Although thyroid cancers are rare in children and adolescents, they are still one of the most frequent cancers in these age groups. They constitute 1% of all malignancies worldwide1 and are heterogeneous in terms of histology, clinical presentation, treatment response, and prognosis. Papillary (PTC) and follicular (FTC) thyroid carcinomas are the two entities referred to as differentiated non-medullary thyroid carcinomas of the follicular epithelium. Differentiated non-medullary thyroid cancer is more common among women, particularly during the fertile part of women's life.
PTC is the most common of the 2 entities, comprising between 60–90% of differentiated thyroid cancer.2 The incidence of PTC for women rises to a peak at 55–65 years, whereas this pattern is less pronounced in men.3 The incidence of FTC increases slowly with age and reaches its highest level around 60 years of age for both men and women.
The only established risk factor for thyroid cancer in humans, besides age and gender, is ionizing radiation.4 Studies of individuals living in the Chernobyl areas and of the survivors of the atomic bombs in Hiroshima and Nagasaki have shown an increased risk of particularly PTC, although only among those exposed as children.4 Sex hormones, iodine deficiency and other factors have been proposed as risk factors for thyroid cancer but findings are inconsistent.5
The prognosis of differentiated thyroid cancer is known to be excellent with a 10-year survival exceeding 90%.6 Comparisons are, however, difficult due to differences in selection of patients and duration of follow-up. Local invasiveness, presence of distant metastases, and increasing age at diagnosis convey a worse prognosis.7 Recurrence or metastases may appear even 20 years after the initial diagnosis. Mazzaferri and Jihang6 followed 1,528 patients for 40 years, but many studies of survival after a diagnosis of differentiated thyroid cancer are limited by small numbers of cases or a short follow-up period.
The primary aim of the present analysis was to study the prognosis of patients diagnosed with differentiated thyroid cancer in relation to histology, age at diagnosis, gender, and calendar year of diagnosis in a large Swedish cohort of 5,554 thyroid cancer patients, with a comparatively long follow-up.
MATERIAL AND METHODS
The Swedish Cancer Registry (SCR) was established in 1958 and covers the entire population (8.4 million inhabitants in 1987). All Swedish residents are assigned a unique national registration number, which makes it possible to follow every individual from birth to death. Reporting to the SCR has been mandatory for clinicians and cytologist/pathologists since 1958. The proportion of all thyroid cancers reported to the SCR for the period up to 1978 was estimated to be 98%8 but coverage is now considered to be close to 100%.9 Cases detected incidentally at autopsy are included in the cancer register but were excluded from our study. The SCR does not register tumors where the death certificate is the only source of information.
The study was based on all cases of thyroid cancer reported to the registry between 1958–87, with follow-up until December 31, 1999. The study was limited to individuals diagnosed up to 1987 because detailed information on histology is not available for patients diagnosed after 1987. Information on stage or treatment of the tumors is not recorded by the cancer registry, nor is detailed information on histopathology. The original histopathological reports of all registered thyroid cancers were therefore examined to establish the histopathological entity of each case.10 Of the 7,906 thyroid cancers reported to the registry, 1,405 (18%) patients with anaplastic and medullary thyroid cancers and 947 (12%) patients with follicular thyroid adenomas were excluded leaving 5,554 individuals diagnosed with differentiated non-medullary thyroid cancer in the study. The few Hürthle cell carcinomas (n = 125) did not allow detailed analyses and they were included in the FTC group. Tumors classified as mixed follicular and papillary were included in the PTC group.
Annual incidence rates per 100,000 were calculated for 5-year age groups and 5-year calendar periods of diagnosis. The relative survival ratio (RSR) was used as the measure of patient survival.11 The advantage of the RSR is that it does not rely on the accurate classification of cause of death. It also provides a measure of the total excess mortality associated with a diagnosis of cancer irrespective of whether the excess mortality is directly or indirectly associated with this diagnosis. Estimated from life tables, the RSR is defined as the observed survival in the patient group (where all deaths are considered events) divided by the expected survival of a comparable group from the general population, which is assumed to be free of the cancer in question. Expected survival was estimated using the Hakulinen method from Swedish population life tables stratified by age, gender and calendar time.12 Expected and relative survival was estimated using software developed at the Finnish Cancer Registry.13 Standard errors were estimated using Greenwood's method14 and 95% confidence intervals (95% CI) constructed on the log cumulative hazard scale. Estimates of relative survival for all ages combined were made by calculating a weighted average of the age-specific estimates where the number first at risk in each age group were used as weights.15
Relative survival was modeled using the life table regression model developed by Hakulinen and Tenkanen.16 The model provides estimates of excess hazard ratios (relative excess risk = RER)17 adjusted simultaneously for potential confounding factors. The model was fitted separately to follow-up Years 1–6, 7–13 and 14–20 because the proportional excess hazards hypothesis is not generally valid over long periods of follow-up. Because the majority of deaths occurred in the first 6 years subsequent to diagnosis, more emphasis was placed on the analysis of these years.
