Enhancement of lymph node metastasis and distant metastasis of thyroid carcinoma

A multivariate analysis of clinical risk factors

Abstract

BACKGROUND

The mechanisms of local and distant metastases are imperfectly understood. The goal of the current study was to add to the body of knowledge regarding local and distant metastases of thyroid malignancies.

METHODS

The authors performed multivariate analysis of 573 patients who underwent surgery between November 1994 and May 2002 for follicular (FTC; n = 100), papillary (PTC; n = 236), or medullary thyroid carcinoma (MTC; n = 237) at a university hospital.

RESULTS

In multivariate analysis, extrathyroidal extension consistently evolved as the key risk factor for both lymph node metastasis and distant metastasis. This correlation was most pronounced in MTC and least pronounced in FTC. The risk of lymph node metastasis also increased with reoperative status in patients with MTC and with primary tumor diameter in patients with MTC (tumor diameter > 10 mm) and patients with PTC (tumor diameter > 20 mm). In the PTC group, lymph node metastasis was more common among patients younger than age 45. In the MTC group, extrathyroidal growth and distant metastasis were associated exclusively with lymph node metastasis. Lymph node metastasis was the only secondary risk factor for distant metastasis. In the analysis of risk factors for distant metastasis in the FTC and PTC groups, no interaction was found between extrathyroidal growth and lymph node metastasis. This finding suggests that extrathyroidal growth and lymph node metastasis of FTC and PTC, and presumably also MTC, represent separate mechanisms and routes of distant metastasis.

CONCLUSIONS

Screening for both local residual disease and distant metastases should be intensified in the high-risk population of patients whose primary tumors exhibit large diameters, extrathyroidal growth, or lymph node metastasis. Cancer 2003;98:712–9. © 2003 American Cancer Society.

DOI 10.1002/cncr.11581

It is widely accepted that cancer results from the accumulation of mutations in the genes that directly control cell birth and cell death.1 This random accumulation of genetic abnormalities allows for a Darwinian selection of cells that have acquired the crucial genetic changes that can drive the development and progression of malignant disease.2 Unrestrained growth and metastasis to lymph nodes and distant organs are the clinical and morphologic manifestations of such genetic changes on the molecular level.3

The various types of thyroid malignancies feature distinct patterns of lymph node and distant metastases. As a rule of thumb, papillary thyroid carcinoma (PTC) preferentially spreads to lymph nodes, follicular thyroid carcinoma (FTC) to distant organs, and medullary thyroid carcinoma (MTC) to both lymph nodes and distant organs. More recent data indicate that a sequential increase in chromosomal complexity underlies the progression to more aggressive phenotypes of thyroid malignancy.2, 4 Chromosomal instability is the driving force behind both tumor progression and tumor heterogeneity, guaranteeing that no two tumors are exactly alike and that no single tumor is composed of genetically identical cells.1 For all types of thyroid malignancy, this heterogeneity may make it difficult to predict the occurrence of lymph node or distant metastasis based on individual molecular risk factors. Consequently, therapeutic decision-making based on established clinical risk factors that reflect the impact of existing mutations in a given tumor most likely will continue to play a preeminent role in clinical practice.

Unfortunately, the mechanisms and routes of local and distant metastases are imperfectly understood, because of the relative rarity of thyroid carcinoma in the general population, the infrequency of routine systematic lymph node dissection in patients with this condition, and the scarcity of distant metastasis of thyroid carcinoma. A thorough understanding of clinical risk factors is needed to identify the subgroup of high-risk patients who will require more extensive surgery and intensive postoperative screening for lymph node and distant metastases. To identify and assess the magnitude of clinical risk factors predictive of lymph node and distant metastases in patients with thyroid carcinoma, institutional multivariate analysis of 573 patients was conducted.

