A National Cancer Data Base report on 53,856 cases of thyroid carcinoma treated in the U.S., 1985-1995

Authors


Abstract

BACKGROUND

The National Cancer Data Base (NCDB) represents a national electronic registry system now capturing nearly 60% of incident cancers in the U. S. In combination with other Commission on Cancer programs, the NCDB offers a working example of voluntary, accurate, cost-effective "outcomes management" on a both a local and national scale. In addition, it is of particular value in capturing clinical information concerning rare cancers, such as those of the thyroid.

METHODS

For the accession years 1985-1995, NCDB captured demographic, patterns-of-care, stage, treatment, and outcome information for a convenience sample of 53,856 thyroid carcinoma cases (1% of total NCDB cases). This article focuses on overall 10-year relative survival and American Joint Committee on Cancer (AJCC) (3rd/4th edition) stage-stratified 5-year relative survival for each histologic type of thyroid carcinoma. Care patterns also are discussed.

RESULTS

The 10-year overall relative survival rates for U. S. patients with papillary, follicular, Hürthle cell, medullary, and undifferentiated/anaplastic carcinoma was 93%, 85%, 76%, 75%, and 14%, respectively. For papillary and follicular neoplasms, current AJCC staging failed to discriminate between patients with Stage I and II disease at 5 years. Total thyroidectomy ± lymph node sampling/dissection represented the dominant method of surgical treatment rendered to patients with papillary and follicular neoplasms. Approximately 38% of such patients receive adjuvant iodine-131 ablation/therapy. At 5 years, variation in surgical treatment (i.e., lobectomy vs. more extensive surgery) failed to translate into compelling differences in survival for any subgroup with papillary or follicular carcinoma, but longer follow-up is required to evaluate this. NCDB data appeared to validate the AMES prognostic system, as applied to papillary cases. Younger age appeared to influence prognosis favorably for all thyroid neoplasms, including medullary and undifferentiated/anaplastic carcinoma. NCDB data also revealed that unusual patients diagnosed with undifferentiated/anaplastic carcinoma before age of 45 years have better survival.

CONCLUSIONS

The NCDB system permits analysis of care patterns and survival for large numbers of contemporaneous U. S. patients with relatively rare neoplasms, such as thyroid carcinoma. In this context, it represents an unsurpassed clinical tool for analyzing care, evaluating prognostic models, generating new hypotheses, and overcoming the volume-related drawbacks inherent in the study of such neoplasms. [See editorial on pages 2434-6, this issue.] Cancer 1998;83:2638-2648. © 1998 American Cancer Society.

The National Cancer Data Base (NCDB), a joint project of the Commission On Cancer (COC) of the American College of Surgeons (ACoS) and the American Cancer Society (ACS), assesses both patient care and treatment outcomes at an institutional, regional, and national level.

Annually, each institution voluntarily submitting data receives a detailed report comparing institutional and national results. Institutions use this information to benchmark and evaluate trends in both cancer care and outcomes. The NCDB, when viewed in combination with other COC programs operating at both the local and national level, represents a working example of a national "outcomes management system." It offers a fine example of how voluntary, collaborative, professional efforts can reduce the morbidity and mortality of cancer through application of an "outcomes management" approach.1

The NCDB began operations in 1989. Through multiple calls for data, cases from 1985 through 1995 have been accessioned and are available for analysis. For the 1995 accession year, the NCDB captured 60% of cancer cases treated in the U.S. By current estimates, a new linkage between the NCDB and the COC's Cancer Program Approval System will increase accessions to nearly 80% of 1998 cancer cases in the U. S. Benefiting from the high degree of professional volunteerism that has typified COC-ACoS-ACS collaborations, the national NCDB office runs on an annual direct-cost budget of only $800,000. The popularity, utility, and cost-effectiveness of this system offers an unsurpassed example of how professional volunteerism, standardization, computing technology, and careful planning can combine to create a valuable tool for benchmarking and improving health care.

