Causality between sarcopenia-related traits and major depressive disorder: A bi-directional, two-sample Mendelian randomized study

Abstract

Observational studies have demonstrated an association between sarcopenia and depression. However, these studies may be influenced by confounding factors, and the causal relationship between sarcopenia and major depressive disorder (MDD) remains unclear. This study aimed to apply the Mendelian randomization (MR) method to address confounding factors and assess the causal effect of sarcopenia on MDD. A two-way, two-sample MR method was employed in this study. Instrumental variables of genome-wide significance level were obtained from the open large-scale genome-wide association study summary data. MR analysis was conducted using inverse variance weighted, MR-Egger, and weighted median methods. The reliability of the results was verified using the heterogeneity test, pleiotropy test, and leave-one-out method for sensitivity analysis. Grip strength (right-hand grip strength: odds ratio [OR] = 0.880, 95% confidence interval [CI] 0.786–0.987, P = .027; left-hand grip strength: OR = 0.814, 95% CI 0.725–0.913, P < .001) and usual walking pace (OR = 0.673, 95% CI 0.506–0.896, P = .007) exhibited a direct causal effect on MDD. MDD had a significant causal effect on appendicular lean mass (β = −0.065, 95% CI −0.110, −0.019, P = .005). There was a causal relationship between sarcopenia-related traits and MDD. Loss of muscle strength, rather than skeletal muscle mass, is correlated with an increased risk of MDD. Furthermore, individuals with MDD are more likely to experience loss of skeletal muscle mass.

1. Introduction

Depression is a severe mental illness that profoundly impacts the physical and mental well-being as well as the overall quality of life of affected individuals.^[1,2] Major depressive disorder (MDD) accounts for more than 90% of depression cases,^[3] is the main cause of suicide^[4] and an important factor inducing various diseases.^[5–7] Over the past 15 years, MDD has become the third leading cause of the global burden of disease.^[8] The number of reported MDD cases worldwide increased by nearly 50% from 1990 to 2017,^[3] highlighting the urgent need to address and manage depression as a critical public health concern.

Sarcopenia is a syndrome diagnosed with loss of muscle mass and/or function.^[9] Grip strength (GS), appendicular lean mass (ALM), and usual walking pace (UWP) are parameters commonly used to evaluate sarcopenia.^[10] Recent studies have focused on how sarcopenia and depression may be related. In a meta-analysis of 19 cross-sectional cohort studies, researchers found that individuals with sarcopenia are significantly more likely to suffer from depression (odds ratio [OR] = 1.57, 95% confidence interval [CI] 1.32–1.86).^[11] Nevertheless, while most observational studies support a positive association between sarcopenia and depression, some have failed to find such a relationship. For instance, in a cross-sectional study of 204 older adults, the association between sarcopenia and depression disappeared after adjusting for confounding factors.^[12] These inconsistent findings may be attributed to various factors such as sample size, ethnic diversity, subjectivity in assessment, individual variations in depression and sarcopenia-related traits, and other factors. Furthermore, observational studies have inherent limitations in establishing the causal relationships between sarcopenia-related traits and depression. Consequently, the role of sarcopenia in the development of depression and its causal nature remains unclear. Evaluating the causal link between sarcopenia and depression may shed light on the prevention and treatment of depression.

In Mendelian randomization (MR), instrumental variables (IVs), which are genetic variants strongly correlated with exposure of interest, were used to investigate the causal link between exposure and outcome. By leveraging genetic variants randomly assigned at conception that are not susceptible to confounding, the MR approach yields more robust and reliable results.^[13] However, to date, no study has applied Mendelian randomization to examine causality between sarcopenia and MDD.

In this study, we aimed to provide valuable insights into the causal nature of the sarcopenia-MDD association by utilizing genetic data.

2. Methods

2.1. Study design

The design flow diagram of this study is shown in Figure 1. In this study, a two-sample MR study design was employed to investigate the causal relationship between sarcopenia and MDD. The exposure variables in this analysis were sarcopenia-related traits, specifically GS, ALM, and UWP. The genome-wide association study (GWAS) data used in this study were collected from different European population.

Figure 1.

Flow diagram of Mendelian randomization study.

To ensure the validity of the results, 3 core assumptions were required to be met in each MR analysis. First, the IVs used should be strongly correlated to the exposure variables; Second, IVs should not be related to confounding factors that could influence both the exposure and outcome variables; Lastly, IVs should only affect the outcome variables through exposure variables, ruling out alternative causal pathways. Pleiotropy, which occurs when an IV affects the outcome through pathways other than exposure, was taken into consideration during the analysis.

