Introduction

Osteoporosis (OP) is a metabolic bone disorder disease characterized by diminished bone density, compromised bone integrity, and deterioration of bone microarchitecture, culminating in heightened susceptibility to fragility fractures1. OP is linked to elevated medical expenses, physical incapacitation, diminished quality of life, and heightened mortality rates2. In the United States, an estimated 10.2 million older adults were diagnosed with OP in 2010, while approximately 43.4 million older adults exhibited low bone density3. With the acceleration of global aging, OP has emerged as a significant public health concern in contemporary society4,5. Hence, it is imperative to ascertain the risk factors associated with OP to facilitate the prompt implementation of interventions and mitigate the occurrence of adverse clinical events related to OP, thereby enhancing the prognosis for the elderly population.

Lung function tests play a crucial role in assessing respiratory health, particularly in detecting conditions like asthma and early-onset airway diseases6. Key metrics include the forced expiratory volume in 1 s (FEV1), forced vital capacity (FVC), peak expiratory flow (PEF), and the FEV1/FVC ratio7,8. These indicators are essential for diagnosing conditions such as chronic obstructive pulmonary disease (COPD), as well as assessing the extent of airway constriction and lung impairment9. Multiple prior studies have investigated the correlation between lung function and bone metabolism; however, the results have shown inconsistency. S Lekamwasam et al. have reported a significant decrease in bone mineral density (BMD) at the total hip and femoral neck with the deterioration of pulmonary function10. A study conducted in China reveals a significant association between declining lung function and reduced bone mass among adults aged 40 to 70 years11. Studies from diverse regions corroborate this finding12,13. In contrast, a study examining the correlation between lung function and BMD in over 900 older adults did not find a statistically significant association after adjusting for confounding variables (age, height, BMI, smoking, alcohol consumption, physical activity, etc.)14. Elucidating the causal relationship between lung function and OP is pivotal for prevention and treatment. Nonetheless, various confounding factors in clinical observational studies often lead to inconclusive findings regarding a causal relationship between lung function and OP.

Mendelian randomization (MR) is a method for evaluating whether there is a causal relationship between exposure factors and outcomes. MR studies have several advantages, making them powerful tools for investigating causal relationships between genes and specific factors. Firstly, MR leverages the natural random distribution of genotypes, similar to Mendel’s laws of genetic segregation. This ensures that genotypes are randomly assigned in the study, helping to reduce the influence of confounding factors. Secondly, MR addresses common issues of reverse causation in observational studies, providing more substantial evidence that genetic variation’s impact on a specific factor is causal15. Thirdly, this approach allows researchers to simulate actual randomized experiments without the need for practical experimental operations, thereby reducing costs and ethical considerations. Lastly, in certain situations, conducting randomized controlled trials on study subjects is either impractical or unethical16,17. In such cases, MR Studies provide an alternative approach to causal inference.

In this study, we conducted a two-sample MR study using publicly available GWAS data, employing FVC, PEF, FEV1, and FEV1/FVC as exposures and OP as the outcome. This study aimed to investigate whether there is a causal connection between lung function and OP. This endeavor aims to lay a theoretical foundation that will contribute to the establishment of effective public health policies, facilitating more efficient mitigation of potential health issues.

Materials and methods

Study design

Anchored in the MR design, this study relies upon three fundamental principles:1 that IVs exhibit a strong correlation with the exposure factors;2 that IVs do not affect outcome via the confounding factors; and3 that IVs can only impact the outcome through exposure factors and are incapable of having a direct correlation with outcome. In our study, lung function served as the exposure factor, and IVs consisted of single nucleotide polymorphisms (SNPs) that.

displayed a robust correlation with lung function (FVC, PEF, FEV1/FVC, FEV1). OP served as the outcome variable in this study. Therefore, we employed a MR study to analyze the causal connection between lung function and OP. The flowchart of the MR design is shown in Fig. 1.

Fig. 1
figure 1

Figure 1 Framework and flowchart of MR. (A)Framework for investigating the causal effect of lung function (FVC, PEF, FEV1/FVC and FEV1) on OP. (B) Flowchart of the sequential steps taken to investigate the causal effect of lung function (FVC, PEF, FEV1/FVC and FEV1) on OP.

Full size image

GWAS data sources

GWAS data for lung function used in our study were obtained from publicly available GWAS from the UK Biobank (http://www.nealelab.is/uk-biobank). These data were meta-analyzed by Shrine N et al. and grounded on the UK Biobank and SpiroMeta Consortium data18. The main metrics examined included FVC, FEV1, FEV1/FVC ratio, and PEF. The GWAS datasets for FVC, FEV1, and FEV1/FVC comprised 321,047 individuals of European ancestry, while the GWAS datasets for PEF included 24,218 individuals of European ancestry. Since there are multiple datasets to choose from in the GWAS database, we selected the datasets with the largest sample size for analysis. The GWAS datasets for OP were gained from the FinnGen Consortium, which comprises 3,203 OP cases and 209,575 controls. Additional information regarding the definition of the OP datasets can be found on the website: https://risteys.finregistry.fi/endpoints/M13_OSTEOPOROSIS. Since the exposure and outcome datasets come from different consortiums, our results are not affected by sample overlap. Additional data details are listed in Table 1.

