Causal relationship between affect disorders and endometrial cancer: a Mendelian randomisation study

Abstract

Background

The aim was to assess the causal relationship between depression and anxiety disorders and endometrial cancer.

Method

We performed two-sample Mendelian randomisation analysis using summary statistics from genome-wide association studies to assess associations of major depressive disorder, anxiety and stress-related disorders with endometrial cancer. The genome-wide association studies(GWASs) data were derived from participants of predominantly European ancestry included in the Genome-wide Association Research Collaboration. Inverse variance-weighted, MR-Egger and weighted median MR analyses were performed, together with a range of sensitivity analyses.

Results

Mendelian randomisation analysis showed no statistically significant genetic responsibility effect of anxiety and stress-related disorders on any pathological type of endometrial cancer. Only the effect of major depressive disorder under the inverse variance weighting method increasing the risk of endometrial endometrial cancer (effect 0.004 p = 0.047) and the effect of major depressive disorder under the MR-Egger method decreasing endometrial cancer of all pathology types (effect −0.691 p = 0.015) were statistically significant. Other Mendelian randomisation analyses did not show a statistically significant effect.

Conclusion

Major depressive disorder(MDD), anxiety and stress-related disorders(ASRD) are not genetically responsible for endometrial cancer. We consider that emotional disorders may affect endometrial cancer indirectly by affecting body mass index. This study provides us with new insights to better understand the aetiology of endometrial cancer and inform prevention strategies.

PLAIN LANGUAGE SUMMARY

This study used public genomic data to analyse association between affective disorders, including depression and anxiety, and endometrial cancer. Genes treated as instrumental variables help us understand the causal link between affective disorders and endometrial cancer through bioinformatics. In addition to this, we added type 2 diabetes, body mass index, polycystic ovary syndrome, and age at menopause for multivariate Mendelian randomisation analyses with the aim of reducing confounding bias. Because we consider these factors may potentially influence the relationship between affective disorders and endometrial cancer. Ultimately we believe that the association between depression and endometrial cancer is not as strong as that of obesity, due to the genetic correlation between depression and obesity.

Keywords:

• Affect disorder
• depression
• anxiety
• risk factor
• Mendelian randomisation

Introduction

Emotional and psychological factors that increase the risk of certain cancers have intrigued researchers in recent years. Numerous cohort studies and meta-analyses have provided high-level evidence for associations between psychological factors and certain types of cancer (Chida et al. 2008). Researchers have proposed a number of mechanisms by which affective disorders may affect cancer risk. In terms of behaviour, patients with affective disorders may have a more unfavourable lifestyle, such as high smoking rates, alcohol abuse, poor diet, and reduced physical activity, which would lead to obesity and increase cancer risk (Murphy et al. 2003, Strine et al. 2008, Beydoun et al. 2009). In terms of molecular mechanisms, affective disorders could increase inflammation,su impair immune function and lead to decreased cortisol levels, which are associated with increased cancer risk (Reiche et al. 2004).

Evidence-based clinical studies suggest that obesity and bad lifestyle are the risk factors with the highest level of association with endometrial cancer, but few studies have examined whether emotional disorders are risk factors for endometrial cancer (Raglan et al. 2019). Considering that endometrial cancer is a common cancer in women, it is thus important to explore the causal relationship between affect disorders and endometrial cancer.

Mendelian randomisation has been successful in confirming the causal relationship between affective disorders and cancer, depressive disorders have been confirmed to increase the risk of prostate cancer but not colorectal cancer (Chen et al. 2020, Wu et al. 2022). In this study, we assessed whether there is a causal association between affective disorders and endometrial cancer through a two-sample Mendelian randomisation (MR) analysis involving depression, anxiety, and endometrial cancer (Long et al. 2023). Our study incorporated adjustments for established endometrial cancer risk factors such as obesity and type 2 diabetes. This adjustment aimed to ascertain the independence of the causal effects of depression and anxiety on endometrial cancer (Burgess and Thompson 2015).

