Gut microbiota mediates the relationship between smoking and fall risk: A 2-step Mendelian randomization study

Abstract

Falls represent a critical public health challenge for the global elderly population, leading to severe health consequences such as fractures, strokes, and even death, while imposing a substantial burden on the healthcare system. While previous studies have identified lifestyle factors (e.g., smoking, alcohol consumption, sleep, and physical activity) as contributors to fall risk, the causal role of the gut microbiota and its mediating mechanisms remain unclear. Mendelian randomization (MR) was employed to analyze the potential mediating effects of 131 gut microbial taxa (genus level) on the associations between smoking, alcohol consumption, sleep, vigorous exercise, and fall risk. First, a 2-sample MR analysis using genome-wide association study summary data was conducted to establish causal links between gut microbiota and fall risk. Subsequently, a 2-step MR approach was applied to investigate the mediating role of gut microbiota in the smoking–fall risk relationship. Smoking significantly increased fall risk (OR = 1.11, P = 8.99 × 10^-4), while alcohol consumption, sleep, and vigorous exercise showed no significant causal associations. Notably, the Ruminococcus gnavus group (OR = 1.04, P = 2.20 × 10^-3) was identified as a risk-enhancing gut microbiota, while Allisonella (OR = 0.97, P = 5.00 × 10^-4) exhibited protective effects. Further analysis confirmed that the R gnavus group mediated 8.63% of the smoking–fall risk association. This study provides the first genetic evidence supporting the mediating role of the gut microbiota, particularly the R gnavus group, in the relationship between smoking and fall risk. These findings offer novel scientific insights for the development of gut microbiota-targeted strategies to prevent falls.

1. Introduction

Falls are a significant global public health issue affecting the health of older adults and specific high-risk populations, imposing a heavy burden on the healthcare systems and socioeconomic resources. According to the World Health Organization, falls account for 684,000 annual deaths globally, ranking as the second leading cause of mortality due to unintentional injury, with incidence rates increasing progressively with age. Severe falls can result in critical injuries, including fractures, strokes, traumatic brain injuries, and fatalities.^[1–6] Additionally, the psychological impact of falls may lead to approximately 40% of older adults voluntarily reducing physical activity, triggering a further decline in physical function, forming a vicious cycle that elevates the risk of recurrent falls.^[7]

In recent years, growing evidence suggests that the gut microbiota is not only essential for digestive health but also plays a key role in neuromuscular function and frailty, potentially modulating fall risk.^[8] A randomized trial has shown that multistrain probiotics reduced falls in patients with liver cirrhosis. Furthermore, frail individuals commonly exhibit gut dysbiosis, characterized by reduced beneficial bacteria (e.g., Prevotella, Bacteroides, and Eubacterium) and increased pathogenic species (e.g., Enterobacteriaceae, Bacteroides fragilis).^[9,10] However, the causal link between gut microbiota and fall risk remains unclear, and microbiota-targeted strategies for fall prevention lack robust evidence.

Several modifiable lifestyle factors, such as smoking, alcohol use, physical inactivity, and poor sleep, are associated with fall risk.^[11,12] These factors may influence gut microbiota composition, suggesting that the microbiota could serve as a mediator between lifestyle and fall outcomes. Nonetheless, mechanistic insights into this pathway are still limited. To address this gap, we employed Mendelian randomization (MR), a causal inference method using single nucleotide polymorphisms (SNPs) as instrumental variables (IVs), to minimize the confounding and reverse causality inherent in observational studies.^[13] As genetic variants are determined at conception, MR mitigates environmental interference and measurement errors, thereby enhancing causal inference robustness.^[14] This study utilized a 2-step MR approach to identify fall-related gut microbiota and elucidate the mediating network among modifiable factors, microbiota, and falls, thereby uncovering the underlying mechanisms and informing novel preventive strategies.

