Risk of thyroid cancer in first-degree relatives of patients with non-medullary thyroid cancer by histology type and age at diagnosis: a joint study from five Nordic countries
- Mahdi Fallah1,
- Eero Pukkala2,3,
- Laufey Tryggvadottir4,5,
- Jörgen H Olsen6,
- Steinar Tretli7,
- Kristina Sundquist8,
- Kari Hemminki1,8
- 1Division of Molecular Genetic Epidemiology, German Cancer Research Center, Heidelberg, Germany
- 2Finnish Cancer Registry, Institute for Statistical and Epidemiological Cancer Research, Helsinki, Finland
- 3School of Health Sciences, University of Tampere, Tampere, Finland
- 4Icelandic Cancer Registry, Reykjavik, Iceland
- 5Faculty of Medicine, University of Iceland, Reykjavik, Iceland
- 6Institute of Cancer Epidemiology, Copenhagen, Denmark
- 7Norwegian Cancer Registry, Oslo, Norway
- 8Center for Primary Health Care Research, Lund University, Malmö, Sweden
- Correspondence to Dr Mahdi Fallah, Division of Molecular Genetic Epidemiology, German Cancer Research Center, Im Neuenheimer Feld 580, Heidelberg 69120, Germany;
- Received 9 November 2012
- Revised 28 February 2013
- Accepted 13 March 2013
- Published Online First 13 April 2013
Background We aimed to estimate lifetime cumulative risk of thyroid cancer (CRTC) in first-degree relatives of patients with non-medullary thyroid cancers (NMTC), including papillary (PTC)/follicular/oxyphilic/anaplastic thyroid carcinoma, by histology and age at diagnosis in patients and their relatives.
Design A population-based cohort of 63 495 first-degree relatives of 11 206 NMTC patients diagnosed in 1955–2009 in Nordic countries was followed for cancer incidence. Standardised incidence ratios (SIRs) were calculated using histology-specific, age-specific, sex-specific, period-specific and country-specific incidence rates as reference.
Results The 0–84-year CRTC in female relatives of a patient with PTC was 2%, representing a threefold increase over the general population risk (SIR=2.9, 95% CI 2.4 to 3.4; Men: CRTC=1%, SIR=2.5, 95% CI 1.9 to 3.3). When there were ≥2 PTC patients diagnosed at age <60 years in a family, CRTC for female relatives was 10% (male 24%). Twins had a 23-fold increased risk of concordant PTC. Family history of follicular/oxyphilic/anaplastic carcinoma increased CRTC in relatives to about 1–2%. Although no familial case of concordant oxyphilic/anaplastic carcinoma was found, familial risks of discordant histology types of NMTC were interchangeably high for most of the types, for example, higher risk of PTC when a first-degree relative had follicular (SIR=3.0, 95%CI 1.7 to 4.9) or anaplastic (SIR=3.6, 95% CI 1.2 to 8.4) carcinoma. The earlier a patient was diagnosed with PTC in a family, the higher was the SIR in his/her younger relatives. There was a tendency towards concordant age at diagnosis of thyroid cancer among relatives of PTC patients.
Conclusions This study provides clinically relevant risk estimates for family members of NMTC patients.
