological findings before surgery. analysis to correct for this increase in rates of malignancy in
Those in the cohort not subjected to surgery were followed patients younger than 30 yr and older than 80 yr at the time
up in the clinic at 6-month intervals for a minimum of 2 yr of presentation (P ⬍ 0.001).
to allow identification of those with changing symptoms or When the goiter type determined clinically at presentation
signs and the need for repeat FNAB. was analyzed, we found the highest rates of malignancy in
Overall, the rate of malignancy in our cohort was 8%. Table subjects presenting with a solitary nodule (n ⫽ 861, 10.8%),
1 displays the number of patients with each of the different compared with those who presented with a diffuse or nod-
types of thyroid cancer found. The number of subjects within ular goiter (n ⫽ 639, 4.2%, P ⬍ 0.001). No significant differ-
each clinical goiter classification subgroup and within each ences were detected when comparing the risk in subjects
cytology diagnostic category are also given. A true-negative with clinically diffuse (n ⫽ 183, 5.5%) or multinodular goiters
cytological diagnosis was made in 1164 (77.6%) patients, (n ⫽ 456, 3.7%, P ⫽ NS).
whereas 105 (7%) subjects were assigned to the true positive Subsequently a binary logistic regression analysis was per-
cytological category. Only one of the 216 (14.4%) patients formed, simultaneously analyzing gender, age, and goiter
with a false-positive result had an original cytological diag- type to determine which factors could be considered inde-
nosis of malignancy, the remainder having indeterminate pendent risk predictors. Significantly increased adjusted
cytological aspirates. A false-negative cytological diagnosis odds ratios for malignancy (Table 2) were detected in male
was found in 15 patients (1%), and three of these harbored patients, those of younger age, and those presenting with
a microcarcinoma. Review of the initial cytology in this clinically solitary lesions. The calculations also took into ac-
group did not indicate an error of interpretation in any case, count the square of the age to correct for the increased risk
and false-negative cytology results were therefore likely to in patients of age younger than 30 or older than 80 yr.
reflect sampling error. The final histological diagnosis for the
Serum TSH concentration at presentation predicts risk
15 patients in this false-negative group was follicular carci-
of malignancy
noma in six, papillary carcinoma in seven, and Hürthle cell
carcinoma in two subjects. Based on these findings, the over- A total of 1183 patients had their serum TSH concentration
all sensitivity and specificity of FNAB for the diagnosis of at presentation measured in a sensitive assay, and of these,

TABLE 1. Number of patients with different types of thyroid cancer diagnosed

Thyroid cancer type No. of patients Clinical goiter type Cytology diagnostic category
Follicular carcinoma 31 D-3, MN-7, SN-21 Thy1–2, Thy2– 4, Thy3–16, Thy4 –7, Thy5–2
Papillary carcinoma 58 D-4, MN-6, SN-48 Thy1– 0, Thy2–7, Thy3–24, Thy4 –13, Thy5–14
Mixed follicular and papillary carcinoma 4 D-1, MN-3, SN-0 Thy1– 0, Thy2– 0, Thy3–2, Thy4 – 0, Thy5–2
Hürtle cell carcinoma 13 D-0, MN-1, SN-12 Thy1–2, Thy2– 0, Thy3–9, Thy4 –2, Thy5– 0
Medullary carcinoma 3 D-0, MN-0, SN-3 Thy1– 0, Thy2– 0, Thy3– 0, Thy4 – 0, Thy5–3
Anaplastic carcinoma 4 D-0, MN-2, SN-2 Thy1– 0, Thy2– 0, Thy3–1, Thy4 – 0, Thy5–3
Non-Hodgkin’s lymphoma 7 D-1, MN-1, SN-5 Thy1– 0, Thy2– 0, Thy3–2, Thy4 – 0, Thy5–5
The number of subjects within each of the different clinical goiter type subgroups and within each cytology diagnostic category are also
displayed. D, Diffuse goiter; MN, multinodular goiter; SN, solitary nodule; Thy1, nondiagnostic cytology; Thy2, nonneoplastic cytology; Thy3,
follicular lesions; Thy4, suspicious of malignancy; Thy5, diagnostic of malignancy according to BTA guidelines (19).

