Nodal disease burden and outcome of medullary thyroid carcinoma

Abstract

Background:

Medullary thyroid carcinoma (MTC) is a rare malignancy with high incidence of cervical lymph node (CLN) metastasis. We investigated the impact of nodal disease burden on survival.

Methods:

We searched the National Cancer Database for MTC patients treated surgically. Impact of nodal metastasis on survival was analyzed using Cox univariable and multivariable regression.

Results:

We identified 2627 patients from 2004 to 2015. Positive CLNs were identified in 1433 (54.5%), and 542 (20.6%) had >10 CLN+. Overall survival was 94.5% and 89.6% at 3 and 5 years. Patients with 11 to 20 CLN+ had significantly worse survival than patients with 1 to 10 CLN+ in univariable and multivariable analyses (HR = 3.56 (2.31-5.50) vs 2.26 (1.60-3.20); P < .0001). The ratio of positive to dissected CLN was associated with overall survival.

Conclusions:

Higher burden of nodal disease is associated with worse survival in MTC. The number of positive nodes could be a valuable prognosticator in addition to the current staging system.

1 |. INTRODUCTION

Medullary thyroid cancer (MTC) is a rare entity that accounts for only 2% to 4% of thyroid cancers in the United States.¹ It is a slow growing but aggressive malignant neoplasm that originates from the calcitonin-producing parafollicular C cells of the thyroid gland.^2,3 Approximately 75% of cases are sporadic and 25% are hereditary, associated with mutations in the rearranged during transfection (RET) proto-oncogene on chromosome 10q.^4,5 MTC is more aggressive than differentiated thyroid cancer and distinct in that MTC cells do not concentrate radioactive iodine and are not sensitive to manipulation of thyroid stimulating hormone.⁶ Surgical treatment is therefore the mainstay of therapy. MTC has a high propensity for lymph node involvement with positive nodes identified in greater than 70% of cases.^2,6,7 The current N classification for MTC consists of N0 (no positive nodes), N1a (positive nodes in the central neck only), and N1b (positive nodes in the lateral neck). It does not take into account the number, size, or laterality of positive nodes. The most current guidelines for treatment of MTC recommend routine central compartment dissection in all cases and lateral neck dissection if the pretreatment calcitonin level is greater than 500.¹ Therefore, a majority of patients will undergo some form of lymph node dissection. Whether the number of lymph nodes involved has any prognostic value is not well studied. We analyzed the National Cancer Database (NCDB) to first study if the burden of nodal disease has an impact on survival, and second, to identify if the ratio of involved nodes to dissected nodes could be used as a prognostic and quality indicator in patients with MTC.

2 |. METHODS

Data used in this study were obtained from the NCDB, a clinical oncology database sourced from hospital registry data of more than 1500 Commission on Cancer accredited facilities and jointly sponsored by the American Cancer Society and American College of Surgeons. NCDB data represent more than 70% of newly diagnosed cancer cases nationwide. Since all patient information is de-identified in the NCDB, IRB approval was not required in accordance with the New York University School of Medicine Institutional Review Board policy.

The NCDB was queried from the years 2004 to 2015 for all patients diagnosed with medullary thyroid carcinoma (MTC). Tumors attributed to the thyroid gland were identified using the International Classification of Diseases for Oncology (ICD-O-3.1) topography code C73.9.⁸ Medullary carcinoma histologies were identified using two ICD-O-3.1 histology codes: 8345/3 and 8510/3.⁹ Patients were excluded if they had T0 tumors, distant metastasis or did not have lymph nodes pathologically evaluated. Additionally, patients with a history of prior or subsequent neoplasms, who were not treated with surgery, had last follow-up or death within 6 months of diagnosis, or had incomplete data were not included. The NCDB does not distinguish between hereditary and sporadic cases of MTC; therefore, both are included in the data set. A flow diagram demonstrating cohort selection can be seen in Figure 1.

FIGURE 1.