With increasing case accrual, the NCDB also has emerged as a superb tool for obtaining contemporaneous data concerning patterns of care and outcomes for rare neoplasms such as thyroid carcinoma. In the U.S. in 1994, newly diagnosed thyroid carcinoma was reported in approximately 13,000 people. The same year, approximately 1000 people died of this disease.2 Although its incidence appears to be increasing, thyroid carcinoma, accounting for only 1% of all reportable malignancies, remains rare both in the U.S. and worldwide.3, 4 This fact, combined with the indolence of this disease and its generally favorable prognosis, serves to undermine the comparative study of different treatments by prospective randomized clinical trials. Current treatment recommendations generally are based on retrospective studies derived from analyses of patients treated in the distant past and collected over several decades.5-12 Recognizing that risk models derived from retrospective analysis of one population, particularly if relatively homogeneous (e.g., all treated at one institution, from one geographic region, from one organization, treated in distant past, etc.), may not predict outcomes for a different population, the value of data derived from a relatively contemporaneous national source such as the NCDB13 cannot be underestimated.

This report summarizes NCDB findings concerning thyroid carcinoma in the U.S., focusing on the period between 1985-1995.

METHODS

This article on thyroid carcinoma continues the tradition of analyzing data voluntarily submitted to the NCDB to determine national patterns of patient care and to assess outcomes. Using standardized definitions and uniform data formats, the NCDB annually collects data for all forms of cancer from hospitals throughout the U.S., the majority of which have COC-approved cancer programs. The methods of the NCDB have been described previously.14

This analysis, based on seven annual calls for data issued since 1989, draws on experience with patients diagnosed between 1985-1995. The NCDB's "calls for data" routinely were transmitted to approximately 2100 hospitals (1340 participating in COC Approval Programs and 760 others), as well as to all known central and state registries and all known software vendors and suppliers.

The NCDB cohort represents data from U.S. institutions located in all 50 states, Puerto Rico, and the District of Columbia. Each hospital receives an annual report comparing its own data with that of the entire NCDB cohort to facilitate assessment of its own care patterns and outcomes.

NCDB cancer registry data are coded according to the schema published in the Data Acquisition Manual,15 the 3rd and 4th editions of the American Joint Committee for Cancer (AJCC) Manual for Staging of Cancer,16, 17 and the second edition of the International Classification of Disease for Oncology (ICD-O-2).18 Formats for data reporting and electronic transmission to the NCDB are uniform. Data is transmitted to the NCDB by diskette or data tapes. To preserve confidentiality, individual patient identifiers such as social security numbers, names, etc. are collected by individual participating registries, but are neither transmitted to nor collected by the NCDB.

Hospital cancer committees or their equivalent supervise the quality of institutional case finding, abstracting, internal reviews of abstracts by registry staff, and hospital-based computer data edits. On submission to the NCDB, all data were checked electronically for valid codes and for logic between data items. Record linkage to detect duplicates was performed using gender, three-digit site code, birth date, and zip code. Cases judged to be duplicates were removed from the analytic file.

To minimize the number of patients with unknown stage in this study, a combination AJCC stage was used. Each patient was assigned the AJCC pathologic stage, if available. If a particular T, N, or M classification was not confirmed pathologically, then the patient was assigned the appropriate AJCC clinical stage, and the "combined stage" was reported. Of the 53,856 patients in this study, 17,299 did not have an assigned pathologic stage. Of these, 5418 did have an assigned clinical stage, leaving 11,881 patients (22.1%) without a known stage. The majority of AJCC unstaged patients were submitted by central registries, which, in the U. S., often do not explicitly collect AJCC stage information.

This analysis covers primary tumors of the thyroid gland (ICD-O-2 site C73.9) coded as one of the five major histologies: papillary (8050/3, 8340/3, and 8260/3), follicular (8330/3, 8332/3, and 8331/3), medullary (8510/3 and 8511/3), Hürthle cell (8290/3), and undifferentiated/anaplastic (8020/3 and 8021/3, respectively).

Relative survival analyses were conducted in the usual fashion, and adjusted using expected survival rates for the U.S. population in 1980. 16 Both descriptive and survival analyses for this article were performed using SPSS software.19

Due to the methodology of the data collection, this article is comprised of descriptive statistics to identify current patterns of care and outcomes. Inferential statistics regarding the multiple comparisons contained herein are not warranted due to the lack of a priori hypotheses and the very large sample size, representing a substantial proportion of the population under study. These facts contravene the assumptions of inferential statistics and could confound results. We advocate a straightforward interpretation of these data and do not offer statistical comparisons. We do consider these data suitable for benchmarking, for describing care patterns, for assessing national community-based outcomes, and for generating new hypotheses.

