Unbalanced bidirectional causal association between thyroid cancer and ER-positive breast cancer: should we recommend screening for thyroid cancer in breast cancer patients?

In recent years, with the advancement of various malignant tumor diagnoses and treatment methods, the survival rate of patients has increased significantly, and the incidence of secondary tumors caused by this has also increased. In the United States, secondary malignancies account for approximately 18% of all tumors [1]. In addition to common risk factors and sequelae of radiotherapy or chemotherapy, this result may also be caused by genetic susceptibility. Compared with the general population, the probability of another tumor developing into a secondary malignant tumor is significantly higher among patients with breast cancer (BC) and thyroid cancer (TC) after surgery or tumor carriers [2–9].

BC was the third most common incident cancer overall in 2007, with an estimated 2.0 million cases. The majority occurred in women and it caused 601,000 deaths in women and 11,000 deaths in men, making it the fifth leading cause of cancer deaths for both sexes combined in 2017 globally [10].

The incidence of TC continues to rise worldwide, and it has become the fifth most common malignant tumor for women in the United States. It is estimated that in 2015, there were more than 62,000 new cases in men and women [11]. It is well known that TC is gender-biased, with up to three times more diagnoses being made in the female population than the male population.

Many previous clinical studies of epidemiological and genetic evidence have proven the potential correlation between BC and TC [2–4, 12–19]. However, metachronous (occurring in succession, or called the precedence) BC and TC have not been identified clearly. The reasons may be due to a single factor or a combination of factors, such as genetic factors, environmental factors, and treatment-related factors [20]. For the current published literature, there is no evidence of a clear causal relationship between BC and TC. This is caused by the research type (most of the studies are retrospective) and the lack of studies with long-term follow-up and a large sample size.

Causal inference in traditional observational epidemiological research is hindered by the possibility of confounding and preserving causality [21]. Mendelian randomization (MR) research is a data analysis method that has been used in recent years in epidemiological etiology inference. It can be used to uncover causal relationships between an exposure and outcome in the presence of such limitations. It is a form of instrumental variable analysis in which genetic variation is used as an interest exposed proxy [22]. In this study, we used MR to explore the causal relationship between TC and BC.

There have been many studies that have reported a possible link between BC and TC, but it is difficult to determine whether there is indeed a link between BC and TC, and in which direction. Nielsen et al. conducted a meta-analysis that confirmed and quantified the increased likelihood of the co-existence of BC and differentiated thyroid cancer [27]. Due to the inevitable heterogeneity in the meta-analysis study that originated from the included literature, and that the included literature was after the discovery of one kind of tumor, Nielsen et al. could not quantify the time between the first tumor and the second tumor. They found that the risk of two tumors becoming each other’s secondary tumors increased, indicating that control bias alone cannot fully explain this connection, and there may be potential pathophysiological risks. Therefore, in order to further clarify whether there is a potential causal relationship between these two cancers, a new analytical research method called Mendelian randomization opened up new ideas for our research.

The Mendelian law of inheritance stipulates those alleles obtained from parents are passed to offspring through random separation. Therefore, it is unlikely that the genotype of the offspring is related to the confounding factors in the population. In addition, when the fertilized egg is formed, the germline genotype of the offspring is already fixed, which precedes the observed variable in time, avoiding the problem of reverse causality. The MR method involves looking for genetic variants associated with exposure and then testing the association between these variants and the results. When the necessary conditions are met, the causal “disambiguation” relationship between exposure and outcome can be estimated [28].

In recent years, a variety of MR research methods have been summarized, all of which use genetic variation to infer the causal relationship between features of interest, such as two-sample MR, two-step MR, multivariable MR, and factorial MR [29, 30]. Bidirectional MR is one of those above in which the tools of exposure and outcome are used to assess whether the “exposure” variable causes the “outcome”, or whether the “outcome” variable causes the “exposure” [31]. First of all, “exposure” and “outcome” variables need to be defined. The MR analysis is first carried out from “exposure” to “outcome”, and then in the opposite direction (i.e., from “outcome” to “exposure”). MR analysis is performed by using SNPs that are robustly related to each trait in the individual GWAS. The principle is to assume that the causal relationship between two variables works through the potential mechanism, and MR analysis can determine the direction of the causal sequence that acts in the mechanism [30]. Therefore, MR promises to be a valuable strategy to examine causality in complex biological/omics networks for disease prevention in the future.

