Risk of thyroid as a first or second primary cancer. A population‐based study in Italy, 1998–2012

Abstract Background The number of patients living after a cancer diagnosis is increasing, especially after thyroid cancer (TC). This study aims at evaluating both the risk of a second primary cancer (SPC) in TC patients and the risk of TC as a SPC. Methods We analyzed two population‐based cohorts of individuals with TC or other neoplasms diagnosed between 1998 and 2012, in 28 Italian areas covered by population‐based cancer registries. Standardized incidence ratios (SIRs) of SPC were stratified by sex, age, and time since first cancer. Results A total of 38,535 TC patients and 1,329,624 patients with other primary cancers were included. The overall SIR was 1.16 (95% CI: 1.12–1.21) for SPC in TC patients, though no increase was shown for people with follicular (1.06) and medullary (0.95) TC. SPC with significantly increased SIRs was bone/soft tissue (2.0), breast (1.2), prostate (1.4), kidney (2.2), and hemolymphopoietic (1.4) cancers. The overall SIR for TC as a SPC was 1.49 (95% CI: 1.42–1.55), similar for all TC subtypes, and it was significantly increased for people diagnosed with head and neck (2.1), colon–rectum (1.4), lung (1.8), melanoma (2.0), bone/soft tissue (2.8), breast (1.3), corpus uteri (1.4), prostate (1.5), kidney (3.2), central nervous system (2.3), and hemolymphopoietic (1.8) cancers. Conclusions The increased risk of TC after many other neoplasms and of few SPC after TC questions the best way to follow‐up cancer patients, avoiding overdiagnosis and overtreatment for TC and, possibly, for other malignancies.


| INTRODUCTION
The number of patients living after a cancer diagnosis is rising in Italy, 1 as well as in several other countries. 2 In particular, one of the largest 10-year increases is foreseen for thyroid cancer (TC) patients (+79% between 2010 and 2020), 1 largely due to the impact of overdiagnosis on the rapidly increasing TC incidence, 3,4 the modest improvements of survival, 5 and substantially stable mortality rates. 3 The vast majority of people living after a TC diagnosis have the same life expectancy as the general population (i.e., they are cured), 6 in particular those who were overdiagnosed. 7 Notably, overdiagnosis of asymptomatic TC accounted for 75% of cases in Italian women and 63% in Italian men, between 1998 and 2012. 3 In Italy, more than 3.4 million people are living after a cancer diagnosis other than TC. 1 They may be at risk of developing a new primary cancer due to several reasons, including common etiologic factors (i.e., environmental exposures, genetics, and lifestyles), late effects of cancer treatments, and, possibly, enhanced surveillance. 8,9 In addition, several studies from Europe and the United States have reported a substantial risk of TC detection during other cancers follow-up. [10][11][12][13] A higher than expected incidence rate of second primary cancer (SPC) in TC patients has been reported and recently confirmed. 13,14 In particular, an increased incidence has been consistently reported for breast cancer, [15][16][17][18][19] kidney cancer, 10,12,[19][20][21] and lymphomas/ leukemias. 10,12,21,22 The present study aimed to provide updated estimates of the risk of SPC after TC, as well as the risk of TC as SPC. In addition, we comprehensively explored the risks of TC as a first or SPC by sex, age, and time since first diagnosis. Evidence on the combination of TC with other cancers, as a first and second primary one, may help clarify whether the association is due to shared genetic or lifestyle risk factors, close anatomic proximity, treatment of the first cancer, or intensity of diagnostic activities.

| Study population
We analyzed data collected in 28 population-based Italian cancer registries (CRs), covering over 22 million inhabitants (39% of the Italian population) (Appendix 1). All included CRs had been active for at least 10 years in the period of interest, that is 1998-2012. 3 First cancer and SPC were classified using international classifications for topography and morphology (ICD10 and ICD-O-3). 23, 24 We analyzed 36 cancer sites or types (Appendix 2), but we excluded non-melanoma skin cancer, cases detected at autopsy, and those known from death certificate only or with follow-up time equal to zero. Third or subsequent malignant tumors were very rare (0.3% of all subjects) and cases diagnosed at the age of 85 years or more were, therefore, excluded. 24 Two cohorts were analyzed: Cohort 1 (SPC after TC): included patients with a TC diagnosis evaluated for the incidence of a SPC other than TC, with overall 276,100 person-years of observation (216,431 in women and 59,669 in men).
Cohort 2 (TC as SPC): included patients with cancers other than TC evaluated for TC incidence. TC cases diagnosed as a first tumor were excluded due to the international rules for multiple primaries definition adopted by Italian CRs. 24