RESULTS

Table 1 summarizes accrual and histology by era (1985-1990 and 1991-1995). Over the 10-year period between 1985-1995, the NCDB captured data on a total of 53,856 cases: 42,686 papillary carcinomas, 6764 follicular carcinomas, 1585 Hürthle cell carcinomas, 1928 medullary carcinomas, and 893 undifferentiated/anaplastic carcinomas. Over the 10-year period, submissions to the NCDB increased considerably, but proportional representation of each histology varied only slightly.

Table 1. Percentage of Thyroid Carcinoma Cases by Diagnosis Year and Histologic Subgroup
 1985-19901991-1995Total
Histologic subgroup
Papillary17,39777.925,28980.242,686
Follicular316914.2359511.46764
Hürthle cell5942.79913.11585
Medullary8293.710993.51928
Undifferentiated/anaplastic3541.65391.7893
Total22,343100.031,513100.053,856

Figure 1 (9K) depicts age distribution for the 53,856 patients by histology.

Figure 1.

Age distribution (in percent) of thyroid carcinoma by histology.

In 1988, with publication of the 3rd edition of the AJCC Manual for Staging of Cancer, both stage definitions and stage groupings for this site changed. Definitions and groupings remained the same through subsequent changes (i.e., same for 3rd, 4th, and 5th editions). Table 2 reveals stage distribution for patients staged according to the 3rd or 4th editions of the AJCC manual.

Table 2. Percentage of 1990-1994 Thyroid Carcinoma Cases by Combined Stagea and Histologic Subgroup
 Combined Stage Group
I (%)II (%)III (%)IV (%)Unknown (%)Total (%) No.
  • a

    This table only includes cases that were staged according to the 3rd or 4th editions of the American Joint Committee on Cancer staging manual.

Histologic subgroup
  Papillary56.914.410.32.116.4100.034,794
  Follicular41.226.76.97.217.9100.05271
  Hürthle cell20.835.19.35.729.0100.01310
  Medullary16.529.626.811.016.0100.01550
  Undifferentiated/anaplastic0.00.00.0100.00.0100.0741
Total51.516.810.34.816.6100.0 
Cases22,4867335449120917263 43,666

As seen in Table 3, over the 10-year period, surgical treatment for the majority of patients with thyroid carcinoma was comprised of near-total or total thyroidectomy. For papillary, follicular, and Hürthle cell carcinoma, ipsilateral lobectomy with isthmectomy was performed in 18.1%, 25.2%, and 24.7% of cases, respectively. Lymph node sampling appeared more popular than formal, comprehensive lymph node dissection; however, lymph node-directed treatment was not practiced routinely for any histology.

Table 3. Percentage of Thyroid Carcinoma Cases by Type of Surgery and Histologic Subgroup
 None/unknownLobectomyNear-totalaTotalb without LNDTotalc with LNDTotald rad/mod LNDOtherTotal
  • LND: lymph node dissection; rad: radical; mod: modified.

  • a

    Near-total thyroidectomy.

  • b

    Total thyroidectomy without lymph node dissection.

  • c

    Total thyroidectomy with limited lymph node dissection.

  • d

    Total thyroidectomy with radical or modified lymph node dissection.

Papillary5.418.116.926.114.67.211.8100.0
Follicular7.825.215.230.56.82.012.5100.0
Hürthle cell5.924.716.132.77.82.510.3100.0
Medullary11.48.96.824.217.619.211.9100.0
Undifferentiated/anaplastic47.813.49.08.25.74.311.6100.0
Total6.618.816.226.513.46.811.8100.0
Cases356210,103870514,27671913653636653,856

Table 4 confirms that, for this site, treatment during the 10-year period was comprised overwhelmingly of surgery or surgery plus radiotherapy (usually iodine-131 [I-131] rather than external beam). Many patients received postoperative hormonal replacement or hormonal suppression therapy with thyroid hormone or analogue, but the NCDB data set fails to capture this information specifically. Systemic chemotherapy plays little role in the current treatment of thyroid carcinoma, and the percentage of patients receiving "other treatment" (surgery or radiation plus chemotherapy, etc.) was low.