The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
4298 J Clin Endocrinol Metab, November 2006, 91(11):4295– 4301 Boelaert et al. • Serum TSH and Prediction of Thyroid Malignancy

those who had their TSH concentration measured in a sen-

sitive assay, a total of 92 patients (7.8%) had a final diagnosis
of thyroid malignancy.
The prevalence of malignancy (n ⫽ 182, 2.8%) was lowest
in subjects with serum TSH below the normal range (⬍0.4
mU/liter). Compared with subjects with below-normal se-
rum TSH, higher rates of malignancy (3.7%, P ⫽ NS) were
present in those with serum TSH in the lower tertile of the
normal range, i.e. 0.4 – 0.9 mU/liter (n ⫽ 322). Even higher
rates were found in patients with serum TSH 1.0 –1.7 mU/
liter (n ⫽ 336, 8.3%, P ⫽ 0.02), compared with TSH less than
0.4 mU/liter, rising to 12.3% (P ⫽ 0.001, compared with low
TSH) for those with serum TSH 1.8 –5.5 mU/liter (i.e. the
highest tertile of the normal range, n ⫽ 316). The highest
prevalences of malignancy (29.6%, P ⬍ 0.001, compared with
low TSH) were evident in those with serum TSH above the
normal range (⬎5.5 mU/liter, n ⫽ 27, Fig. 1B).
Binary logistic regression analysis simultaneously analyz-
ing gender, age, goiter type, and TSH concentration con-
firmed significantly increased odds ratios for malignancy in
males, those of younger age, those with solitary nodules, and
those with serum TSH greater than 0.9 mU/liter (Table 3). A
simultaneous likelihood ratio test of the effect of all these
factors gives ␹28 ⫽ 59.8 (P ⬍ 0.001), indicating the combi-
nation of these factors in the prediction of malignancy to be
very valuable.
Because the majority of patients had a serum TSH con-
centration within the normal range (n ⫽ 974), the same anal-
FIG. 1. A, Prevalence of malignancy in relation to patients’ age in ysis was repeated, excluding subjects with subclinical thy-
years, demonstrating increased prevalence in patients at the ex- roid dysfunction (n ⫽ 209). The goiter type assessed at
tremes of age. B, Prevalence of malignancy according to the serum
TSH concentration measured at presentation in 1183 subjects with
clinical examination and the serum TSH concentration were
normal serum fT4 concentrations, indicating increased prevalence in again identified as independent predictors of the risk of
those with higher TSH. The dashed vertical lines denote the normal diagnosis of thyroid malignancy, whereas patients’ age and
reference range for serum TSH. Subjects with TSH measurements gender were not found to predict independently the presence
within the normal range were divided into tertiles of similar size. The of malignancy (Table 4).
