Introduction

Sarcopenia is a progressive syndrome which is characterized by a decline in muscle mass and strength function of the whole body, accompanied by a decline in quality of life and an increase in mortality1. It occurs commonly as an age-related process in older people. Sarcopenia significantly impacts daily activities, functional status, contributes to increased disability, and affects quality of life in older populations2. It also elevates healthcare costs3, and imposes a substantial burden on both individual health and the social economy4. It is estimated that there are currently about 50 million people with sarcopenia in the world, and this number is expected to reach 500 million by 2050 as the world population ages rapidly5. With growing life expectancy, the prevalence of sarcopenia will continuously increase during the next decades6. The incidence of sarcopenia is predicted to increase to > 200 million affected older adults worldwide over the next 40 years3, highlighting the urgency for understanding risk factor.

Iron as an essential trace element has very important biological functions in the body7, is an essential micronutrient for many biochemical processes such as oxidative energy metabolism, electron transfer reactions, gene regulation, binding and transport of oxygen8, it plays a pivotal role in cell survival and proliferation9. Iron in the body is in a dynamic balance of constant absorption, utilization, storage, recycling, namely iron homeostasis10. Iron dyshomeostasis can lead to diseases related to iron metabolism, resulting in damage of organism, including iron deficiency and iron overload11. Several studies reported that iron deficient caused anemia, neurocognitive dysfunction, and impaired functional capacity energy metabolism abnormality12,13,14, whereas iron overload resulted in osteoporosis, neurodegeneration, cardiovascular diseases and hepatic disease15,16,17,18. Iron overload is prone to occur with age19. Ferritin, the primary protein responsible for iron storage20, often exhibits increased levels as one ages21. While the majority of iron is stored in the liver and spleen22, skeletal muscle also contains iron, albeit in smaller amounts. Excessive iron in the body, or iron overload, could potentially have adverse effects on skeletal muscle health23.

Iron status and its relationship with sarcopenia have been investigated in previous studies, yielding inconsistent findings. For instance, an observational study reported lower serum iron levels in individuals with sarcopenia compared to those without sarcopenia24. Conversely, other studies have demonstrated a significant association between serum ferritin and transferrin saturation with reduced grip strength, while serum iron did not show such an association25,26. These discrepancies in previous findings may be attributed to limitations inherent in observational studies27, including susceptibility to potential confounding factors and the challenge of establishing causal relationships between iron status and sarcopenia.

Mendelian randomization (MR) is a method that can overcome problems of unmeasured confounding and reverse causation typical of conventional observational epidemiology28, it assess causal inference of an exposure on an outcome by using genetic variants as instrumental variables for the exposure29. genetic variants are randomly allocated at conception, so they can be exploited to simulate randomization30.

The aim of this study is to employ a bi-directional two sample Mendelian randomization (MR) analysis, utilizing four iron-related biomarkers as clinical indicators of iron status, to comprehensively assess the causal association between iron status and sarcopenia. Additionally, a multivariable Mendelian randomization (MVMR) analysis will be conducted to determine the potential dominant role of any specific trait in the causation process. To our current knowledge, a paucity of Mendelian randomization (MR) investigations exists to scrutinize the causal effects pertaining to iron status and sarcopenia.

Materials and methods

Overall study design

In this MR study, we utilized a series of analyses approach to investigate the association between iron status and sarcopenia. Univariable Mendelian randomization (UVMR) analysis and Multivariable Mendelian randomization (MVMR) analysis were employed. The analysis was conducted using summary-level data from published genome-wide association studies (GWASs), and all studies included in cited GWASs had been approved by relevant review boards and obtained informed consent from participants. The present MR analyses were conducted in accordance with the STROBE-MR guidelines31. The study design is illustrated in Fig. 1.

Figure 1
figure 1

An overview of the study design includes three key assumptions.MR satisfies the following three conditional assumptions32: (1) there is a strong association between instrumental variables and exposure factors; (2) no confounding factors exist in the association between exposure and outcome, in other words, there is no genetic pleiotropy; and (3) the instrumental variables do not have a direct effect on outcome and can only influence outcome through exposure factors.

Full size image

Data sources

The source of exposure data for our study on iron status-related indicators was obtained from a meta-analysis of three genome-wide association studies from Iceland, the UK, and Denmark33. The GWAS summary data included blood levels of ferritin (N = 246,139), total iron binding capacity (N = 135,430), iron (N = 163,511), and transferrin saturation (N = 131,471). These four iron-related biomarkers were each rank-based inverse normal transformed to a standard normal distribution, separately for each sex, and adjusted for age using a generalized additive model. Furthermore, the UK cohort was adjusted for menopausal status, ABO blood group, BMI, smoking levels, alcohol levels, and iron supplementation status.