In addition, reverse MR analysis was conducted to investigate the possibility of reverse causality. During the analysis, MDD was treated as the exposure variable, and sarcopenia-related traits (GS, ALM, and UWP) were considered as the outcome variables.

2.2. Data sources

2.2.1. Sarcopenia-related traits.

Sarcopenia-related traits genetic summary data were gained from the IEU open GWAS project (https://gwas.mrcieu.ac.uk/), a comprehensive genetic correlation summary database. The data from this project were subjected to the Mendelian randomization analysis using the R package. Data related to GS and UWP were sourced from the UK Biobank (UKB). This dataset included left-hand grip strength (LGS) data from 461,026 Europeans, right-hand grip strength (RGS) data from 461,089 Europeans, and walking pace from 459,915 Europeans. This study contained a total of 9851,867 single nucleotide polymorphisms (SNPs).^[14]

Summary of genetic data related to ALM were derived from a large GWAS on ALM published by Yufang Pei et al in 2020.^[15] This study included 450,243 participants aged 48 to 73 years from the UKB cohort, comprising 244,730 women and 205,513 men. ALM was quantified using appendiceal fat-free mass measured using the bioelectrical impedance analysis. This study included 18,071,518 SNPs.

2.2.2. Major depressive disorder.

Genetic variation data for depression data were obtained from a large meta-analysis of 2 independent GWAS datasets for MDD by David M. Howard et al.^[16] All the participants were of European ancestry. This meta-analysis involved 170,756 patients with depression and 329,443 controls. The details of exposure and outcome GWAS data in this study are presented in Table S1, Supplemental Digital Content, http://links.lww.com/MD/J750.

2.3. IV selection and validation

To select eligible SNPs as IVs, we performed quality control based on the 3 key assumptions of Mendelian randomization using the summary GWAS data of the exposure factors. First, we extracted SNPs that exhibited a strong association with exposure (P ≤ 5E−08). To account for linkage disequilibrium, we set a threshold of r² < 0.001 with a window size of 10,000 base pairs.^[17] Finally, the selected SNPs mentioned above were extracted from the outcome variables, while SNPs strongly associated with the outcome were removed. These procedures were implemented using the “extract instruments” function, “extract outcome data” function, and “harmonize data” function of the TwoSampleMR package, respectively.

Additionally, in a two-sample MR Study, when the correlation between genetic variants and exposure factors is insufficient, weak IVs may lead to an underestimation of the association strength, reduced test power, and bias in the causal association. Hence, we employed an F-test to evaluate the impact of weak IVs and excluded SNPs with an F-statistic of <10.^[18] To ensure the robustness of the results, the MR pleiotropy residual and outlier test (MR-PRESSO) was conducted to examine and remove any outliers if present, followed by reanalysis.

2.4. Analysis of Mendelian randomization

In this study, 3 different MR methods were employed: inverse variance weighted (IVW), MR-Egger, and weighted median (WM). The IVW method, used as the primary outcome analysis method, assumes that all genetic variants are valid IVs. If the results were heterogeneous, a random-effects model was employed.^[19] The WM method calculates the median of the distribution function obtained by ranking all individual SNP effect values by weight.^[20] The MR-Egger regression method does not impose a regression line that passes through the origin, and causality is assessed through the slope coefficient.^[21] These 2 methods complemented the IVW method.

2.5. Sensitivity analysis

To ensure the reliability of the results, the heterogeneity test, pleiotropy test and leave-one-out sensitivity test (LOO) were employed for sensitivity analysis. The Cochran Q test was used to evaluate heterogeneity, which was considered significant if the Cochran Q test P-value was <0.05. The MR-Egger intercept term was conducted to evaluate the presence of horizontal pleiotropy. If the Egger intercept approach zero, it indicates the absence of horizontal pleiotropy in the study. Conversely, In the case of the Egger intercept is nonzero and P < .05, it suggests the presence of pleiotropy.

2.6. Reported results and software

The results of the study were reported as odds ratios (ORs) for dichotomous variables and beta values for continuous variables to facilitate interpretation. We performed the TwoSampleMR package (version 0.5.6) and the MR-PRESSO (version 1.0) package in the R programming language for data analysis (version 4.2.0) for data analysis.

2.7. Ethics

As the data utilized in this study were publicly available published data, additional ethical scrutiny was not needed.

3. Results

3.1. Determination of IVs

In this study, 142, 129, 515, and 48 SNPs were identified as the IVs for RGS, LGS, ALM, and UWP, respectively. In addition, in the reverse causality study, 40, 39, 39, and 37 SNPS that were significantly associated with MDD but not with RGS, LGS, UWP, and ALM, respectively, were identified as IVs,. The F-statistics of the included SNPs were all >10, indicating that there was no weak instrument bias in the study, thus confirming the reliability of the results.