Incidence of thyroid cancer, the most common endocrine malignancy, is increasing with over 213 000 new thyroid cancer cases diagnosed in the world per year (48 020 new cases in the USA; 6.4% annual percentage increase during 1997–2008).1–3 The most common thyroid cancer is papillary thyroid carcinoma (PTC; 50–80%), followed by follicular (FTC; 10–40%) and medullary thyroid carcinomas (MTC; 3–10%).4–,6 Familial risk of thyroid cancer is known to be highest among all cancer sites, for which the increased risk extends beyond the nuclear family.7 ,8 A strong genetic component is already known for MTC, so it is not discussed further in this study.9–11 Although non-medullary thyroid cancers (NMTC) are mostly sporadic, the familial form unrelated to Mendelian cancer syndromes (eg, familial adenomatous polyposis and Cowden's syndrome), is well-documented and thought to cause more aggressive disease.2 ,12 At least five per cent of follicular cell-derived well-differentiated thyroid carcinomas (PTC and FTC) have family history of thyroid cancer.13 Oxyphilic (Hurthle cell) carcinoma is a less common variant of follicular cell-derived carcinoma of the thyroid that may present as a low-grade tumour or as a more aggressive type.14 Undifferentiated or anaplastic carcinoma, which occurs more commonly in the elderly, makes up one to five per cent of all thyroid malignancies, but it is one of the most aggressive solid tumours in humans and accounts for most of the cancer deaths from thyroid cancer.15–18 The histological patterns of anaplastic carcinoma include giant cell, spindle cell and squamoid cell tumours, which frequently coexist and are not predictive of patients’ outcome.15
Family history is a readily available risk factor for which advice and management may bring medical and psychosocial benefits. However, in order to provide evidence-based advice, the counsellors and the caregivers along with the entire medical referral system need to be aware of the true familial risks, particularly for the cancers such as NMTC that are not covered by the current familial risk management guidelines.
A limited number of population-based epidemiological studies have been able to quantify the familial relative risk of histology types of MTC and NMTC, but none had a large sample size to analyse rare histology types, familial associations of histology types with each other or the familial risk stratified by age at diagnosis.10 ,11 ,19 One NMTC study reported only the risk of PTC when probands had any thyroid cancer (combined histology types, not PTC type alone) and the other one reported the familial risk of adenocarcinomas together with anaplastic carcinoma.11 ,19 A Norwegian study estimated the familial risk of thyroid cancer (in general, not concordant histology) among first-degree relatives with different histology types of thyroid.20 None of the case-control studies focusing on familial cancers have been able to address the histology-specific familial risk estimates.21–28 Moreover, all previous studies only provide relative risk in terms of standardised incidence ratio (SIR) which needs to be translated to a tangible risk estimate for clinicians. Therefore, a merged dataset was created by pooling the nationwide family-cancer data from five Nordic countries to be able to systematically and comprehensively quantify the familial risks of all concordant histology types of thyroid cancer and to elucidate the familial associations between different (discordant) histology types. Our main goal was to present the familial risks also in terms of cumulative risk, which is more tangible for clinicians, patients and their relatives.
Materials and methods
Our large dataset consisted of pooled family-cancer data from five Nordic countries including thyroid cancer patients from Denmark, Finland, Iceland, Norway and Sweden with their unbiased genealogy records and high quality cancer data. Information on all NMTC cases in the pooled dataset (n=11 867) and their relatives (n=63 495) was used for this study. The Nordic countries have population-based registers, through which any thyroid cancer patient can be identified with the cancer status (and histology type) in their parents, siblings or children. With the exception of Iceland, sibships could be ascertained only in the offspring generation (those with identified parents). According to the ‘Survey of Nordic Cancer Registries 2000’, the cancer registries in the Nordic countries, Sweden, Norway, Finland, Iceland and Denmark are operating in populations that are rather alike.29 They are all nationwide benefitting from a unique personal identification system. Furthermore, a very close contact and co-operation between the registry departments in these countries has ensured many similarities. The registries have thus often been used in comparative and multicentre studies including all or some of the Nordic countries. The data characteristics of each country are shown in table 1 and some additional information is presented in the online supplementary information.
SIRs were used to compare the cancer risks for individuals with identified parents and family history of cancers in their relatives compared to the risk in their peers in the general population. Follow-up was started for parents of the NMTC cases at the birth of the NMTC child, for the others at birth, immigration or 1st of January of the country-specific registration start year (table 1), whichever came latest. Follow-up was terminated on the diagnosis year of the first primary cancer, death, emigration or the closing date of the study, (31 December 2008/2009; table 1). The SIRs were calculated as the ratio of observed to expected number of cases (indirect method of standardisation). The sex-specific, age-specific (0–4, …, 80–84, 85+ years), period-specific (5-year bands), cancer site-specific, morphology-specific and country-specific background population incidence rates provided by the cancer registries was used as the reference groups (strata-specific cancer incidence rate in the background population). The expected numbers were calculated from the strata-specific cancer incidence rate in the background population multiplied by the corresponding person-years for proband-positive subjects in the study population. p Values and CIs (95% CI) were calculated assuming a Poisson distribution. SAS software V.9.2 (by SAS Institute Inc, Cary, North Carolina, USA) was used for the data analysis. Use of the anonymised family-cancer data and study protocol was approved by the Norwegian Regional Medical Ethical Committee (South-East) and Regional Ethics Committee at the Lund University.
