STATE-OF-THE-ART REVIEW IN ENDOCRINOLOGY
Twin study method: Unlocking genetic and environmental interactions
2025 Volume 72 Issue 10 Pages 1061-1068
Details
2025 Volume 72 Issue 10 Pages 1061-1068
Twin studies offer a powerful approach for disentangling genetic and environmental influences on human traits. By comparing monozygotic twins with identical genetic makeup, researchers can more accurately identify the environmental contributions to phenotypic variation. Structural equation modeling provides a theoretical framework for estimating the relative contributions of additive genetic effects, shared environmental factors, and unique environmental influences. This model allows researchers to determine the most suitable parameters for explaining the observed data in twin populations. In addition, bioinformatic tools enable in-depth analyses of phenotypically discordant monozygotic twin pairs, helping to uncover both environmental sensitivities and genetic predispositions. This review examines the advantages and limitations of twin study methodologies in research on endocrine disorders, lipid metabolism, and thyroid function. Findings from twin cohorts have enhanced our understanding of heritability, environmental modifiers, and epigenetic factors, offering valuable insights into gene–environment interactions. Overall, twin studies remain critical tools in genetics, endocrinology, and obesity research. In the future, genetic information may enable the development of optimal personalized environments, ultimately providing valuable insights that contribute to physical, mental, and social well-being throughout the lifespan.
The twin study method is a robust research approach that enables precise analysis of the genetic and environmental factors influencing human phenotypes. Identical twins with perfectly matched genomic DNA sequences, provide an ideal model under the assumption that their genetic factors are entirely identical. Consequently, any phenotypic discordance observed in identical twins can be attributed to environmental factors (Fig. 1). Although genetic and environmental factors can also be analyzed in the general population, such studies are inherently less rigorous because of the heterogeneity of both factors within broader groups. However, general population analysis has its merits, especially for examining rare phenotypes or conducting large-scale studies, which can be challenging in twin studies owing to their limited sample sizes.
Genetic factors refer to the DNA base sequence, which remains constant and unmodifiable throughout an individual’s life. In contrast, environmental factors encompass all influences that can theoretically be altered, regardless of practical feasibility. These factors include living conditions, education, economic circumstances, and epigenetic modifications (Fig. 1). The epigenome is broadly defined to include mechanisms such as DNA methylation, histone modifications, microRNAs, glycosylation, and other factors that influence processes ranging from gene expression to protein function. A comprehensive analysis should consider the wide range of environmental influences on phenotypic outcomes.
Two primary methods are used in twin studies: structural equation modeling analysis and statistical tests commonly employed in bioinformatics. The former estimates the theoretical contribution rates of genetic and environmental factors to a given phenotype, whereas the latter identifies specific genetic and environmental factor candidates.
Structural equation modeling analysis, a form of multivariate analysis, is an analytical approach used to test hypotheses regarding the causal relationships for a given event. In twin studies, this method is employed to theoretically estimate the contributions of genetic and environmental factors to observed variables in monozygotic and dizygotic twin pairs [1]. The analysis used the observed data to construct influence models of the genetic and environmental factors that best fit the distribution of the observed variables. The basic model for analyzing a single variable assumes that three primary factors affect the observed phenotypes in each twin pair: A, representing additive genetic factors; C, representing shared (common) environmental factors; and E, representing unique environmental factors specific to each individual. Although dominant genetic factors can also be included, it is standard to begin with these three factors. In this model:
• A is assumed to be identical in monozygotic twins and 50% identical in dizygotic twins. Therefore, the correlation coefficient for A is 1.0 for monozygotic twins and 0.5 for dizygotic twins.
• C represents the environmental factors shared by both twins and is, by definition, perfectly correlated (correlation coefficient = 1.0) in both monozygotic and dizygotic twins.
• E represents environmental factors unique to each twin and is uncorrelated between them.