Table 4. Percentage of Thyroid Carcinoma Cases by Treatment Combination and Histologic Subgroup
 SurgerySurgery + radiationRadiationOtherNoneTotal
Histologic subgroup
Papillary56.338.31.60.03.7100.0
Follicular53.638.63.30.14.4100.0
Hürthle cell61.632.52.10.03.8100.0
Medullary71.816.84.11.45.9100.0
Undifferentiated/anaplastic17.434.833.12.512.2100.0
Total56.037.32.50.14.0100.0
Cases30,18620,108132774216153,856

Figure 2 (8K) depicts 10-year relative survival rates by histology. Patients with undifferentiated/anaplastic carcinoma (only 1.7% of the 53,856 cases) had a dismal prognosis, with only 14% surviving 10 years. However, for the vast majority of patients with thyroid carcinoma, 10-year survival rates appear excellent.

Figure 2.

Relative survival by histology of thyroid carcinomas diagnosed between 1985-1990.

Table 5 emphasizes the influence of age on the prognosis of patients with thyroid carcinoma, with both 5-year and 10-year survival rates for all histologies (i.e., not just papillary and follicular carcinomas) influenced by this factor.

Table 5. Thyroid Carcinoma 5-Year/10-Year Relative Survival (%) by Histologic Subgroup and Age
 < 4545-5960-6970-7980+
Histologic subgroup%%%%%
  • a

    Insufficient numbers for reliable 10-year relative survival calculation.

Papillary98/9796/8592/6586/4782/a
Follicular98/9892/6685/5783/a62/a
Hürthle cell97/a91/9092/092/7053/a
Medullary86/8481/8072/6370/a68/a
Undifferentiated/anaplastic55/a13/a13/a13/a14/a

To simplify presentation of histology specific data pertaining to 5-year stage-stratified survival, prognostic models, and other factors, we have subdivided the remainder of this section.

Papillary and Follicular Carcinoma ("Differentiated Thyroid Carcinoma")

Many pathologists fail to include an assessment of grade for papillary and follicular carcinoma. However, Table 6 confirms the contention by some that tumor grade strongly influences prognosis.

Table 6. Thyroid Carcinoma 5-Year/10-Year Relative Survival (%) by Histologic Subgroup and Histologic Grade
 Histologic Grade
123
Histologic subgroup% (n)% (n)% (n)
Papillary98/86 (2365)93/73 (760)70/68 (254)
Follicular97/47 (878)85/65 (200)59/52 (167)

Figures 3 (5K) and 4 (5K) depict 5-year stage-stratified survival according to AJCC 3rd and 4th edition staging. Only slight differences in prognosis between the two histologies were noted, but the prognosis for Stage III and Stage IV patients with follicular thyroid carcinoma appeared marginally worse.

Figure 3.

Relative survival of papillary thyroid carcinomas diagnosed between 1985-1990 and staged according to definitions from the 3rd/4th editions of the American Joint Committee on Cancer (AJCC) staging system.

Figure 4.

Relative survival of follicular thyroid carcinomas diagnosed between 1985-1990 by American Joint Committee on Cancer (AJCC) stage.

Tables 7 and 8 illustrate treatment-related 5-year survival comparisons for each of the following categories of papillary thyroid carcinoma and for each of the following categories of follicular carcinoma: T-1, N-0, M-0; T-4, N-0, M-0; any T, N-0, M-0; any T, N-1, M-0; any T, any N, M-0; and any T, any N, M-1. Only patients followed for at least 5 years and with all TNM factors coded according to 3rd and 4th edition AJCC definitions were included in this analysis. No particular surgical treatment appeared to generate compellingly superior 5-year survival, but we expect this to change with longer follow-up. However, even with the large number of total cases some subsets were problematically small. Cases accessioned between 1985-1988 usually were staged according to the 2nd edition of the AJCC staging system, and because of incompatibility with the 3rd and 4th edition staging system, could not be included in this analysis. It is for this reason we report only 5-year stage-stratified survival rates despite our possession of 10-year data.