Similar to breast tissue, both benign and malignant thyroid tissue are highly responsive to circulating estrogen [32]. In fact, elevated circulating hormone levels are associated with TC [33, 34]. Hyperestrogenism (elevated endogenous estrogen) during reproductive years is associated with an increased prevalence of TC in women of reproductive age; however, HRT (Hormone replacement therapy) or other exogenous estrogen exposure are not linked to TC [34–36]. Additionally, estrogen may serve as a link between the co-occurrence of autoimmune thyroid disorders and BC, which predominantly affect women [37]. Immune tolerance and the development of autoimmunity is largely controlled by the AIRE gene [38], estrogen is a key regulator of this gene, and reduced activity from elevated estrogen contributes to autoimmune susceptibility [38–42]. Overall, sex hormones, primarily estrogen have a connection to not only BC, but also thyroid malignancy.

In our study, MR analysis based on a large sample size GWAS study revealed a potential causal relationship between BC and TC; that is, TC may be a protective factor for BC occurrence (especially in ER-positive BC), and ER-positive BC is a risk factor for TC occurrence. Since subgroup analysis based on whether ER is positive had produced different results, we believe that ER has a potential pathophysiological role in the occurrence, development, and secondary tumors of breast and TC. However, ER itself is one of the main molecular targets in BC pathogenesis and is expressed in approximately 70% of invasive BC. It is a steroid hormone receptor and a transcription factor that, when activated by estrogen, activates oncogenic growth pathways in BC cells. Therefore, using endocrine agents to downregulate ER signaling is the primary systemic therapy for ER-positive BCs [43]. On the aspect of TC, estrogen plays its growth-promoting role through classical genomic and non-genomic pathways mediated by membrane-bound ER [44]. It is also a potent growth factor both for benign and malignant thyroid cells, which may explain the sex difference in the prevalence of TC [45]. Therefore, ER-positive may play an important role between BC and TC.

This study has the following advantages and limitations. First, we chose the largest GWAS database on BC and TC to ensure that our results are true and reliable. Secondly, we used MR for the first time to conclude that genetic prediction found that TC is a protective factor for BC, and BC is a risk factor for TC. This also provides a clear explanation for some previous epidemiological investigations and provides a basis for TC screening for BC patients. Thirdly, this study is the first to perform subgroup analysis by estrogen receptor, detailing the relationship between BC and TC in the subgroup above mentioned. However, the GWAS data in this study are all derived from European populations. Therefore, it is not clear whether the results of this study are applicable to populations or races in other regions. In addition, the data of BC patients in this study uses only whether ER is positive for subgroup analysis. The lack of progesterone receptor-positive data (and triple-negative BC) makes it impossible to conduct a more accurate subgroup analysis of BC patients. We believe that if more sample sizes and pathological types of GWAS data can be included in the future, the results of MR will be more directional and provide a clear direction for basic research to explore the potential relationships.

This study used a bidirectional two-sample MR approach to explore the potential causal relationships between BC and TC. It was found that patients with TC are less likely to have ER-positive BC; on the contrary, people with ER-positive BC are more likely to have TC. It is recommended that patients with BC be screened regularly for TC.

None.

F.Z. and H.W. wrote the main manuscript text, and S.L. performed data analysis and prepared figures and tables. F.Z. provided the idea and funding. J.S. provided technical support. C.F. and Y.W. performed data extraction and processing. All authors reviewed the manuscript. The authors hereby confirm that neither the manuscript nor any part of it has been published or is being considered for publication elsewhere. We acknowledge that all authors participated sufficiently in the work and take public responsibility for its content.

This study was supported by the 345 Talent Project of Shengjing Hospital; National Natural Science Foundation of China (81600115); Liaoning Provincial Department of Science and Technology Applied Basic Research Program (joint project). The funder had no role in study design, data collection or analysis or in the presentation or publication of the results.

All data were from the Integrative Epidemiology Unit Open genome-wide association studies (GWAS) Project website (https://gwas.mrcieu.ac.uk/↗). The original contributions statement presented in the study are included in the article/Material. Further inquiries can be directed to the corresponding authors.