A total of 5,554 cases were included in the study (Table I) of which 65% were classified as PTC (n = 3,588) and 35% as FTC (n = 1,966). In all, 75% of the cases were found in women, and the female dominance was slightly more pronounced in PTC than in FTC (Table I). The mean age at diagnosis was 50 years for PTC and 56 years for FTC. The incidence of PTC started to increase already before the age of 20, increased steadily until 65 years of age, and decreased thereafter (Fig. 1). A similar pattern was seen for FTC although less pronounced. The sharp increase in female PTC seen during the fertile part of life was not seen for female FTC. The incidence of PTC increased over the time span of the study (Fig. 2), particularly for females (p < 0.0001, for the interaction between gender and calendar period). Little change in the incidence of FTC was observed over calendar time.
Table I. Characteristics of Patients Diagnosed with Papillary (PTC) and Follicular (FTC) Thyroid Cancer
Mean age at diagnosis, years
Mean follow up, years
The mean follow-up time for the PTC patients was 17 years and for the FTC patients 15 years (Table I). The unadjusted relative survival estimates suggest a more favorable prognosis for patients diagnosed with PTC compared to those diagnosed with FTC (Table II). Much of this difference can be explained by the confounding effect of age. Women experience a superior survival to men (Table II), although the magnitude of the difference varied according to age and histopathological subgroup. Patients with mixed papillary and follicular tumors did not differ in survival compared to PTC patients (data not shown). These 2 groups were therefore treated as one entity.
Table II. Cumulative 5-, 10-, 20- Year Relative Survival Ratio of Papillary (PTC) and Follicular (FTC) Thyroid Cancer by Gender and Age at Diagnosis in Sweden 1958–871
The relative survival compares the observed number of deaths with the expected mortality in and age-, gender-, and calendar period matched comparison group.
Before adjusting for potential confounding factors, it was estimated that patients diagnosed with FTC experienced 90% higher excess mortality than patients diagnosed with PTC during the first 6 years of follow-up. After adjusting for age, gender and calendar period of diagnosis, there was no evidence of a difference in excess mortality between patients diagnosed with FTC and PTC during the first 6 years after diagnosis (Table III). Figure 3 shows the pattern of excess mortality as a function of time since diagnosis for patients at the reference level of each of the variables in the regression model (males, aged 50–59, diagnosed with PTC during 1958–67). Estimates of the number of excess deaths for other categories of patients can be obtained by multiplying the estimated excess mortality rates by the appropriate RERs. For example, estimates for the number of excess deaths during the first 6 years for patients diagnosed 1978–87 can be obtained by multiplying the numbers in Figure 3, by 0.44 (see Table III). The reduction, however, in the number of excess deaths during the first 6 years of follow-up for patients diagnosed 1978–87 compared to 1958–67 is offset by a 31% increase in the number of excess deaths during years 14-20. This is an example of nonproportional excess hazards and the reason why separate models were estimated for each period of follow-up.
Table III. Estimated Relative Excess Risks (RER) and 95% Confidence Intervals for each of 3 Periods of Follow-Up1
1–6 years RER 95% CI
7–13 years RER 95% CI
14–20 years RER 95% CI
All estimates are simultaneously adjusted for all other factors.
The deviance is a measure of model goodness-of-fit (under the hypothesis that the model fits, the deviance should follow a chi-square distribution with the specified degrees of freedom.
Somewhat contrary to our perceptions based on clinical practice, we observed no difference in excess mortality between patients diagnosed with PTC and FTC during the years immediately after diagnosis (where the majority of deaths occur). There was, however, evidence that patients diagnosed with FTC experienced higher excess mortality during the period 7–20 years after diagnosis (Table III). In general, women experienced lower excess mortality than men, although there was evidence that gender-specific differences in excess mortality are modified by other factors. Under the assumption of no effect modification (i.e., the estimates presented in Table III) it was estimated that females experienced 40% lower excess mortality than men during the first 6 years after diagnosis, a difference that was statistically significant. Closer examination, however, revealed larger differences in survival between males and females among young patients compared to old patients, particularly among those diagnosed with FTC (Table IV). In fact, among individuals diagnosed with FTC at age 70 years or older there was no evidence of a difference in survival between males and females. The estimates in Table IV were estimated from an age-gender-histopathology interaction term, which was the only statistically significant interaction in any of the models. Hürthle cell carcinomas had a less favorable prognosis compared to FTC and PTC, but the difference was not statistically significant, mainly because of the few cases identified (n = 125).
