Two of the main features of thyroid malignancy are its rarity and its excellent prognosis in the overwhelming number of cases. On the other hand, these factors induce some specific difficulties in epidemiologic studies of thyroid cancer.

Apart from special aspects for thyroid malignomas,  the validity of epidemiological data in general depends on a number of factors, which may cause pitfalls.

Each scientist who implements or analyzes studies in this field should bear in mind those fundamental difficulties, otherwise erroneous conclusions or fatal epidemiologic misinterpretations may be drawn. This chapter therefore opens with some basic methodological reflections.

Ideally, for epidemiological research, complete data on the entire population should be available. At least, it must be assured that the data sample is representative of the study population in every relevant aspect.

For this aim,  well-organized structures for data acquisition and transmission to centralized agencies are necessary. Mandatory reporting of all diagnosed malignancies, cancer deaths, routine autopsies, and comparison of these data with information from the official register of births and deaths by a centralized cancer registry would be beneficial for oncological epidemiology. This ideal situation does not exist, even in many industrial nations with highly developed health care systems, due in part to ethical considerations as well as concerns about the protection of privacy.  For example,  in Germany,  not until 1995 was legislation passed to allow physicians to report newly diagnosed malignancies to the registry (however, reporting is not mandatory). Meanwhile, most German federal states have established central cancer registries, but even at present the registration is not nationwide.  In the USA,  the National Cancer Data Base (NCDB)  records approximately 60% of all cancer cases but, again, the majority of the information is provided on a voluntary basis.

Due to the limitations of centrally acquired data or even their complete absence in many countries,  epidemiological studies are often based on clinical data,  which are flawed by selection biases.  The continual improvement of diagnostic modalities is another factor complicating data comparison, e.g., the improved sensitivity of ultrasonography since the 1980s makes the historical comparison of incidence data highly problematic. The reported increasing incidence of thyroid cancer, especially in the early stages, is at least in part due to improved diagnostic techniques, which lead to earlier detection _[69].

The slow growth rate of thyroid carcinomas causes another difficulty for epidemiological research.  There are often decades between tumor induction and clinical manifestation,  and in a substantial number of cases the disease remains undetected during life, and tumor mortality is generally low. In 1998 for example, in the United States 17,200 new cases of thyroid cancer from a total of 1,228,600 new cancer cases were estimated, with a male:female ratio of 1:3.7.

In contrast,  only 1,200 deaths from thyroid cancer were estimated _[101].  For this reason, in contrast to highly malignant tumors such as lung or pancreatic cancer,  mortality statistics do not reflect the incidence of thyroid cancer and statistics of the incidence of clinically manifest tumors do not describe the prevalence in the population.

Data from such mortality rates, incidence rates or survival rates of malignomas such as thyroid cancer should be described in terms of “age-standardized rates”  rather than “crude rates”.  The “age-standardized mortality”  describes the expected number of deaths per 100,000 for a reference population of a given age and sex distribution.  Due to the age dependence of cancer incidence and mortality, comparison of those data in populations with different age distributions, e.g., in different regions or even within the same region are not possible without age standardization, since the age structure of the population changes over time.  The reported age-standardized figures depend on the individual reference population, which may underlie regional or historical variations. For thyroid cancer in Wales/United Kingdom (1985–1996, all ages) for example, the crude incidence rates were 1.04 for males and 2.94 for females. The corresponding figures underlying the European age-standardized rates (EASR) were 0.97 and 2.52,  and underlying the world age-standardized rates (WASR)  0.75 and 2.06, respectively _[92].

Data on the frequency of thyroid cancer can be obtained by systematic thyroid diagnostics in large representative populations. However, the incidence of thyroid nodules in endemic goiter areas is substantially greater than the incidence of thyroid cancer: up to 30–50% nodules, of which only 5–10% are malignant _{[65, 87]}. Adequate further examination (i.e., scintigraphy, cytology and histology) of all detected nodules is of critical importance, but due to the immense resources and costs such studies have remained scarce. Serial autopsy studies are another means of providing prevalence data for thyroid cancer, especially for clinically occult carcinomas; however, the data obtained from those studies do not represent the rate of clinically apparent and relevant malignancy.