The lifetime cumulative risk (assumed to be 0–84 years) was calculated based on the following formulas: age-specific annual incidence rate=number of cases for each 5-year age group divided by person-years of that age group (0–4, …, 80–84 years); age-specific cumulative rate = 5×age group-specific annual incidence rate; lifelong cumulative rate = summation of all age-specific cumulative rates; and lifelong cumulative risk=1−exp (– lifelong cumulative rate).
Due to the number of subgroup testing, an adjustment for the multiple testing could be used to estimate significance levels in this large study. However, this correction would be too conservative because the tests are correlated. Instead, we provide the p values in addition to the 95% CI to show that after any type of correction for multiple testing, many of the associations would still remain significant. This study benefited from a large sample size providing enough power to detect SIR≥1.5 for the relative risk of NMTC patients. For instance, in relatives of PTC patients, the power to detect SIR≥1.5 for endocrine tumours in general was 99%, thyroid cancer in general 96%, NMTC 94%, PTC 85%, and for many non-endocrine cancers such as stomach, colorectal, pancreas, melanoma and non-melanoma skin, breast, prostate, kidney, urinary bladder, brain and non-Hodgkin's lymphoma, the power was about 100%. However, for other rare histology types the power was lower.
The lifetime cumulative risk of thyroid cancer (CRTC) in female relatives of a patient with PTC was 2.1% (table 2), which represents about a threefold increase over the general population risk (Women: SIR=2.9, 95% CI 2.4 to 3.4; Men: CRTC=0.8%, SIR=2.5, 95% CI 1.9 to 3.3; table 3). When there were ≥two PTC patients diagnosed at age 30–59 years in a family, CRTC for relatives was 10.1% in women and 24.1% in men (table 2). Twins had about a 23.1-fold increased risk of concordant PTC (table 3) and they were diagnosed at the same age group.
When there was a follicular carcinoma patient in a family, CRTC in relatives was 1.8% in women and 1.0% in men (table 2), which represents about a threefold increase over the general population risk (table 3). When there was an early onset follicular carcinoma patient diagnosed before age 30 years in a family, CRTC in female relatives was increased to 2.7% (table 2; 6.3–27.4 times higher depending on sex of the patient and age of the relative, table 4).
Having an oxyphilic carcinoma patient diagnosed at age 30–59 years in a family increased CRTC in female relatives to 1.5% (table 2). Having an anaplastic carcinoma patient in a family increased CRTC in relatives to 1.5–3.9% depending on age at diagnosis and sex (table 2). No familial case of concordant oxyphilic or anaplastic carcinoma was found in our data, but familial risks of discordant histology types of NMTC were interchangeably high for most of the types.
Discordant histology types
Even familial risks of discordant histology types of NMTC were increased (table 3). For instance, when a relative had PTC, risk of oxyphilic carcinoma in female relatives was 6.1-fold. When a relative had FTC, risk of PTC was threefold. When a relative had anaplastic carcinoma, risk of PTC increased 4.1-fold in women.
Discordant cancer sites
Family history of PTC increased risk of non-thyroid endocrine cancers (such as parathyroid cancer) and non-melanoma skin cancer up to about twofold in both sexes and kidney (1.4-fold) and prostate (1.2-fold) cancer in men (table 3). Having a patient with oxyphilic thyroid carcinoma in a family was associated with a 10-fold higher risk of Hodgkin's lymphoma in female relatives of the patient (only two cases) and a 4.2-fold increased risk of non-Hodgkin's lymphoma in male relatives. When a patient had anaplastic carcinoma, risk of pharyngeal cancer in their male relatives increased to about 5.6 times (only three cases).