Table IV. Estimated Relative Excess Risk (RER) for Females Compared to Males During the First 6 Years of Follow-Up from the Model Containing an Interaction Between Age, Gender and Histopathology
RER female/male (age group)
In this large population based cohort of 5,554 patients it was shown that the incidence of PTC increases over calendar period and that incidence of PTC among women is high during and after their fertile part of life. The divergence over calendar period in PTC and FTC incidence could reflect etiology but could also be due to misclassification because methods of diagnosing and classifying thyroid cancer have changed over time. Up until the beginning of the 1980s, morphological signs of papillary growth pattern were needed to classify a thyroid tumor as PTC. Today detection of psammon bodies or “ground-glass” nuclei is sufficient for a diagnosis of PTC, even in the absence of the typical papillary changes.
A validity control of the cohort, using the classification introduced in 1974,18 has been carried out previously.10 Among 220 tissue samples obtained for review, the original histological type was confirmed in 98% of the PTC and in 54% of the FTC. The proportion of falsely classified FTC was higher in the early part of the study period. Approximately 10% of the samples originally classified as FTC were re-classified as benign thyroid lesions and 30% were re-classified as PTC. This misclassification most certainly influences our findings, but it does not completely explain the increase of female PTC.
Ionizing radiation is the only established risk factor for differentiated thyroid cancer and PTC has a stronger association to ionizing radiation than FTC.4 Females are also known to be more susceptible to ionizing radiation.4 It has been discussed if this gender difference could be due to hormonal factors. It could also be that women have a better access to health care facilities that influences detection of thyroid cancer. There is also a strong effect modification of age at exposure. The increased incidence of PTC could partly be due to a previous extensive use of radiotherapy for benign conditions in the neck region.4 During the 1930–50s it was common to treat benign lesions, such as tonsillitis or tuberculosis that affected lymph nodes, with radiotherapy. It has been proposed previously that the increase of PTC could be due to such treatments.2, 19 Intensified and better screening modalities are most likely also influencing incidence over time.
Female sex steroid hormones are known to influence the risk of breast cancer.20 The striking gender differences in the incidence of differentiated thyroid cancer, particularly for PTC, suggests a possible influence of hormone related factors because the higher overall incidence of female PTC is established through a sharp increase during the fertile years of life. An increased risk of PTC has been seen among current users of oral contraceptives,21 whereas no such increase was evident among former users. Estrogens are known to induce thyroid cancer in mice,22 providing further support for the hypothesis that incidence is hormone related. The increase of PTC over calendar time may have a similar etiological basis as the increase in female breast cancer related to increasing use of oral contraceptives and hormone replacement therapy.20 Although, from a pooled analysis of case-control studies no association between hormone replacement therapy and thyroid cancer could be shown.21
We found a better prognosis for patients diagnosed during the latter part of the study period, among those diagnosed at younger ages, and in women. As a result of the more widespread use of fine-needle biopsy and radioisotope thyroid scanning, diagnostic intensity increased during the study period. Through earlier detection the increased medical surveillance could induce a lead-time bias that affects prognosis. A possible explanation for the improvement in survival over time could thus be that more cancers were found at an early stage.23, 24
The age at which excess mortality increases has been discussed and a less favorable prognosis noted in individuals diagnosed after 40 years of age,7 whereas others claim that tumors behave more aggressively after the age of 60 years.24 In the present study, survival declined with increasing age at diagnosis although there was evidence of a large difference between patients >40 years of age at diagnosis compared to those <40 years of age.
The fact that women have a higher incidence of PTC, particularly during the fertile part of life, but at the same time a better prognosis than men, is intriguing. The female prognostic advantage was particularly evident in patients younger than 50 years at diagnosis. Several studies have shown that female thyroid cancer patients experience superior survival in all age groups,2, 7, 25, 26 in contrast to one other report where no significant difference was seen.27 We observed a superior survival for females, but noted that the magnitude of the difference in survival between males and females depended on both histology and age at diagnosis. Our data suggest that there may exist a class of female thyroid tumors that are diagnosed during the fertile part of life and associated with a superior prognosis. Sex hormones may play a role in the etiology of these tumors. A similar pattern has been suggested for breast cancer associated with hormone replacement therapy.28, 29
In the clinical setting, patients with PTC are thought to have a better prognosis than those diagnosed with FTC. After adjusting for gender, age and calendar period, however, this advantage is not evident within the first 6 years of diagnosis. During the period 7–20 years after diagnosis, there is evidence that patients with PTC have a better prognosis than those diagnosed with FTC, although total excess mortality is relatively low during these years.
The strengths of the current study are the large population-based cohort with a long follow-up and detailed information on histopathology. This gave us the opportunity to address questions such as prognosis over time, gender differences and the influence of age. A weakness of the study is the lack of information on the stage of the tumors at diagnosis, because stage is an important prognostic factor,28 as well as the misclassification of FTC discussed above.
The higher incidence and corresponding superior survival among females compared to males is puzzling and needs further evaluation. The fact that the gender differences appear to be greater during the fertile part of life suggests that sex hormones may play an important role in the etiology of thyroid cancer.