2. Methods

2.1. Overall study design

This study used genome-wide association study (GWAS) summary data and MR analysis to systematically investigate the mediating role of 131 gut microbial taxa (genus level) in the associations between smoking, alcohol consumption, sleep, physical activity, and fall risk (Fig. 1). Specifically, a 2-sample MR analysis was conducted to assess the causal effects of 131 gut microbial genera on fall risk. Next, MR was used to evaluate the association between lifestyle factors and fall risk. Finally, a 2-step MR analysis was performed to explore the mediating role of fall-related gut microbiota in the association between lifestyle factors and fall risk.

Figure 1.

(A and B) Study design schematic.

To ensure reliable causal inference in MR analysis, the selected IVs must satisfy 3 core assumptions: relevance: IVs are strongly associated with exposure; exclusion restriction: IVs influence the outcome solely through exposure, with no alternative pathways; independence: IVs are independent of confounders affecting the exposure–outcome relationship.^[15]

2.2. Data sources for lifestyle and gut microbiota GWAS

The gut microbiota GWAS data were sourced from the largest meta-analysis by the MiBioGen consortium,^[16] which integrated GWAS and 16S rRNA sequencing data from 18,340 participants across 11 countries (24 cohorts), covering 211 bacterial taxa (9 phyla, 16 classes, 20 orders, 35 families, and 131 genera). The genus represented the lowest taxonomic level, including 12 unclassified genera. A total of 119 genera were included in the analysis (details are provided in File S1: Table S1, Supplemental Digital Content, https://links.lww.com/MD/Q96). Lifestyle-related GWAS summary data^[17–19] are detailed in Table 1.

Table 1.

Data sources for lifestyle factors.

Trait	Sample size	Ancestry	Unit	PMID
Vigorous physical activity	261,055	European	≥ 3 vs 0 d/wk	29899525
Sleep duration	128,266	European	60 min	30846698
Smoking	607,291	European	SD in prevalence of smoking	30643251
Alcoholic drinks per week	335,394	European	Alcoholic drinks/week	30643251

PMID = PubMed ID.

2.3. Fall risk GWAS data

Fall-risk GWAS data were obtained from the UK Biobank (UKB) and FinnGen databases. The UKB dataset includes 89,076 cases and 362,103 controls (451,179 individuals of European ancestry).^[20] For replication, FinnGen data (“falls and fall tendency”; ICD codes: R29.6, W00-W19) comprising 102,835 cases and 335,679 controls (total 377,277 Europeans) were analyzed.^[21] Ethical approval or informed consent was waived as all data were obtained from public genetic repositories.

2.4. Selection of valid IVs

For lifestyle factors as exposures, SNPs were selected at P < 5 × 10^-8, with clumping thresholds of r² < 0.01 and a 1000 kb window to minimize linkage disequilibrium. For gut microbiota exposure, a relaxed threshold (P < 1 × 10^-5, r² < 0.01, 1000 kb clumping) was applied to ensure sufficient IVs.^[22] This approach is widely adopted in microbiome-related MR studies due to the limited number of genome-wide significant SNPs identified in current microbial GWAS. Although using a relaxed threshold may increase the risk of weak instrument bias, we addressed this concern by calculating the F-statistic for each SNP using the formula: F = [R² × (N − K − 1)]/[K × (1 − R²)], where N is the sample size of the exposure, K is the number of instruments, and R² is the proportion of explained variance. Palindromic SNPs and allele-mismatched SNPs between the exposure and outcome datasets were excluded. Only SNPs with F-statistics > 10 were retained, in line with conventional criteria to mitigate weak instrument bias.^[23]

2.5. Statistical analysis

Five MR models were employed for causal inference: inverse variance weighted (IVW), weighted median estimation, MR-Egger regression, simple mode, and weighted mode.^[24] IVW, with high statistical power, served as the primary method (fixed effects IVW by default; multiplicative random effects IVW if heterogeneity was detected), whereas the other 4 models provided supplementary analyses. In 2-step MR analysis, the relationships between exposure (X) to mediator (M) and from mediator (M) to outcome (Y) must be considered. The mediation effect was calculated as βxm × βmy, with the mediation proportion defined as (βxm × βmy)/ βxy.^[25]

2.6. Two-step MR analysis

When the gut microbiota serves as an exposure, the relaxed threshold for selecting IVs may compromise the reliability of MR-based causal inferences. To address this, we performed 2 MR analyses using independent GWAS datasets for fall risk (UKB and FinnGen), followed by a meta-analysis to enhance the robustness of the results. Subsequently, 2-sample MR was applied to assess causal links between lifestyle and fall risk, while 2-step MR examined the mediating role of identified fall-associated gut microbiota in these relationships.