Risk trend by age
The incidence of thyroid cancer varied by age, histology type and family history of thyroid cancer. Incidence of PTC in the background population, which is mostly composed of non-familial cases, increased from childhood to adulthood with a rather steady state from age 40 years (4.0/100 000 person-years) to 80 years (4.6; figure 2A). Familial PTC had a higher incidence (figure 2B), especially when two or more first-degree relatives had PTC, and its first incidence peak was at age 20–24 years (280.9 when ≥two relatives had PTC; 9.7 when only one relative had PTC) and it stayed high in the entire adulthood with highest incidence at age 45–49 years for those who had multiple PTC in their family (591.7). Incidence of FTC in the general population increased from childhood (0.2) to elderly ages (2.5), while familial FTC had its first surge at age 25–29 years (20.4) and the last and highest peak was at age 75–79 years (23.6). Among all histology types in the general population, anaplastic carcinoma had latest age at onset with most cases occurring from age 60 years (1.3) to 84 years (2.5). In general, oxyphilic carcinoma had the lowest incidence among these types with a gradual increase from age 15 years (0.1) to 84 years (0.6).
The CRTC was somehow associated with age at diagnosis of the patient with NMTC in the family and age of family members (table 2). In general, CRTC was increasing by age of the person (table 2), but the younger the patient was diagnosed with PTC in a family, the higher was the relative risk (SIR) in younger relatives, so that the SIR for thyroid cancer in same-sex relatives of early onset PTC patients (diagnosed before age 30 years) had a decreasing trend from young (age 0–29 years: SIR=13.4) to older (age 30–59 years: SIR=3) ages in women (table 4). When there were two or more later-onset PTC patients (diagnosed at age 30–59 years) in the family, risk of thyroid cancer at a similar age group in their male relatives was markedly higher (405.1-fold). Risk of concurrent PTC at age 30–59 years among female twins was 37.4-fold (table 4). Age-specific results of both sexes combined are presented as online supplementary table S1, which is an extension to table 4.
Risk by sex
In the general population, lifetime CRTC was higher in women than that in men (0.6% vs 0.3% with no overlapping 95% CI; table 2). The same applied to CRTC when a family member had PTC (2.1% vs 0.8%). However, in terms of relative risk (SIR), male relatives had a higher familial risk when the NMTC patients in their family were also of the same sex, for example, SIR of thyroid cancer for like-sex versus unlike PTC in a family is 3.2 versus 2.8 in women, while it is 5.2 versus 2.1 in men although their 95% CIs overlapped (table 4).
Combining nationwide family-cancer datasets of five Nordic countries with homogenous cancer registries,30 enabled us to provide clinically useful information on familial risk of histology types of thyroid cancer, familial associations between different histology types, and trend of the familial risk by age at diagnosis of NMTC patients and age and sex of their relatives. According to GLOBOCAN 2008, incidence of thyroid cancer slightly varies between Nordic countries (world age standardised rate from minimum 3/100 000 person-years in Sweden and Denmark to maximum 5 in Finland), but in general their incidence is equal to the world's average (northern Europe=3.1; world=3.1).1 Our data materials are highly valid, alike and comparable without too many artefacts or confounders.29 A Y-chromosomal and mitochondrial DNA analysis to find out the population structure in contemporary Sweden showed that the haplogroup frequencies of the counties closest to Finland, Norway, Denmark and the Saami region in the north exhibited large similarities to the neighbouring populations, resulting from the formation of the Swedish nation during the past millennium.31 A large-scale study on correlation between the genetic and geographical structures in Europe confirmed a common genetic ancestry in northern and western Europe with Finland as an exception in the principal component analysis; however, a map of the genetic structure of Europeans by Nelis et al showed some parts of the Finnish population have a Swedish origin reflecting the 5–10% Swedish speaking part of the Finnish population residing mostly in the western and southern parts of Finland.32 ,33 In addition, Finns comprise the largest group of immigrants to Sweden, from where nearly 40% of our data come. According to our ad hoc analysis plan, in this paper, only results of the pooled dataset are presented as regional differences of sporadic and familial risks of NMTC in the Nordic countries are subject to random variation due to small case numbers.
