Introduction
Thyroid carcinoma is the most common malignancy of the endocrine system, accounting
for 1% of all newly diagnosed cancers [1] and 80–90% of them are papillary thyroid carcinoma
(PTC) [2]
[3]. The prognosis is favorable when thyroid cancers are treated
appropriately. Cervical lymph node metastasis (LNM) has been detected in
27–46% of patients with newly diagnosed PTCs, and
5–20% of them have locoregional recurrences after initial surgery,
often involving cervical lymph nodes and making them candidates for secondary
surgical procedures [4]
[5]. The presence of cervical LNM increases the
risk of locoregional recurrence. Prophylactic central neck dissection reduces
locoregional recurrence but is associated with a higher risk of postoperative
complications. Therefore, it is recommended to be preferred by high-volume surgeons
in high-risk patients with advanced primary tumors [6]. The structural incomplete response to initial therapy results in
persistent locoregional disease, occuring in 2–6% of ATA
low-risk, 19–28% of intermediate-risk, and 67–75% of
high-risk patients [7]
[8]. The revised 2009 ATA Management Guidelines
for Patients with Thyroid Nodules and DTC recommend surgical management of
locoregional disease in the absence of distant metastases [9].
Thyroglobulin (Tg) is a tumor marker for DTC, if anti-Tg antibodies are negative
[10]
[11]. But routine preoperative measurement of serum Tg or anti-Tg
antibodies is not recommended [12].
Nevertheless, it is a reliable marker in patients treated with total thyroidectomy
(TTx) and even more consistent – for those – who had remnant
ablation or treated with radioactive iodine (RAI). On routine follow-up measures
after surgery with or without RAI, serum Tg measurements and neck ultrasonography
are performed [12]. Tg levels increase with
increasing TSH serum levels and tumor burden in patient with metastatic disease
[13]
[14]. To our knowledge, few studies have investigated this relation,
probably due to the difficulties in estimating the tumor burden with quantitative
analysis [13]
[15]. While considerably high Tg levels do suggest distant metastasis. The
relationship between localization, tumor burden of the metastases, and serum Tg
levels is not clear for the moderately increased levels, which almost always
indicates locoregional metastases of different neck compartments. Some studies have
investigated the Tg cutoff level for distant organ metastasis [16]
[17],
but to our knowledge, none assessed the variations in serum Tg levels with regard
to
the involved cervical lymph node compartment in patients with locoregional
metastasis. Thus, we aimed to verify the effectiveness of Tg levels to differentiate
the localizations of the metastatic lymph node in the neck independent of tumor
burden, RAI treatment, and serum TSH levels in patients with
persistent/recurrent locoregional metastases.
Patients and Methods
Study population and inclusion criteria
This retrospective study included patients who underwent therapeutic neck
dissection for persistent/recurrent locoregional LNM in a tertiary care
university hospital between January 2006 and August 2020. All data were
collected from electronic and manual medical records. University Ethical
Committee on human research approved the study (Project Number:
İ1–56–21).
Simultaneous serum Tg, TSH and anti-Tg levels before lymph node dissection (LND)
were recorded. Results from neck ultrasonography (US), computed tomography (CT)
and histopathological exam were analyzed to determine the localization, number
and largest diameter of metastatic lymph nodes. The inclusion criteria were as
follows: 1) TTx with or without LND for PTC as an initial surgery, 2)
persistent/recurrent metastatic cervical lymph node/s found
radiologically (US/CT) and proven by Tg washouts and/or FNAC
after initial surgery during follow-up period, 3) an appropriate neck dissection
for the proven locoregional metastasis and 4) biochemical and structural
complete response after neck dissection with or without RAI treatment. Exclusion
criteria were as follows: 1) radiologically detectable residual thyroid tissue
before neck dissection, 2) anti-Tg antibody positivity, 3) distant metastasis at
the time of neck dissection, 4) inoperable cervical LNM, 5)<1 month time
period between thyroidectomy and neck dissection to measure the dip (lowest) Tg
level, 6) radiologically detectable (macroscopic) (US/CT) cervical
fibro-muscular tissue invasion with LNM, 7) biochemical or structural incomplete
response, at the final evaluation performed after neck dissection, and 8)
reported residual tissue radiologically before surgery or having had remnant
thyroid tissue in histopathological reports.
