Familial aggregation of esophageal cancer

Esophageal cancer has a substantial health burden in China, especially in the northern regions. Although family history has been established to be associated with esophageal cancer risk, some quantitative aspects of the association remain unclear or imprecise, including the age-specific magnitude of familial aggregation, the risks associated with the number and type of affected relatives, and how the familial risk combines with other known risk factors. A study conducted by Zhou et al[1] on the familial association of esophageal cancer answered some of the questions above. The study has several strengths including its large sample size of 33,008 participants with a high proportion of family history (15.2% having an affected first-degree relative), collection of data on known risk factors and detailed family history, and pathologically confirmed esophageal cancer diagnoses. Major findings of the study. The study had two major findings. First, the esophageal cancer familial risk ratio associated with an affected first-degree relative was 1.65, with considerable precision (95% confidence intervals: 1.47–1.84). Second, the familial risk ratio increased to 4.05 (albeit with wide confidence intervals) if the first-degree relative was diagnosed <35 years, and decreased with the increasing of diagnosed age of the relative. These findings suggest that esophageal cancer has a substantial familial aggregation, which appears to decrease with age. If this familial aggregation was due entirely to polygenic risk factors, then on average those individuals in the upper quartile of risk would be at >10 times the risk of those in the lower quartile. The strength of this risk gradient is much greater at a younger age, such as a >60-fold interquartile risk ratio for having esophageal cancer diagnosed <35 years.[2] Limitations of the study. The study only investigated family history in relatives defined by familial relatedness, but not by the specific type of relatives, and in particular parents and siblings. A difference in familial risk between parents and siblings would suggest different inheritance modes of the susceptibility genes. The absence of a difference would suggest that the genes are likely to act dominantly or additively, while the existence of a difference would suggest that the genes are likely to act recessively, given that recessively inherited genes result in higher risks in siblings than in parents and offspring. For several common cancers like breast cancer (especially if diagnosed at a young age), there is a higher familial risk for having an affected sibling.[3] However, there are some caveats related to this; see below. Another limitation is that the study only investigated age-specific familial aggregation in terms of the diagnosis age of the relatives, but not that of the cases. For breast cancer, the decreasing trend is more obvious with the diagnosis age of the cases.[4] The resource of Zhou et al[1] would be useful for investigating if esophageal cancer familial aggregation also depends on the diagnosis age of the cases. On the other hand, the study did not analyze the familial risk ratio associated with having relatives diagnosed >50 years. Nevertheless, based on their results, the familial risk ratio with having a first-degree relative diagnosed >50 years must be <1.65, which further supports the evidence that esophageal cancer familial aggregation decreases with age. The familial aggregation of several common cancers, like breast cancer, also decreases with age.[3,4] This pattern suggests that familial factors are more important at younger ages (e.g., the breast cancer familial variance explained by BRCA1 and BRCA2 pathogenic variants is greater for young women[3]), and studying young cases could be more fruitful for identifying novel familial risk factors, especially susceptibility genes or variants. The study collected data on several known risk factors including sex, smoking, and alcohol consumption and adjusted for them in the analysis. What can be further investigated includes the change in familial risk before and after adjusting for the risk factors, and whether the magnitude of the familial risk differs across levels of known risk factors. The former has etiological implications of whether known risk factors help explain the familial aggregation, and the latter has implications for risk prediction and prevention — whether family history and known risk factors combine multiplicatively in increasing risk. If yes, the absolute risk attributed to known risk factors will be greater for people with a family history, and relevant interventions would be more fruitful for these people. Implications of the study findings. For etiology: Segregation analyses of esophageal cancer in Chinese populations have suggested that high-penetrant recessively inherited major genes explain esophageal cancer familial aggregation.