Cohort selection from the National Cancer Database

Patient demographics, clinicopathologic factors, and survival outcomes were recorded. Clinical staging and pathologic staging were compared and incidence of upstaging and downstaging was calculated for each stage. Nodal disease burden was evaluated based on (a) location: central neck vs lateral neck; (b) number of positive nodes: 0, 1 to 10, 11 to 20, and greater than 20 positive nodes; and (c) ratio of positive to dissected nodes both as a continuous variable and by dividing into categories of 0%, 1% to 25%, 26% to 50%, and 51% to 100%. Three Cox multivariable regression models were created using each of these nodal factors to evaluate the impact of nodal disease on survival. Each model included the same demographic (age, sex, race, comorbidity), clinicopathologic (T classification, lymphovascular invasion), and treatment factors (surgery, radiation, chemotherapy, surgical margins, and number of lymph nodes pathologically evaluated). Models differed only by method of nodal disease evaluation and were compared by Akaike information criterion (AIC) and concordance index (c-index). AIC and c-index are measures of goodness of fit of a model, with larger c-index and smaller AIC values indicating a better fit.¹⁰

We investigated if the nodal ratio model is influenced by the number of nodes dissected independently; meaning, if the same ratio of positive nodes predicts a different outcome based on the number of nodes dissected. In an exploratory multivariable model, this interaction did not have an influence on the outcomes; therefore, interaction terms were not included in the multivariable models.

3 |. RESULTS

We identified 2627 patients who satisfied inclusion criteria from 2004 to 2015. Demographic data are displayed in Table 1. Median age at time of diagnosis was 52 years (18-90) and the majority of patients were female (1527, 58.1%). Most patients did not have comorbidities (2207, 84.0%). More than half of patients with documented clinical nodal staging (1107, 64.2%) were staged as N0. Approximately half of patients had advanced stage MTC (stages III-IV, 1384, 55.3%). Positive cervical lymph nodes (CLNs) were identified in 1433 (54.5%) of cases, and 542 (20.6%) had greater than 10 positive nodes. The majority of patients were treated with unimodality surgery (2145, 81.7%), followed by surgery with adjuvant radiotherapy (332, 12.6%, Table 2). Most surgeries obtained negative margins (2113, 84.2%).

TABLE 1.

Clinicopathologic factors of 2627 patients with medullary thyroid carcinoma

	N	Percentage
Total	2627	100%
Age: median ± SD	52 ± 15.2
Sex
Female	1527	58.1%
Male	1100	41.9%
Race
White	2054	78.2%
Black	222	8.5%
Hispanic	204	7.8%
Asian	75	2.9%
Other	72	2.7%
Comorbidities
No comorbidities	2207	84.0%
Comorbidity present	420	16.0%
AJCC clinical stage
Stage 1	569	21.7%
Stage 2	370	14.1%
Stage 3	235	8.9%
Stage 4	384	14.6%
Unknown	1069	40.7%
Pathologic T classification
pT1	909	34.6%
pT2	544	20.7%
pT3	484	18.4%
pT4	199	7.6%
Unknown	491	18.7%
Clinical N classification
cN0	1107	42.1%
cN1A	100	3.8%
cN1B	326	12.4%
cN1 NOS	191	7.3%
Unknown	903	34.4%
Pathologic N classification
pN0	1195	45.5%
pN1A (central compartment only)	365	13.9%
pN1B (lateral nodes involved)	687	26.2%
pN1 NOS	380	14.5%
Number of positive nodes
No nodes positive	1194	45.5%
1–10 nodes	891	33.9%
11–20 nodes	326	12.4%
>20 nodes	216	8.2%
Ratio of positive to dissected nodes
0%	1194	45.5%
1% to 25%	521	19.8%
26% to 50%	444	16.9%
51% to 100%	468	17.8%
Lymphovascular invasion
No invasion	661	25.2%
Invasion	388	14.8%
Unknown	1578	60.1%

TABLE 2.