Table 7. Five-Year Relative Survival Rates of Papillary Thyroid Carcinoma Patients Diagnosed between 1985-1990 by AJCC TNM Status (3rd edition) and Type of Surgery
 Type of surgery
TNM statusNone/unknown (No.) %Lobectomy (No.) %Near-totala (No.) %Total w/o LNDb (No.) %Total with LNDc (No.) %Total with rad/mod LNDd (No.) %
  • AJCC: American Joint Committee on Cancer; w/o: without; LND: lymph node dissection; rad: radical; mod: modified.

  • a

    Near-total thyroidectomy.

  • b

    Total thyroidectomy without lymph node dissection.

  • c

    Total thyroidectomy with limited lymph node dissection.

  • d

    Total thyroidectomy with radical or modified lymph node dissection.

No. of cases11582848162224964
  T = 1      
  N = 092.610010099.799.595.7
  M = 0      
No. of cases56126183205179165
  T = 4      
  N = 085.090.992.699.094.991.2
  M = 0      
No. of cases250159411781672661192
  Any T      
  N = 094.210099.099.899.296.5
  M = 0      
No. of cases201298456289760600
  Any T      
  N = 196.698.499.093.299.3100
  M = 0      
No. of cases4581984164320521340734
  Any T      
  Any N10010099.9100100100
  M = 0      
No. of cases313948815952
  Any T      
  Any N63.078.082.274.584.695.3
  M = 1      
Table 8. Five-Year Relative Survival Rates of Patients with Follicular Thyroid Carcinoma Diagnosed between 1985-1990, by AJCC TNM Status (3rd edition) and Type of Surgery
 Type of surgery
TNM statusNone/unknown (No.) %Lobectomy (No.) %Near-totala (No.) %Total w/o LNDb (No.) %Total with LNDc (No.) %Total with rad/mod LNDd (No.) %
  • AJCC: American Joint Committee on Cancer; w/o: without; LND: lymph node dissection; rad: radical; mod: modified.

  • a

    Near-total thyroidectomy.

  • b

    Total thyroidectomy without lymph node dissection.

  • c

    Total thyroidectomy with limited lymph node dissection.

  • d

    Total thyroidectomy with radical or modified lymph node dissection.

No. of cases161296170286
  T = 1      
  N = 099.510096.691.7100102
  M = 0      
No. of cases73826412110
  T = 4      
  N = 096.393.684.082.491.819.7
  M-0      
No. of cases4840723337311226
  Any T      
  N = 088.096.492.495.010090.5
  M = 0      
No. of cases161523183126
  Any T      
  N = 150.987.794.882.882.657.7
  M = 0      
No. of cases6245825938013546
  Any T      
  Any N88.398.393.997.695.380.3
  M = 0      
No. of cases20111635910
  Any T      
  Any N23.648.449.953.199.357.1
  M = 1      

Similar analyses also were attempted for each of the categories of treatment listed in Table 4, but overly small subsets precluded meaningful survival analysis.

Of the many non-TNM prognostic models offered for differentiated thyroid carcinoma, perhaps the easiest to apply is that offered by Cady et al., the AMES system.5 Because "major capsular invasion" versus "minor capsular invasion" by follicular carcinoma constitutes a criterion for division into "high risk" versus "low risk" cases according to this system and because the NCDB data set fails to code this, the AMES system could not be evaluated for follicular cases. Figure 5 (5K) summarizes results for 5-year survival of the AMES "high risk" (6.4%) versus AMES "low risk" (93.6%) papillary thyroid carcinoma cases. This figure is based on the 9369 cases for which both 5-year follow-up and all factors for AMES calculations were available.

Figure 5.

Relative survival of thyroid carcinomas diagnosed between 1985-1990 by AMES risk levels.

Other Histologies

Figure 6 (5K) documents that 5-year survival for patients with Hürthle cell carcinoma closely parallels that of follicular cases, thus supporting the popular view that Hürthle cell carcinoma simply is a variant of follicular carcinoma. However, disparate 10-year overall survival was noted (76% vs. 85%).