3.2. Genetically predicted sarcopenia-related traits on the risk of MDD

The heterogeneity test revealed significant heterogeneity among the selected IVs (Table S5, Supplemental Digital Content, http://links.lww.com/MD/J752), therefore, a random-effects model with the IVW method was employed.

Figure 2 and Table S2, Supplemental Digital Content, http://links.lww.com/MD/J753 show the results of causality of sarcopenia-related traits on the risk of MDD. The scatter plot and the forest plot of SNPs related to sarcopenia-related traits and their MDD risk are shown in Figures S1, Supplemental Digital Content, http://links.lww.com/MD/J754 and S2, Supplemental Digital Content, http://links.lww.com/MD/J755.

Figure 2.

Estimated causal relationship of sarcopenia-related traits to MDD using IVW, MR-egger and WM methods. ALM = appendicular lean mass; LGS = left-hand grip strength; MDD = major depressive disorder; OR = odds ratio; RGS = right-hand grip strength; SNP = single nucleotide polymorphism; UWP = usual walking pace.

The results of the genetic prediction analysis demonstrated a significant correlation between GS and the risk of MDD. IVW analysis showed that an increase in genetically determined RGS and LGS significantly reduced the risk of MDD (RGS: OR = 0.880, 95% CI 0.786–0.987, P = .027; LGS: OR = 0.814, 95% CI 0.725–0.913, P < .001). The weighted median (WM) method also yielded significant results (RGS: OR = 0.871, 95% CI 0.772–0.983, P = .025; LGS: OR = 0.847, 95% CI 0.746–0.961, P = .01). However, the causal estimates from MR-Egger regression were attenuated (RGS: OR = 0.790, 95% CI 0.517–1.207, P = .277; LGS: OR = 0.702, 95% CI 0.449–1.099, P = .124). After adjusting for outliers using MR-PRESSO, the results of all the 3 methods remained consistent. Additionally, the Egger intercept P-values showed no horizontal pleiotropy for the selected IVs (RGS: intercept = 0.001, P = .602; LGS: intercept = 0.001, P = .505), further supporting the main causal estimates (Table S4, Supplemental Digital Content, http://links.lww.com/MD/J756). LOO showed no substantial differences when individual SNPs were removed, reinforcing the reliability of the results (Fig. S4, Supplemental Digital Content, http://links.lww.com/MD/J758).

Genetically predicted walking pace was also found to be significantly associated with the development of depression. An increase in walking pace was correlated with a lower incidence of depression (OR = 0.673, 95% CI 0.506–0.896, P = .007). Similar causal relationships were obtained by using the WM method (OR = 0.648, 95% CI 0.497–0.844, P = .001) and the MR-Egger method (OR = 0.276, 95% CI 0.080–0.953, P = .047). The results remained significant after adjustment for MR-PRESSO. The MR-Egger intercept (intercept = 0.008, P = .15) indicated no evidence of horizontal pleiotropy (Table S4, Supplemental Digital Content, http://links.lww.com/MD/J756). LOO further confirmed the reliability of the results (Fig. S4, Supplemental Digital Content, http://links.lww.com/MD/J758).

No significant causal relationship was found between ALM and MDD (OR = 0.981, 95% CI 0.9531.0–09, P = .188). Similar risk estimates were obtained using WM (OR = 0.994, 95% CI 0.957–1.032, P = .748) and MR-Egger regression methods (OR = 0.962, 95% CI 0.899–1.028, P = .251). The Egger intercept suggested the absence of horizontal pleiotropy (intercept = 0.0005, P = .518) (Table S4, Supplemental Digital Content, http://links.lww.com/MD/J756). LOO did not identify any individual SNPs that strongly influenced the overall results (Fig. S4, Supplemental Digital Content, http://links.lww.com/MD/J758). The visualized results of the sensitivity analysis are presented as a funnel plot (Fig. S3, Supplemental Digital Content, http://links.lww.com/MD/J760).

3.3. Genetically predicted MDD on sarcopenia-related traits

To examine reverse causality, this study investigated the causality of genetically predicted MDD on sarcopenia-related traits. Owing to the observed heterogeneity (Table S5, Supplemental Digital Content, http://links.lww.com/MD/J762), the IVW method employed a random-effects model. Neither the IVW method nor the other methods provided evidence supporting the causal effect of MDD on GS (Fig. 3, and Table S3, Supplemental Digital Content, http://links.lww.com/MD/J764). The results were visualized using scatter plot and forest plot, as shown in Fig. S5, Supplemental Digital Content, http://links.lww.com/MD/J769 and S6, Supplemental Digital Content, http://links.lww.com/MD/J770. The Egger intercept P-values manifested no significant horizontal pleiotropy (RGS: P = .177, LGS: P = .190) (Table S4, Supplemental Digital Content, http://links.lww.com/MD/J756).