MATERIALS AND METHODS

Patient Selection

Between November 1994 and May 2002, 713 consecutive patients with FTC (n = 134), PTC (n = 283), or MTC (n = 296) underwent surgery at Martin-Luther-Universität Halle-Wittenberg (Halle/Saale, Germany) for primary (24, 86, and 112 patients, respectively) or recurrent tumors (110, 197, and 184 patients, respectively). These 713 consecutive patients make up the original study base. Due to the unavailability of tumor diameter data, mainly for reoperative patients, the study population was reduced to 573 patients (100 with FTC, 236 with PTC, and 237 with MTC) who were eligible for the analysis of clinical risk factors. The study population for risk factor analysis represented 80.4% of the original study base.

Extent of Surgery

All 713 patients received at least a standard subtotal thyroidectomy at our institution, provided that the procedure had not been performed elsewhere previously. Total thyroidectomy was or had been performed for 684 patients (96%). In addition, 622 patients (87%) underwent standard systematic lymph node dissection in at least 1 lymph node compartment: 596 (84%) underwent systematic dissection of the central lymph node compartment, which extends vertically from the hyoid bone to the thoracic inlet and horizontally between the carotid sheaths.5 The lateral cervical compartments were dissected in 321 (ipsilateral; 45%) and 245 patients (contralateral; 34%), and the mediastinal compartment was dissected in 127 patients (18%). Before surgery, informed consent was obtained for each procedure mentioned. All specimens underwent pathologic examination.

Pathologic Examination and Tumor Staging

After macroscopic evaluation by the pathologist, the entire thyroid gland was divided vertically to separate the left and right lobes. The thyroid halves then were sectioned horizontally from the superior to the inferior pole, as described previously.6 After fixation in formalin, the entire thyroid gland was embedded in paraffin. Soft tissue and lymph nodes were processed separately. Conventional staining (hematoxylin and eosin) and, when appropriate, thyroglobulin and calcitonin immunohistochemistry were performed on every surgical specimen, using the standard avidin-biotin complex–peroxidase approach. Tumors were categorized using the World Health Organization classification system.7 Staging was performed according to the American Joint Committee on Cancer staging system8 and the International Union Against Cancer (UICC) TNM classification system.9 Although a diagnosis of lymph node metastasis always required pathologic confirmation, the requirement was waived for distant metastasis when there was unequivocal evidence obtained using ultrasonography (liver), computerized tomography, magnetic resonance imaging, positron emission tomography, scintiscan, or a combination of these methods.

Statistical Analysis

Categoric and continuous data were tested on univariate analysis using the two-tailed Fisher exact test and the two-tailed exact Mann–Whitney–Wilcoxon rank sum test. Separate conditional logistic regression models then were fitted for each tumor type (FTC, PTC, and MTC) and clinical endpoint (lymph node metastasis and distant metastasis). Independent parameters entered in the initial models were categorized tumor diameter; extrathyroidal extension; operation status (reoperation vs. primary surgery); gender; categorized age; and, for the endpoint of distant metastasis, lymph node metastasis. Subsequently, conditional backward stepwise procedures were performed, using the SPSS software package (Version 11.0; SPSS Inc., Chicago, IL), on the initial models that encompassed the same set of independent parameters under study. In these backward stepwise logistic regression analyses, the least statistically significant parameters were removed one at a time until all the remaining parameters were deemed significant (optimized models). Values of the Spearman correlation coefficient (Spearman rho) were calculated to assess correlations between categoric and continuous parameters. The level of significance was set at P = 0.05.

RESULTS

Patient Selection by Primary Tumor Diameter

To better define the potential for, and the impact of, selection bias due to the exclusion of patients lacking data on primary tumor diameter, univariate analysis of the pathologic characteristics of patients with and without such information was performed (Table 1). Irrespective of tumor type, thyroid lesions of unknown diameter more frequently exhibited extrathyroidal growth and tracheal invasion. It seems that these conditions necessitated piecemeal removal of lesions at primary surgery, making the exact determination of primary tumor diameter unfeasible in these cases. In addition, PTC and MTC lesions of unknown diameter were associated more frequently with lymph node metastasis. MTC lesions of unknown diameter also were associated more frequently with distant metastasis (Table 1). Lymph node metastasis and distant metastasis of lesions of unknown diameter in this setting may have been enhanced by the frequent extrathyroidal extension of these malignancies.