Treatment of 2627 Patients with Medullary Thyroid Carcinoma

	Number of patients	Percentage
Total	2627	100%
Treatment regimen
Surgery only	2145	81.7%
Surgery and radiotherapy	332	12.6%
Surgery plus unknown adjuvant therapy	104	4.0%
Surgery and chemotherapy	23	0.9%
Surgery plus chemoradiotherapy	23	0.9%
Surgical margins
Negative margins	2113	80.4%
Positive margins	392	14.9%
Unknown	122	4.6%
Nodes examined
1 to 10 nodes	1083	41.2%
11 to 20 nodes	467	17.8%
21-30 nodes	324	12.3%
> 30 nodes	753	28.7%

In comparing clinical and pathologic staging, 27.3% (302) of patients who were staged clinically as N0 were upstaged to N+ (Table 3). Only a small minority were down staged from cN+ to pN0 (15, 1.8%). The majority of patients who were staged as cN1a remained pN1a (63, 63%), 17% were upstaged to N1b. The majority of patients who were staged as cN1b remained pN1b (266, 81.6%)

TABLE 3.

Comparison of clinical and pathologic nodal staging for patients with medullary thyroid carcinoma

	Initial clinical stage
	cN0		cN1A		cN1B		cN1, NOS
Final pathologic Stage	Number	Percent	Number	Percent	Number	Percent	Number	Percent
pN0	805	72.7%	1	1.0%	5	1.5%	9	4.7%
pN1A	144	13.0%	63	63.0%	6	1.8%	19	9.9%
pN1B	100	9.0%	17	17.0%	266	81.6%	97	50.8%
pN1, NOS	58	5.2%	19	19.0%	49	15.0%	66	34.6%

Median length of follow-up was 65 months (interquartile range: 38-94 months). Overall survival for the entire cohort was 94.5% and 89.6% at 3 and 5 years, respectively. In univariable analysis, older age, male sex, comorbidities, advanced T classification, lymphovascular invasion, use of adjuvant therapy, and positive surgical margins were associated with worse overall survival. All remained significantly associated with worse survival in multivariable regression with the exception of lymphovascular invasion (HR = 1.19, 95% CI = 0.73-1.92, P = .49) and positive surgical margins (HR = 1.21, 95% CI = 0.92-1.59, P = .18).

In multivariable regression, presence of one or more positive CLNs remained associated with worse survival compared to N0 disease (HR = 2.41, 95% CI = 1.72-3.38, P < .001) with N1b disease having slightly worse survival compared to N1a disease (HR = 2.17, 95% CI = 1.45-3.26 for N1a vs HR = 2.38, 95% CI = 1.62-3.51 for N1b, P < .001, Figure 2). Patients with 11-20 positive CLNs had significantly worse survival than patients with 1-10 positive CLNs (HR = 3.56, 95% CI = 2.31-5.50 for 11 to 20 CLN+ vs HR = 2.26, 95% CI = 1.60-3.20 for 1 to 10 CLN+, P < .0001, Figure 3). The ratio of positive to dissected CLN was a strong, independent predictor of survival when controlling for all other factors (HR = 3.06, 95% CI = 2.21-4.40 for 51%-100% CLN+ vs HR = 1.66, 95% CI = 1.11-2.50 for 1%-25% CLN+, P < .001, Figure 4). This finding was true both when the nodal ratio was calculated as a categorical variable and as a continuous variable (multivariable HR = 1.11, 95% CI = 1.08-1.15, P < .001, per 10% increase in node ratio). Node ratio had a strong negative correlation with 5-year overall survival (correlation coefficient: −0.89, P < .001). In comparing three multivariable survival models based on nodal location, number of nodes, and ratio of positive to dissected CLNs, the model utilizing ratio of positive to dissected CLNs had the strongest association with survival (AIC = 4359, c-index = 0.8291, Table 4), while the model utilizing nodal location (N1a vs N1b) had the weakest association with survival (AIC = 4373, c-index = 0.8278). The results of the best multivariable model are displayed in Table 5.

FIGURE 2.

Overall survival of patients with medullary thyroid carcinoma stratified by nodal location [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 3.

Overall survival of patients with medullary thyroid carcinoma stratified by number of positive nodes [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 4.

Overall survival of patients with medullary thyroid carcinoma stratified by ratio of positive to dissected lymph nodes [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 4.