The quality of all diagnostic data is also of critical importance. It is dependent on the technical equipment and the experience of the clinician and the training of the pathologist in the challenging task of classifying the subtypes of thyroid tumors. Classification schemes used for clinical staging as well as histological typing are important for making comparisons. In cancer registry data banks, epidemiological data (e.g., for incidence and mortality) are often only available following the International Classification of Disease (ICD) code, where all types of thyroid cancer are encoded with one common code number. Thus, no statements could be given concerning the considerable epidemiological differences for the individual histological subtype on the basis of those data banks. Since the classification of tumors and staging have repeatedly changed historically, appropriate reclassification is necessary for any comparison of current with older data, if the epidemiology of histological subtypes is examined.

The current histological World Health Organization (WHO)  classification was published 1988 (2nd edition)  and superseded the 1st edition published in 1974 _{[53, 54]}. The decision in the 1st edition to regard a tumor with even minor papillary component as a papillary carcinoma (despite the presence of follicular patterns)  has been upheld.  In previous schemes,  the predominant formation determined the final classification of the tumor, leading to a larger number of follicular carcinomas in these older reports. Differences between the 1st and the 2nd edition are: the removal of the subtypes of undifferentiated carcinomas, a revision of the nonepithelial and miscellaneous tumors,  and the recognition that the great majority of tumors previously diagnosed as “small-cell carcinomas” of the thyroid are malignant lymphomas. Like the 1st edition, the 2nd edition does not consider certain subtypes as tumor entities in their own right (e.g.,  tall cell,  columnar cell,  diffuse sclerosing or diffuse follicular variant of papillary carcinoma, insular carcinoma, Hürthle cell carcinoma), although they show substantial epidemiological differences from the classic types of papillary and follicular carcinomas.

The tumor-node-metastasis (TNM)  classification has also been revised repeatedly. The differences between the 3rd and the 4th edition _[126] also influence the prognostic scoring in risk-group stages, which is derived from the TNM system.  Lateralization and multicentricity of the primary tumor were abandoned in favor of tumor size, and the regional lymph node (N) classification was simplified. After a 2nd revision in 1992 of the 4th edition and a supplement in 1993, the 5th edition was published in 1997. It reconciled the systems of the American Joint Committee on Cancer (AJCC) and the Union International Contre Cancer (UICC)  _[115].  Another change was the definition of “occult papillary carcinoma” or “papillary microcarcinoma”: older publications allow a maximal diameter of 1.5 cm,  and the 5th edition of the TNM classification defined tumors smaller than 1 cm as T1 carcinomas.  In 2002,  the 6th edition was introduced _[116].  It implies considerable modifications (e.g.,  borders for the T-stage classification, dependency of the T4 state from clinical information from the surgeon, lymph node level) and is not compatible with the former TNM classification, which is why previous study data involving TNM stages cannot be transformed. Even a recently published supplement _[135] could not completely eliminate these criticisms.

Data quality concerning the prognosis of a given thyroid cancer and the prognosis itself not only depend on the tumor biology or the properties of the individual patient, but also on the quality and duration of aftercare of the patients (especially on the sensitivity of diagnostic methods for detection of recurrences), on demographic factors and on the therapy carried out. Due to the overall favorable prognosis of differentiated thyroid carcinoma, slight variations of relapse-free intervals or survival rates between different histological subtypes, stages or special risk-groups and changes of these factors in the time course or due to new forms of therapy can only be evaluated if large patient groups are observed over a long time period. On this point, the existence of long-term follow-up programs with adequate data documentation is of benefit.

###

R. Görges
H.-J. Biersack , Professor Dr. med.
Klinik und Poliklinik für Nuklearmedizin
Rheinische Friedrich-Wilhelms-Universität
Sigmund-Freud-Strasse 25
53127 Bonn, Germany