2.7. Sensitivity analyses

To further validate the accuracy and stability of the findings, sensitivity analyses were conducted using heterogeneity tests, horizontal pleiotropy tests, and leave-one-out validation.^[26,27] Cochran Q test was used to assess heterogeneity among IVs. When heterogeneity was detected (P < .05), the multiplicative random effects IVW method was applied; otherwise, the fixed effects IVW method was used. Horizontal pleiotropy was estimated using the intercept of the MR-Egger regression. Leave-one-out analysis iteratively excludes each IV to evaluate the robustness of the collective effect of the remaining IVs, identifying whether MR associations are driven by individual SNPs.^[28]

3. Results

3.1. Selection of valid IVs

Following the IV selection criteria, GWAS data from 4 lifestyle factors and 119 gut microbial genera were filtered, yielding 1772 qualified IVs. All SNPs exhibited F-statistics of > 10, indicating no weak instrument bias. The details of the selected IVs are provided in File S1: Table S2, Supplemental Digital Content, https://links.lww.com/MD/Q96.

3.2. Causal effects of lifestyle factors on fall risk

Using a 2-sample MR, we explored the causal effects of 4 lifestyle factors (smoking, alcohol consumption, sleep, and vigorous physical activity) on fall risk. IVW analysis (Fig. 2) revealed that smoking significantly increased fall risk [OR = 1.11, 95% CI = 1.05–1.19, P = 8.99 × 10^-4], while weekly alcoholic drinks [OR = 1.09, 95% CI = 0.95–1.26, P = .23], sleep duration [OR = 0.98, 95% CI = 0.74–1.29, P = .89], and vigorous physical activity [OR = 1.03, 95% CI = 0.59–1.81, P = .92] showed no significant associations. The MR-Egger test results indicated no horizontal pleiotropy among the included IVs (File S1: Table S3, Supplemental Digital Content, https://links.lww.com/MD/Q96). Leave-one-out sensitivity analysis revealed that the sequential removal of individual SNPs did not substantially alter causal effect estimates (File S2: Figure 1, Supplemental Digital Content, https://links.lww.com/MD/Q97).

Figure 2.

Forest plots for Mendelian randomization (MR) analysis. (A) Forest plot of MR studies examining the impact of lifestyle on fall risk; (B) Forest plot of MR studies examining the impact of gut microbiology on fall risk.

3.3. Causal effects of gut microbiota on fall risk

The initial MR analysis identified 11 gut microbiota taxa significantly associated with fall risk (File S1: Table S3, Supplemental Digital Content, https://links.lww.com/MD/Q96, Fig. 3). Six microbial genera were associated with increased fall risk based on IVW estimates: Ruminococcus gnavus group (OR = 1.05, 95% CI: 1.02–1.09, P = 6.46 × 10⁻⁴), Desulfovibrio (OR = 1.05, 95% CI: 1.01–1.09, P = 2.25 × 10^-2), Butyrivibrio, Collinsella, Ruminococcaceae UCG009, and Parasutterella (ORs = 1.03–1.07, all P < .05). Conversely, 5 genera demonstrated protective associations: Allisonella (OR = 0.96, 95% CI: 0.93–0.98, P = 7.04 × 10^-4), Bacteroides (OR = 0.91, 95% CI: 0.86–0.97, P = 5.03 × 10^-3), Victivallis, Peptococcus, and Terrisporobacter (ORs = 0.95–0.97, all P < .05). After FDR correction, significant associations remained for R gnavus group and Allisonella (P = .04), while associations for other genera were attenuated. Sensitivity analyses revealed no evidence of horizontal pleiotropy or heterogeneity (File S1: Table S4, Supplemental Digital Content, https://links.lww.com/MD/Q96, Fig. 4).