number of patients in each group is given beneath the graph. *, P ⬍
0.05; **, P ⬍ 0.01; ***, P ⬍ 0.001, compared with TSH less than 0.4 The same binary logistic regression analysis was also re-
mU/liter. peated in the 291 patients with indeterminate cytology; 235
subjects of these patients had their serum TSH concentration
182 were found to have subclinical hyperthyroidism, i.e. TSH measured in a sensitive assay. When analyzing the serum
concentration less than 0.4 mU/liter (normal range 0.4 –5.5 TSH concentration as a continuous variable, we again de-
mU/liter) with normal serum fT4 and fT3 concentrations. tected a significant adjusted odds ratio of 1.17 (P ⫽ 0.04) with
Nine hundred seventy-four subjects had a serum TSH con-
centration within the normal range, and these were subdi- TABLE 3. Independent risk predictors of diagnosis of thyroid
malignancy defined by binary logistic regression analysis
vided into three tertiles of similar size (TSH 0.4 – 0.9 mU/ considering gender, age, goiter type, and serum TSH concentration
liter, n ⫽ 322; TSH 1.0 –1.7 mU/liter, n ⫽ 336; TSH 1.8 –5.5 simultaneously
mU/liter, n ⫽ 316). Twenty-seven patients were noted to
display subclinical hypothyroidism, i.e. TSH measurement Adjusted 95% confidence
Variable P value
odds ratio interval
greater than 5.5 mU/liter with normal serum fT4. Among
Male gender 1.80 1.80–3.10 0.036
Decreasing age (overall 1.08 1.01–1.15 0.025
TABLE 2. Independent predictors of the diagnosis of thyroid annual decrement)
malignancy defined by binary logistic regression analysis of Solitary nodule 2.53 1.50– 4.28 0.001
gender, age, and goiter type TSH, mU/liter
Less than 0.4 1.00
Adjusted 95% confidence 0.4 – 0.9 1.31 0.45–3.81 0.622
Variable P value
odds ratio interval
1.0 –1.7 2.72 1.02–7.27 0.046
Male gender 1.63 1.01–2.64 0.046 1.8 –5.5 3.88 1.48–10.19 0.006
Decreasing age (overall 1.09 1.03–1.15 0.004 ⬎5.5 11.18 3.23–38.63 ⬍0.001
annual decrement)
The serum TSH concentration was analyzed as a categorical vari-
Solitary nodule 2.77 1.77– 4.33 ⬍0.001
able. Patients with subclinical thyroid disease (serum TSH below or
The adjusted odds ratio, 95% confidence interval, and P value for above normal) are included in this analysis. A total of 1183 patients
each variable are displayed. were included in this analysis.

The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
Boelaert et al. • Serum TSH and Prediction of Thyroid Malignancy
TABLE 4. Independent risk predictors of diagnosis of thyroid
malignancy defined by binary logistic regression analysis
considering gender, age, goiter type, and serum TSH concentration
at presentation simultaneously
Variable
Adjusted 95% confidence
odds ratio
Male gender 1.54
Decreasing age (overall 1.06
annual decrement)
Solitary nodule 2.79
TSH, mU/liter
0.4 – 0.9 1.00
1.0 –1.7 2.05
1.8 –5.5 2.91
This analysis was confined to subjects with serum TSH concen-
trations within the normal range. Serum TSH was analyzed as a
categorical variable. A total of 974 patients were included in this
analysis.
increasing TSH. When the TSH was analyzed as a categorical
variable, a significantly increased risk (adjusted odds ratio
6.3, P ⬍ 0.02) was again evident for those with serum TSH
concentrations greater than 5.5 mU/liter, compared with
those with serum TSH less than 0.4 mU/liter. No signifi-
cantly increased adjusted odds ratios were detected in pa-
tients with serum TSH concentrations within the normal
range (divided into tertiles), reflecting the smaller sample
size.
The presence of antibodies to thyroid peroxidase was de-
termined in 1229 patients. A titer of greater than 1:100 was
considered to be positive and was evident in 160 patients.
The rate of malignancy was significantly higher (11.9 vs.
6.7%, P ⫽ 0.02) in patients with detectable thyroid peroxidase
antibodies, compared with those in whom antibodies were
absent (n ⫽ 1069). Binary logistic regression analysis, simul-
taneously analyzing the presence of antibodies and the se-
rum TSH concentration, did not identify antibody status to
be an independent predictor of prognosis for thyroid ma-
lignancy (adjusted odds ratio 1.19, 95% confidence interval
0.6 –2.35, P ⫽ 0.6).