Figure 3.

Estimated causal relationship of MDD to sarcopenia-related traits using IVW, MR-egger and WM methods. ALM: appendicular lean mass; LGS = left-hand grip strength; MDD =major depressive disorder; OR = odds ratio; RGS =right-hand grip strength; SNP = single nucleotide polymorphism; UWP =usual walking pace.

Interestingly, using both IVW (β = −0.065, 95% CI −0.110, −0.019, P = .005) and WM (β = −0.087, 95% CI −0.126, -0.049, P = 8.51E−06) methods, it was found that genetically predicted MDD was significantly related to a lower ALM index (Fig. 3, Table S3, Supplemental Digital Content, http://links.lww.com/MD/J764, Fig. S5, Supplemental Digital Content, http://links.lww.com/MD/J769 and S6, Supplemental Digital Content, http://links.lww.com/MD/J770). The Egger intercept was near to zero (P = .63) (Table S4, Supplemental Digital Content, http://links.lww.com/MD/J756). LOO indicated no significant difference in causality between the 2 groups when a single SNP was removed, further supporting the robustness of the results (Fig. S8, Supplemental Digital Content, http://links.lww.com/MD/J774). Figure S7, Supplemental Digital Content, http://links.lww.com/MD/J772 shows the funnel plot as a visual result of sensitivity analysis.

MDD was significantly correlated with slower walking pace in the IVW (β = −0.031, 95% CI −0.062, −0.001, P = .005) and WM (β = −0.040, 95% CI −0.068, −0.011, P = .005) methods, However, the MR-Egger method showed the opposite relation (β = 0.094, 95% CI −0.061, 0.248, P = .241), although the results were not significant (Fig. 3, Table S3, Supplemental Digital Content, http://links.lww.com/MD/J764, Figs. S5, Supplemental Digital Content, http://links.lww.com/MD/J769 and S6, Supplemental Digital Content, http://links.lww.com/MD/J770).

4. Discussion

MR analysis of sarcopenia and MDD was performed for the first time in this study. Sarcopenia-related traits and depression are causally related, according to our findings. Specifically, lower GS and slower walking speed were associated with higher rates of MDD, while MDD was related to a decrease in ALM.

Grip strength is commonly used as a measure of overall muscle strength^[22] and is a key characteristic of sarcopenia.^[10] Observational studies have consistently demonstrated that GS and depression are negatively correlated with each other. For instance, a prospective cohort study involving participants from the UKB reported a 7% reduction in depression risk for every 5 kg increase in GS. Moreover, individuals in the middle and lowest GS tertiles had 11% and 24% higher risks of depression over a 2-year follow-up period, respectively.^[23] Similarly, a study involving 34,129 adults from different countries found that a higher prevalence of depression among individuals with weak GS than those without (8.8% vs 3.8%; P < .001).^[24] Additionally, increased GS has been shown to improve cognitive function in people with MDD.^[25] In agreement with these observational studies, our results suggest that a higher genetically predicted GS is related to a reduced occurrence of depression. However, depression did not appear to influence the GS.

ALM is a crucial physiological indicator used to estimate muscle mass, and a low ALM is a key component in the diagnosis of sarcopenia.^[9,15] A cross-sectional study conducted in South Korea demonstrated that adolescent girls with low muscle mass were at an increased risk for depression.^[26] However, a study involving Korean adults,which used skeletal muscle mass as a measure of sarcopenia failed to show significant association with depression.^[27] In our MR analysis, ALM did not significantly contribute to MDD occurrence. However, in the other direction, MDD has a significant causal relationship with ALM, indicating that individuals with MDD are more likely to experience a reduction in skeletal muscle mass.

A bidirectional relationship has been demonstrated between walking speed and depression in previous observational studies. As a measure of physical function, the walking pace can quickly, safely, and accurately assess sarcopenia and predict associated adverse outcomes with it.^[10] A cross-sectional study of 796 participants found that slow walking speed was relevant to depression in older adults.^[28] Another study indicated that depression or depressive status is a risk factor for slow walking among older adults.^[29] Our study showed a negative correlation between genetically predicted usual walking speeds and the occurrence of depression. However, evidence regarding the influence of depression on walking speed was insufficient, as the results of the MR-Egger, IVW, and WM analyses were contradictory.