As illustrated in Fig. 2, a path diagram typically represents these relationships. Using this model, parameters a, c, and e—representing the theoretical contribution rates of genetic, shared environmental, and unique environmental factors, respectively—are calculated based on the data. These parameters were estimated using optimization methods to best fit the observed data. Different models, including the ACE model (accounts for all three factors), AE model (excludes shared environmental factors), CE model (excludes additive genetic factors), and E model (assumes only unique environmental factors), were evaluated to determine the best fit using model selection criteria, such as the Akaike Information Criterion (AIC). Notably, the E parameter was consistently included in the models to account for measurement errors and other residual factors.
For monozygotic twins, environmental factors affecting phenotypes and genetic susceptibility to these environmental factors can be analyzed by comparing the environmental influences in phenotypically discordant twin pairs (Fig. 3A) or by examining the differences in genetic backgrounds between phenotypically concordant and discordant pairs (Fig. 3B). The methods used for such analyses often include linear regression models, similar to those employed in genome-wide association studies (GWAS), to compare genetic backgrounds. Conversely, when comparing environmental factors in phenotypically discordant pairs, statistical tests appropriate for the data type should be applied. For example, a paired t-test may be used to analyze quantitative environmental factors, whereas other methods may be employed depending on the purpose of the analysis. These approaches allow the calculation of effect sizes for specific genomic sequences (e.g., polygenic risk scores [2]) and environmental factors, including epigenomic factors, using epigenome-wide association studies (EWAS) [3], rather than estimating contribution rates based on theoretical models.
(A) Within-pair comparisons are used to assess the influence of environmental factors on phenotypic variation.
(B) Between-pair comparisons help identify differences in environmental sensitivity across various genetic backgrounds.
(C) Findings from one monozygotic twin cohort can be validated using another cohort of co-twins.
By focusing on monozygotic twins, researchers can distinguish between individuals whose phenotypes are discordant owing to heightened genetic sensitivity to environmental factors and those whose phenotypes are primarily determined by genetic factors (Fig. 4). This approach enables the identification of individual differences in the effect of environmental factors on traits, thereby providing valuable insights into the variability of environmental influences on phenotypes.
Studies involving monozygotic twins also allowed the validation of findings through reproducibility testing within populations with identical genetic backgrounds (Fig. 3C). Specifically, a population consisting of one twin from each monozygotic pair and a population consisting of co-twins from the same pair represent genetically identical human populations. If the results obtained in one population, such as those from GWAS findings, can be reproduced in another population, the reliability of the results is greatly enhanced. This methodology offers research opportunities unattainable in studies involving the general population.
Numerous case reports of endocrine disorders in twins have been published. However, this review focuses not on case reports but rather on studies that analyze twin populations. Below is a summary of the relevant research. These studies primarily estimate the heritability of diseases and examine cases in which monozygotic twins are discordant for a particular disease. Analyses comparing discordant monozygotic twins have proven effective in identifying various environmental factors that influence disease manifestations in individuals with shared genetic backgrounds. Once identified, these environmental factors provide important insights that contribute to the advancement of personalized preventive medicine through targeted interventions.
Twin studies have estimated the heritability of type 1 diabetes to be approximately 30–40% and that of type 2 diabetes to be approximately 60–70% [4-6]. However, GWAS have shown that diabetes-related genetic variants are largely consistent across different ethnicities and sexes, indicating that sex differences in heritability are relatively small or limited in scope [7]. A study investigating adipose tissue in monozygotic twin pairs discordant for type 2 diabetes reported differences in the expression of several genes, including IRS1 and VEGFA. Additionally, some CpG sites exhibiting differential DNA methylation in diabetic individuals from general population case-control studies have shown methylation differences in discordant twin pairs [8]. Similarly, another study examining adipose tissue in monozygotic twin pairs discordant for type 2 diabetes found decreased expression of specific microRNAs (miR-30 and let-7 families) in affected individuals, accompanied by increased DNA methylation at CpG sites associated with the promoters of these microRNAs [9]. A study involving an oral glucose tolerance test (OGTT) in monozygotic and dizygotic twins applied structural equation modeling, which revealed high heritability estimates for IGF-I (0.65), IGFBP-3 (0.71), and insulin secretion (0.56), whereas insulin sensitivity exhibited a lower heritability (0.14). Among 45 twin pairs discordant for type 2 diabetes, IGF-I and IGFBP-3 levels were comparable between co-twins [10]. Another study reported that the relationship between body mass index (BMI) and metabolite changes following OGTT differed between monozygotic and dizygotic twins. Among monozygotic twin pairs discordant for BMI and liver fat, the twins with higher BMI exhibited lower insulin sensitivity, highlighting pronounced within-pair differences in metabolomic profiles [11]. The relatively low heritability of insulin sensitivity suggests that this trait is influenced by a complex interplay of non-genetic environmental factors and may be modifiable to some extent through acquired or lifestyle-related interventions.