Methods

Study design and data sources

This was a Mendelian randomisation study in which exposures and outcomes of interest were selected to observe their causal effects. In this study, the exposures were ‘Major depressive disorder (MDD)’ and ‘anxiety and stress-related disorders (ASRD)’. We used summary statistics from the largest publicly available GWAS for MDD and ASRD gene associations (Howard et al. 2019, Meier et al. 2019). We selected the largest genome-wide association study of endometrial cancer to date, which divided patients into the following groups according to endometrial cancer pathology: those with endometrial cancer pathology, those with endometrioid tissue endometrial cancer pathology, and those with nonendometrial tissue endometrial cancer pathology. We chose the genome-wide association study involving the largest SNP sequencing database available in the IEU-GWAS to determine the associations of type 2 diabetes, age of menopause, body mass index, and polycystic ovary syndrome with outcomes (Hemani et al. 2018, Elsworth et al. 2020, Lyon et al. 2020). Table 1 provides the summary GWAS data used in this study. Please refer to the supplementary file for diagnostic criteria and sources of traits. Because all analyses herein were based on publicly available summary data, no ethical approval from an institutional review board was required for this study.

Table 1. Summary of GWAS^a statistics information.

Author, year	Setting	Trait	Sample size	Ancestry
Meier et al.2019	Integrative Psychiatric Research	Lifetime diagnoses of anxiety and stress-related disorders	31,880(Male 44.4%)	European
Howard et al.2019	23andMe, UK Biobank, PGC^b	Major depressive disorder	807,753^c	European
O’Mara et al.2018	ECAC, E2C2, UK Biobank^d	Endometrial cancer^e	121,885(Male 0%)	European
FinnGen	FinnGen	Polycystic ovary syndrome	118,228^b	European
Neale	Neale Lab	Body mass index	336,107	European
Ben Elsworth	MRC-IEU	Age at menopause (last menstrual period)	143,819	European
Wood et al. 2016	UK Biobank	Type 2 diabetes	120,286	European

^aGWAS, genome-wide association studies; ^bPGC, Psychiatric Genomics Consortium; ^cRatio of males and females not available in the original literature; ^dECAC, Endometrial Cancer Association Consortium; E2C2, the Epidemiology of Endometrial Cancer Consortium; ^eEndometrial cancer includes all histologies, endometrioid histology, and nonendometrioid histology.

Sample independence

The population for this study was the European population. To ensure that the population did not overlap as much as possible, we selected summary of GWAS data from different GWAS cooperation organisations for different population cohorts.

Statistical analysis

All analyses were performed by using the TwoSampleMR (0.5.6) and MendelianRandomization (0.7.0) R packages (Hemani et al. 2018b). This study was approved by the Medical Ethics Committee of the Affiliated Hospital of Qingdao University, and the approval number is QYFY WZLL 28410. We extracted usable SNPs from exposure data (MDD and ASRD GWASs), and the criteria were a p value less than 1 × 10⁻⁶ and being uncorrelated (10,000 kilobase pairs apart and r² ≤0.001). We obtained data on SNP beta or odds ratio (OR) effects and corresponding standard errors from the exposure and outcome GWASs. If the data of the selected SNP could not meet the needs of the MR analysis (for example, there was no OR or beta value), then the SNP was deleted.

For single-variable MR analysis, we assessed the causal effect of the exposure on outcomes using inverse variance weighted (IVW) MR, MR–Egger regression, and MR median weighted methods (Rosoff et al. 2021). In addition, we used the MR–Egger intercept test and MR-PRESSO test to evaluate the level of pleiotropy of the IVs and the Cochrane Q test to evaluate the heterogeneity of the IVs. The alpha values of the above tests were all set to 0.05.

For MVMR analysis, we determined the IVs by using SNPs in each of the GWASs that met the single-variable MR selection criteria. We considered several traits clinically associated with endometrial cancer: body mass index, polycystic ovary syndrome, age at menopause, and type 2 diabetes (Raglan et al. 2019, Guo et al. 2022, Kurki et al. 2023). By adjusting for risk factors for endometrial cancer, it was possible to see whether affect disorders are independent risk factors for endometrial cancer (Burgess and Thompson 2015). We removed duplicate and correlated SNPs (within 10,000 kilobase pairs; r² ≥0.001) and then extracted the SNP effects and corresponding standard errors from the exposure and outcome GWASs.