Figure 6.

Relative survival of Hürthle cell thyroid carcinomas diagnosed between 1985-1990 and staged according to the 3rd/4th editions of the American Joint Committee on Cancer (AJCC) staging system.

Figure 7 (5K) shows 5-year relative survival for cases of medullary carcinoma. The NCDB data set does not contain a separate item for distinguishing familial from sporadic cases.

Figure 7.

Relative survival by AJCC stage of medullary thyroid carcinomas diagnosed between 1985-1990 and staged according to the 3rd/4th editions of the American Joint Committee on Cancer (AJCC) staging system.

All cases of undifferentiated/anaplastic carcinoma are, by definition, Stage IV. Figure 2 (8K) summarizes 10-year relative survival for these patients. Age < 45 years and small tumor size (and possibly the absence of extrathyroidal spread) appear overrepresented among those patients surviving ≥ 5 years.

DISCUSSION

To our knowledge, this 1985-1995 cohort of 53,856 thyroid carcinoma cases represents the largest contemporaneous, stage-stratified series yet reported. The ability of a national, community-based, electronic registry system such as the NCDB to collect large numbers of patients with rare tumors such as thyroid carcinoma constitutes yet another virtue of this cost-effective benchmarking system. As diagnostic methods, imaging techniques, and treatment paradigms shift over time, stage distribution, patterns of care, and outcomes can be expected to shift, hence the importance of using contemporaneous data for cancer benchmarking.

The number of patients followed for 5 years and available for stage-stratified analysis was undermined by the change in AJCC staging for this site, which occurred with publication of the 3rd edition of the AJCC manual in 1988. Reliable conversion of cases staged according to the previous edition was not feasible given the NCDB's current data set. Hence, only cases staged according to 3rd/4th edition definitions were used in reporting stage-stratified results. It is for this reason we can offer only 5-year stage-stratified results despite 10 years of data.

Some factors for this site remained relatively stable over time. Gender distribution, ethnic distribution, and income level distribution previously have been discussed in an article devoted to describing NCDB cases accessioned in 1992.20 Briefly, females predominate for all histologies: papillary (74%), follicular (74%), medullary (62%), and undifferentiated/anaplastic (64%). Ethnic distribution of papillary and follicular thyroid carcinoma roughly paralleled that for the NCDB as a whole; however, case ratios for Hispanics and Asians appeared slightly higher than expected. The lack of significant change in these items dissuaded us from reiterating findings already described in the previous report.

Although the majority of clinicians recognize the effect of age on the prognosis of patients with follicular and papillary thyroid carcinoma, these NCDB data suggest an effect on other thyroid histologies as well. The effect of this factor on prognosis for medullary and undifferentiated/anaplastic thyroid carcinoma appears to be noteworthy.

Risk models derived from one population, particularly if that population shares certain characteristics (e.g., all treated at one institution, from one geographic region, from one organization, treated in distant past, etc.), may not predict outcomes for a different population.21 Hence, national data such as those described here assumes particular importance in assessing various prognostic models. A number of different prognostic models have been proposed for patients with "differentiated thyroid carcinoma" (i.e., papillary or follicular): the AMES system,5 the AGES system,8, 9 the MACIS system,10 the Clinical Class system,11 the Mazzaferri system,12 the GAMES system,22 and, of course, the TNM system.6, 16, 17 In all these models, a relatively small proportion of patients fit into the higher risk subgroup(s). The superimposable survival curves for AJCC TNM Stages I and II for papillary and follicular thyroid carcinomas in this series suggests an possible opportunity for simplifying the current TNM prognostic system further. Unfortunately, certain elements of other prognostic systems were either not yet included in the 1985-1995 NCDB data set (e.g., ploidy, degree of capsular invasion, etc.) or not yet reported consistently by pathologists (e.g., grading as utilized in the Mayo Clinic AGES system). This fact makes evaluating these systems using NCDB data problematic. However, we were able to evaluate the AMES system as it applied to cases of papillary thyroid carcinoma, and found the survival differences between "high risk" and "low risk" groups consistent with the earlier report by Cady and Rossi.5 Additional data elements collected as part of an ACoS Patient Care Evaluation Study of thyroid carcinoma cases diagnosed in 1996 should facilitate evaluation of other prognostic models in the future.

