The present study aimed to explore the causal association between childhood BMI and risk of EC in adulthood. To achieve this, four MR statistical techniques were used, including IVW, weighted median, weighted mode and MR Egger regression [11]. Our results implied that childhood BMI causally increased susceptibility to EC.
The relationship between childhood BMI and EC has not been well investigated. Previous observational studies have suggested that childhood or adolescent obesity might be a risk of EC [16, 17]. A large (n = 155,505) Danish prospective cohort study with 35 years of follow-up showed that the childhood (age 7–13) BMI was non-linearly associated with all endometrial cancers [16]. Furthermore, participants with higher BMI-gain across all child’s ages (6.28-14.0 years) were found to have increased prevalence of EC in later life [17]. In line with previous findings, we found that childhood BMI was causally related to the risk of EC. On the contrary, one MR study showed a non-significant correlation between EC and childhood body size [29], which was collected by asking adult participants to describe themselves as thinner, plumper or about average at age 10 [29]. Therefore, recall bias might be introduced. Another MR study also reported that the childhood obesity was not related with EC [30], which was inconsistent with our findings. This discrepancy might be explained by the 15 loci associated with childhood BMI used in that study had been updated to 25 genetic variants used in our study [31, 32]. Besides, selection of variables and data sources were also different in previous MR studies compared to our study [29, 30].
The association between early-life BMI and risk of EC was considered not independent from adult BMI [29]. However, recent study identified 25 genome-wide significant loci associated with childhood BMI and addressed that genes influencing childhood and adult BMI were not completely overlapped [32]. Potential age-specific differences or stronger effects of these genetic loci on childhood rather than adult BMI were suggested [32]. Thus, the relationship between childhood BMI and EC remains uncertain and needs further examination. More attention should be paid to early life intervention such as childhood weight control considering its causal link to the risk of EC during the life-span.
Nowadays, the first-line management of children obesity is lifestyle intervention on eating habits and physical activities [33], particularly for children under 12 years old. However, it often fails to achieve significant and enduring weight-reduction [10, 34]. If necessary, anti-obesity medications or surgeries are also options. Notably, a 3-year observational study of behavioral intervention on children at 6–16 years old revealed that the BMI z-score was reduced at least 0.5 units in 58% of the severely obese children at 6–9 years old compared with only 2% of the adolescents at 14–16 years old [35]. It was also worth noting that 92% of these severely obese adolescents were already obese at age 7 [35], addressing the importance of early intervention on severe obesity in childhood.
MR is employed to decrease the inherent biases of observational studies but is susceptible to bias amplification, which occurs when a single genetic variant is associated with multiple phenotypes, potentially leading to bias in causal inferences [36]. Incorporating a wide array of genetic variations into MR studies can increase statistical power; however, it also raises the probability of introducing pleiotropic variants that are not valid instrumental variables, which requires sensitivity analyses [18]. In our study, the MR analysis employed satisfied three assumptions [11]: 1) there had links between the genetic variants used as instrumental variables and childhood BMI; 2) the instrumental variables were not associated with potential confounders; 3) the instrumental variables only affected EC through childhood BMI, rather than other pathways. To solve the problem of pleiotropy, we applied methodologies including weighted median estimator and MR-Egger regression. Although there was a lack of consistence in the outcomes from different methods, the similar results from both the weighted median estimator and IVW approach bolstered the reliability of these associations.
This study has several strengths. The nature of MR analysis decreased bias from unobserved confounding of childhood BMI and EC. Also, the large-scale sample size used in the MR analysis increased the statistical power for reliable estimation of causal effects. However, our study only included endometrioid histology type of EC, and the study participants were all European. Considering BMI differs among ethnic groups, it might impact the generalizability of our findings. Further MR studies with improvement on these aspects are needed to untangle the relationship between childhood BMI and EC.
In conclusion, applying a two-sample MR study, we found childhood-BMI causally contributed to an increased risk of EC using meta-analyses data from GWAS. Our results suggested the importance of weight control for obese children to reduce their risk of EC in adult stage.