| Statistical analysis
Person-years at risk (PY) were computed by first cancer site, histological type of TC, (i.e., papillary, follicular, medullary, and poorly differentiated including anaplastic), geographic area (North, Center, and South and Islands), sex, age group (0-4 years, …, 80-84 years), and calendar-year group (1998-2002, 2003-2007, and 2008-2012). Observed cases included incident cancers reported to CRs during the above-defined person-years at risk. The expected number of cancer cases was computed by multiplying the cumulative person-years of observation by the specific incidence rates for the strata in which person-years were distributed. Observed SPC incidence among cancer patients was compared with expected numbers by means of standardized incidence ratio (SIR). Byar's approximation was applied to the exact Poisson distribution to calculate 95% confidence interval (CI).
To minimize the impact of "intensive screening" in concurrence with the first cancer diagnosis, [24][25][26] which may detect other pre-existing tumors, observed and expected cases during the synchronous period (2 months) were shown only in the analysis by follow-up duration. All the other SIRs were, therefore, computed excluding observed cases and person-years in the first 2 months after the first tumor. Excess absolute risk (EAR) was computed (with 95% CIs) subtracting the expected number of subsequent cancers in the general population from observed number; the difference was then divided by the PY and the number of cancer cases in excess (or deficit) was expressed per 1,000 PY. All analyses were conducted using the "MP-SIR" session of SEER*Stat 8.3.6. 27

| RESULTS
The study populations included 1,406,694 patients who had been diagnosed with cancer below the age of 85 years, between 1998 and 2012 (38,535 patients with TC and 1,368,159 with any other primary cancer) (Appendix 1), followed for a maximum of 15 years (median follow-up <7 years). After TC diagnosis, the SIR of other cancers diagnosed in <2 months-period (i.e., synchronous) was high (1.8, 95% CI: 1.4-2.3 in women and 2.7, 95% CI: 2.0-3.6 in men), even if they represented only 4% of all the tumors (Figure 1). At 12-59 months after TC diagnosis, the SIRs in both sexes were between 1.1 and 1.3, and they remained 1.2 even 5 years after TC. The SIR for TC <2 months after other neoplasm (8% of all such cases) was 2.6 in women and 6.6 in men, and it gradually decreased up to 1.3 (95% CI: 1.2-1.4) in women and 1.5 (95% CI: 1.3-1.7) in men after 5 or more years since first cancer diagnosis ( Figure 1).
In Italy, the overall SIR of second metachronous (i.e., diagnosed after >2 months since TC) cancers after TC was 1.16 (95% CI: 1.12-1.21) ( TC as SPC was diagnosed more frequently than in the general population: overall SIR = 1.49 (95% CI: 1.42-1.55), 1.42 in women and 1.67 in men (Table 3). SIR ranged between 1.35 for follicular TC and 1.61 for medullary TC. SIR of TC after female breast cancer was 1.3 (1.2-1.4) and increased SIRs were also found after acute lymphoid leukemia (SIR = 6.1), bone cancers (4.3), kidney cancers (3.2), Hodgkin lymphomas (2.8), F I G U R E 1 Risk a of second primary cancer after thyroid cancer and risk of thyroid cancer as second tumor by time since the first tumor. Italy, 1998-2012. Obs, observed cases. a Measured as standardized incidence ratio (SIR) and 95% confidence intervals (CIs). b Excluded from subsequent analyses (4% of cancer subsequent to TC and 8% of TC after another malignancy)