Summary genetic association estimates for sarcopenia were obtained from MRCIEU GWAS database. We used appendicular lean mass (N = 450,243), right hand grip strength (N = 461,089), left hand grip strength (N = 461,026), low hand grip strength (60 years and older) (N = 256,523) and walking pace (N = 459,915) as genetically predicted sarcopenia-related traits. In particular, low hand grip strength, which is defined by the European Working Group on Sarcopenia in Older People (EWGSOP) as grip strength < 30 kg in men and < 20 kg in women, was used as a measure of sarcopenia in a meta-analysis comprising 256,523 individuals aged 60 years or older from 22 independent cohorts of European descent, including the UK Biobank, the US Health and Retirement Study, and the Framingham Heart Study, among others, with 18.9% of participants (46,596 individuals) exhibiting muscle weakness. Details of the included traits are displayed in Table 1.

Table 1 Details of studies and datasets used in the study.
Full size table

As both exposure and outcome data were partially obtained from UKBs, we assessed sample overlap rate and type 1 error rate using a web-based application (https://sb452.shinyapps.io/overlap/34 to ensure the validity of our results. The results showed that the overlap rate in this study ranged from 8.7 to 15.6%, and all the type 1 error rates were less than 0.05, which suggested that our subsequent analyses were reliable and robust.

Instrumental variable selection

To meet the relevance assumption, the first of the three key assumptions, instrumental variants should be associated with the exposure factors28. The single nucleotide polymorphisms (SNPs) associated with exposures were extracted at a genome-wide significance level (p < 5 × 10–8) from the GWAS datasets35. Afterwards, independent SNPs for exposures were obtained by linkage disequilibrium (LD) clumping with a threshold r2 < 0.001 and an allele distance > 10,000 kb36. We then extracted the SNPs and corresponding statistics from the GWAS dataset of outcomes, removing the SNPs with a minor allele frequency (MAF) < 0.0137. We employed proxy SNPs with a high correlation coefficient (R2 > 0.8) as a substitute for the missing SNPs. Further, we harmonized the data by removing all palindromic SNPs38. To fulfill the second MR assumption, we inquired for each IV and its proxy traits referring to PhenoScannerV2 database (http://www.phenoscanner.medschl.cam.ac.uk/) and discarded the SNPs surrogating for these confounding traits at a threshold of r2 > 0.8039,40. Accordingly, these rigorously selected SNPs were used as IVs for the following analyses.

Instrumental strength

We computed the proportion of phenotypic variation that is explained by all SNPs (i.e., R2-values) in our MR analysis using the formula R2 = 2 × β2 × EAF × (1 − EAF)/[2 × β2 × EAF × (1 − EAF) + SE2 × 2 × N × EAF × (1-EAF)] where β represents the effect estimate of the genetic variant in the exposure GWAS and EAF is the Allele 1 frequency, SE is the standard error and N is the sample size41,42. Then we calculated F-statistic to evaluate the instrumental strength of our SNPs for each trait in explaining phenotypic variation using the formula F = [(N − k − 1)/k] × [R2/(1 − R2)]43 where N is the sample size, k is the total number of SNPs that are selected for MR analysis, and R2 is the total proportion of phenotypic variation that is explained by all the SNPs in our MR analysis. An F-statistic > 10 suggests that the combined SNPs is a sufficiently strong instrument to explain phenotypic variation, while a F-statistic ≤ 10 implies a weak instrument43.

Statistical analysis

Univariable Mendelian randomization analyses

We undertook a bi-directional MR study to estimate the causal associations between four iron status and sarcopenia-related traits using three MR methods, inverse variance weighted (IVW), MR-Egger, weighted median (WM)44. The IVW method uses a meta-analysis approach to combine the Wald ratios of the genetically causal effects of each SNP, relying on the assumption that all SNPs are valid IVs with no evidence of directional pleiotropy37. So, it is considered to provide an estimate with the highest power and the best precision and is used as major analysis45,46. Considering the acknowledged variances in iron homeostasis across genders47, we applied UVMR to extend our inquiry into the correlation between the four iron status indicators and appendicular lean mass (ALM), leveraging available gender-stratified datasets. We calculated a Bonferroni-corrected p threshold, by dividing 0.05 by the number of tests, which assumes each test is independent48,49,50. In this study, the Bonferroni-corrected p threshold for both forward and reverse Mendelian randomization analyses are 0.0025(0.05/20). We considered a p value less than Bonferroni-corrected p threshold as being statistically significant51, and that larger than Bonferroni-corrected p threshold but less than 0.05 was suggestive of statistical significance in the univariable MR analysis52. Odd ratios (ORs) and corresponding 95% confidence intervals (CIs) were calculated for estimating causal effects of iron status on low hand grip strength.