Thus, in this study we assumed that serum Tg levels in the patients represented
tumor burden caused by LNM since the thyroid resection is total, and possible
remnant tissue is completely ablated, patients with distant organ metastases are
excluded. Tumor burden caused by metastatic lymph nodes were represented with
the largest diameter of LNM and number of LNM in histopathology and/or
imaging studies. To evaluate the localization of metastatic node/s,
preoperative US reports/maps, CT reports, and/or
histopathological results after surgery were used.
Localization of LNM and neck dissections were classified and performed
respectively, according to the American Head and Neck Society and the American
Academy of Otolaryngology classification [18]. Our sample was classified in accordance with the type of
dissection central (C) LNM group for central compartment neck dissection,
lateral (L) LNM group for lateral compartment neck dissection 3)
central+lateral (C+L) LNM group, for those with LNM in both
central and lateral neck compartments and both compartments had been dissected
in the same session. Lateral compartment dissection was referred right, left or
both sides.
Thyroglobulin was measured with immunochemiluminometric DxI 800 assay (Beckman
Coulter, USA). The Tg assay used for the analyses has a functional sensitivity
of 0.1 mg/l with 95% confidence.
Statistical analysis
Statistical analysis was performed using SPSS version 22.0 and R version 4.1.0.
Descriptive statistics were presented as counts and percentages for categorical
variables and standard and mean deviations for continuous variables. Normality
assumption was tested by Kolmogorov – Smirnov test with Lilliefors
Significance Correction. Natural logarithm transformation was applied to the
outcome variable to achieve a normal distribution and homogeneous variances.
Chi-square test was performed for clustered data and non-parametric analysis of
clustered receiver operating characteristics (ROCs), which were performed to
account for intracluster correlation among multiple measurements from the same
patient. An ROC curve was used to decide the diagnostic value of serum
thyroglobulin to distinguish localization of LNM. To increase the specificity
and sensitivity of the cutoff Tg level ROC analysis was performed both in the
whole sample and in two groups suppressed (<0.1 mU/l) and
non-suppressed (>0.1 mU/l) according to TSH.
The areas under the curves (AUC) and 95% confidence intervals for all
variables were quantified. Cutoff values were selected using with
Youden’s index. Measurements were dichotomized according to cutoff
values, and specificity, sensitivity and negative and positive predictive values
were calculated. To account for the clustered structure of the data, linear
mixed-effects models were used to analyze factors that affected serum Tg
variation. Serum Tg level was the dependent variable. Simultaneous TSH level,
the largest diameter of LNM, the number of LNM, and the condition of receiving
RAI after the first surgery independent variables in the model to eliminate
their effects on Tg. The Bonferroni correction was applied to control Type I
error rate. A p-value of<0.05 was considered significant.
Baseline sample characteristics
We enrolled 222 patients diagnosed with PTC who underwent neck dissection for
persistent/recurrent locoregional LNM, between January 2006 and August
2020 at University Hospitals. After assessing the eligibility criteria and data,
143 patients underwent neck dissections for a total of 172 for
persistent/recurrent locoregional PTC. Of them, 19 underwent neck
dissection twice, and 5 for three times due to recurrence. Flow diagram of the
sampling is shown in [Fig. 1].
Fig. 1 Flow diagram of the sampling recurrent/persistent PTC
patients (n=222).; PTC: Papillary Thyroid Carcinoma, Tg:
Thyroglobulin, LNM: Lymph Node Metastasis.
