A two-sample MR was utilized to assess the causal association of glucosamine on cognition and sarcopenia-related traits. Briefly, glucosamine served as the exposures, while two indicators of cognition and seven sarcopenia-related traits served as the outcomes. This two-sample MR study consisted of two phases of analysis. In the first phase, we first examined the causal effects of glucosamine on 2 cognitive traits and 7 sarcopenia-related traits. Then, we investigated whether these causal effects were independent of OA and BMI. In the second phase, we assessed the mediating roles of CRP and BMR in the causal associations between glucosamine and cognitive function outcomes and sarcopenia-related traits. MR was performed under three crucial assumptions (1). genetic IVs are associated with glucosamine (the relevance assumption) (2); genetic IVs are independent of confounding variables (the independence assumption) (3); genetic IVs only affect outcomes above through glucosamine (the exclusion restriction assumption) (21).

This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) (22). Summary statistics were obtained from the reputable publicly available GWAS conducted by European participants. Initially, mendelian randomization (MR) analyses. The included research and databases obtained the ethical committee’s approval and informed consent from participants. The GWAS datasets utilized in this investigation are detailed in Supplementary Table 1 . The ethical approval for the GWAS studies can be found in the referenced studies provided below. As instrumental variables (IVs), single-nucleotide polymorphisms (SNPs) strongly associated with glucosamine at genome-wide significance were selected (p < 5 ×10^-6). The flowchart is shown in Figure 1 and Figure 2 .

Genetic variants significantly associated with glucosamine were extracted from the largest Open GWAS data of the Medical Research Council Integrative Epidemiology Unit (MRC-IEU) consortium, which comprised 89,339 cases and 372,045 controls, available through the UK Biobank (https://gwas.mrcieu.ac.uk/datasets/ukb-b-11535/). A baseline touchscreen questionnaire was used to acquire habitual glucosamine information. The participants were asked, “Do you regularly take any of the following?” The touchscreen questionnaire allowed participants to pick various answers from a variety of Supplementary Materials (Mineral and other dietary supplements). We classified glucosamine users as follows: 0 no; 1 yes.

Genetic variants associated with glucosamine at genome-wide significance (p < 5×10^-6) were identified and extracted as potential IVs. Then, we performed the clumping procedure (linkage disequilibrium r2 < 0.001 within kb = 10,000) to ensure that genetic variants were independent of each other. To assess weak instrumental bias, the F statistics were calculated using the following formula: F = R² (N–k–1)/[(1–R²) k], where N represents the sample size and k denotes the number of included SNPs, R² represents the proportion of variability that can be attributed to each SNP. Besides, R² can be calculated by the formula: R² = 2 × β² × EAF × (1 – EAF), where β indicates the estimated effect size of the IVs and EAF represents the effect allele frequency. A genetic variation used as a weak IV is considered when the F-statistic is less than 10, as it can potentially introduce bias to the results. Details of GWAS studies are provided in the Supplementary Table 1 .

Cognition outcomes included cognitive performance and fluid intelligence score (FIS) traits. Cognitive performance data were extracted from the largest publicly available GWAS in 257,841 European-ancestry participants from COGENT Consortium (35 sub-cohorts) and UK Biobank, with adjustment for sex, age, and population stratification (23). GWAS data for FIS data were from the Medical Research Council Integrative Epidemiology Unit (MRC-IEU) consortium, available through the UK Biobank. Seven traits were selected for valid predictors of sarcopenia: appendicular lean mass (ALM), whole body fat-free mass (WBFM), low hand grip strength, facial aging (FA), moderate to vigorous physical activity levels, usual walking pace and DNA methylation GrimAge acceleration (24). Among them, grip strength was used to measure muscle strength, and WBFM and ALM were used to measure muscle mass. ALM is predominantly determined by skeletal muscle and has been shown to make up ≥75% of skeletal muscle in the body, whereas WBFM comprises smooth muscle, cardiac muscle, and skeletal muscle (25). The GWAS meta-analysis of 256,523 Europeans aged 60 years and older from 22 cohorts was utilized to extract grip strength data; the criteria for this analysis were established by the European Working Group on Sarcopenia in Older People (EWGSOP) consortium (grip strength <30 kg Male; <20 kg Female) (26).