Heterogeneity, pleiotropy, and sensitivity analysis

We applied the Pleiotropy RESidual Sum and Outlier (MR-PRESSO) analysis53 to provide outlier-adjusted estimates of causal associations. This involved removing one or more pleiotropic outlying SNPs and re-conducting the MR analyses. To detect whether the observed causal estimates were biased by reverse causality, we applied Steiger filtering. Furthermore, in order to assess potential heterogeneity and pleiotropy biases, we conducted heterogeneity, pleiotropy, and sensitivity analyses. The Cochran's Q test was used to evaluate heterogeneity between instrumental variables in the MR, with random-effect models used if the p value of the Cochran's Q test was less than 0.0554. We also performed leave-one-out sensitivity analyses to assess the influence of each SNP on the overall MR estimates55. If one or more SNPs were found to significantly alter the overall MR estimates, it would be removed and the MR analyses were re-performed. Lastly, we used the MR-Egger intercept method, specifically the mr_pleiotropy_test function in R TwoSampleMR package, to evaluate the pleiotropy of our effect estimates.

MVMR analysis

Given the interrelated nature of the four iron biomarkers established in prior research33, it was imperative to conduct MVMR analysis to elucidate the primary drivers behind the causal associations observed between iron-related biomarkers and sarcopenia-related traits. Unlike UVMR analysis, MVMR analysis assumes that the IVs are strongly associated with at least one exposure, although not necessarily with each. In the forward analysis, we excluded TIBC from the subsequent MVMR analysis due to collinearity with other iron-related biomarkers. Instead, we employed a combination of SNPs as an integrated proxy for the three iron-related biomarkers, ensuring convergence in our analysis.

To address potential confounding factors, particularly the reduction in physical activity (PA) associated with anemia in individuals with iron deficiency, we conducted additional analyses utilizing MVMR. Specifically, each of the four iron status indicators was adjusted for physical activity. Details of the physical activity data source are provided in Table 1.

Statistic power

Moreover, we used a webpage-based application, the online sample size and power calculator (https://sb452.shinyapps.io/power/), to estimate the statistical power for detecting causal effects of iron status on sarcopenia-related traits56. The power calculator uses simulations to estimate the power for a given set of parameters, providing researchers with valuable information for designing MR studies with sufficient statistical power.

Statistical tools

All statistical analyses and visualization for results were performed using R statistical software (version 4.2.2, R Foundation for Statistical Computing, Vienna, Austria; https://www.R-project.org) with the TwoSampleMR, LDlinkR, presso, and forestplot Packages.

Ethics statement

Ethical approval and consent to participate in the original genome-wide association studies (GWASs) were obtained from relevant review boards.

Results

Instrumental variables for iron status

We obtained 48 sets of SNPs serving as IVs when performing the UVMR analysis (Supplementary Table S1). We calculated the F-values of 48 sets of SNPs and found that they ranged from 38.1 to 521.5 (Tables 2, 3, Supplementary Table S2), which suggests that there is no bias caused by weak instrumental variables in this study.

Table 2 Univariable Mendelian randomization estimates of iron status on sarcopenia-related traits.
Full size table
Table 3 Univariable Mendelian randomization estimates of sarcopenia-related traits on iron status.
Full size table

UVMR analysis

As shown in Fig. 2, genetically predicted ferritin has a significant causal effect on appendicular lean mass (β = − 0.051, 95% CI − 0.072, − 0.031, p = 7.325 × 10–07), indicating that for each standard deviation (SD) increase in ferritin levels, there is an associated decrease in ALM by approximately 0.051 kg. This effect was observed with a high level of statistical power (95.6%). In contrast, no significant causal effects were observed for ferritin on the other four traits of sarcopenia. The associations were still significant after Bonferroni correction (p < 0.0025).

Figure 2
figure 2

Forest plot for causal associations of iron status with ALM in UVMR analysis. Ferritin, serum iron, TIBC, TSAT are scaled to an SD increase. Effects (beta) represents change in kg ALM. IVW inverse variance weighted, WM weighted median, ALM appendicular lean mass, TIBC total iron binding capacity, TSAT transferrin saturation, CI confidence interval, UVMR univariable Mendelian randomization.

