Introduction

Although relatively rare, the incidence of thyroid cancer more than doubled between 1984 and 2004 in the United States [1]. Between 1995 and 2004, thyroid cancer was the third fastest growing cancer diagnosis, behind only peritoneum, omentum, and mesentery cancers and “other” digestive cancers [1]. Similarly dramatic increases in thyroid cancer incidence have also been observed in Canada, Australia, Israel, and several European countries [2–9]. The factors underlying this epidemic are not well understood. In the apparent absence of increases in known risk factors, scientists have widely speculated that changing diagnostic practices may be responsible [10, 11].

The primary known risk factor for thyroid cancer is radiation exposure. Potential sources of exposure include radiation used in diagnostic and therapeutic medicine, as well as radioactive fallout from nuclear explosions. However, neither source appears to have increased over the past two decades in the United States. Radiation therapy to the head and neck for benign childhood conditions, once common in the United States, declined after the early 1950s [12]. Similarly, atmospheric testing of nuclear weapons in the United States ceased in 1963 with the signing of the Limited Test Ban Treaty. The effect of such nuclear testing on thyroid cancer rates, though not entirely clear, is thought to be limited [13–15].

The rise in thyroid cancer incidence might be attributable to increased detection of sub-clinical cancers, as opposed to an increase in the true occurrence of thyroid cancer [10]. Thyroid cancer incidence within the United States has been rising for several decades, yet mortality has stayed relatively constant [1, 10]. The introduction of ultrasonography and fine-needle aspiration biopsy in the 1980s improved the detection of small nodules and made cytological assessment of a nodule more routine [16, 17]. This increased diagnostic scrutiny may allow early detection of potentially lethal thyroid cancers. However, several studies report thyroid cancers as a common autopsy finding (up to 35%) in persons without a diagnosis of thyroid cancer [18–21]. This suggests that many people live with sub-clinical forms of thyroid cancer which are of little or no threat to their health.

Though increased detection is the leading hypothesis to explain the increase in thyroid cancer incidence, few studies to date have examined measures of health care access in relation to thyroid cancer. In this study, we identify recent trends (1980–2004) in thyroid cancer incidence in Wisconsin according to patient factors and tumor characteristics. In addition, we conducted ecologic analyses by county to evaluate the association between thyroid cancer incidence and socioeconomic indicators of health care access. We hypothesized that thyroid cancer incidence would be elevated in communities with higher income, education, and health insurance levels, each of which may reflect a higher likelihood of having various symptoms examined for thyroid irregularities.

Materials and methods

Data sources

Incident thyroid cancer cases (ICD-O code 73.9) in Wisconsin between the years 1980 and 2004 were obtained from the Wisconsin Cancer Reporting System. Lymphomas with a thyroid primary site code were not included. Available data provided by the registry for each case included year of diagnosis, race, gender, age group (<20 years, 85+ years, and 5-year categories in between), county of residence at diagnosis, stage of disease at diagnosis, histology code, tumor size, and type of diagnostic confirmation. Tumor size was not a required field for reporting to the registry prior to 1995, thus many cases were missing these data. Age was provided in grouped categories to protect the confidentiality of the cases. The study was determined to be exempt from institutional review board review by the University of Wisconsin Health Sciences Human Subjects Committee.

Histology codes were provided according to The International Classification of Diseases for Oncology, 3rd Edition [22]. Thyroid cancers were grouped into histological code categories as papillary (8050, 8052, 8130, 8260, 8340–8344, 8450, 8452), follicular (8290, 8330–8332, 8335), medullary (8345, 8346, 8510), and other/not specified. Stage at diagnosis was categorized according to the SEER Summary Stage Manual code [23] as localized, regional, distant, and unknown.

Information regarding socioeconomic status for each Wisconsin county was obtained from the 2000 US Census. Variables obtained for each county were the percent of the population aged 25 years and over with a 4 year college degree and the median household income. The percent of the county population with health insurance was obtained from 2006 Wisconsin County Health Rankings Full Report [24]. This measure was based on 1998–2004 data from the Family Health Survey, a statewide random-sample telephone survey of household residents [25]. Respondents were asked whether they had health insurance coverage of any kind at the time of the survey, including but not limited to prepaid plans, HMOs, Medicare or Medicaid. The percent of respondents answering no in each county was averaged over the 7-year period and subtracted from unity. This statistic was available for 71 of 72 counties; the percent insured in Menominee County was not available due to insufficient sample size.