FA serves as a prominently visible indicator of aging. The summary statistics for facial aging of GWAS were obtained from the UK Biobank, which contained a total of 423,999 participants of European ancestry (194,391 males and 229,601 females, aged 40–69). Facial aging (FA) is evaluated with non-subjective perceived age through a touchscreen questionnaire. 103,300 individuals reported appearing to be their age. A total of 312,062 individuals reported appearing younger than their biological age, while 8,630 individuals reported appearing older than their biological age. The participants were asked, “Do people say that you look?” and allowed to pick the answer from a variety of Supplementary Materials . The researchers coded the participants in accordance with their actual age and perceived age. Participants were assigned codes as follows: 0 indicated that they appeared older, 1 stated that they appeared younger, and 0.5 indicated that they appeared to be their age (third-party observations, including those of non-participants and non-researchers, were conducted without the knowledge of the participant’s actual ages). The summary statistics for moderate to vigorous physical activity levels were derived from a GWAS conducted on 377,234 UK Biobank participants (27). Just as low physical performance is one of the characteristics of sarcopenia, clinical walking speed is a fast, safe, and highly reliable test for sarcopenia. Usual walking pace data were obtained from the UK Biobank with 459,915 European individuals. DNA GrimAge acceleration is a measure of accelerated biological aging. It is established by combining chronological age, sex, and surrogate biomarkers based on DNA methylation for seven plasma proteins and smoking pack-years. The data for DNA methylation GrimAge acceleration was derived from McCartney et al. GWAS study with 34,467 subjects (28). Details of GWAS studies are provided in the Supplementary Table 1 .

Oral glucosamine has traditionally been advised for managing knee and hip OA (29). Therefore, OA was selected and identified as a major confounder. Summary statistics for OA were obtained from the largest publicly available GWAS with 484,598 participants (39,515 OA cases and 445,083 controls), which utilized the UK Biobank data (30). We also assessed the effect of body mass index (BMI) as a confounder. BMI data were derived from the largest publicly available GWAS by the Genetic Investigation of Anthropometric Traits (GIANT) consortium, which included 681,275 European individuals (31). C-reactive protein (CRP) levels may be involved in altered cognitive function and accelerated aging. Besides, the development of sarcopenia usually brings about a decrease in the basic metabolic rate (BMR) of the organism and a decrease in trunk activity. Therefore, CRP and BMR were selected and identified as the mediators in the effect of glucosamine on cognition and sarcopenia‐related traits. CRP data were obtained from the largest publicly available GWAS with 575,531 participants of European ancestry, which was conducted in the UK Biobank participants (427,367 European descent) and the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium (575,531 European descent) (32). Summary statistics for basal metabolic rate were derived from the MRC-IEU consortium, available through the UK Biobank, which included 454,874 European participants. Details of GWAS studies are provided in the Supplementary Table 1 .

MR estimates were reported as odds ratios (ORs) with 95% confidence intervals (CIs). We evaluated the causal relationship of glucosamine with two cognition traits and seven sarcopenia-related traits by applying univariable MR analysis (UVMR), with estimates represented as β0. Then, we applied multivariable MR (MVMR) to evaluate the causal effect of glucosamine on cognition and sarcopenia-related traits with both adjustments for OA and BMI to determine whether glucosamine was causally associated with the cognition and sarcopenia-related outcomes independent of OA and BMI.

We applied fixed effects inverse variance weighted (IVW-FE) as the major analysis for UVMR. To enhance the robustness of IVW-FE results, we also employed MR Egger, maximum likelihood, weighted median, and weighted mode methods for additional validation. The MR-Egger method aids evaluating horizontal pleiotropy and bias caused by weak instruments (33). A weighted median approach could effectively mitigate the impact of outlier SNPs on the results, thereby enhancing robustness (33). The weighted mode approach provides dependable estimates when the majority of similar individual-instrument causal effect estimates are derived from genuine instrumental variables, even if a large portion of them is invalid (33). Sensitivity analyses were conducted to assess potential biases and underlying assumptions in the MR analysis, aiming to validate and refine the results. Heterogeneity was assessed using the Cochran Q test. There is no heterogeneity if the p-value of Cochran’s Q test is greater than 0.05. Once heterogeneity (p < 0.05) has been identified, use the multiplicative random effects IVW (IVW-MRE) method to ascertain the causal influence. Besides, the MR pleiotropy residual sum and outlier (MR-PRESSO) test and MR-Egger were performed to assess possible horizontal pleiotropy. Horizontal pleiotropy is described as some instruments’ additional biological effects that influence the outcome independently of the exposure. The presence of horizontal pleiotropy can be detected using the MR-PRESSO global test, while the MR-PRESSO outlier test is utilized to exclude outlier SNPs and assesses the impact of excluding these SNPs on causal estimations (34). The MR-Egger intercept test is additionally employed to assess the horizontal pleiotropy (35). A leave-one-out method was employed to identify high-influence points and whether a particular SNP influenced the significant results.

