Discussion
Main Findings
The meta-analysis suggested that higher BW was relative to the risk of many cancers including breast cancer, testicular cancer, colorectal cancer, prostate cancer, ovarian cancer while higher BW was the protective factor of endometrial cancer. But only breast cancer and testicular cancer had a non-linear relationship with BW. With the increase of BW, the slope of the dose‐response curve of breast cancer increased. For testicular cancer, the slope increased before 3000g and decreased above 3000g. Our results were consistent with the conclusions of most previous articles. Current meta-analyses of BW and breast cancer reported that higher BW was relative to breast cancer53-55. Some found that the relationship was linear53, 55 but some indicated that it was nonlinear54. One meta-analysis also found a U-shaped distribution between BW and subsequent risk for testicular cancer, but the risk of corresponding BW group was not specifically reported56. While another meta-analysis showed that there was no significant correlation in high BW 57.
Strengths
Our study has several advantages. Firstly, it’s the first study to systematically elucidated the relationship between BW and the risk of multiple cancers by combining dose-response meta-analysis and mendelian randomization analysis. In dose-response meta-analysis, we synthesized several articles to calculate the pooled RR and the changing trend of cancer risk with the increase of dose. In Mendelian randomization analysis, it could avoid the interference of confounding factors to calculate the effect value from the causal level. Secondly, when we collected studies in dose-response meta-analysis, we calculated the missing data of some studies instead of excluding them which made more articles were included in our study. Thirdly, we used three models for dose-response curve fitting to explore the optimal dose-response relationship. We changed the percentage of three nodes in the restricted cubic spline model according to different cancers data to ensure the accuracy of the fitting effect. Fourthly, we chose the newest GWAS studies of BW and cancers based on a large consortium which provided enough sample size to perform statistical calculations. And according to the power value and F statistic, we ensured that under the current sample size and genetic variation interpretation our MR result was statistically persuasive.
Limitations
Yet there are still some limitations in our study. Firstly, when we gathered data on BW with the risk of breast cancer, we found moderate heterogeneity among articles (I2=48%). But in the sensitivity analyses by removing one study at a time, we did not observe obvious fluctuation of the result, with a range from 1.08 (95% CI, 1.02-1.15) to 1.13 (95% CI, 1.06-1.19). In addition, except for breast cancer and testicular cancer, studies included in other cancers were fewer so it may increase publication bias. But all of them passed the Egger’s test (Povarian=0.41, Pcolorectal=0.31, Pendometrial=0.68, Pprostate=0.59). Secondly, most of our research articles came from high-income countries especially for the Mendelian randomization analysis which the population was all from Europe. It may limit the generalization of the results. Therefore, it should be considered carefully when applied the results.
Interpretation
Several studies had shown that high BW was related to the increase of intrauterine estrogen 58, 59. Significantly, excessive intrauterine estrogen exposure was often considered as the pathogenesis of both breast cancer and testicular cancer 60, 61. In the process of individual growth, the development of the mammary gland and testis was regulated by estrogen. Trichopoulos et al. early reported that breast cancer may originate from the uterus in 199062. The study suggested that prenatal exposure to estrogen is the highest at any other time in a woman’s life. After 4 weeks of pregnancy, most estrogens in the maternal body were produced by the placenta that was significantly higher than before pregnancy which estrogen was produced by the ovary 63. One study also found that the occurrence of breast cancer may be due to the imbalance of the self-renewal function of normal stem cells. Estrogen could promote cell proliferation, which also made it a cancer promoter, thus affecting cell growth and mutation 64, 65. During the fetal period, excessive estrogen can inhibit the secretion of Miller inhibitory substance (MIS) and the development of Leydig cells which produced testosterone, thereby affecting the development of testicles and then promoting the incidence rate of testicular cancer61. On the other hand, intrauterine growth retardation (IUGR) may lead to a lower BW which is associated with cryptorchidism and maldescended testis 66, 67. Both of them were considered to be risk factors for testicular cancer68. Therefore, low BW also increases the risk of testicular cancer.
However, in our two-sample MR, we provided no evidence to support the association between BW and cancer which meant that BW may not affect tumorigenesis as an independent factor. We only found that BW had a casual effect of invasive mucinous ovarian cancer. But the incidence rate of this subtype of ovarian cancer was very low69, and most of them were metastatic cancer. These positive results did not affect the total conclusion of the irrelevant relationship between BW and ovarian cancer. Due to the lack of corresponding GWAS articles, we did not analyze the causal relationship between testicular cancer and BW.
Our MR analysis contradicted the traditional observational results. The MR analysis used single nucleotide polymorphisms (SNPs) relative to the exposure to infer a causal relationship between the exposure and the outcome. Compared with an observational study, it could avoid potential confounding and reverse causality. In addition, our MR results showed consistency when using different MR methods and the sensitivity analysis under sufficient statistical power. Therefore, we were sure that our MR results can better illustrate the relationship between BW and cancer.
There were some reasons to explain the contradiction between meta-analysis and MR analysis. Firstly, BW was correlated with other characteristics such as birth length, body mass index, menarche age and so on which would also affect the risk of cancers. One newest meta-analysis of BW and cancer risk found that after adjusting the potential confounding factors the relationship was null or weak52. Silva et al. reassessed the relationship between birth size and breast cancer risk from 32 studies and found that birth length is the strongest independent indicator of breast cancer risk. After adjusting for birth length, the association with BW disappeared70. Meanwhile, as mentioned above, low BW may be caused by fetal intrauterine dysplasia which was also the risk factor of cryptorchidism 66, 67. All suggested that BW may not be a direct factor in the occurrence of testicular cancer. Secondly, compared with genetic factors, BW is more affected by maternal nutrition and hormones. The growth of the fetus depends on the nutrition provided by the mother through the placenta. As the report of Horikoshi M et al.45, the genes they identified associated with BW could only explained 15% of the weight variation. Therefore, the increase of BW may be related to the risk of cancer through other non genetic factors. But so far, It is difficult to distinguish the role of genetic and non genetic factors in the relationship between BW and the developing of cancers. What’s more, Horikoshi M et al. excluded individuals with extreme BW when collecting the data which we thought to be relative to cancer risk. To some extent, we lacked genetic variation for extreme BW that may affect the results.