A structural equation modeling study of Swedish twins estimated the heritability of Hashimoto’s disease at 65% and Graves’ disease at 63%, with approximately 8% genetic overlap between the two conditions [12]. Additionally, the risk of comorbid Hashimoto’s disease and type 1 diabetes is influenced by both genetic and environmental factors [13]. The heritability of thyroid function parameters in the Brisbane twin cohort was estimated to be 70.8% for TSH, 67.5% for FT4, 59.7% for FT3, and 48.8% for TPOAb [14]. Similarly, a UK twin-cohort study reported heritability estimates of 65% for TSH, 39% for FT4, and 23% for FT3 [15]. Another study of Dutch twins examined the influence of genetic and fetal environmental factors on T4 levels in the early postnatal period [16]. Furthermore, NISAb and PenAb levels are influenced by environmental factors in individuals with specific genetic backgrounds [17]. The heritability of TPOAb levels was estimated to be 61% in males and 72% in females, whereas the heritability of TgAb levels was 39% in males and 75% in females [18]. It is well established that even the same gene can exhibit sex-specific transcriptional regulation, and that identical environmental conditions can result in differing gene expression profiles and endocrine responses between males and females [19]. Sex differences in heritability are likely attributable not only to variations in sex hormone environments but also to a range of biological mechanisms, including sex chromosome composition, epigenetic regulation, and interactions with the immune system.
A Danish twin cohort study found that among discordant twin pairs, those affected by hypothyroidism exhibited a relatively higher mortality rate, particularly among dizygotic twins [20]. Similarly, in twin pairs discordant for hyperthyroidism, the affected twins showed increased mortality rates, with a more pronounced effect observed among dizygotic twins [21]. A study investigating the relationship between birth weight and thyroid-specific autoantibodies in twins concluded that differences in birth weight had a minimal impact on the emergence of these autoantibodies [22].
In recent years, research has highlighted the association between epigenetic factors and thyroid function as part of environmental influences. Specifically, distinct DNA methylation differences have been observed in monozygotic twin pairs with discordant TgAb levels [23]. Additionally, genomic and epigenomic factors affecting FT4 levels have been systematically reported [24].
Studies on thyroid nodules have indicated that 67% of individual variability in thyroid nodules is attributable to genetic factors, whereas 33% is due to environmental influences [25].
An intriguing study of monozygotic twins discordant for hyperthyroidism demonstrated that affected individuals had an increased risk of receiving disability pensions and a decrease in income, highlighting the application of twin studies beyond the medical field [26, 27].