Although total thyroidectomy remains the standard for patients with medullary carcinoma,23 the surgical treatment of papillary and follicular ("differentiated") thyroid carcinoma remains controversial.23, 24 Initial autopsy studies suggesting a very high degree of pathologically confirmed contralateral disease supported the routine use of total thyroidectomy.25 However, the significance of this finding became suspect when autopsy studies in the U.S. showed a high incidence (3-6%) of asymptomatic, occult, intraglandular foci of this clinically rare tumor (with world standard reported incidence rates of usually 2-5 per 100,000 population) in patients dying of other causes,26 and a very high autopsy incidence (14-24%) when glands were analyzed using fine step-sectioning techniques.27, 28 A large clinical series from the Memorial Sloan-Kettering Cancer Center (MSKCC) with 20-year follow-up revealed excellent results (i.e., 99% 20-year survival and only 4% local recurrence) associated with the treatment of patients with favorable prognostic factors (defined as "low risk" according to the MSKCC classification) by complete ipsilateral lobectomy and isthmectomy and long term hormonal suppression.29 Other series with long follow-up, using different risk group definitions, also have supported the selective use of complete lobectomy rather than routine total thyroidectomy.9 These findings echo those of smaller, earlier studies.30-32 However, other major studies document fewer recurrences and lower mortality for patients treated with total/near-total thyroidectomy and I-131.6, 11, 12 The multiplicity of risk stratification schemes (with varying definitions of "low risk," "favorable prognosis," etc.), the variable use of I-131, and variable practice with respect to long term hormonal suppression complicate a stratified analysis of treatment-related outcomes. Because complications with lobectomy are lower than with total thyroidectomy,33 because effective salvage therapy in the presence of recurrence in the opposite lobe is possible, and because mortality has been shown by multiple studies to be low, several experts recommend that patients at low risk for recurrence initially be treated with simple lobectomy-isthmectomy and long term hormonal suppression.5, 9, 10, 29

These NCDB data provide a snapshot of current surgical opinion concerning the "lobectomy versus total" controversy. Overall, only 18% of patients with papillary carcinoma and 25% of patients with follicular carcinoma are treated by ipsilateral lobectomy. Even for patients with T1N0M0 tumors, only 35% of patients with papillary carcinoma and 42% of patients with follicular carcinoma are so treated. Tables 7 and 8 document that, with 5 years of follow-up, more extensive surgical resection fails to generate convincingly superior survival for any subgroup. Naturally, given the indolence of this disease, prolonged follow-up is required for a true assessment of the survival impact of variation in surgical treatment.

The NCDB data set fails to include descriptions of the dose and frequency of I-131 treatments, but confirms that such treatments are used rarely as the primary therapeutic modality. Similarly, the NCDB data set fails to include details of adjuvant hormonal suppression with thyroid hormone. The aforementioned ACoS Patient Care Evaluation Study of thyroid carcinoma cases diagnosed in 1996 will examine care patterns and outcomes with respect to I-131 and hormonal suppression.

Five-year, stage-stratified relative survival curves for patients with Hürthle cell carcinoma appear similar to those for patients with follicular neoplasms, thus lending support to the widely held contention that these histologic types behave similarly.34 The World Health Organization and the majority of pathologists classify Hürthle cell neoplasms as a variant of follicular neoplasms.26 However, the 9% difference in survival at 10 years (76% for Hürthle cell carcinoma vs. 85% for follicular carcinoma), emphasizes the hazard of generalizing based on 5-year results for this indolent tumor, and prognosis indeed may be worse.

The majority of authorities recommend total thyroidectomy for patients with medullary carcinoma of the thyroid.23, 35, 36 The vast majority of patients in this NCDB cohort were treated by total thyroidectomy, and only 8.9% of patients were treated by ipsilateral lobectomy.