Full size image

For TIBC, our observational analysis revealed a negative association with appendicular lean mass (β = − 0.020, 95% CI − 0.037, − 0.002, p = 0.028), though with modest statistical power (47.3%). Our reverse Mendelian randomization study also found a negative TIBC-appendicular lean mass correlation (β = − 0.031, 95% CI − 0.052, − 0.010, p = 0.004) with higher power (97.3%). Additionally, TSAT was positively associated with appendicular lean mass (β = 0.022, 95% CI 0.002, 0.043, p = 0.035) with 51.3% power, indicating its potential protective role. Notably, the p-values for these associations fell between the Bonferroni-corrected threshold and 0.05, suggesting that additional studies with larger sample sizes are needed to confirm the observed effect.

No significant causal associations were observed between the remaining iron status and sarcopenia-related traits in both the forward and reverse MR analyses. The results regarding causal associations between the four iron-related biomarkers and sarcopenia-related traits by UVMR analyses based on three MR methods are demonstrated in Tables 2 and 3. Upon stratifying the dataset by gender, our investigation revealed no statistically significant associations between four iron status indicators and ALM (Supplementary Table S2).

Heterogeneity, pleiotropy, and sensitivity analysis

We obtained estimates that were consistent with our original results after removing outliers detected by MR-PRESSO analyses, demonstrating the stability of our findings after correcting for the presence of pleiotropic effects. Our investigations employing steiger filtering revealed no presence of reverse causation among the examined SNPs, ensuring the reliability of the inferred causal direction. In addition, we evaluated the potential impact of pleiotropy by utilizing the MR-Egger intercept, which revealed no indication of any such influence on our estimates. However, we noted moderate heterogeneity in the analysis of low hand grip strength by TIBC. Moreover, in our gender-stratified analyses, heterogeneity was observed in several associations: Serum Iron (female) and TIBC (female) with ALM (female), as well as TSAT (female) with ALM (female). Additionally, heterogeneity was evident in the analysis of ferritin (male) with ALM (male) (Supplementary Tables S3–S5). Furthermore, our leave-one-out sensitivity analyses did not reveal any significant changes in effect estimates when any one SNP was removed (Supplementary Figs. S1–S47), suggesting that our findings were not driven by any one particular SNP.

MVMR analysis

The MVMR analyses using the IVW method demonstrated a significant inverse association between higher ferritin levels and ALM (β = − 0.068, 95% CI − 0.12, − 0.017, p = 9.658 × 10–03), as illustrated in Fig. 3 and Table 4. Interestingly, this effect appears to be the predominant driver of the associations observed between iron status and sarcopenia-related traits, as adjustment for serum iron (β = − 0.019, 95% CI − 0.095, 0.057, p = 0.623) and TSAT (β = 0.051, 95% CI − 0.006, 0.108, p = 0.078) had negligible impact on the observed effect. Consistent with our UVMR findings, no significant associations were found between iron status and hand grip strength (left or right), low hand grip strength, or walking pace in our MVMR analyses.

Figure 3
figure 3

Forest plot for causal associations of iron status with ALM in MVMR analysis. Ferritin, serum iron, TSAT are scaled to an SD increase. Effects (beta) represents change in kg ALM. ALM appendicular lean mass, TSAT transferrin saturation, IVs numbers of instrumental variable, CI confidence interval, MVMR multivariable Mendelian randomization.

Full size image
Table 4 Causal effects of multiple iron status on sarcopenia-related traits based on IVW MVMR model.
Full size table

After adjusting for PA, the analysis revealed that each SD higher genetically predicted ferritin was associated with lower ALM (β = − 0.054, 95% CI − 0.092, − 0.015, p = 0.006). However, TIBC (β = 0.033, 95% CI 0.072, 0.005, p = 0.090) and TSAT (β = 0.011, 95% CI − 0.008, 0.03, p = 0.262), which demonstrated suggestive statistical significance with ALM in UVMR analysis, did not exhibit significance with ALM after correction for the PA (Supplementary Table S6). All the data used in MVMR analysis are detailed in Supplementary Table S7–S11.

Discussion

In this study, we employed a comprehensive analytical approach to investigate the relationship between iron status and sarcopenia. By analyzing pooled data from genome-wide association studies (GWAS) conducted on European populations, our study aimed to establish a causal relationship between iron status and sarcopenia-related traits. While our sex-subgroup analysis did not reveal any association between the four iron status indicators and ALM, our analysis of the overall dataset, conducted through UVMR and MVMR, consistently demonstrated that genetically predicted serum ferritin levels exerted a significant causal effect on ALM.