Population data used in incidence rate calculations were obtained from the 1980, 1990, and 2000 US Census via CDC Wonder [26]. Population counts for inter-decade incidence rate calculations were estimated by straight-line interpolation between decade end-points. Population estimates for 2001–2004 were obtained from the National Center for Health Statistics’ interim population projections via CDC Wonder [27]. Annual incidence rates per 100,000 persons were calculated for the time period 1980–2004 and subgroups as defined by gender, age group, stage at diagnosis, cancer histology, and tumor size. Incidence was also calculated by county for the 1995–2004 time period for use in ecologic analyses. Thyroid cancer mortality rates were obtained from CDC Wonder [28, 29]. All incidence and mortality rates were age-adjusted to the US 2000 Standard Population [30].

Statistical analyses

The Estimated Annual Percent Change (EAPC) is the average rate of change in a rate over several years and is used to measure trends over time [31]. The EAPC values were calculated by fitting a least squares regression line to the natural logarithm of the incidence rates using the calendar year as a regressor variable [32]. The slope of the line was tested for a statistically significant difference (p < 0.05) from zero. The EAPC is equal to 100 * (eslope − 1). The 95% confidence intervals (CIs) were obtained using the standard error of the slope. Separate regression analyses were performed for the periods 1980–2004, 1980–1990, and 1990–2004.

Pearson correlation coefficients weighted by population counts in each county were calculated between county income, education, and health insurance variables and thyroid cancer incidence [33]. Weighted maximum likelihood regression was used to estimate the linear association between thyroid cancer incidence and county variables, weighted by county population [34].

The spatial scan statistic was used to test for spatial randomness of the 1995–2004 thyroid cancer incidence rates by county. This method scans the entire spatial area for disease clusters without a priori specification of their size or location and uses a significance test based on the Poisson distribution that compensates for the multiple tests inherent in the process [35].

Age-adjusted incidence rates were calculated using Excel (Microsoft Corporation, Redmond, WA). All regression analyses and correlation calculations were conducted using SAS statistical software (SAS Institute Inc., Cary, NC). The spatial map of Wisconsin incidence rates was produced in ArcView v3.3 (ESRI, Redlands, CA). The spatial scan statistic was performed using SaTScanTM v7.0.2 [36].

Results

A total of 6,898 cases of thyroid cancer were reported to the Wisconsin Cancer Reporting System between 1980 and 2004. Most of the cases were white (95.2%) and female (73.2%). Of the thyroid cancers, 74.1% were papillary carcinomas, 13.9% were follicular carcinomas, 3.7% were medullary carcinomas, and 8.4% were other or not specified carcinomas. Over half of cases were diagnosed while under the age of 50 years. Most cases (94.3%) were histologically confirmed.

The age-adjusted thyroid cancer incidence rate in Wisconsin rose 1.9-fold from 4.3 per 100,000 in 1980 to 8.0 per 100,000 in 2004 (Fig. 1). The trend in incidence was relatively flat prior to 1990, but rose with a 4.0% EAPC (95% CI: 3.3–4.6) between 1990 and 2004 (Table 1). Age-adjusted thyroid cancer mortality remained stable throughout this period at about 0.5 per 100,000 (Fig. 1).

Fig. 1
figure 1

Age-adjusted thyroid cancer incidence and mortality in Wisconsin, USA, 1980–2004

Full size image
Table 1 Trends in age-adjusted thyroid cancer incidence in Wisconsin, USA, 1980–2004
Full size table

Trends by patient characteristics

Both females and males experienced a rise in thyroid cancer incidence (Fig. 2a). The rate of increase was slightly higher in females, such that the female-to-male ratio increased from 2.0 in 1980 to 3.0 in 2004. The EAPC during the 1990–2004 period for females was 4.3% (95% CI: 3.6–5.0) compared to 3.4% (95% CI: 2.3–4.5) for males (Table 1). While the incidence rate was highest in persons aged 60 years or older, increases in incidence between 1980 and 2004 were observed across all age ranges (Fig. 2b; Table 1).