Goodness of fit of survival models

	Univariable		Multivariable
	AIC	C-Index	AIC	C-Index
Node location (central neck vs lateral neck)	4637	0.6979	4373	0.8278
Number of positive nodes (0, 1–10, 11–20, and >20)	4631	0.7001	4364	0.8290
Ratio of positive to dissected nodes (0%, 1%-25%, 26% to 50%, 51% to 100%)	4616	0.7157	4359	0.8291

Note: Larger c-index and smaller AIC values indicate a better fit.

Abbreviations: AIC, Akaike information criterion, c-index, concordance index.

TABLE 5.

Univariable and multivariable analysis of clinicopathologic and treatment factors and overall survival in medullary thyroid carcinoma

	Univariable regression			Multivariable regression
	HRa	95% CI	P value	HR	95% CI	P value
Age ≥ 60	2.88	2.32-3.58	<.001	2.66	2.12-3.35	<.001
Male sex	2.24	1.80-2.80	<.001	1.51	1.20-1.90	<.001
Race (reference: White)
Black	0.95	0.65-1.41	.81	1.06	0.71-1.57	.79
Hispanic	0.51	0.29-0.90	.02	0.48	0.27-0.84	.01
Asian	0.75	0.35-1.58	.45	0.51	0.24-1.11	.09
Other	0.61	0.27-1.38	.24	0.77	0.34-1.75	.54
Comorbidity present	1.83	1.42-2.36	<.001	1.46	1.12-1.91	.005
T stage (reference: T1)
T2	2.03	1.34-3.07	<.001	1.63	1.07-2.49	.02
T3	6.02	4.19-8.64	<.001	2.98	2.01-4.42	<.001
T4	10.35	7.09-15.10	<.001	3.43	2.22-5.32	<.001
Unknown	1.18	0.77-1.82	.44	0.80	0.51-1.24	.32
Lymph node ratio (reference: 0%)
1% to 25%	2.66	1.84-3.85	<.001	1.66	1.11-2.50	.01
26% to 50%	4.73	3.34-6.70	<.001	2.48	1.68-3.67	<.001
51% to 100%	7.19	5.22-9.90	<.001	3.06	2.12-4.40	<.001
Lymphovascular invasion (reference: no invasion)
Invasion present	3.45	2.17-5.50	<.001	1.19	0.73–1.92	.49
Unknown	1.70	1.13-2.56	.01	1.38	0.91-2.09	.13
Treatment (reference: surgery only)
Surgery with adjuvant radiotherapy	4.30	3.40-5.46	<.001	2.09	1.62-2.70	<.001
Surgery with adjuvant chemotherapy	8.00	4.23-15.14	<.001	2.85	1.44-5.63	.003
Surgery with adjuvant chemoradiotherapy	6.76	3.68-12.44	<.001	2.70	1.42-5.13	.002
Surgery with unknown adjuvant treatment	1.75	1.06-2.87	.03	1.40	0.84-2.33	.19
Surgical margins (reference: negative)
Positive	3.74	2.97-4.71	<.001	1.21	0.92–1.59	.18
Unknown	1.61	1.02-2.52	.04	0.90	0.56-1.44	.67
Lymph nodes examined (reference 1–10)
11 to 20 nodes examined	1.44	1.03-2.01	.03	0.95	0.67-1.36	.80
21 to 30 nodes examined	1.66	1.16-2.38	.003	0.94	0.64-1.38	.75
> 30 nodes examined	2.22	1.70-2.89	<.001	1.08	0.80-1.47	.62

Abbreviations: HR, hazard ratio; CI, confidence interval. Bold P values indicate a statistically significant association.

^aHR represents the association with each 10% increase in ratio of positive to dissected nodes (range 0%-100%).