Baseline characteristics of the whole sample are shown in [Table 1]. Mean age was 47.0±13.1
years, and 73.3% (n=126) were females. The mean age of diagnosis
was 38.3±12.5 years, and disease duration at the time of neck dissection
was 3.7±4.0 years. 90.1% (n=155) of the sample had stage
I PTC, and the rest (9.9%, n=17) had stage II PTC. Age, sex, age
at diagnosis, disease duration at the time of neck dissection, largest diameter
of primary tumor and TSH levels were not significantly different between the LNM
groups ([Table 1]).
Table 1 Descriptive and clinical characteristics of the
study samples (n=172).
|
C LNM n=47
|
L LNM n=99
|
C+L LNM n=26
|
p
|
|
Female
|
32 (68.1%)
|
74 (74.7%)
|
20 (76.9%)
|
0.658
|
|
Stage*, n (%)
|
|
|
|
0.834
|
|
I
|
43 (91.5%)
|
88 (88.9%)
|
24 (92.3%)
|
|
|
II
|
4 (8.5%)
|
11 (11.1%)
|
2 (7.7%)
|
|
|
Age, mean±SD
|
48.0±11.3
|
47.9±13.2
|
42.4±14.5
|
0.247
|
|
Age at diagnosis, mean±SD
|
39.4±11.3
|
38.5±12.6
|
34.2±14.1
|
0.989
|
|
Disease duration**,
mean±SD
|
3.91±3.45
|
3.87±4.32
|
2.91±3.89
|
0.582
|
|
Diameter of primary tumor (mm), mean±SD
|
16.88±12.03
|
17.14±10.45
|
20.76±11.88
|
0.311
|
|
Number of LNM, mean±SD
|
2.81±2.68
|
3.34±3.61
|
6.00±3.45a
|
<0.001
|
|
The largest diameter of LNM (mm), mean±SD
|
11.07±5.79
|
13.94±6.38b
|
13.51±7.21
|
0.046
|
|
Tg, μg/l, mean±SD
|
1.43±2.59c
|
3.70±5.78
|
8.60±15.94
|
<0.001
|
|
TSH (mU/l), mean±SD
|
1.53±2.91
|
2.30±8.06
|
0.74±1.23
|
0.201
|
LNM: Lymph Node Metastasis; C: Central; L: Lateral; C+L:
Central+Lateral; Tg: Thyrogloblin.; aDifferent from C
and L (p<0.001).; bDifferent from C
(p=0.043).; cDifferent from L and C+L
(natural logarithm was applied, p=0.006 and p<0.001,
respectively).; *There were no patients with stage
III or IV thyroid cancer.; **Years, at the
time of neck dissection.
The number of metastatic nodes was higher in the C+L LNM group than those
of the other groups (p<0.001). While that of C LNM and L LNM groups did
not differ from each other (p=1.000) ([Table 1]). C+L LNM group had similar, mean largest diameter
of metastatic nodes with C LNM and L LNM groups, whereas L LNM group had higher
mean largest diameter than C LNM group (p=0.043) ([Table 1]).
The mean pre-neck dissection Tg levels were
1.43±2.59 μg/l for C LNM,
3.7±5.78 μg/l for L LNM, and
8.60±15.94 μg/l for C+L LNM groups.
While the difference between Tg levels of L LNM and C+L LNM groups was
not significant (p=0.183), C LNM group exhibited significantly lower
mean Tg level than those of other groups (L LNM p=0.006 and C+L
LNM p<0.001) ([Table 1]).
RAI was administered to approximately 80% (n=137) of the patients
after initial surgery and 66% (n=113) after neck dissection for
persistent/recurrent locoregional disease.
The effect of lymph node metastases localization on Tg level
We analyzed the effect of LNM localization on Tg level using mixed-effects model.
The number of metastatic nodes, the largest diameter of the metastatic nodes and
serum TSH levels were significantly associated with higher Tg levels
(p=0.028, p=0.015, p<0.001, respectively), whereas
receiving RAI after initial surgery is associated with lower Tg levels (
p=0.033).