Fig. 2
figure 2

Age-adjusted thyroid cancer incidence by gender (a), age (b), stage at diagnosis (c), histology (d), and tumor size (e; papillary only, 1995–2004) in Wisconsin, USA, 1980–2004

Full size image

Trends by diagnostic classification

Increases in both localized- and regional-staged thyroid cancers were observed between 1980 and 2004 (Fig. 2c; Table 1). Little change was observed in distant-staged cancers during this time period. The increase in regional-staged thyroid cancer appeared to begin during the 1980–1990 period although the EAPC did not reach statistical significance (Table 1; p = 0.12). However, during 1990–2004, incidence of localized cancers rose at a faster rate (EAPC = 5.4, 95% CI: 4.7–6.1) than regional-staged (EAPC = 3.4, 95% CI: 1.7–5.2).

The rising trend in incidence between 1980 and 2004 appeared to be due almost entirely to increases in cancers of papillary histology, which experienced a 2.6-fold increase between 1980 and 2004 (Fig. 2d). The EAPCs for papillary cancers were 3.0 (95% CI: 1.2–4.9) during 1980–1990 and 5.2 (95% CI: 4.5–5.9) during 1990–2004 (Table 1). Follicular cancers decreased in incidence during the 1980–1990 period, but this was balanced by an increase in incidence during the 1990–2004 period such that no overall trend between 1980 and 2004 was observed (Fig. 2). Among papillary thyroid cancers, increases in incidence between 1995 and 2004 were observed for all tumor sizes (Fig. 2e; tumor size data unavailable prior to 1995). The vast majority of the absolute increase in incidence consisted of tumors ≤5 cm in size. A large percentage of papillary cancers reported since 1995 (18.8%) were missing tumor size data. Between 1995 and 2004 the incidence of papillary thyroid cancers of unknown size increased at a similar rate (EAPC = 4.6, 95% CI: −0.4–9.9) as cancers with tumor size less than or equal to 5.0 cm (EAPC = 6.0, 95% CI: 4.4, 7.5).

Geographic variation

Substantial geographic variation was observed in thyroid cancer incidence within Wisconsin (Fig. 3). The spatial scan statistic indicated that the geographic pattern was not likely due to random variation (p = 0.001). Incidence ranged 7.8-fold by county, from 1.7 per 100,000 in Sawyer County to 13.3 per 100,000 in Menominee County.

Fig. 3
figure 3

Geographic variation in thyroid cancer incidence by county in Wisconsin, USA, 1995–2004

Full size image

A moderate positive correlation by county was observed between thyroid cancer incidence and indicators of socioeconomic status and health care access. Counties with higher median household income tended to have higher age-adjusted thyroid cancer rates during the period 1995–2004 (Fig. 4a; Pearson correlation = 0.25, p = 0.04). For each increase of $10,000 in county median income, thyroid cancer incidence increased on average by 0.5 per 100,000 persons. Similarly, counties with a higher percentage of residents with a college degree tended to have higher thyroid cancer incidence (Fig. 4b; Pearson correlation = 0.24, p = 0.04). For each increase of 10% in percentage of county residents with a college degree, thyroid cancer incidence increased on average by 0.5 per 100,000 persons. A stronger correlation was observed between thyroid cancer incidence and percent of county residents with health insurance (Fig. 4c; Pearson correlation = 0.41, p < 0.001). Exclusion of the outlier, Iron County (with only 79% insured), did not substantially affect the correlation (Pearson correlation = 0.40, p < 0.001). For each increase of 5% in percent of county residents with health insurance, thyroid cancer incidence increased on average by 1.4 per 100,000 persons.