Table 1. Patient Selection by Availability of Primary Tumor Diameter Data

Variable	FTC (%)			PTC (%)			MTC (%)
	Data available (n = 100)	Data unavailable (n = 34)	Pa	Data available (n = 236)	Data unavailable (n = 47)	Pa	Data available (n = 237)	Data unavailable (n = 59)	Pa
FTC: follicular thyroid carcinoma; PTC: papillary thyroid carcinoma; MTC: medullary thyroid carcinoma; LN: lymph node. a Two-tailed Fisher exact test.
T classification
pT1	1	9		31	10		29	23
pT2	57	26		39	22		50	21
pT3	25	17		6	10		5	11
pT4	17	48	0.001	24	59	< 0.001	17	45	< 0.001
LN metastasis
pN1	19	18	1.00	37	70	< 0.001	64	86	0.001
Distant metastasis
M1	17	29	0.14	7	13	0.23	17	32	0.011
Tracheal invasion
Present	5	27	0.001	3	15	0.003	3	11	0.03

Univariate Analysis of Lymph Node Metastasis and Distant Metastasis

Irrespective of tumor type, lymph node and distant metastases (Tables 2, 3) consistently were associated with larger primary tumor diameters and extrathyroidal extension. Distant metastasis (Table 3) also was associated with lymph node metastasis for all tumor types. Furthermore, some tumor type–specific correlations were noted (Table 2). Lymph node–positive patients with PTC were markedly younger, on average, than their lymph node–negative counterparts. In addition, lymph node–positive patients with MTC were more prevalent on reoperation, most likely due to detection bias caused by the persistence in all reoperative patients of hypercalcitoninemia, a condition associated with residual lymph node and distant metastases.

Table 2. Univariate Analysis for Lymph Node Metastasis of Thyroid Carcinoma

Variable	FTC (%)			PTC (%)			MTC (%)
	No LN metastasis (n = 81)	LN metastasis (n = 19)	Pa	No LN metastasis (n = 148)	LN metastasis (n = 88)	Pa	No LN metastasis (n = 85)	LN metastasis (n = 152)	Pa
FTC: follicular thyroid carcinoma; PTC: papillary thyroid carcinoma; MTC: medullary thyroid carcinoma; LN: lymph node. a Two-tailed Fisher exact test and two-tailed exact Mann–Whitney–Wilcoxon rank sum test, respectively.
Tumor diameter (mm)
Median	30	45		13	23		8	21
25–75 percentile	20–50	30–65	0.004	5–25	15–40	< 0.001	3–20	15–34	< 0.001
Extrathyroidal extension	12	37	0.02	13	43	< 0.001	0	26	< 0.001
Primary surgery	16	26	0.33	32	30	0.77	26	69	< 0.001
Male gender	30	32	1.00	25	35	0.10	41	48	0.34
Age at operation (yrs)
Median	54	57		54	46		42	45
25–75 percentile	42–64	51–67	0.44	42–61	31–66	0.04	30–56	33–60	0.09

Table 3. Univariate Analysis for Distant Metastasis of Thyroid Carcinoma

Variable	FTC (%)			PTC (%)			MTC (%)
	No distant metastasis (n = 83)	Distant metastasis (n = 17)	Pa	No distant metastasis (n = 220)	Distant metastasis (n = 16)	Pa	No distant metastasis (n = 197)	Distant metastasis (n = 40)	Pa
FTC: follicular thyroid carcinoma; PTC: papillary thyroid carcinoma; MTC: medullary thyroid carcinoma; LN: lymph node. a Two-tailed Fisher exact test and two-tailed exact Mann–Whitney–Wilcoxon rank sum test, respectively.
Tumor diameter (mm)
Median	30	49		15	35		15	30
25–75 percentile	20–50	33–60	0.006	7–27	19–58	0.005	7–27	19–50	< 0.001
Extrathyroidal extension	12	41	0.008	21	69	< 0.001	7	65	< 0.001
LN metastasis	15	41	0.02	35	75	0.002	57	100	< 0.001
Primary surgery	17	24	0.50	31	25	0.78	41	45	0.73
Male gender	27	47	0.14	28	44	0.25	45	48	0.86
Age at operation (yrs)
Median	53	58		51	58		45	44
25–75 percentile	41–64	54–66	0.12	38–61	29–70	0.51	33–57	31–62	0.89