Results

We studied the causal relationship between MDD and ASRDs in three pathological endometrial cancer types. There was no statistically significant heterogeneity or horizontal pleiotropy in the other categories of analysis. Table 2 presents results of the MRPRESSO assessment of horizontal pleiotropy. Table 3 shows the results of the three MR methods. Only the effect of major depressive disorder under the inverse variance weighting method increasing the risk of endometrial endometrial cancer (effect 0.004 p = 0.047) and the effect of major depressive disorder under the MR-Egger method decreasing endometrial cancer of all pathology types(effect −0.691 p = 0.015) were statistically significant. Other Mendelian randomisation analyses did not show a statistically significant effect.

Table 2. Results of horizontal pleiotropy assesment by MR-PRESSO method.

Exposure	Outcome	Estimate	Sd	p-Value
ASRDs	EC(EH)	0.002	0.007	0.78
	EC(NEH)	−0.010	0.018	0.61
	EC(AH)	0.002	0.007	0.82
MDD	EC(EH)	0.150	0.117	0.205
	EC(NEH)	0.099	0.239	0.678
	EC(AH)	0.036	0.092	0.70

Sd: standard deviation; ASRDs: anxiety and stress-related disorders; EC: endometrial cancer; EH: endometrioid histology; NEH: nonendometrioid histology; MDD: major depressive disorder; AH: all histology.

Table 3. Single-variable MR results of ASRDs and MDD on the risk of three subtypes of EC.

Exposure	Outcome	Method	Effect	SE	95% CI	p Value
ASRDs	EC(EH)	IVW	0.002	0.007	−0.013–0.016	0.84
	EC(EH)	MR-Egger	0.013	0.076	−0.136–0.162	0.865
	EC(EH)	Weighted median	0.002	0.01	−0.016–0.021	0.793
ASRDs	EC(NEH)	IVW	−0.009	0.018	−0.044–0.025	0.595
	EC(NEH)	MR-Egger	0.142	0.337	−0.519–0.802	0.674
	EC(NEH)	Weighted median	−0.003	0.023	−0.048–0.042	0.9
ASRDs	EC(AH)	IVW	0.002	0.008	−0.013–0.017	0.794
	EC(AH)	MR-Egger	0.161	0.049	0.064–0.258	0.001
	EC(AH)	Weighted median	0	0.008	−0.016–0.017	0.964
MDD	EC(EH)	IVW	0.004	0.002	0–0.008	0.047*
	EC(EH)	MR-Egger	−0.253	0.374	−0.987–0.48	0.498
	EC(EH)	Weighted median	0.004	0.003	−0.002–0.01	0.176
MDD	EC(NEH)	IVW	−0.001	0.005	−0.01–0.009	0.9
	EC(NEH)	MR-Egger	0.588	0.791	−0.962–2.138	0.457
	EC(NEH)	Weighted median	0	0.007	−0.013–0.014	0.983
MDD	EC(AH)	IVW	0.001	0.002	−0.002–0.005	0.443
	EC(AH)	MR-Egger	−0.691	0.284	−1.248–−0.134	0.015*
	EC(AH)	Weighted median	0.003	0.002	−0.002–0.008	0.191

MR: Mendelian randomisation; SE: Standard Error; CI: confidence interval; ASRDs: anxiety and stress-related disorders; EC: endometrial cancer; EH: endometrioid histology; IVW: inverse variance weighting; NEH: nonendometrioid histology; MDD: major depressive disorder; AH: all histology; *: the p value is less than 0.05, which is statistically significant.