Fig. 4
figure 4

Age-adjusted thyroid cancer incidence (1995–2004) versus median household income (a), percent of residents with a college degree (b), and percent of residents with health insurance (c), by county, Wisconsin, USA. Circle sizes are proportional to county population. Pearson correlations are 0.25 (p = 0.04), 0.24 (p = 0.04), and 0.41 (p < 0.001), respectively

Full size image

Discussion

The incidence of thyroid cancer in Wisconsin rose dramatically, almost doubling, between 1980 and 2004. The increase in incidence was largely limited to the time period 1990–2004, in which an overall estimated annual change of 4.0% was observed. The increasing trend was present in men and women, across all age groups. However, the increase primarily consisted of localized- and regional-staged cancers, the vast majority of which were of papillary histology. Among papillary cancers, increases in incidence were observed for all tumor sizes. Significant geographic variation was correlated with the health care access indicators of income, education, and health insurance.

Wisconsin is comparable in many demographic factors to national norms. A state of slightly more than five million residents, Wisconsin has similar levels of educational attainment and median income as the nation, though it is less diverse with 87% of residents identifying as non-Hispanic whites in 2000 compared to the national average of 69% [37, 38]. Approximately 32% of Wisconsin residents live in rural areas, placing it in the top two-fifths of states nationally [37]. Wisconsin also has a higher percentage of residents with health insurance coverage. Roughly 9% of the overall population of Wisconsin is uninsured, compared to the national average of about 15% [39]. The annual Wisconsin County Health Rankings report has noted that the more affluent counties surrounding the Milwaukee metropolitan area consistently rank near the top of the state for health determinants and health outcomes [24]. Menominee County, home to the Menominee Indian Reservation and the highest percentage (25%) of families living under the poverty line [38], consistently ranks poorly.

Trends of thyroid cancer incidence in Wisconsin largely reflect those observed nationally [1, 10]. In 2004, SEER [1] reported an overall age-adjusted incidence of 9.8 per 100,000, representing a 2.3-fold increase since 1980. In 2002, diagnoses were more common in women than in men by a ratio of 2.7–1, and papillary cancers represented 88% of all thyroid cancer diagnoses [10]. The increase in incidence between 1973 and 2002 in the US consisted almost entirely of papillary cancers [10]. We observed a similar female: male ratio (3.0 in 2004) in Wisconsin, and a similar dominance of papillary cancers in the rise in incidence.

Davies [10] observed that the rise in papillary cancer was largest among tumor sizes ≤1 cm, accounting for 49% of the rise in overall papillary cancer incidence between 1988 and 2002. Similarly, we found that approximately 50% of the rise in papillary cancer in Wisconsin between 1995 and 2004 could be attributed to tumor sizes ≤1 cm. However, we also observed substantial increases in cancers >2 cm in size. Whereas Davies attributed only 13% of the increase in incidence to cancers >2 cm, 39% of the increase in Wisconsin could be attributed to these larger cancers. Notably, substantial increases in incidence for tumors larger than 1 cm has also been reported in other countries [4, 40]. Unfortunately, almost 20% of papillary cancers reported in Wisconsin since 1995 were missing tumor size data. It is difficult to estimate the potential for bias that this may introduce, but it is possible that size data for larger tumors may be preferentially reported, thereby explaining the difference between our results and those observed nationally.

The increasing preponderance of small, localized papillary cancers is consistent with the hypothesis that increasing detection capabilities have contributed to increasing incidence. In order to further explore this hypothesis, we analyzed the relationship between thyroid cancer incidence and various community-level socioeconomic factors related to health care access. Persons with higher health care utilization may be more likely to be referred for diagnostic workup of the thyroid in response to the often vague and relatively common symptoms of thyroid disorders, such as menstrual irregularities, fatigue, and nervousness [41]. In general, thyroid cancer incidence across counties was moderately correlated with indicators of higher socioeconomic status and health care access. Of the factors we examined, percentage of county residents with health insurance was most strongly correlated with incidence (Pearson correlation = 0.41, p < 0.001). This supports the hypothesis that advancing diagnostic capabilities may be contributing to the observed increase in thyroid cancer incidence.