Our UVMR and MVMR analyses provide evidence that increased serum ferritin levels may have a detrimental causal effect on ALM. This finding is in line with previous observational studies showing iron overload associates with adverse muscle outcomes26,57,58. About 30% of the body's iron is stored in the form of ferritin or hemosiderin, so serum ferritin is a good indicator of the body's iron reserves59. High ferritin levels indicate iron overload and saturation of transferrin, allowing non-transferrin bound iron to accumulate and catalyse reactive oxygen species generation. Oxygen-free radicals could cause mitochondrial RNA peroxidation, which further induces the opening of mitochondrial permeability transition pores (mPTP), leads to cytochrome C release into the cytoplasm, caspase-3 activation, and finally skeletal muscle cell apoptosis60,61,62. In addition, some researchers have found that iron overload may affect the function of muscle satellite cells through ferroptosis, affecting the repair of damaged skeletal muscle63. Additionally, animal research showed that with increase in iron load in skeletal muscles, skeletal muscle mass decreased while muscle cells were atrophied, Akt-FOXO3 was activated, and atrogin-1 and MuRF1 (ubiquitination marker genes associated with muscle cell atrophy) levels were upregulated64. All of the above underlying mechanisms could explain the relationship between serum ferritin and skeletal muscle mass in the extremities. The lack of associations for handgrip strength, walking pace and ferritin also implies that higher ferritin may preferentially induce muscle mass loss rather than strength or physical performance decline.

We observed that TIBC exhibited a potential risk association with sarcopenia, leading to a negative impact on appendicular lean mass. On the other hand, our reverse MR result indicates that decreased appendicular lean mass could elevate TIBC. One possible explanation is that under normal circumstances, ferritin can be disassembled by autophagy, releasing iron for cellular processes65. Autophagy is impaired in skeletal muscles with aging66. Inappropriate sequestration of iron into ferritin, or a failing in the breakdown of ferritin that ultimately reduces the available free iron in the cell, causing functional iron deficiency, which then affects the normal energy metabolism of skeletal muscle, leading to skeletal muscle atrophy3. The above underlying mechanisms could similarly explain the significant association of reduced TSAT with the decrease in appendicular lean mass. This is because elevated serum ferritin associated with decreased TSAT is often typical of functional iron deficiency. Notably, under the Bonferroni correction significance level, no correlation was observed between TIBC and TSAT with appendicular lean mass. These findings imply that future investigations should include larger GWAS datasets and consider conducting meta-analyses using data from multiple sources to provide further insights into the relationship between these variables.

Interestingly, no significant associations were found between serum iron and sarcopenia traits in our study. To date, limited research has been conducted to investigate this particular relationship. Bartali et al.67 conducted a longitudinal study involving 698 participants but failed to identify a significant link between serum iron levels and physical function. Similarly, a prior systematic review68 also failed to demonstrate a significant relationship between serum iron and sarcopenia. A reason could be that these markers reflect iron availability in the short term, while ferritin indicates long-term iron storage and may better predict chronic health risks.

The present study possesses several notable strengths. To the best of our knowledge, this study represents the first attempt to explore the causal associations between iron status and sarcopenia using Mendelian randomization, leveraging large-scale genome-wide association study (GWAS) data. The implementation of MR design stands as a significant strength, as it effectively mitigates residual confounding and other biases, thereby enhancing the strength of causal inferences drawn69. Our employment of UVMR and MVMR analyses surpasses previous observational studies, as we have leveraged summary data derived from GWASs featuring an extensive sample size and a vast number of SNPs. Furthermore, the outcomes obtained are characterized by robustness and reliability, demonstrated by the absence of heterogeneity or pleiotropic effects.

However, several limitations are inherent in our study. Primarily, the genetic variant data primarily relied upon GWASs conducted on individuals of European descent, which may restrict the generalizability of our findings to the broader population. Nonetheless, the restriction of participant descent serves to minimize the potential confounding effects stemming from population admixture. Secondly, it is important to note that while efforts were made to calculate type 1 error rates below 0.05, the possibility of weak instrumental variable bias resulting from sample overlap could not be entirely eliminated, as both exposure and outcome data were partially obtained from UKBs. Lastly, iron deficiency and iron overload may have distinct effects on muscle mass, and understanding these differences is vital for accurate interpretation. However, the inability to assess non-linear relationships hampers our understanding of the underlying mechanisms driving the observed associations. The use of summary-level data limits our ability to capture potential threshold or saturation effects, as it only allows for the estimation of average linear causal effects.

Conclusion

In summary, our MR study offers novel insights into the potential role of elevated serum ferritin as a causal factor associated with decreased appendicular lean mass. While our results might suggest that strategies to reduce ferritin levels could potentially influence muscle atrophy, it is important to note that these findings are preliminary. Obtaining more detailed data in the future is necessary to conduct nonlinear analyses and elucidate the relationship between iron status and sarcopenia further.