In our study, NMTC in general showed about threefold familial risk when a parent, child or singleton sibling had a history of PTC (table 3). Depending on age and sex, the risk in those with multiple affected relatives (6–400-fold) or twins (about 20–40-fold) was much higher (table 4). Further histology-specific analysis in our study for the first time showed about fivefold familial risk for FTC when a first-degree relative had a history of concordant histology type. We also found that having a close relative with FTC increased the risk for less common concordant histology types, and most common type, PTC, up to threefold.
Although there were some exceptions probably due to the small sample size, in general, there was a tendency towards concordant age at diagnosis of thyroid cancer among relatives of PTC cases, in line with our previous finding for some more common cancers such as colorectal and prostate cancers.34 This could be due to (1) concordant exposure to environmental risk factors for instance among siblings, (2) genetical predetermination of age of cancer in a family, (3) the fact that a certain histology of cancer occurs mostly in particular age groups and our analysis was age-specific and histology-specific or (4) a combination of all these potential reasons.
The most important findings of this study are summarised in table 2, where sex-specific and age-specific cumulative risks are presented by the number of patients with a particular histology in a family and their age at diagnosis as well as the population risk. One should note that the cumulative risk of NMTC ignores the competing cause of death, so this is the pure risk estimate in the absence of death due to other causes than NMTC in this study. The risk in familial cases could be directly compared with the risk in the general population in table 2, however, tables 3 and 4 show relative risks (SIRs) with more adjustments for other potential confounding factors, such as country and calendar period of diagnosis. In addition, we presented a number of cases and 95% CIs for lifetime cumulative risks in table 2 that should be taken into account when one wants to use this table in clinical practice. For instance, the 24% risk of thyroid cancer in male family members of multiple patients with papillary carcinoma diagnosed at age 30–59 years is based on only seven cases, leading to a rather wide 95% CI, ranging from 10% to 36%. Altogether it represents a very high-risk group in which more intensive primary and secondary preventive measures could be recommended to family members, especially when they are aged 30–60 years. Results of a recent study favoured ultrasonography screening of first-degree relatives of even patients with apparently sporadic multicentric PTCs, especially among women.35
The genetic and biological bases for the familial predisposition of NMTC (germline mutations) are only recently beginning to emerge.36 ,37 An Icelandic study revealed that two common variants, located on 9q22.33 near FOXE1 (TTF2) gene (rs965513) and 14q13.3 near NKX2–1 (TTF1) gene (rs944289), contribute to an increased risk of papillary and follicular thyroid cancers.38 These risk alleles were associated with decreased concentrations of thyroid stimulating hormone (TSH), and the 9q22.33 allele was associated with a decreased concentration of thyroxine (T4) and an increased concentration of triiodothyronine (T3). They estimated that the risk of thyroid cancer in individuals who are homozygous for both variants is 5.7-fold greater than that of non-carriers. They also found three other common variants (rs966423 on 2q35, rs2439302 on 8p12 and rs116909374 on 14q13.3) associated with low TSH levels and NMTC risk.39 They speculated that the consequence of the low concentration of TSH may be less differentiation of the thyroid epithelium, leading to a predisposition to malignant transformation; however, not all sequence variants related to low TSH levels associate with the risk of the disease. In a study investigating a susceptibility locus for PTC on chromosome 8q24, gene expression analysis indicated that AK023948 was significantly downregulated in most PTC tumours.40
Although the present study was the largest of its kind and there were more than 300 cases of very rare histology types of thyroid cancer, no familial pair with concordant histology was found for histology types such as oxyphilic or anaplastic. A germline mutation of GRIM-19 was found in a patient with familial oxyphilic carcinoma.41–43 However, lack of direct link for concordant rare histology types could be due to rareness of their familial form and/or non-specific report of histology of NMTC at the beginning of cancer registries (eg, 1950s–1980s). Instead, we found their associations to other discordant histology types of NMTC. For instance, risk of the most common type of thyroid cancer, PTC, was about fourfold increased when a female relative had anaplastic carcinoma. Also a non-significant tendency was found for a higher risk (10 times) of FTC in men when a relative had anaplastic carcinoma. There was also a non-significant tendency for a twofold increased risk of NMTC/PTC when a female relative had oxyphilic carcinoma. The high risk of discordant histology types of NMTC might be due to common genetic and/or environmental risk factors. Checking mutations found in one histology type of thyroid cancer in patients with different but associated histology types of this cancer (significant associations in table 3) may reveal a common pathway leading to the interchangeably increased familial risk between different histology types. Altogether, these findings may suggest an aggregation of related susceptibility genes responsible for a common causal pathway rather than different pathways. The best hypothesis to explain these findings is that the genetic component in liability for NMTC is polygenic but affecting the same pathway.
