This study found consistent evidence for a positive association between elevated TG levels and EC risk, both when using quartiles and dichotomous values of TG as well as when conducting stratified analysis by parity or menopausal status.
A series of contradictory observations have emerged regarding the relation between lipids and EC. In a prospective study of 5,209 subjects, no association was found between TC and any female cancer, while others have found both negative and positive associations between TC and EC risk [26
]. Most recently, a prospective cohort study including 31,473 women found no association between TC, LDL, or HDL and EC risk, but there was a positive association of TG levels and EC risk (RR: 2.34 (95% CI: 1.04-5.28)) [6
Our findings of an association between TG and EC risk corroborate this most recent study, but we also found an association between TC and EC, albeit not when using medical cut-offs [6
]. Certain nuclear receptors, such as perioxisome activated receptors (PPAR), have been found to modulate lipid levels [6
]. One form of PPAR, PPAR gamma, is expressed in adipose tissue and is activated by endogenous ligands, such as fatty acids, and is involved in lipid metabolism [31
]. In EC, as among several other cancers, it has been noted that PPAR gamma expression is downregulated, which suggests that, since PPAR-gamma has also been found to regulate expression of p21 in EC cells, down-regulation of PPAR-gamma supports EC while preventing the clearing of TG [30
]. Such a mechanism would suggest that increased triglyceride levels may possibly cause a decrease in PPAR-gamma levels, causing adverse effects on regulation of cell cycle, as is present in cancerous cells. PPAR-alpha, another form of PPAR, is also expressed in adipose tissue and facilitates catabolism of TG [33
]. Furthermore, in mice fed ligands for PPAR-alpha, increased expression of proteins such as CDK-1 and CDK-4 in murine hepatic cells was noted, which are over-expressed in cancer cells [31
]. As fatty acids are recognized ligands for PPAR-alpha and it has been noted that PPAR-alpha levels are elevated in endometrial cancer, it could be speculated that PPAR-alpha may be a possible explanation to the association between TG and EC risk [31
Following the studies investigating the association between the metabolic syndrome and EC risk [35
], we performed a glucose-stratified analysis. Despite the rather consistent positive association between serum glucose levels and EC risk, our study did not find evidence for effect modification. The association between TG and EC remained for those with normal glucose levels, but the lack of an association among those with high glucose is likely due to the small number of cases with high glucose levels. Besides the PPAR mechanism mentioned above, it was recently suggested that dendritic cells, responsible for initiation and maintenance of immune response, from tumour-bearing mice and persons with cancer have high amounts of TG compared with those from tumour free mice and healthy individuals [38
]. These findings suggest that manipulating lipid levels in dendritic cells may improve immune responses in cancer and illustrate the importance of assessing reverse causality when assessing the link between lipid components and risk of cancer. We did not find evidence for reverse causation, however to verify the findings by Herber et al. [38
], we lack information on tumour stage in this database.
In our study, upon stratification by BMI, the positive association between TG and EC risk was attenuated for overweight women; this may be due to the difference in baseline levels of triglyceride-associated EC risk between overweight and non-overweight individuals. Our results have to be interpreted carefully, as there is already a higher EC risk associated with obese individuals, which is not represented in our calculated hazard ratios, but can be seen in .
We found that the association between TG and EC risk was slightly greater in women who were nulli-parous at baseline, which may suggest that parity may affect the association of TG and EC. However, as this was observed for only one quartile, it may not be conclusive and TG may be a proxy of another metabolic marker such as obesity. Several studies have noted the protective effect of parity on EC risk [39
]. As EC is associated with high levels of unopposed estrogen, the protective effect of parity may be due to the high levels of progesterone that are present during the course of pregnancy [19
]. Protection due to parity may also arise from the shedding of malignant cells through childbirth [40
]. Experiments in mice have shown a decrease in elevated TG levels in mice injected with progesterone, which while speculative, could explain the protective effect of parity on the association of TG with greater EC risk [41
Additionally, we noted a stronger association between TG and EC risk for postmenopausal women compared to premenopausal women. However, this observation was limited to the highest quartile of TG. In addition, one has to be aware that we did not have detailed information on menopausal status and used age as a proxy. Among premenopausal women, obesity is related to an increased number of anovulatory menstrual cycles [42
]. Our observation of an association between menopause and EC risk via TG is in line with previous research, which found menopause to be positively associated with increased TG, possibly related to increased hepatic production of TG-rich particles [43
]. The increase in TG and possible subsequent downstream increases in PPAR-alpha and cancer-associated proteins is a potential mechanism of association between possible effect-modification of menopause status at baseline and association between TG and EC risk.
The large number of women with prospective measurements of lipid biomarkers, all measured at the same clinical laboratory, is a major advantage of our analysis. The AMORIS population covers a range of SES and ethnicity similar to that of the general working population of Stockholm county. During the study period the population selected for AMORIS had an all-cause mortality about 14% lower than in the general population of Stockholm county when taking age, gender, and calendar year into account [45
], but this does not affect the internal validity of our study. Another strength of this prospective study was its ability to account for parity, menopausal status, and BMI as potential confounders. However, only limited data on BMI was available which confined the statistical power when also studying BMI (n=31,792) and age was used as a proxy for menopausal status. The latter approximation has been used previously and is based on a previously estimated median age at menopause of 50 [12
]. All models were adjusted for fasting status, however it was missing for about 10% of the study population. Lack of possible confounding variables, such as age at menarche, oral contraceptive use, measurements of different hormone levels, and hormonal use are limitations of this study.