Figure 3.

Scatterplot of MR studies examining the effect of 11 gut microbes on fall risk. MR = Mendelian randomization.

Figure 4.

MR leave-one-out sensitivity analysis for 11 gut microbes and fall risk. MR = Mendelian randomization.

3.4. Replication and meta-analysis confirm key microbial associations with fall risk

To enhance the robustness of our findings, we conducted a replication MR analysis using an independent GWAS dataset for fall risk. Although the replication estimates were not statistically significant: likely due to reduced statistical power and phenotype definition discrepancies: the directional consistency supported the initial findings. A subsequent meta-analysis of the discovery and replication datasets identified 7 gut microbiota taxa with consistent causal associations with fall risk. Among them, R gnavus group (OR = 1.04, P = 2.20 × 10^-3), Butyrivibrio (OR = 1.02, P = 1.76 × 10^-2), and Ruminococcaceae UCG009 (OR = 1.03, P = 1.76 × 10^-2) were associated with increased risk. Conversely, higher abundance of Allisonella (OR = 0.97, P = 5.00 × 10^-4), Bacteroides (OR = 0.97, P = 5.00 × 10^-4), Victivallis, and Peptococcus was protective (P < .05 for all). Associations for Desulfovibrio, Terrisporobacter, Collinsella, and Parasutterella were not replicated and did not retain significance in the meta-analysis (Fig. 2B).

3.5. Two-step MR reveals a mediating role of R gnavus

In prior analyses, we established causal links between smoking and the 7 gut microbiota taxa and fall risk. Subsequent MR analysis revealed a genetic association between smoking and the R gnavus group (OR = 1.21, P = 8.35 × 10^-3), with no evidence of horizontal pleiotropy or heterogeneity among the IVs (File S2: Figure 2, Supplemental Digital Content, https://links.lww.com/MD/Q97). Two-step MR analysis further indicated that the R gnavus group mediated 8.63% of the association between smoking and fall risk, representing a relatively small proportion.

4. Discussion

This study is the first to systematically explore the causal and mediating roles of the gut microbiota in the relationship between lifestyle factors (particularly smoking) and fall risk using MR. Our findings demonstrate that smoking significantly increases the fall risk, with the gut microbiota potentially playing a critical mediating role. Specifically, the R gnavus group was identified as a key mediator in the smoking–fall risk association, accounting for 8.63% of the total effect, thereby offering a potential target for future fall prevention strategies.

Our results confirm a direct genetic causal link between lifestyle factors (especially smoking) and fall risk, which is consistent with prior observational and clinical studies. Recent evidence indicates a dose–response relationship between smoking intensity and recurrent fall risk, with individuals smoking ≥ 20 cigarettes/day exhibiting a higher risk than lifelong nonsmokers.^[29] Smoking also increases frailty risk.^[30,31] Tobacco smoke contains toxic compounds that impair the mitochondrial oxygen supply, directly inducing skeletal muscle dysfunction, atrophy, and sarcopenia.^[32] Smoking upregulates myostatin and muscle atrophy F-box expression and suppresses muscle protein synthesis.^[33] Chronic inflammation, oxidative stress, and DNA methylation induced by smoking further compromise physical function, leading to muscle wasting, weight loss, and gait impairment.^[34] Additionally, smoking accelerates cellular senescence via telomere attrition, which directly contributes to systemic aging.^[35] Collectively, these mechanisms reinforce the detrimental role of smoking in falls. Although alcohol consumption, sleep, and vigorous exercise showed no significant causal associations in our study, their complex and potentially nonlinear effects observed in epidemiological research warrant further investigation.