Previous series of patients with undifferentiated/anaplastic carcinoma of the thyroid document occasional long term survivors, but the overall prognosis remains dismal. Survivors appear primarily to be patients with completely resected, small areas of anaplastic carcinoma, often within areas of differentiated thyroid carcinoma.37 True assessment of the volume or size of an area of anaplastic transformation in such cases is difficult. The fact that 48% of NCDB patients receive no primary surgical treatment implies that the majority of patients with this diagnosis are judged to have incurable disease. However, the 14% 10-year relative survival rate of NCDB patients, and the plateau of the survival curve, document that occasional cases can be cured. The 55% 5-year relative survival of patients age < 45 years, admittedly based on few patients, encourages speculation that a definable subgroup of patients with undifferentiated/anaplastic thyroid carcinoma might perhaps be viewed as curable. Independent pathology review certainly would be desirable with respect to these 5-year survivors with anaplastic carcinoma, but the NCDB does not yet have a workable mechanism for doing this.

The NCDB currently accessions nearly 60% of incident cancers in the U.S. Though a new linkage with the COC Approvals Program, an 80% accession rate for 1998 is predicted. In contrast, the population-based Surveillance, Epidemiology, and End Results system, used for estimating U.S. cancer incidence rates, accessions approximately 14% of incident cancers and, judging from U.S. census data, is neither racially nor geographically representative of the current U.S. population.13 A joint initiative by the ACS and the ACoS to add a population-based component to the NCDB currently is being explored. Such a component would permit the NCDB to track cancer incidence as well as care patterns and treatment outcomes. In the context of thyroid carcinoma, such a national system might prove useful in monitoring health effects attributable to I-131 contamination from U.S. atmospheric nuclear testing in the 1950s, a topic recently addressed by the National Cancer Institute,38, 39 and other projects.

The aforementioned ACoS Patient Care Evaluation (PCE) Study of thyroid carcinoma cases diagnosed in 1996 recently completed accrual and baseline data will be reported within a year. The study includes data items not included in the NCDB data set and should help answer some questions left unaddressed by this study (e.g., the effect of I-131 ablation and subsequent I-131 therapy and hormonal suppression). We currently plan to track the 1996 PCE cohort of patients prospectively over the next 20-30 years using the NCDB.

Conclusions

Overall 10-year relative survival for patients with papillary thyroid carcinoma is 93% and is 85% for patients with follicular thyroid carcinoma. Although 5-year overall and stage-stratified relative survival for patients with Hürthle cell carcinoma closely matches that of patients with follicular carcinoma, overall survival at 10 years is 9% lower, suggesting a marginally worse prognosis.

At 5 years, the current AJCC TNM staging system fails to prognostically discriminate between Stage I and Stage II papillary carcinoma patients and between Stage I and Stage II follicular carcinoma patients. Should this observation hold over time, a simplification of the current AJCC staging system may be warranted.

Total thyroidectomy ± lymph node sampling/dissection represents the dominant method of surgical treatment rendered to patients with papillary and follicular neoplasms. Approximately 38% of such patients receive I-131 ablation/therapy. Only 2-3% of such patients are treated with I-131 alone. At 5 years, patients with T-1 N-0 M-0 disease (as well as patients with Stage I and Stage II disease) survive comparably whether treated by simple lobectomy or by more extensive procedures. The lack of compelling treatment-related differences in survival for patients with more advanced disease emphasizes the importance of longer follow-up for patients with these neoplasms. Such reports currently are planned. With longer follow-up, the survival impact of variations in treatment should become detectable.

Age at diagnosis appears to influence the prognosis of not only patients with papillary, follicular, and Hürthle cell carcinomas, but the prognosis of patients with medullary and undifferentiated/anaplastic carcinomas as well.

The 14% 10-year relative survival of patients with undifferentiated/anaplastic carcinoma, the plateau of the survival curve, and the 55% 5-year relative survival of patients age < 45 years at time of diagnosis suggest that a subset of these patients may be cured of their disease.

The NCDB system permits analysis of care patterns and survival for large numbers of contemporaneous U.S. patients with rare neoplasms such as thyroid carcinoma. In this context, it represents an unsurpassed clinical tool for analyzing care, evaluating prognostic models, generating new hypotheses, and overcoming the volume-related drawbacks inherent in the study of such neoplasms.

Ancillary