After adjusting for the possible confounding factors, such as number of
metastatic nodes, the largest diameter of metastatic nodes, serum TSH levels and
RAI treatment, the mixed-effects model revealed that mean pre-neck dissection Tg
levels of patients with L LNM was significantly higher than those with C LNM
(with a mean difference 1.76±1.26 μg/l
95% CI 1.12–2.76) (p=0.014). In addition, metastases on
central+lateral compartment rather than central compartment increased Tg
level to a mean±SD 2.53±1.38 μg/l
(95% CI 1.33–4.79) (p=0.002). However, Tg levels was not
significantly associated on lateral compartment metastasis and
central+lateral compartment metastasis (p=0.13).
Biochemical response after additional therapy for persistent/recurrent
locoregional disease was not associated with RAI treatment after neck dissection
(for whole sample p=0.853, C LNM group p=0.119, L LNM group
p=0.117, C+L LNM group p=0.255).
Diagnostic ability of thyroglobulin level to predict localization of lymph
node metastases
ROC curve were used to determine the cutoff Tg level that could discriminate
metastatic compartments ([Tables 2]
[3]). To discriminate C LNM from L LNM, the
optimal cutoff was 1.05 μg/l with the AUC of
0.651±0.049 (p=0.003, 95%
CI=0.555–0.747). To discriminate C LNM from L LNM and
C+L LNM, the optimal cutoff was 1.05 μg/l with
an AUC of 0.672±0.045 (p<0.001, 95%
CI=0.584–0.760) ([Table
2]). The sample was divided into two groups according to the TSH value
as<0.1 mU/l (suppressed, n=79) and>0.1
mU/l (non-suppressed, n=94). Tg levels were not significantly
different between C LNM and other groups with suppressed TSH. To discriminate C
LNM from L LNM in the group with non-suppressed TSH, the optimal cutoff was
1.06 μg/l with the AUC of 0.703±0.060
(p<0.001, 95% CI=0.586–0.819). Sensitivity,
specificity, positive predictive value (PPV) and negative predictive value (NPV)
were 70.7%, 70.4%, 83.7% and 52.8% for this
cutoff, respectively. To discriminate C LNM from L LNM and C+L LNM, the
optimal cutoff was 1.05 μg/l with an AUC of
0.714±0.058 (p<0.001, 95%
CI=0.603–0.824) Sensitivity, specificity, PPV and NPV were
74.7%, 70.4%, 87.7%, and 50% for this cutoff,
respectively ([Table 3]).
Table 2 The results of ROC curve analysis in the whole
group (n=172).
|
C LNM vs. L LNM (95% CI)
|
C LNM vs. L and C+L LNM (95% CI)
|
|
Cutoff (μg/l)
|
1.05
|
1.05
|
|
AUC
|
0.651±0.049 (0.555–0.747)
|
0.672±0.045 (0.584–0.760)
|
|
Sensitivity
|
56.6% (46.8–65.9%)
|
61.9% (53.2–69.9%)
|
|
Specificity
|
72.3% (58.2–83.0%)
|
72.3% (58.2–83.0%)
|
|
Positive predictive value (PPV)
|
81.1% (70.4–88.6%)
|
85.7% (77.1–91.4%)
|
|
Negative predictive value (NPV)
|
44.2% (33.6–55.3%)
|
41.4% (31.4–52.3%)
|
LNM: Lymph Node Metastasis; C: Central; L: Lateral; CL:
Central+Lateral; CI: Confidence interval; AUC: Area under the
ROC curve.