Prediction of risk of thyroid malignancy in
individual patients
Based on the independent risk factors identified from the
binary logistic regression analysis, we calculated formulae to
predict the risk of thyroid malignancy from simple clinical
and laboratory markers in individual patients. These calcu-
lations were based on the cohort of 1183 patients who had
their TSH concentration measured in a sensitive assay; the
adjusted odds ratios analyzing serum TSH as a categorical
variable suggested that TSH could be satisfactorily used as
a continuous variable in these formulae. The risk of malig-
nancy was calculated through the formula: P ⫽ 1/(1 ⫹ e⫺x),
TABLE 5. Calculation of the probability of diagnosis of malignancy based on simple clinical and biochemical parameters
Patient Gender Age (yr)
1 Male 16
2 Female 40
3 Male 75
Examples of risk predictions for three patients of differing age, gender, goiter type, and serum TSH concentration are displayed.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
J Clin Endocrinol Metab, November 2006, 91(11):4295– 4301 4299
with e being the antilogarithmic transformation (e ⫽ 2.71828)
and x representing a calculation taking into account the pa-
tient’s age and gender, the goiter type, and serum TSH. For
the probability of malignancy, a numeric value for x was
P value
obtained through the following calculation: x ⫽ ⫺1.266 ⫹
interval 1.029 (type) ⫺ 0.662 (gender) ⫺ 0.085 (age) ⫹ 0.00089 (age2)
0.85–2.79 0.155 ⫹ 0.247 (TSH concentration). The goiter type was coded as
0.99–1.14 0.08 1 for diffuse or multinodular goiter and 2 for solitary nodule.
1.57– 4.94 ⬍0.001 The patient’s gender was represented by 1 for males and 2
for females. Examples of risk calculations based on this for-
mula are given in Table 5.
1.01– 4.17 0.048 To explore the specific impact of the TSH concentration at
1.49–5.71 0.002
presentation on the predicted risk, we calculated the risk of
malignancy in a hypothetical patient with differing TSH
concentrations (see Table 6). These results illustrate that the
predicted probability of diagnosis of malignancy increases
from less than 10% for serum TSH concentrations at the lower
end of the normal range up to 25% if the same patient has a
TSH concentration at the upper end of the normal range.
Discussion
This study confirms that FNAB is a highly effective
method for identifying those patients with thyroid enlarge-
ment who require surgery because of the presence of ma-
lignancy. For the first time, we have demonstrated that the
serum TSH concentration at presentation, even when within
the normal range, is an independent predictor of the presence
of thyroid malignancy. Our study confirms that patients’
gender and age, as well as the goiter type defined clinically,
are further independent predictors of the presence of thyroid
malignancy. Finally, this study demonstrates that simple
clinical and biochemical criteria can be systematically inte-
grated into the risk stratification of patients referred with
thyroid enlargement.
The sensitivity (88%) and specificity (82%) of FNAB as well
as the low false-negative rate for the diagnosis of malignancy
(1%) in this large UK cohort compare favorably with other
large series (15, 22, 23). Similarly, our rate of nondiagnostic
FNAB samples (6.2% after repeated biopsies) was compara-
ble with that reported by others as between 5 and 20% (15,
17, 18).
Measurement of serum TSH, which is a highly sensitive
determinant of thyroid dysfunction, is the recommended
biochemical test in the initial evaluation of patients present-
ing with thyroid enlargement (11, 19). A subgroup of 1183
subjects within our cohort had a serum TSH measurement in
a sensitive assay performed, and subsequent fT4 (as well as
fT3 measurements in subjects with below-normal TSH con-
centrations) confirmed the absence of overt thyroid dysfunc-
tion in those with TSH measurements outside the reference
range. Subclinical hyperthyroidism was present in 15% (n ⫽
TSH concentration Calculated risk of diagnosis
Clinical goiter type
(mU/liter) of malignancy (%)
Solitary nodule 4.0 49.6
Multinodular goiter 0.5 3.2
Diffuse goiter 6.0 31.2
4300 J Clin Endocrinol Metab, November 2006, 91(11):4295– 4301
TABLE 6. Calculation of probability of diagnosis of malignancy
based on simple clinical and biochemical parameters
Age Clinical TSH concentration
Gender
(yr) goiter type (mU/liter)
Female 40 Solitary nodule 0.3
Female 40 Solitary nodule 0.5
Female 40 Solitary nodule 1.0
Female 40 Solitary nodule 3.0
Female 40 Solitary nodule 5.0
Female 40 Solitary nodule 6.0
The impact of changing TSH concentrations on the calculated risk
is demonstrated by varying the presenting TSH measurement and
keeping the other parameters unchanged.