[5] However, genetic studies have yet to find any major genes for esophageal cancer. There might be two reasons for the inconsistency. First, the findings of recessive genes from the segregation analyses might be explained by birth cohort effects that siblings have a higher background cancer incidence than parents, siblings are more likely to be diagnosed due to increased screening, and/or siblings share some non-genetic and lifestyle risk factors. Segregation analyses have also suggested recessive genes for cancers like breast cancer, which might be due to these effects too.[3] Second, the esophageal cancer segregation analyses only considered a single major gene model. The increase in familial risk ratio with the number of affected first-degree relatives observed by Zhou et al[1] suggests that there are many genetic risk groups, consistent with a polygenic susceptibility model and the evidence from genome-wide association studies (GWAS) of esophageal cancer. Segregation analyses have found that a model including major genes and polygenes is the optimal genetic susceptibility model for cancers like breast cancer.[3] On the other hand, the familial risk ratio associated with having one affected first-degree relative was 1.6 times the ratio associated with having one affected second-degree relative (on the log scale), <2, suggesting that part of the familial factors is likely to be non-genetic. This is plausible, especially given that the participants were from high-risk areas where environmental factors have been established to be related to esophageal cancer risk. For risk prediction and prevention: Under the polygenic susceptibility model,[6] the familial risk of 1.65 implies that 31% of individuals in the population have an increased risk (familial risk ratio >1.0), and these individuals account for 69% of all individuals affected; half of the esophageal cancer cases come from the 16% with the highest risk in the population. These have obvious public health implications—targeting high-risk individuals can significantly reduce the esophageal cancer burden at the population level. A related question is: how to identify the individuals who are at high risk? Obviously, individuals having just one affected first-degree relative do not have a substantially increased risk unless the relative is diagnosed at a young age (e.g., <35 years). High-risk individuals are those with a strong family history. Segregation analyses can be used to develop cancer risk models for identifying high-risk individuals based on multigenerational family history, genotypes, and other factors, and a successful example is the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm for estimating a woman's future risk of developing breast cancer. The same could be applied to esophageal cancer. Polygenic risk score (PRS) has been proposed to be able to identify high-risk individuals. For esophageal cancer, PRSs developed from the population of European ancestry almost have no discrimination accuracy, with an area under receiver operating characteristic curve (AUC) of ∼0.55.[7] Such AUC is equivalent to a familial variance of 0.03; therefore, these PRSs only explain ∼3% of the esophageal cancer familial variance of 1.0 under the assumption that European populations have the same familial risk ratio of 1.65. This suggests that more esophageal cancer genetic variants are yet to be found. The performance of esophageal cancer PRS in Chinese populations is unknown, as, to the best of our knowledge, there is no valid PRS specific to Chinese populations published. Breast and colorectal cancers have some of the largest cancer GWAS in the world, but PRSs developed from these studies only explain ∼20% on average of the familial variance in these cancers.[8,9] Assuming the same proportions apply to esophageal cancer, esophageal cancer PRS would have an AUC of 0.62. However, combining PRS with family history is expected to improve disease discrimination accuracy, as observed for breast cancer.[10] Note that, there is a maximum AUC that can be achieved by knowing all esophageal cancer genetic risk factors, which is 0.76 under the assumption that all familial factors are genetic. Another issue of PRS is its age dependency. For common cancers including breast and colorectal cancers which have an age-deceasing familial aggregation, the risk gradient of their PRSs also decreases with age, except the current best PRS for breast cancer.[8,9] The absence of clear age dependence in the risk associated with breast cancer PRS implies that this PRS explains a small proportion of familial aggregation at a young age, and it is unlikely to predict death due to breast cancer, that is, women with a higher PRS are not “eliminated.”[3,8,9] It would be interesting to know the age dependence when more esophageal cancer PRSs are developed in the future. Funding S.L. is a National Health and Medical Research Council (NHMRC) Emerging Leadership Fellow (GNT2017373). This work was conducted when S.L. was a Victorian Cancer Agency Early Career Research Fellow (ECRF19020). J.L.H. is an NHMRC Senior Principal Research Fellow. Conflicts of interest None.