Table 3 The results of ROC curve analysis in the group
with non-suppressed TSH (n=94).
|
C LNM vs. L LNM (95% CI)
|
C LNM vs. L and C+L LNM (95% CI)
|
|
Cutoff (μg/l)
|
1.06
|
1.05
|
|
AUC
|
0.703±0.060 (0.586–0.819)
|
0.714±0.058 (0.603–0.824)
|
|
Sensitivity
|
70.7% (58–80.8%)
|
74.7% (63.8–83.2%)
|
|
Specificity
|
70.4% (51.5–84.2%)
|
70.4% (51.5–84.2%)
|
|
Positive predictive value (PPV)
|
83.7% (71–91.5%)
|
87.7% (77.2–93.5%)
|
|
Negative predictive value (NPV)
|
52.8% (37–68%)
|
50% (34.9–65.2%)
|
LNM: Lymph Node Metastasis; C: Central; L: Lateral; CL:
Central+Lateral; CI: Confidence interval; AUC: Area under the
ROC curve.
Discussion
To the best of our knowledge, this is the first study that examined the relationship
between serum Tg levels and the location of metastatic neck lymph nodes in PTC
patients. We found that the increase in TSH and the tumor burden (defined as the
largest diameter and number of metastatic nodes) resulted in higher serum Tg levels
which was consistent with the literature [13]
[14]
[15]. The patients who had recurrences in the
lateral compartment had significantly higher Tg levels, when compared to those with
the central lymph node metastasis.
Rosenbaum-Krumme et al. showed a positive correlation of serum Tg and tumor mass in
their study which quantitatively examined the relationship between tumor burden and
Tg via a mathematical formulation mainly by using serum Tg level [15]. Authors of a French study showed the
significant relationship between Tg and tumor burden, in a group of 75 PTC patients
who had undergone thyroidectomy and completely ablated thyroid residue.
‘Tg/TSH’ was used to eliminate the effect of TSH and tumor
burden defined as number of metastatic lymph nodes and their total surface
area/volume [13]. However, they
assumed that the effect of TSH on Tg is linear. Although both parameters are
positively correlated, it is unclear if it was linear. Consistent with the
literature we found that TSH level is positively correlated with Tg level before
neck dissection. We believe the mixed-effects regression model provides a sensitive
adjustment for the possible confounders. Our results showed a correlation between
involved compartment and serum Tg levels prior to LND in
persistent/recurrent locoregional PTC cases. C LNM resulted in significantly
lower serum Tg levels than lateral or both central and lateral LNM in spite of the
possible confounding effects of serum TSH levels, tumor burden and RAI
treatment.
The limitations of our study were its retrospective design, lack of volumetric
measurements for metastatic lymph nodes, and the relatively small sample size of the
C+L LNM group compared to the other groups. The mean Tg level of C+L
LNM was twofold higher than that of L LNM; but, the difference was not significant
between the groups which may be due to the small sample size of C+L LNM
group. In addition, it was not possible to detect a cutoff Tg level in the
suppressed group in ROC analyzes because there were many overlapping thyroglobulin
values.
Neck US for all compartments in experienced hands and Tg wash outs are the gold
standards for preoperative diagnosis of recurrent disease [19]
[20]
[21]. Preoperative metastatic
lymph node mapping by an experienced sonographer, becomes even more important since
the risk of complications is higher in repeated neck dissections. The serum Tg
levels may be used to help the sonographers for the localization of LNM. The cutoff
values for whole sample we found are not high diagnostic accuracy due to significant
overlap in the TSH suppressed group and a high variance in Tg values. The Tg values
for TSH suppressed group were not significantly different between C-LNM and other
groups. In contrast, in the non-suppressed group, the cutoff to discriminate C LNM
from L LNM and C+L LNM has a higher diagnostic accuracy.
Our findings suggest that non-suppressed Tg level above
1.05 μg/l indicates lateral or central+lateral LNM,
while levels below 1.05 μg/l indicates central LNM.
Discordance between LNM localization and Tg level may lead the sonographer
and/or clinician for further investigation of the lateral compartment. If
non-suppressed Tg level is above 1.05 μg/l and lateral LNM
was not detected on neck sonography, a second look ultrasonography may be useful.
The pathophysiological mechanism under this significant difference is is difficult
to elucidate, but it may be related to complex lymphatic drainage of different neck
compartments.