Furthermore, the confounding analysis utilized Phenoscanner V2 (http://www.phenoscanner.medschl.cam.ac.uk/) (36). We have identified certain diseases or physical ailments that correlate highly with IVs at a significance level of p < 5 × 10^-6. Subsequently, we condensed and examined pertinent data pertaining to IVS, GWASes, and illnesses. This approach not only assists in identifying important factors that need to be accounted for in MVMR, but it also enables us to explore the process of mediation and potential causal pathways.

MVMR provides insights into the extent to which the mediator explains the relationship between the exposure and the outcome (21). In addition, MVMR allows for the simultaneous examination of exposure while considering potential confounding factors. The IVW method was used for the MVMR. Moreover, we performed two‐step MR to explore whether CRP levels and BMR mediate the causal associations between glucosamine and cognition and sarcopenia‐related outcomes. The first step was to use UVMR to estimate the causal effect of glucosamine on mediators (CRP and BMR, separately), with each estimate represented as β1. The second step was to estimate the causal effect of the mediators on outcomes (cognition and sarcopenia‐related traits), with adjustment for glucosamine using MVMR (MVMR estimate represented as β2). The entire effect of glucosamine on outcomes can be split into direct (not mediated by mediators) and indirect effects (effect mediated by the mediators). The following conditions exist: 1) If the coefficients β0, β1, and β2 are all statistically significant, it suggests that a causal relationship between the exposure and outcome variables. Furthermore, this relationship may be influenced to some extent by mediating factors. The mediation proportion of mediator (CRP and BMR) in the causal association between glucosamine and outcomes was calculated as the product of β1 and β2 divided by β0; 2) If the coefficient β0 is not statistically significant, yet the coefficients β1 and β2 are both statistically significant, it suggests that the mediator variable mainly influences the impact; 3) If the coefficient β0 is statistically significant and either β1 or β2 is not statistically significant, then indicates that the mediating effect does not play a role in mediating the relationship between exposure and the result (37).

A significance level of 0.05 was implemented for the p-value. To adjust for multiple testing (several exposures), the MR effect estimates were statistically significant at less than five percent using the Benjamini-Hochberg false discovery rate (FDR). All analyses were carried out using packages “TwoSampleMR,” “MendelianRandomization,” “MR-PRESSO,” and “Phenoscanner” in R version 4.3.1.

After analyzing and evaluating the related information about glucosamine-associated SNPs, GWASes, and Diseases through Phenoscanner, we found some potential confounders or outcomes, mainly including arm fat-free mass, arm predicted mass, chronotype, morning or evening person, forced expiratory volume in 1-second, forced vital capacity, and qualifications, etc. ( Figure 4 ).

We performed MVMR Analysis by both adjusted OA and BMI to further assess the associations between glucosamine and outcomes. After adjusting both OA and BMI, we found that taking glucosamine continued to have a significant effect on cognitive performance (OR = 1.87, 95% CI = 1.23-2.83, p = 3.42E-03), FIS (OR = 4.88, CI = 1.87-12.73, p = 1.18E-03), ALM (OR = 3.59, 95% CI = 2.29-5.63, p = 2.66E-08), WBFM (OR = 1.66, 95% CI = 1.33-2.08, p = 7.38E-06), usual walking pace (OR = 1.53, 95% CI = 1.34-1.76, p = 9.46E-10), and moderate to vigorous physical activity levels (OR = 1.40, 95% CI = 1.12-1.74, p = 3.25E-03) ( Supplementary Table 9 ).

We further explored the mediating roles of CRP and BMR in the improvement of cognition and sarcopenia‐related traits by glucosamine using MVMR coupled with a two-step MR method, respectively ( Supplementary Table 10 ). The protective effects of glucosamine on ALM, WBFM, usual walking pace, and moderate to vigorous physical activity levels were continued after adjusting for CRP. However, the effects of glucosamine on cognitive performance, FIS, FA, DNA methylation GrimAge acceleration, and low hand grip strength were weakening after adjusting for CRP, suggesting this association was largely affected by CRP. The proportion mediated by CRP in the associations between glucosamine and outcomes was 4% for ALM, 1.4% for WBFM, 5% for usual walking pace, 3% for moderate to vigorous physical activity levels ( Table 1 ).

Moreover, the protective effects of glucosamine on cognitive performance, ALM, WBFM, usual walking pace, and moderate to vigorous physical activity levels were continued after adjusting for BMR. However, the effects of glucosamine on FIS, FA, and DNA methylation GrimAge acceleration were weakening after adjusting for BMR, suggesting this association was largely affected by BMR. Specifically, the proportion mediated by BMR in the associations between glucosamine and outcomes was and 3% for cognitive performance, 3% for FIS, 39% for ALM, 71% for WBFM, and 3% for usual walking pace ( Table 2 ).