126) of our patients, a prevalence that is higher than that
generally reported (0.5– 6.3%) in population-based surveys
but consistent with reported prevalences in subjects with
nodular thyroid disease (24 –26). The rate of subclinical hy-
pothyroidism in our patients who had sensitive TSH mea-
surements was 2.3% (n ⫽ 27), a prevalence that is somewhat
lower than that reported by others, again in population-
based surveys (2, 24, 27).
We demonstrated for the first time that the risk of diag-
nosis of malignancy rises in parallel with the serum TSH
concentration at presentation, and further analysis indicated
significantly increased odds ratios for the presence of ma-
lignancy in patients with TSH greater than 1.8 mU/liter after
adjustment for patients’ gender, age, and goiter type (Table
3). It is well documented that TSH has a trophic effect on
thyroid cancer growth, which is most likely mediated by TSH
receptors on tumor cells (28, 29), and furthermore that TSH
suppression is an independent predictor of relapse-free sur-
vival from differentiated thyroid cancer (30). We propose
that the risk increase associated with serum TSH concentra-
tions in the upper half of the normal range, and even more
strikingly in those whose TSH measurements were above
normal, may at least in part be mediated by this trophic effect
of TSH. An alternative explanation is that patients with lower
TSH concentrations were developing autonomous function,
which is itself associated with lower rates of malignancy (15,
23, 32).
Twenty-seven patients had above-normal serum TSH con-
centrations (⬎5.5 mU/liter), and antimicrosomal antibodies
were present in 66.6% (n ⫽ 18 subjects) of those 27, consistent
with the diagnosis of autoimmune thyroiditis in these pa-
tients. When we assessed the association between the pres-
ence of thyroid antibodies and final outcome, we detected
significantly higher rates of malignancy (P ⫽ 0.02) in those
with detectable antimicrosomal antibodies, although further
statistical analysis, taking into account the serum TSH con-
centration, did not identify patients’ antibody status as an
independent risk predictor. Previous studies have found the
presence of antithyroglobulin and antithyroid peroxidase
antibodies not to be useful in distinguishing benign from
malignant lesions (33).
An increased risk of underlying malignancy in men, com-
pared with women, has been demonstrated previously (5, 7,
23), although a few smaller studies indicated patients’ gender
not to be helpful in predicting risk of carcinoma (34, 35). More
recently, a retrospective case series of 1009 patients evalu-
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
Boelaert et al. • Serum TSH and Prediction of Thyroid Malignancy
ating the natural history of cytologically benign thyroid nod-
ules using ultrasonography indicated that patient’s gender
Calculated risk
did not predict nodule growth (36). In our series, signifi-
cantly increased rates of malignancy were detected in men,
of diagnosis of
malignancy (%) regardless of age and goiter type, and logistic regression
8.1 analysis identified patients’ gender as an independent pre-
8.4 dictor of thyroid malignancy (adjusted odds ratio 1.63, P ⫽
9.4
0.046, Table 2). We also identified patients’ age as an inde-
14.6
21.9 pendent risk predictor for malignancy (P ⫽ 0.004, Table 2)
26.4 after correcting for gender and clinically determined goiter
type, in accord with previous studies (5, 21, 23, 37).
Notably, we found that goiter type assessed by physical
examination alone was significantly associated with final
diagnosis. It is well recognized that examination of the neck
is poorly specific in the classification of goiter, and ultra-
sound examination often reveals the presence of multiple
nodules in patients judged clinically to have a single nodule
or diffuse thyroid enlargement (5, 23, 38). Because ultra-
sonography was not routinely performed in our practice [or
that of many other centers (13, 14)] at the time of data col-
lection and because the aim of this study was to identify
simple clinical or biochemical predictors of risk of malig-
nancy, we chose to analyze findings with respect to clinical
examination alone. Strikingly, our results indicate that the
presence of a clinically solitary lesion was associated with
increased malignancy rates, compared with those patients
presenting with clinically diffuse or multinodular goiters.