An analysis of twin data from 20 countries estimated the heritability of BMI, a key indicator of obesity, to be 73%. It has also been reported that after middle age, the heritability of BMI tends to be higher in women than in men [28]. This sex difference may reflect varying degrees of obesity suppression due to environmental or social constraints, suggesting that social factors may contribute to the observed pattern. Another study showed that genetic factors influencing BMI change from childhood to adolescence, with particularly pronounced variations occurring from early to mid-childhood, during which new genetic factors begin to exert influence [29]. Additionally, studies comparing genetically predicted BMI with actual measurements in monozygotic and dizygotic twin pairs have sought to identify individual differences in environmental factors contributing to obesity [30]. Another twin-pair study investigating psychological and environmental factors found a significant association between increased BMI and physical well-being. However, no clear associations were observed between BMI and symptoms of depression, anxiety, self-esteem, life satisfaction, or social well-being [31]. Numerous other twin studies have examined the genetic and environmental contributions to BMI. A multinational study of European twin cohorts reported consistently high heritability of lipid profile parameters. Specifically, heritability estimates ranged from 61% to 83% for HDL cholesterol, 62% to 75% for LDL cholesterol, and 48% to 71% for triglycerides. Although not statistically significant, a trend toward higher heritability in women was observed [32]. Comparative data from twin pairs suggest that metabolic syndrome may be genetically linked to accelerated epigenetic aging, independent of physical activity, smoking, and alcohol consumption [33]. Furthermore, gene expression analysis of subcutaneous white adipose tissue in twins revealed a partial genetic influence on networks governing fatty acid and carbohydrate transport and metabolism [34]. A higher n-6:n-3 ratio was significantly associated with increased hepatic fat content within monozygotic twin pairs, suggesting that modifying this ratio may be beneficial for the prevention and treatment of nonalcoholic fatty liver disease (NAFLD) [35]. Additionally, in BMI-discordant monozygotic twin pairs, complement dysregulation was observed in obese twins independent of liver fat accumulation [36]. Pathway analysis of differentially expressed genes in the subcutaneous adipose tissue of BMI-discordant monozygotic twins indicated that reduced mitochondrial branched-chain amino acid (BCAA) catabolism and increased inflammation in adipose tissue contribute to unhealthy obesity [37]. Moreover, the expression of androgen-related genes in adipose tissue was significantly higher in the heavier twins of a monozygotic pair, with a positive correlation between fat mass percentage and gene expression observed within pairs, although no association was found with serum androgen levels [38]. Collectively, numerous twin studies have contributed to our understanding of the factors influencing obesity and lipid metabolism.
In the adrenal system, covariance structure analysis has shown that the heritability of individual differences in cortisol production decreases between 12 and 17 years of age, whereas the heritability of cortisol metabolism tends to increase with age [39]. Additionally, the heritability of glucocorticoid receptor activity has been found to be low, suggesting that environmental influences play a predominant role [40].
Although research on the pituitary gland is limited, a study involving twin athletes found intra-pair similarities in the exercise-induced increase in growth hormone (GH) and prolactin (PRL), as well as in cortisol variability. However, no significant differences were observed in the increase in adrenocorticotropic hormone (ACTH) or cortisol concentration. These findings suggest that genetic factors influencing the physiological adaptation of pituitary hormones are diverse [41].
Twin studies represent a uniquely powerful approach for rigorously disentangling the genetic and environmental contributions to all human traits—not only in endocrinology but across all fields of medicine and human biology (Graphical Abstract). To support such investigations, various twin registries have been established worldwide. Although many large-scale twin cohorts in Nordic countries and elsewhere [42-44] were designed with behavioral genetics in mind and structured as birth cohorts, our team at The University of Osaka maintains a volunteer-based twin registry focused primarily on adults. This registry includes data from over 500 twin pairs and serves as a valuable resource for phenotypic and multi-omics information, enabling a broad range of life science-oriented research approaches [45].
By leveraging these resources, we hope to demonstrate that twin-based methodologies can be applied widely across academic disciplines. In the near future, for instance, it may be possible to analyze an individual’s genome at birth and predict environmental conditions that would contribute to healthier and more fulfilling outcomes for specific traits. Such advancements have the potential to transform society by promoting physical, mental, and social well-being throughout an individual’s lifespan.
None of the authors have any potential conflicts of interest associated with this research. Mikio Watanabe is a member of Endocrine Journal’s Editorial Board.
This study did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