| DISCUSSION
In Italy, TC patients have a 16% increased risk of experiencing a SPC, in comparison with the general population, slightly more than the corresponding figure in the United States (10%) 14,21 and South Korea (6%), 28 but less than in Denmark (31%), 10 Japan (44%), 11 and Switzerland (58% in men and 36% in women). 13 SIR in TC patients diagnosed in 1998-2012 corresponded to an excess incidence rate above 1 per 1,000 patients per year (i.e., 100/100,000). In addition, SIR of TC diagnosis after another cancer was 1.49, consistent with estimates in the United States (+50%). 12 Other authors have reported higher risks of papillary TC than of follicular TC after several neoplasms, in particular renal or breast cancers and leukemias/lymphomas. 19 However, in the present study, risks of SPC after papillary TC and of papillary TC as SPC are only slightly higher than corresponding risks for all TC (i.e., all other TC). Most notably, although rare, our data suggested that follicular or medullary TC were not associated with an excess of all SPC, while, as SPC, these two TC subtypes had similar SIRs than papillary or other TC. On the other hand, numbers for such types are quite small and possible associations, for example, as those in multiple endocrine neoplasia for medullary TC, may have not been evidenced.
Several studies reported an elevated risk of SPCs from the use of radioactive iodine (RAI) therapy in TC patients, 29,30 in particular in young patients. 31 This may explain the elevated incidence of acute lymphoid leukemia and other hemolymphopoietic neoplasms after TC treated with RAI. 30 However, increased risks after TC have been reported for different lymphoid and hematological neoplasms, 32,33 as well as for solid tumors. 30 In three large cohorts of TC cases diagnosed until 1995, 34 an increased risk of several solid tumors and leukemia's was found with increasing cumulative activity of administered iodine-131 ( 131 I). However, when contrasting those exposed and not exposed to 131 I, significantly increased risks of SPC were only seen for bone and soft tissue cancers. 34 The association between soft tissue sarcomas and TC (as a first or second neoplasm) that emerged in our study is consistent with recent results reported in the USA. 12,14 Despite the methodological flaw comparing SIRs after TC and for TC as SPC, the bidirectional association between TC and hemolymphopoietic, kidney, and bone/soft tissue cancers is remarkable and consistent across sex and age groups. The highest SIRs were observed for bone cancer and acute lymphoid leukemia after TC, although with broad CI. Moreover, there were moderate bidirectional associations between thyroid and prostate (1. 4  T A B L E 1 (Continued) in our study SIR of breast cancer after TC was 1.2, did not change with time since TC diagnosis, and was similar to what has been reported in Europe 10 and USA, 21 but lower than in Japan (2.0) 11 and Korea (2.5). 35 The consistent associations found between TC and other tumors may suggest the contribution of common risk factors, such as obesity, including genetic predisposition. 36,37 Nevertheless, in the present study the association of TC with the two most common overweight-related neoplasms (i.e., colorectal and endometrial cancers) was limited, if any, and in line with previous results from comparable population-based studies. 10,12,14 For some cancer types (e.g., breast or kidney cancer), the link with TC consistently persisted beyond 5 years since diagnosis 14,19,27 and can be driven, at least in part, by genetic predisposition and treatment exposures. 16,36 To disentangle potential biological associations between first and subsequent tumors seems beyond the capability of CR data, since they are probably mediated by complex interactions between genetic predispositions, lifestyle, and intensity of surveillance. Even if a role of previous radiotherapy on thyroid gland (a radiosensitive organ) was a possible explanation of a small proportion of excess risk for TC as SPC, the finding of a 50% excess risk for TC as SPC may largely be explained by increased medical surveillance in oncologic patients and be accompanied by overdiagnosis. In fact, in a context of overdiagnosis estimated to be at least 70% of TC in Italy 3 as elsewhere, 9 most of the excess risk of TC as SPC (but possibly also of SPC after TC) may be due to the intensity of diagnostic activities in cancer patients. 7,38 In our study, the increased risk of TC after upper aerodigestive tract cancers (SIR = 2.1, Table 3), but not the opposite (SIR = 0.8, Table 1), supports the relationship between TC and diagnostic procedures of the neck. In Korea, the country with the highest TC incidence worldwide, SIRs for kidney cancer after TC increased with calendar period and paralleled TC increase in the general population. 28 For kidney cancer, the impact of enhanced screening (i.e., through abdominal ultrasound and imaging) has also been reported both in Korea 39 and elsewhere. 40,41

| Strengths and weaknesses
The major strengths of this study are the population-based design, the availability of the largest well-documented TC series (>38,500 cases) in Europe, and the corresponding cohort of patients with other cancer types (1.4 million cases). Few other studies, 10,11,14,32 could similarly explore the bidirectional association between TC and other cancers. To the best of our knowledge, only another study has compared the association of TC with other cancers using population-based data with specific attention to tumor histology and latency. 19 Our study is the first one capable of computing SIR separately for different TC histological types (i.e., follicular, medullary, and poorly differentiated) and more than 30 cancer types as a first or SPC.
Among the limitations, it should be mentioned that although Italian CRs were considered complete and accurate and cover one third of the population, 42 they do not cover the whole country. They may miss part of the substantial TC incidence variability among Italian regions. 3 Moreover, international comparisons of SIRs of second cancers suffer from the lack of a standardized cut-off between metachronous (e.g., subsequent) and synchronous tumors. In the present study, we excluded cancers occurring in the 2 months after diagnosis (4% of those diagnosed after TC and 8% of TC as SPC). Their inclusion would have had a negligible impact for the overall SIR estimates. The same definition was used by some authors, 21,28,43 but other cut-offs were also used (such as 6 months, 13,26 12 months, 12,17 or 5 years 14 ), suggesting the need of caution in comparisons between studies. Unfortunately, Italian cancer registries do not systematically collect information on stage at diagnosis or treatment (e.g., RAI), and we could not take these important factors into account. 12,34,44 The length of follow-up (<15 years, median <7 years) precluded the assessment of longer term incidence of SPC. Finally, as a result of the large number of comparisons made, some SIRs may be spuriously statistically significant, calling for caution in the interpretation of results.

CONSEQUENCES AND CONCLUSIONS
With the growing number of people living after a cancer diagnosis, the number of those developing a SPC will also increase, generating further medical and financial burdens for patients, families, and society. 45 Our findings may have potential implications for screening other neoplasms in patients with certain types of malignancies. In particular, TC patients need a comprehensive support, including surveillance for treatment side effects. However, overdiagnosis and overtreatment of TC should be avoided, particularly among younger women. 46,47