Notably, this study identified multiple gut microbiota taxa with significant or suggestive causal links to fall risk, particularly R gnavus and Allisonella. The R gnavus group increases fall risk through multifaceted pathways involving inflammation, metabolic dysregulation, and neuromuscular dysfunction.^[36–45] Although it produces short-chain fatty acids to maintain intestinal homeostasis under normal conditions, its overabundance disrupts the mucosal barrier, triggering systemic metabolic disturbances and chronic inflammation that exacerbate obesity, diabetes, cardiovascular diseases, and osteoporosis. These disruptions propagate through interconnected “gut–brain,” “gut–bone,” and “gut–muscle” axes, ultimately impairing neuromuscular coordination, postural control, cognitive function, bone strength, and muscle performance: all critical contributors to fall risk. In contrast, Allisonella exhibits protective effects by producing propionate-dominated short-chain fatty acids, enhancing insulin sensitivity, preserving skeletal muscle mass, and promoting neurocognitive function, thereby reducing fall susceptibility.^[46–49] Elucidating these microbial mechanisms will advance microbiome-targeted interventions to mitigate falls and improve the overall health. Our novel finding of the R gnavus group’s mediating role of R gnavus in smoking-related fall risk suggests its potential involvement in chronic inflammation and other pathways affecting neuromuscular integrity.

Methodologically, this study benefits from the MR’s ability to minimize confounding and reverse causality through genetic instruments, thus strengthening causal inference. Replication using the UKB and FinnGen datasets, followed by meta-analysis, further enhanced the robustness of the results. However, this study has several limitations. First, all GWAS summary statistics used were derived from European populations, which may limit generalizability to other ancestries. Validation in more diverse cohorts is needed. Second, microbiota exposures were analyzed at the genus level, potentially overlooking species- or strain-specific effects. Higher-resolution taxonomic and functional data could provide deeper insights. Third, due to the limited statistical power of microbiome GWAS, we used a relatively relaxed P-value threshold (P < 1 × 10^-5) to select microbial IVs, which may introduce the risk of weak instrument bias. To mitigate this issue, we calculated the F-statistics for all selected SNPs to assess the strength of their association with the exposure. An F-statistic >10 is commonly considered indicative of sufficient instrument strength and helps reduce bias arising from weak instruments. In this study, all included SNPs had F-statistics > 10, thereby enhancing the reliability of our causal estimates. Fourth, the replication analysis in the FinnGen cohort was not statistically significant, despite consistent effect directions. Cohort differences in phenotype definitions and sample characteristics may have contributed to heterogeneity. Fifth, the estimated mediation proportions were relatively small, indicating that gut microbiota may account for only a minor part of the pathway linking lifestyle factors to fall risk. Other mechanisms, such as inflammation or neuromuscular decline, may also be involved.

5. Conclusion

In summary, this study provides the first genetic evidence supporting the mediating role of the gut microbiota, particularly the R gnavus group, in the association between smoking and fall risk. These findings offer novel insights into fall prevention, suggesting that microbiota-targeted interventions may emerge as a key strategy to reduce fall risk in older adults. Future research should focus on elucidating the specific mechanisms of these microbial interactions and validating the efficacy of interventions in clinical trials.

Acknowledgments

We thank the studies or consortiums referenced and included in the present analysis for providing public datasets and managing summary statistics.

Author contributions

Conceptualization: Jihong Zhang, Xiaoyi Yan.

Data curation: Jihong Zhang, Xiaoyi Yan.

Formal analysis: Jihong Zhang, Xiaoyi Yan.

Investigation: Bin Zhang.

Methodology: Bin Zhang.

Project administration: Bin Zhang, Han Zhang.

Resources: Han Zhang.

Software: Han Zhang, Shushun Miao, Qianqian Du.

Supervision: Shushun Miao, Qianqian Du, Qinghai Zhang.

Validation: Shushun Miao, Qianqian Du, Qinghai Zhang, Jianwei Du.

Visualization: Qinghai Zhang, Jianwei Du.

Writing – original draft: Qinghai Zhang, Jianwei Du.

Writing – review & editing: Jianwei Du.

Supplementary Material

medi-104-e44723-s001.xlsx ^{(308KB, xlsx)}

medi-104-e44723-s002.docx ^{(443.9KB, docx)}