Sarcopenia and depression have been linked in several observational studies and meta-analyses. According to a meta-analysis of 15 observational studies, sarcopenia is strongly associated with depression.^[30] Similarly, a cross-sectional study of 330 older adults from Thailand found that high depression scores were an important risk factor for sarcopenia in this population.^[31] However, a study by Dursun Hakan Deliba et al indicated that there was no significant association between the 2 among older people.^[12] Based on our results, sarcopenia and depression may be causally linked. Sarcopenia and depression share similar characteristics, such as age-related and genetically influenced disorders. They may interact through various biological pathways, including neurotrophic factors, oxidative stress, inflammation, and others.^[32,33] There are several neurotrophic factors that are expressed in muscles, including the brain-derived neurotrophic factor (BDNF) and neurotrophin-3. They are associated with both muscle growth/repair and mood regulation.^[32] Mediation analysis suggested that depressive symptoms influence sarcopenia through dietary inflammation.^[34] The accumulation of reactive oxygen species can contribute to decreased muscle function, and studies have shown that protein oxidation is independently associated with poor muscle strength.^[35] Depressive disorders can be associated with changes in the brain structure and function caused by oxidative stress.^[36] A meta-analysis found that 8-OHdG and F2-isoprostanes, which are oxidative stress markers, were significantly increased in individuals with depression.^[37] Lifestyle factors such as physical activity and nutrition can also influence the progression of depression.^[38] Based on our findings, we recommend increased attention to screening for depression in individuals with low GS and slow walking speed as well as interventions aimed at improving GS and walking speed to potentially reduce the occurrence and development of depression. Additionally, individuals with depression should be monitored for muscle mass and protein intake to prevent sarcopenia onset. However, the pathophysiological mechanisms of sarcopenia and depression need to be further explored.

The advantages of this study are as follows: First, unlike the previous research on the relationship between sarcopenia and depression that rarely includes MDD patients in specific studies, we supplemented previous observational studies with two-way MR analysis, providing evidence for a causal link between sarcopenia and MDD. Second, the study benefited from a large sample size and utilized MR with exposure and outcome data from the European population, which helped minimize confounding bias. The determination of causal relationship results in this study is as reliable as that of randomized controlled trials, with greater feasibility.

However, there are some limitations to be considered. First, despite employing the random effects model and conducting pleiotropy tests, the study still exhibited high heterogeneity and residual bias that could not be completely eliminated. Second, the samples used in this study were exclusively from Europe, which limits the generalizability of the findings.

5. Conclusion

To summarize, the study demonstrated a causal relationship between increased genetically predicted muscle strength and walking pace and a reduced risk of MDD, whereas the reverse causality was not significant. Additionally, it was found that the occurrence of MDD was not significantly related to ALM. However, MDD had a significant causal relationship with ALM. Therefore, our study suggests that loss of muscle strength, rather than muscle mass, plays a role in the increased risk of MDD. Furthermore, individuals with MDD are more likely to experience decreased in skeletal muscle mass.

Acknowledgments

We are extremely grateful to the researchers and providers for the publicly available data used in this study.

Author contributions

Conceptualization: Yu Zhang, Mingquan Li.

Formal analysis: Yu Zhang, Mengfan Yang.

Funding acquisition: Mingquan Li.

Methodology: Yu Zhang, Mingquan Li.

Software: Yu Zhang, Mengfan Yang.

Validation: Yu Zhang.

Visualization: Mengfan Yang.

Writing – original draft: Yu Zhang.

Writing – review & editing: Mingquan Li.

Supplementary Material

medi-102-e35071-s001.docx ^{(12.7KB, docx)}

medi-102-e35071-s002.docx ^{(14.9KB, docx)}

medi-102-e35071-s003.docx ^{(16.1KB, docx)}

medi-102-e35071-s004.docx ^{(124.7KB, docx)}

medi-102-e35071-s005.docx ^{(268.9KB, docx)}

medi-102-e35071-s006.docx ^{(13.5KB, docx)}

medi-102-e35071-s007.docx ^{(355.2KB, docx)}

medi-102-e35071-s008.docx ^{(101.6KB, docx)}

medi-102-e35071-s009.docx ^{(14.9KB, docx)}

medi-102-e35071-s010.docx ^{(16KB, docx)}

medi-102-e35071-s011.docx ^{(116.9KB, docx)}

medi-102-e35071-s012.docx ^{(168KB, docx)}

medi-102-e35071-s013.docx ^{(242.4KB, docx)}

medi-102-e35071-s014.docx ^{(75.6KB, docx)}