Moreover, regression analysis confirmed the goiter type as-
sessed clinically to be an independent predictor of risk (odds
ratio 2.77, P ⬍ 0.001, Table 2) after correcting for patients’ age
and gender. Previous studies performed by us and others
have indicated increased cancer risk in nodules deemed sol-
itary by palpation (5, 9, 23, 31, 39), although others have
found similar rates of malignancy when comparing solitary
lesions and multinodular goiters defined clinically (8).
Experts advocate thyroidectomy, regardless of cytology
results in subjects with a high clinical suspicion of malig-
nancy (5, 21, 23). However, few studies have attempted to
integrate clinical or laboratory parameters systematically
with the results of FNAB into the selection process of patients
who need surgery. One previous report (31) aimed to de-
velop clinical criteria capable of predicting malignancy in 149
patients with an FNAB diagnosis of follicular neoplasm and
concluded that features such as gender, nodule size, and
character of the gland by palpation can be systematically
integrated into the decision analysis before surgical referral
(31). We demonstrate that the presenting TSH concentration
is an additional useful parameter in the prediction of prob-
ability of underlying malignancy. The results displayed in
Table 6 illustrate that the calculated risks based on our for-
mula rise dramatically with increasing TSH concentrations if
the other parameters (patients’ age, gender, and goiter type)
remain constant.
We acknowledge the following limitations to this study: 1)
the final diagnosis of benign or malignant disease was not
confirmed histologically in all but was based on a minimum
of 2 yr of clinical follow-up and repeat FNAB in the majority;
2) we did not analyze serum TSH measurements for the
whole cohort because the sensitive TSH assay became avail-
able only after 1988; and 3) patients’ goiter types were clas-
Boelaert et al. • Serum TSH and Prediction of Thyroid Malignancy
sified clinically based on findings obtained by physical ex-
amination alone.
Based on our findings, we have, for the first time, been able
to calculate a formula predicting a diagnosis of thyroid ma-
lignancy, taking into account patient’s age, gender, goiter
type evaluated clinically, and the serum TSH concentration
at presentation, factors that we found to be independently
associated with likelihood of diagnosis of thyroid malig-
nancy. This risk prediction, based on simple clinical and
biochemical parameters, can be used as an adjunct to results
of FNAB to identify those who require further investigation
and/or surgical intervention. Prospective studies are re-
quired to define the role of this risk prediction in refining
clinical management of the larger patient population.
Acknowledgments
Received March 9, 2006. Accepted July 18, 2006.
Address all correspondence and requests for reprints to: Professor
J. A. Franklyn, Division of Medical Sciences, Queen Elizabeth Hospital,
Edgbaston, Birmingham B15 2TH, United Kingdom. E-mail:
j.a.franklyn@bham.ac.uk; or Dr. K. Boelaert, Division of Medical Sci-
ences, Queen Elizabeth Hospital, Edgbaston, Birmingham B15 2TH,
United Kingdom. E-mail: k.boelaert@bham.ac.uk.
This work was supported by the Wellcome Trust (to K.B.), the Re-
search Committee of the former West Midlands Regional Health Au-
thority, and the Research Committee of the University Hospital Bir-
mingham Charities (to J.H., R.L.H., J.C.W., J.A.F., and M.C.S.). These
sponsors had no role in design and conduct of the study, data collection/
analysis, or role in preparation or approval of the manuscript.
K.B., J.H., R.L.H., J.C.W., M.C.S., and J.A.F. have nothing to declare.
References
1. Vander JB, Gaston EA, Dawber TR 1968 The significance of nontoxic thyroid
nodules. Final report of a 15-year study of the incidence of thyroid malignancy.