Multivariate Analysis of Lymph Node Metastasis and Distant Metastasis

The superiority of multivariate analysis over univariate analysis lies in the ability of multivariate analysis to efficiently estimate measures of association while simultaneously controlling for a number of confounding factors. In logistic regression analysis, the odds ratio as a measure of association will overestimate the relative risk when the event rate of the outcome of interest (e.g., lymph node metastasis or distant metastasis) exceeds 10%.10 Therefore, the odds ratios obtained in the current investigation cannot be interpreted as relative risks.

In multivariate analysis, extrathyroidal extension consistently evolved as the primary risk factor for both lymph node metastasis and distant metastasis (Tables 4, 5). This correlation was most pronounced in MTC and least pronounced in FTC, paralleling each tumor type's propensity for lymph node spread.

Table 4. Logistic Regression Analysis for Lymph Node Metastasis of Thyroid Carcinoma

Variable	FTC		PTC		MTC
	OR	P	OR	P	OR	P
FTC: follicular thyroid carcinoma; PTC: papillary thyroid carcinoma; MTC: medullary thyroid carcinoma; OR: odds ratio; NA: not assessable. a All cases of lymph node metastasis were accompanied by extrathyroidal growth. b Optimized model: Only variables with a P value of ≤ 0.1 remaining after backward stepwise logistic regression are displayed.
Initial model
Lesion diameter (mm)
	1.0	reference	1.0	reference	1.0	reference
	5.2	0.14	1.4	0.46	6.9	< 0.001
	7.2	0.09	3.2	0.003	7.5	< 0.001
Extrathyroidal extension, pT4 (vs. pT1–3)	2.9	0.10	4.3	< 0.001	NAa	NAa
Reoperation (vs. primary surgery)	0.6	0.49	1.2	0.66	4.4	< 0.001
Male gender (vs. female)	0.8	0.64	1.7	0.11	1.8	0.08
Age < 45 yrs at surgery (vs. ≥ 45 yrs)	1.6	0.46	2.3	0.009	1.6	0.16
Optimized modelb
			1.0	reference	1.0	reference
			1.4	0.42	6.3	< 0.001
			3.4	0.001	6.7	< 0.001
Extrathyroidal extension, pT4 (vs. pT1–3)	4.1	0.02	4.1	< 0.001	NAa	NAa
Age < 45 yrs at surgery (vs. ≥ 45 yrs)			2.2	0.01
Reoperation (vs. primary surgery)					4.2	< 0.001

Table 5. Logistic Regression Analysis for Distant Metastasis of Thyroid Carcinoma

Variable	FTC		PTC		MTC
	OR	P	OR	P	OR	P
FTC: follicular thyroid carcinoma; PTC: papillary thyroid carcinoma; MTC: medullary thyroid carcinoma; OR: odds ratio; NA: not assessable. a For all cases of distant metastasis, the primary lesion had diameter > 10 mm. b All cases of distant metastasis were accompanied by lymph node metastasis. c Optimized model: Only variables with a P value of ≤ 0.1 remaining after backward stepwise logistic regression are displayed.
Initial model
FTC lesion diameter (mm)
	1.0	reference	1.0	reference	NAa	NAa
	2.5	0.44	0.3	0.22	1.0	reference
	3.6	0.27	1.0	0.97	1.4	0.45
Extrathyroidal extension, pT4 (vs. pT1–3)	2.7	0.13	6.6	0.003	15.8	< 0.001
Lymph node metastasis, pN1 (vs. pN0)	3.1	0.08	2.9	0.13	NAb	NAb
Reoperation (vs. primary surgery)	1.1	0.86	2.0	0.28	0.8	0.68
Male gender (vs. female)	1.7	0.40	1.8	0.29	0.7	0.39
Age < 45 yrs at surgery (vs. ≥ 45 yrs)	0.6	0.42	0.9	0.88	1.2	0.67
Optimized modelc
Extrathyroidal extension, pT4 (vs. pT1–3)	3.9	0.03	5.6	0.004	15.8	< 0.001
Lymph node metastasis, pN1 (vs. pN0)	3.1	0.07	3.2	0.07	NAb	NAb