In Multivariate Mendelian randomisation, using the IVW method to assess the genetic responsibility of MDD for endometrial cancer, we found no statistically significant effect of MDD for endometrioid endometrial cancer (OR −0.14, p = 0.25). MDD showed no statistically significant effect on the other two subtypes of endometrial cancer (NEH and AH). In the case of multivariate Mendelian randomisation analyses, only genes associated with age at menopause and BMI had statistically significant associations with increased risk of endometrial cancer. Table 4 lists the results of the MVMR analysis.

Table 4. Multivariate Mendelian randomisation results of the three pathology types.

	EC(EH)^a			EC(NEH)^b			EC(AH)^x
Exposure	NO. SNPs^d	effect	p	NO. SNPs	effect	p	NO. SNPs	effect	p
PCOS^e	319	−0.04	0.08	319	0.04	0.39	319	−0.02	0.31
T2D^f	319	0.04	0.14	319	0.11	0.06	319	0.05	0.05
MDD^g	319	−0.09	0.53	319	0.21	0.47	319	−0.14	0.25
BMI^h	319	0.57	<0.01	319	0.44	<0.01	319	0.55	<0.01
AAM^i	319	0.21	<0.01	319	0.34	0.01	319	0.20	<0.01

^aEC, endometrial cancer; EH, endometrioid histology; ^bNEH, nonendometrioid histology; ^cAH, all histology; ^dNumber of single nucleotide polymorphisms; ^ePolycystic ovary syndrome; ^fType 2 diabetes; ^gMajor Depression Disorder; ^hBody mass index; ⁱAge at menopause.

Discussion

In this study, after performing univariate and multivariate MR analyses, we found no direct independent causal association between MDD and ASRDs and endometrial cancer. We evaluated the causal effects of ASRDs and MDD on endometrial cancer at the level of genetic responsibility using MR analysis. The key finding of the study is the lack of a direct independent causal relationship between Major Depressive Disorder (MDD), Anxiety and Stress-Related Disorders (ASRDs), and endometrial cancer. This challenges previous assumptions and underscores the need to differentiate between correlation and causation in disease association studies. The use of both univariate and multivariate MR analyses strengthens this conclusion. It is worth noting that the statistical significance of MR-Egger regression is around 0.05. Since MR-Egger is better than the IVW method in providing accurate causal effect estimates in the case of weak instrument variables bias (Bowden et al. 2015). This study uses the threshold of p < 1*10⁻⁶ to select instrumental variables, and there is a possibility that the selected variables are weak instrumental variables. Therefore, the statistical results of the MR-Egger method may provide this study with a causal effect estimate that excludes weak instrumental bias effects.

MR analysis is a popular method to strengthen the determination of causality, so this study may provide more convincing conclusions about causality than clinical epidemiological observational studies (Sekula et al. 2016). However, this study also has many limitations. Since the effect of ASRDs on endometrial cancer was not significant and the sensitivity analysis did not provide clear evidence of horizontal pleiotropy, we did not perform MVMR analysis for ASRDs. During the instrumental variable screening, a p value of 1 × 10⁻⁶ (less than the conventional genome-wide significance threshold of p < 5 × 10⁻⁸) was used for ASRD-associated SNPs, which may reduce test power (Buzdugan et al. 2016). This methodological choice, while potentially increasing the number of SNPs for analysis, might also increase the risk of false positives, thereby affecting the study’s conclusions. In terms of evaluating the details of horizontal pleiotropic effects and heterogeneity, although we did not find statistically significant horizontal pleiotropy and heterogeneity by using diversity and appropriate statistical methods, there is still a risk of underestimation as mental disorders and cancer are complex and heterogeneous traits. Because anxiety, depression and other mental disorders are often comorbid, instrumental variables may be related to multiple clinical traits, and the diagnosis of mental disorders and cancer may be different in different regions and populations, which also brings the heterogeneity (Verbanck et al. 2018). There are many methods for genetic scoring (Wang et al. 2015). The genetic score chosen in this study is the OR value. The choice of genetic score may also affect the stability of MR analysis results (Böckerman et al. 2019). This study relied on GWAS data from European populations, which resulted in the results being applicable only to European populations, and future GWAS studies should further expand to include more diverse populations to ensure the applicability of the results on a global scale.