Ann Intern Med 69:537–540
2. Tunbridge WM, Evered DC, Hall R, Appleton D, Brewis M, Clark F, Evans
JG, Young E, Bird T, Smith PA 1977 The spectrum of thyroid disease in a
community: the Whickham survey. Clin Endocrinol (Oxf) 7:481– 493
3. Marqusee E, Benson CB, Frates MC, Doubilet PM, Larsen PR, Cibas ES,
Mandel SJ 2000 Usefulness of ultrasonography in the management of nodular
thyroid disease. Ann Intern Med 133:696 –700
4. Sherman SI 2003 Thyroid carcinoma. Lancet 361:501–511
5. Hegedus L, Bonnema SJ, Bennedbaek FN 2003 Management of simple nod-
ular goiter: current status and future perspectives. Endocr Rev 24:102–132
6. Shibata Y, Yamashita S, Masyakin VB, Panasyuk GD, Nagataki S 2001 15
years after Chernobyl: new evidence of thyroid cancer. Lancet 358:1965–1966
7. Mazzaferri EL 1993 Management of a solitary thyroid nodule. N Engl J Med
328:553–559
8. Belfiore A, La Rosa GL, La Porta GA, Giuffrida D, Milazzo G, Lupo L,
Regalbuto C, Vigneri R 1992 Cancer risk in patients with cold thyroid nodules:
relevance of iodine intake, sex, age, and multinodularity. Am J Med 93:363–369
9. Franklyn JA, Daykin J, Young J, Oates GD, Sheppard MC 1993 Fine needle
aspiration cytology in diffuse or multinodular goitre compared with solitary
thyroid nodules. BMJ 307:240
10. Kumar H, Daykin J, Holder R, Watkinson JC, Sheppard MC, Franklyn JA
1999 Gender, clinical findings, and serum thyrotropin measurements in the
prediction of thyroid neoplasia in 1005 patients presenting with thyroid en-
largement and investigated by fine-needle aspiration cytology. Thyroid
9:1105–1109
11. Cooper DS, Doherty GM, Haugen BR, Kloos RT, Lee SL, Mandel SJ, Maz-
zaferri EL, McIver B, Sherman SI, Tuttle RM 2006 Management guidelines
for patients with thyroid nodules and differentiated thyroid cancer. Thyroid
16:1–33
12. Schlumberger MJ 1998 Papillary and follicular thyroid carcinoma. N Engl
J Med 1998 338:297–306
JCEM is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the
endocrine community.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 05:27 For personal use only. No other uses without permission. . All rights reserved.
J Clin Endocrinol Metab, November 2006, 91(11):4295– 4301 4301
13. Bennedbaek FN, Perrild H, Hegedus L 1999 Diagnosis and treatment of the
solitary thyroid nodule. Results of a European survey. Clin Endocrinol (Oxf)
50:357–363
14. Bennedbaek FN, Hegedus L 2000 Management of the solitary thyroid nodule:
results of a North American survey. J Clin Endocrinol Metab 85:2493–2498
15. Gharib H, Goellner JR 1993 Fine-needle aspiration biopsy of the thyroid: an
appraisal. Ann Intern Med 118:282–289
16. Castro MR, Gharib H 2005 Continuing controversies in the management of
thyroid nodules. Ann Intern Med 142:926 –931
17. Carmeci C, Jeffrey RB, McDougall IR, Nowels KW, Weigel RJ 1998 Ultra-
sound-guided fine-needle aspiration biopsy of thyroid masses. Thyroid 8:283–
289
18. Danese D, Sciacchitano S, Farsetti A, Andreoli M, Pontecorvi A 1998 Diag-
nostic accuracy of conventional versus sonography-guided fine-needle aspi-
ration biopsy of thyroid nodules. Thyroid 8:15–21
19. British Thyroid Association and Royal College of Physicians 2002 Guidelines
for the management of thyroid cancer in adults. www.british-thyroid-asso-
ciation.org
20. Gittoes NJ, Miller MR, Daykin J, Sheppard MC, Franklyn JA 1996 Upper
airways obstruction in 153 consecutive patients presenting with thyroid en-
largement. BMJ 312:484