Categorized tumor diameter was a secondary risk factor for lymph node metastasis of thyroid carcinoma (Table 4), except in the FTC group, where statistical significance was not observed, apparently due to the rarity of lymph node metastasis of FTC. The risk of lymph node metastasis significantly increased with increasing primary tumor diameter beyond a certain threshold diameter (10 mm in MTC and 20 mm in PTC). Lymph node metastasis of PTC was more common among patients younger than age 45 years compared with patients older than age 45 years. In patients with MTC, lymph node metastasis occurred exclusively in the presence of extrathyroidal growth and was significantly associated with reoperation. The only interaction identified in the multivariate analysis of lymph node metastasis involved primary tumor diameter and operative status in the MTC group, with larger tumor diameters being prevalent in the reoperative setting; while smaller tumors often are eradicated during primary surgery, larger and therefore more advanced tumors often require another operation (at a more specialized treatment center) to normalize serum calcitonin levels.

Lymph node metastasis was the only secondary risk factor for distant metastasis of thyroid carcinoma (Table 5), although statistical significance was not attained. In the MTC group, some parameters could not be examined with multivariate analysis (Tables 4, 5), because extrathyroidal growth and distant metastasis never were encountered in the absence of lymph node metastasis. In the analysis for distant metastasis, no interaction was found in either the FTC or PTC group between primary tumor diameter and extrathyroidal growth (data not shown). This finding lends support to the hypothesis that extrathyroidal growth and lymph node metastasis represent two separate mechanisms and routes of distant metastasis, with the former (direct pathway) occurring through the direct accession of primary tumor cells to parathyroidal blood vessels and the latter (indirect pathway) occurring through the passage of lymphatic tumor cells at the venous angle into the venous blood system. The hypothesis that distant metastasis can occur via an indirect pathway involving lymph node dissemination also is supported in univariate analysis by the significantly larger numbers of involved lymph nodes removed from patients who were diagnosed with distant metastases compared with patients who were not (FTC: 3.6 vs. 1.2 positive nodes, P = 0.05, r = 0.17; PTC: 7.7 vs. 3.5 positive nodes, P = 0.003, r = 0.19; MTC: 17.9 vs. 5.0 positive nodes, P < 0.001, r = 0.45). All logistic regression models (Tables 4, 5) were quite robust as evidenced by the similarity of results obtained by the initial model and the optimized model generated by the backward stepwise procedure.

DISCUSSION

To our knowledge, the current analysis is the first comprehensive multivariate evaluation of the clinical risk factors for various types of thyroid carcinoma that was based on direct measurement of histopathologic endpoint variables (lymph node metastasis and/or distant metastasis) rather than outcome (local recurrence or cause-specific death). Decreased survival can result from local recurrence, distant metastasis, or a combination of the two; clear-cut determination of cause of death may not always be straightforward. By focusing on clinicopathologic variables, we avoid incurring biases associated with differential surgical treatment, such as less extensive surgery in elderly and morbid populations.

Limitations of Observed Estimates of Effect Due to Misclassification and Selection Bias

Any assessment of clinicopathologic variables hinges on the correct classification of the parameters under investigation, whether they be independent or dependent. For the past decade, compartment-oriented lymph node dissection11 has facilitated the unbiased assessment of lymph node metastasis. Due to the high overall rate of systematic lymph node dissection (87%) in the current series, the frequency of misclassification of lymph node status should be negligible. Therefore, the observed estimates of effect (odds ratios) for the endpoint of lymph node metastasis (i.e., the magnitudes of the effects of individual parameters on lymph node metastasis) are likely to be correct. Ascertainment of distant metastasis did not necessarily require histopathologic confirmation. Consequently, some malignancies with occult distant metastases (M1) may have been misclassified as being devoid of distant metastases (M0), despite the liberal use of sophisticated imaging modalities. Misclassification, as caused by underascertainment, renders the two groups under comparison more alike, such that the observed estimates of effect for the endpoint of distant metastasis tend to underestimate the true effect, the magnitude of which is unknown.12 For the endpoints of lymph node metastasis and distant metastasis, the estimates of effect obtained for extrathyroidal growth are likely to underestimate the true effect, because of the preferential exclusion of more advanced malignancies, for which information regarding primary tumor diameter was not always available (Table 1). In summary, the calculated estimates of effect for the endpoint of lymph node metastasis are more likely to be accurate, whereas estimates for the endpoint of distant metastasis are expected to underestimate the true effects associated with the parameters being examined.