In general, there is a significant link between emotional distress and cancer, including endometrial cancer. The results of this study challenge this hypothesis of association. A large retrospective cohort study using electronic medical record data demonstrated that depression increases subsequent cancer diagnoses, including endometrial cancer (Mössinger and Kostev 2023). However, observational epidemiological studies may be affected by confounding factors and incorrectly estimate causal effects. Obesity and sleep both increase the risk of endometrial cancer, and obesity, sleep and emotional distress are closely linked (Setiawan et al. 2012, Jordan et al. 2017, 2021, Nock et al. 2020). This means that previous clinical studies have been unable to rule out the confounding effects of emotional distress and other endometrial cancer risk factors because emotions and lifestyle influence each other. This Mendelian randomisation analysis used effective instrumental variables to overcome the impact of confounding factors other than emotional distress on endometrial cancer. In addition, many clinical studies based on follow-up observations of patients after surgical treatment of endometrial cancer have found that patients develop anxiety or depression, and this emotional distress is caused by the impact of the cancer rather than a disease which represents genetic responsibility.

Type 2 diabetes, age at menopause, and obesity were all identified as risk factors for endometrial cancer in an MR study (Nead et al. 2015, O’Mara et al. 2018, Masuda et al. 2020). Such results are consistent with clinical experience (Raglan et al. 2019). At present, we believe that weight loss can reduce the risk of endometrial cancer. This study corrects the bias of clinical epidemiological observations. We believe that the focus of clinicians should be shifted to weight management. Since the public data used in this study were from European populations, extrapolation to global populations requires caution. Weight control has been repeatedly identified in clinical studies as having a significant correlation with endometrial cancer prevention and prognosis (Li et al. 2019). Many fertility-related lifestyle factors are associated with endometrial cancer, and older childbearing age, breastfeeding, and full-term pregnancy reduce the risk of endometrial cancer (Setiawan et al. 2012, Jordan et al. 2017, 2021). Genetic variants associated with higher BMI are associated with an increased risk of endometrial cancer, but this increased risk is not as significant as the increased risk of endometrial cancer associated with clinical obesity (Prescott et al. 2015). Obesity increases endometrial proliferation by increasing unopposed oestrogen, hyperactivity of insulin and insulin-like growth factor 1 signalling in the endometrium further promotes the growth of metabolically active tissues in obese populations, and obesity-associated pro-inflammatory adipokines increase the risk of endometrial cellular DNA breaks and mutation, all processes that increase the risk of endometrial cancer, but when overweight and obesity are controlled, these risks are reduced (Onstad et al. 2016). The results of Mendelian randomisation analyses can overcome the effects of confounders in observational studies to directly observe the association between exposure (MDD and ARDS) and outcome (endometrial cancer). The results of this study imply that even though an association between mood disorders and endometrial cancer is observed in clinical practice, it does not mean that treating depression and anxiety reduces the risk of endometrial cancer. Future prospective well-designed cohort or randomised controlled trial studies are still needed to confirm the relationship between emotion disorders and endometrial cancer (Davey Smith et al. 2017). In addition to this, this study did not focus on enough endometrial cancer-related lifestyle and environmental factors, and in the future, Mendelian randomisation analyses of lifestyle, environment and endometrial cancer could be performed as well as a formal mendelian randomisation mediation analysis could be applied that would more robustly quantify the contribution of potentially treatable intermediaries (Richmond et al. 2016).

Acknowledgments

We want to acknowledge the participants and investigators of the FinnGen study. We would like to thank the University of Bristol’s Department of Epidemiology for developing a handy software package for Mendelian randomisation analysis. We thank AJE.com for providing English writing editing services.

Data availability statement

The data used in this study are publicly available and both the data and the methods of acquisition have been provided in Appendix A.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.

Appendix A

The methods and data sources of this study are uploaded to the Github open source website.

https://github.com/Gengshuomengdeer/Causal-relationship-between-affect-disorders-and-endometrial-cancer-A-Mendelian-randomisation-study