21. Hamming JF, Goslings BM, van Steenis GJ, van Ravenswaay CH, Hermans
J, van de Velde CJ 1990 The value of fine-needle aspiration biopsy in patients
with nodular thyroid disease divided into groups of suspicion of malignant
neoplasms on clinical grounds. Arch Intern Med 150:113–116
22. Hamburger JI 1994 Diagnosis of thyroid nodules by fine needle biopsy: use
and abuse. J Clin Endocrinol Metab 79:335–339
23. Hegedus L 2004 Clinical practice. The thyroid nodule. N Engl J Med 351:
1764 –1771
24. Canaris GJ, Manowitz NR, Mayor G, Ridgway EC 2000 The Colorado thyroid
disease prevalence study. Arch Intern Med 160:526 –534
25. Hollowell JG, Staehling NW, Flanders WD, Hannon WH, Gunter EW, Spen-
cer CA, Braverman LE 2002 Serum TSH, T(4), and thyroid antibodies in the
United States population (1988 to 1994): National Health and Nutrition Ex-
amination Survey (NHANES III). J Clin Endocrinol Metab 87:489 – 499
26. Parle JV, Franklyn JA, Cross KW, Jones SC, Sheppard MC 1991 Prevalence
and follow-up of abnormal thyrotrophin (TSH) concentrations in the elderly
in the United Kingdom. Clin Endocrinol (Oxf) 34:77– 83
27. Surks MI, Ortiz E, Daniels GH, Sawin CT, Col NF, Cobin RH, Franklyn JA,
Hershman JM, Burman KD, Denke MA, Gorman C, Cooper RS, Weissman
NJ 2004 Subclinical thyroid disease: scientific review and guidelines for di-
agnosis and management. JAMA 291:228 –238
28. Carayon P, Thomas-Morvan C, Castanas E, Tubiana M 1980 Human thyroid
cancer: membrane thyrotropin binding and adenylate cyclase activity. J Clin
Endocrinol Metab 51:915–920
29. Ichikawa Y, Saito E, Abe Y, Homma M, Muraki T 1976 Presence of TSH
receptor in thyroid neoplasms. J Clin Endocrinol Metab 42:395–398
30. Pujol P, Daures JP, Nsakala N, Baldet L, Bringer J, Jaffiol C 1996 Degree of
thyrotropin suppression as a prognostic determinant in differentiated thyroid
cancer. J Clin Endocrinol Metab 81:4318 – 4323
31. Tuttle RM, Lemar H, Burch HB 1998 Clinical features associated with an
increased risk of thyroid malignancy in patients with follicular neoplasia by
fine-needle aspiration. Thyroid 8:377–383
32. Mann K 1998 Evaluation of risk in autonomously functioning thyroid nodules.
Exp Clin Endocrinol Diabetes 106(Suppl 4):S23–S26
33. DeGroot LJ, Hoye K, Refetoff S, Van Herle AJ, Asteris GT, Rochman H 1977
Serum albumin and antibodies in the diagnosis of thyroid cancer. J Clin
Endocrinol Metab 45:1220 –1223
34. McHenry CR, Walfish PG, Rosen IB 1993 Non-diagnostic fine needle aspi-
ration biopsy: a dilemma in management of nodular thyroid disease. Am Surg
59:415– 419
35. Raber W, Kaserer K, Niederle B, Vierhapper H 2000 Risk factors for malig-
nancy of thyroid nodules initially identified as follicular neoplasia by fine-
needle aspiration: results of a prospective study of one hundred twenty pa-
tients. Thyroid 10:709 –712
36. Alexander EK, Hurwitz S, Heering JP, Benson CB, Frates MC, Doubilet PM,
Cibas ES, Larsen PR, Marqusee E 2003 Natural history of benign solid and
cystic thyroid nodules. Ann Intern Med 138:315–318
37. McHenry C, Smith M, Lawrence AM, Jarosz H, Paloyan E 1988 Nodular
thyroid disease in children and adolescents: a high incidence of carcinoma. Am
Surg 54:444 – 447
38. Tan GH, Gharib H, Reading CC 1995 Solitary thyroid nodule. Comparison
between palpation and ultrasonography. Arch Intern Med 155:2418 –2423
39. Mazzaferri EL, Los Santos ET, Rofagha-Keyhani S 1988 Solitary thyroid
nodule: diagnosis and management. Med Clin North Am 72:1177–1211