Clinical Implications

The current investigation underscores the paramount importance of extrathyroidal growth in both lymph node metastasis and distant metastasis. The risk of lymph node metastasis increased significantly with increasing primary tumor diameter, with an upsurge in rates beyond a threshold diameter of 10 mm in the MTC group and 20 mm in the PTC group. (The data for the FTC group may have been inadequate for attaining statistical significance.) In the MTC group, all cases of lymph node metastasis were accompanied by extrathyroidal growth, while in the PTC group, the risk of lymph node metastasis (Table 4) increased significantly with extrathyroidal growth (odds ratio, 4.3). A similar size effect of extrathyroidal PTC on lymph node metastasis (odds ratio, 4.3) was observed in a smaller multivariate analysis (n = 158).13 In a univariate analysis involving 2376 German patients with differentiated thyroid carcinoma, the frequency of lymph node metastasis increased sharply, from 10–20% to more than 50%, in the presence of extrathyroidal growth.14

Extrathyroidal growth also increased the risk of distant metastasis (Table 5), with odds ratios of 2.7, 6.6, and 15.8 in the FTC, PTC, and MTC groups, respectively. (As was discussed earlier, these odds ratios may underestimate the true effect of extrathyroidal growth.) These results suggest that the growth of thyroid malignancies through the thyroid capsule produces additional opportunities for tumor cell dissemination not only through the lymphatic system, but also through the vascular system of the parathyroidal soft tissues. This hypothesis also is supported by the high rates of distant metastasis (approximately 40% at diagnosis and 60% thereafter in patients with undifferentiated thyroid carcinoma, which almost always has extended beyond the thyroid capsule at the time of first diagnosis.15 Although less important than extrathyroidal growth, lymph node metastasis nonetheless may represent an alternative mechanism of distant metastasis in both PTC and MTC. The close univariate correlation between lymph node metastasis and distant metastasis in patients with MTC has been reported previously in a subset of the current study population.16 Such a dose-effect relation is indicative of a causal association, as are biologic credibility (e.g., consistency with the precepts of anatomy and consistency with histopathologic findings in selected patients)16 and the strength of the association (i.e., the magnitude of the odds ratio).12 In the current investigation, gender and categorized age were not significantly associated with lymph node metastasis or distant metastasis, except for the well-known correlation between young age and lymph node metastasis in patients with PTC.17 In most published series, residual lymph node metastasis is associated with local recurrence,18 whereas distant metastasis is correlated with poorer survival rates.19, 20

Conclusions

Extrathyroidal growth consistently emerged as the most significant risk factor for lymph node metastasis and distant metastasis. Clinical risk factors can be thought of as representing intermediate steps in the causal chain between molecular mutations of genes in tumor cells and histopathologic outcome, such as lymph node or distant metastasis. Controlling for such intermediate variables can obscure or ‘adjust away’ the associations between genetic mutations and histopathologic endpoints.12 Therefore, clinical risk factors that reflect the manifestations of genetically encoded risks will continue to play a pivotal role in predicting local and distant metastasis of thyroid carcinoma and, consequently, in managing patients. Screening for both local residual disease and distant metastasis should be intensified in the high-risk population of patients whose primary thyroid malignancies exceed the threshold diameter, extend beyond the thyroid capsule, or are lymph node–positive. Determination of the extent of cervical surgery should involve consideration of the sequence of lymphatic drainage of the thyroid gland (i.e., the order of involvement of lymph node compartments and UICC lymph node groups).5, 21–24