4 |. DISCUSSION

The current staging system for MTC takes into account only the presence and location of positive CLNs. Prior studies have shown that positive CLNs are present in greater than 70% of patients at the time of diagnosis of MTC, and that in patients with at least one positive node in the central compartment the risk of positive nodes in the lateral compartment is greater than 70%.¹¹ Approximately 50% of patients are therefore staged as having N1b disease; however, these patients do not have equivalent disease burden or long-term survival. This reflects a significant limitation of the current staging system. Leggett et al looked at the prognostic value of lymph node yield and metastatic lymph node ratio in patients with MTC using data from the SEER database and showed that for node-positive patients, increased lymph node yield is associated with improved survival and higher metastatic lymph node ratio is associated with poorer survival.⁴ Machens et al reported that MTC patients with 1-10, 11-20, and >20 CLN metastases had small, intermediate, and high lifetime risks for lung metastases, highlighting the prognostic impact of nodal burden.¹² An earlier study by the same group confirmed the association between the number of positive nodes in the central and lateral neck compartments and suggested that the distinction between surgically curable and incurable disease hinges on the absolute number of lymph node metastases and whether they are in the contralateral vs ipsilateral neck rather than the central vs lateral neck.¹¹

In the current study utilizing the NCDB, we report the largest cohort of MTC patients studied and demonstrate that patients have significantly worse survival when greater than 10 CLNs are positive. Our findings are in line with previous studies,^13,14; however due to the size of our sample, we were able to perform more rigorous statistical analyses and control for other factors influential in survival. In univariable and multivariable analyses, the burden of disease characterized by the number of positive CLNs as well as the ratio of positive to dissected CLNs was a better prognosticator of survival than the current staging model, which takes into account only location of nodal disease. The association of nodal ratio with survival was significant at all levels as evidenced by the correlation of nodal ratio and survival when calculated as a continuous variable.

Based on the current guidelines not all patients undergo lateral neck dissection, and there is no established guideline for a minimum lymph node yield to ascertain adequacy of the lymph node dissection. Therefore, the pathologic stage is the same for patients who had one positive node vs multiple positive nodes in one compartment. Furthermore, there is no metric to determine the adequacy of treatment. In this study, we suggest that examining the impact of the number of positive nodes as well as the ratio of positive to dissected nodes on survival might provide more meaningful prognostic information for patients who undergo surgery for the treatment of MTC. Our finding that the ratio of positive to dissected lymph nodes has the strongest association with survival is consistent with data reported in the studies mentioned earlier, reinforcing the conclusions of previous authors that the number of positive lymph nodes provides valuable information regarding risk of distant metastasis, chance of surgical cure, and overall survival. These results could potentially help inform a revised staging classification that incorporates the number of nodes into the N classification. Number and laterality of positive lymph nodes is part of the existing staging systems for all subsites of squamous cell carcinoma of the head and neck and does not introduce a degree of complexity that makes the staging system difficult to understand. Such a staging system for MTC based on metastatic lymph node number and metastatic lymph node ratio was recently proposed by Chen et al and might offer a basis for initiating this change.¹⁵

Similar to all studies utilizing large databases, our study has several important limitations that must be recognized. The NCDB does not differentiate between central vs lateral neck dissection or between ipsilateral vs contralateral neck dissection. Additionally, many relevant clinical factors such as calcitonin values, carcinoembryonic antigen values, germline mutation status, or specific pathologic features such as perineural invasion are not reported. Therefore, patients who may undergo lateral neck dissection due to high calcitonin levels cannot be differentiated from those undergoing lateral neck dissection due to clinically positive nodal disease. Although these two groups of patients may represent distinct clinical entities, they are grouped into the same cohort in this study. Further studies utilizing either prospective data or retrospective cohorts will be necessary to differentiate further between subgroups of patients with MTC. The NCDB also does not report patterns of failure, therefore our analysis was limited to overall survival. As with any database study, the potential for miscoding of cases must be considered. However, given that MTC is a unique and distinct diagnosis, this risk is low. In spite of these shortcomings, we believe the findings of this study, considering its size and scope, could be useful for clinicians who treat medullary thyroid cancer patients.

5 |. CONCLUSION

MTC is an indolent but persistent disease with a high rate of cervical metastases. The current staging system reflects only the location of positive nodes in the central vs lateral neck. Nodal disease burden confers significant prognostic value in predicting survival following surgery for patients with MTC and could be considered in the pathologic staging of disease.