Few studies have investigated the relationship between thyroid cancer and socioeconomic status or health care access. In a clinic-based study in Houston, Ghori et al. [42] found that thyroid cancer patients from lower income neighborhoods were more likely to have advanced stage cancers. A study of occupation in Canada found no relationship among inferred education, income, or occupational prestige and thyroid cancer incidence [43]. In Switzerland, a strong gradient of thyroid cancer mortality rates was observed by social class, with lower social class experiencing elevated mortality [44].

While the results of this study support the hypothesis that increasing diagnostic scrutiny underlies the rise in thyroid cancer incidence, other theories have been proposed. Baker and Bhatti [45] have voiced concern with the rapid adoption of computed tomography (CT) for a wide array of diagnostic purposes in the United States. Due to the many two-dimensional X-ray slices (up to 128) used to create a three-dimensional image, CT scans impart higher levels of radiation compared to earlier imaging technologies. Alternatively, Liu et al. [2] suggested that secular changes in female hormone use and reproductive patterns in Canada may partially explain the increasing thyroid cancer incidence in young and middle-aged women, as various studies have indicated a relation between reproductive factors and thyroid cancer [46]. We were unable to explore these hypotheses in this study.

A number of limitations to our study need to be considered. Our comparisons of subgroup-specific incidence rates are based on magnitude differences and not formal statistical comparisons. For certain subgroups, such as large, non-papillary, and distant-staged cancers, analyses were based on small numbers of annual cases. The potential effects of changing histological classification procedures should also be noted. We found a decrease over time in the number of thyroid cancers classified as “other” histology. It is possible that similar cancers once reported as “other” histologies may later have been reported as papillary, follicular, or medullary histologies. This could have contributed in part to the rising incidence of papillary tumors, but given the small number of tumors classified as other (<1 per 100,000 in 1980), the potential effect of changing histological classification on papillary tumor incidence is small.

Rates may also have been affected by the estimated population counts for inter-census years. We used straight-line interpolation between census years, which may not accurately reflect the true population. However, it is not likely that these factors had a substantial influence on the long-term trends we observed in thyroid cancer incidence.

For this study, cancer data were obtained from the statewide mandatory tumor registry. A substantial portion of Wisconsin cancer patients living in counties that border other states receive treatment out of state. In this study, thyroid cancer incidence appeared lowest in the northwest corner of the state, much of which borders Minnesota. It is possible that case reporting to Wisconsin’s cancer registry is less complete for border counties. In order to limit the effect of underreporting in these border counties, all ecologic analyses were restricted to the 1995–2004 time period, when data exchange agreements between Wisconsin’s cancer registry and Minnesota hospitals were implemented. Notably, the rise in incidence observed in Wisconsin continued to occur well after 1992 when data exchange agreements were first initiated and thus is not likely attributable to increasing completeness among border counties. In fact, after excluding the 10 counties that directly border Minnesota, the estimated annual percent change in thyroid cancer incidence remained at 2.8% for the 1980–2004 period.

As with any ecologic study, we are unable to draw conclusions regarding the causal relations between health care access, diagnostic techniques, and thyroid cancer incidence. The analyses were based on county-level data, as data on thyroid cancer incidence for census-block or smaller areas were not available. We do not have data on individuals’ income, education, and health insurance status, nor do we have information on the utilization of ultrasonography and fine-needle aspiration biopsy. Future studies are necessary to directly evaluate the role of changing diagnostic techniques in thyroid cancer incidence trends.

In summary, the incidence of thyroid cancer in Wisconsin is increasing dramatically among men and women of all ages. The increase in incidence is largely, but not entirely, due to localized papillary cancers of small size. A moderate ecologic correlation was observed by county between community-level socioeconomic indicators of health care access and thyroid cancer incidence. These results are consistent with the proposed role of improved diagnostic techniques in increasing incidence. The possible over-diagnosis of thyroid cancer may contribute to unnecessary medical care and loss of quality of life. Since mortality rates for thyroid cancer have not substantially changed in conjunction with these increased incidence rates of early cancers, these data strongly support the urgent need to distinguish thyroid cancer cases that have the greatest malignant potential.