The present study, which is the largest yet published, benefited from the nationwide data from five countries with unbiased family history. However, for Denmark, Finland and to some extent Norway, accounting for about half of index cases, the second generation was no older than in their mid-50s, thus resulting in few familial cases in old-age groups. Moreover, the histology type for older periods of time (less specific codes) in such a long follow-up study may not be as accurate as recent years, which in turn can be the source of a potential bias toward underestimation of SIRs for concordant histology types. To rectify this issue (to some extent) we have additionally provided association of more specific histology codes with less specific (umbrella) histology codes (eg, reporting risk of thyroid cancer in relatives of patients with a specific histology cancer) and adjustment for period of diagnosis has been done to take into account the change of incidence over time as well. Of course, role of the surveillance bias (more intensive diagnostic approach for family members of an affected case that may lead to the overdiagnosis of indolent cancers) could not be ruled out for the weak associations.
For the familial association of concordant common histology types of NMTC and for association of NMTC with common discordant cancer sites, we had high study power. However, for other rare histology types the power was lower, but as the results of further stratification by sex, age and histology were in line with the familial risk of NMTC in general, we believe that results of stratified analyses for rare histology types are valid and important for the clinicians, NMTC patients and their family members.
Families, usually, live in the same area therefore they share the same risk factors. The most well-established risk factor for thyroid carcinoma is ionising radiation exposure and for anaplastic carcinoma is the thyroid goiter. Although lack of the spouse correlation for thyroid cancer with shared adulthood44 and existence of parent-child familial association with no time-wise shared childhood environment do not support a strong environmental role, such a potential role could not be completely ruled out in this familial study. In the context of interaction between genetic and environmental factors, the distinction between these two components would be even more difficult. However, the strength of this study is that estimated familial risks could be used in the clinic regardless of the exact underlying reason for them.
In conclusion, results of this five-country joint study provide clinically relevant risk estimates for family members of NMTC patients by histology and age at diagnosis of patients and relatives. In this study using an unbiased population-based family-cancer data, we were able to quantify the high familial risk for thyroid cancer in relatives of patients with PTC (3–400 times) and FTC (about 5 times; reported for the first time). We also found increased risks for different histology types of NMTC, which may show a common oncogenic pathway or environmental risk factor for various types of NMTCs. The familial component of NMTC seems to be polygenic as we found different familial risks for different histology subtypes of NMTC. Significantly high SIRs for discordant histology types between each other may propose that potential genes or environmental factors involved in the pathogenesis of different histology types of NMTC may affect the same pathway. Our finding may suggest that the role of genetic or environmental factors in the differentiation of histology types of NMTC is not as strong as its role in the development of NMTC as such. Although relative risk of the familial NMTC was higher in young adulthood than in elderly ages, familial NMTC was not limited to young ages. In general, there was a tendency towards concordant age at diagnosis of thyroid cancer among relatives of PTC patients.
Contributors Conception and design: KH, MF. Provision of study materials: KS, EP, LT, JHO, ST. Analysis plan: MF, EP. Assembly of data and analysis: MF. Interpretation of results and manuscript writing: MF, KH. Commenting on manuscript and final approval of manuscript: all authors.
Funding This work was supported by the Nordic Cancer Union, Swedish Council for Working Life and Social Research, and German Cancer Aid. No funding agency or sponsor had any role in any part of the report.
Competing interests None.
Ethics approval The Norwegian Regional Medical Ethical Committee (South-East) and Regional Ethics Committee at the Lund University.
Provenance and peer review Not commissioned; internally peer reviewed.