Table 1 Overview of the data sources of phenotype used in the MR Study.
Full size table

Selection and validation of SNPs

We chose SNPs significantly linked to lung function (FVC, FEV1, FEV1/FVC ratio, and PEF) at the genome-wide significance level (p < 5 × 10−8). To ensure all the SNPs for exposure were not in linkage disequilibrium (LD), we conducted a linkage disequilibrium analysis, considering the.

LD correlation coefficient (r2 < 0.001) and base pair distance between the two SNPs (kb > 10,000). Next, we removed palindromic and ambiguous SNPs. Then, we utilized LDlink webs tool (https://ldlink.nih.gov/ accessed on March 10, 2024) to evaluate the selected SNPs. Subsequently, we excluded SNPs potential confounders for OP such as type 2 diabetes19,20, body mass index21, osteoarthritis22, educational attainment23, alcohol consumption, smoking cessation24 and coronary artery disease25, as well as SNPs directly associated with OP (BMD). We computed the F statistic (F = R2(n - 1 - k)/(1 - R2)k) for each SNP to assess the existence of weak instruments; SNPs with F > 10 were selected for further analysis26. Finally, we utilized MR-PRESSO and MR radial methods to detect and remove outliers, thus reducing heterogeneity. These rigorously selected SNPs acted as the ultimate IVs for the following MR analysis.

MR analysis

To investigate the potential causality between lung function (FVC, FEV1, FEV1/FVC PEF) and OP, we conducted a MR analysis employing three approaches: inverse variance weighted (IVW), MR-Egger, and weighted median. The IVW approach was applied as the main statistical analysis approach, however, it may yield biased estimates when there is directional or unbalanced horizontal pleiotropy27. The MR-Egger regression allows for the correction of pleiotropy but diminishes statistical power28. The weighted median approach requires that at least half of the weight in the analysis stems from variables that act as valid instruments29. These three approaches are widely regarded as the most scientifically rigorous and often utilized methods to ensure the robustness of the MR analysis.

Sensitivity analysis

Sensitivity analysis is an essential method for assessing potential biases in MR analysis. It encompasses two parts: a heterogeneity test and a pleiotropy test. To evaluate potential heterogeneity, we used Cochran’s Q test to explore any variations in our study28. MR-Egger regression was used to evaluate directional pleiotropy30. Additionally, the MR-PRESSO method was employed to evaluate horizontal pleiotropy31. Furthermore, we excluded SNPs exhibiting pleiotropic outliers employed both MR-PRESSO and MR radial methods. Upon identifying potential outliers, we removed them and subsequently re-ran the MR analysis to assess the reliability of our findings32. Additionally, we conducted a “leave-one-out” analysis to ascertain whether the results were influenced by any single SNP. As implied by its name, SNPs were systematically excluded one by one, followed by the re-execution of the MR analysis33. Finally, we calculated the statistical power of our MR analyses using the tool provided at https://shiny.cnsgenomics.com/mRnd.

Statistical analyses

All data analyses were conducted utilizing the “two-sample MR” R package (version 0.5.6) in the R software (version 3.6.7) software. A significance level of p < 0.05 was considered statistically significant. Given the publicly accessible nature of the data utilized in our study, ethical approval for their utilization was unnecessary.

Results

Selection and validation of SNPs

After removing palindromic and ambiguous SNPs and SNPs associated with potential confounders and outcome, we finally obtained 175 SNPs for FVC, 157 SNPs for PEF, 230 SNPs for FEV1/FVC, and 202 SNPs for FEV1. The F statistics for all these IVs were above 10, indicating no weak instrumental bias in our MR study. SNPs strongly associated with confounding factors, such as type 2 diabetes, smoking cessation, alcohol consumption, osteoarthritis, body mass index, educational attainment, and BMD were eliminated. Detailed information about the SNPs is listed in Supplementary Tables 1–4. Since the SNPs included in the analysis overlapped across the four lung function metrics, we used a Venn diagram to better illustrate the extent of this overlap (Fig. 2, Supplementary Table 5).

Fig. 2
figure 2

Venn diagram of overlapping SNPs with 4 metrics of lung function.

Full size image

MR analysis

No causal effect of FVC (OR [95% CI]: 1.246 [0.991–1.567], p = 0.060), PEF (OR [95% CI]: 0.889 [0.719-1.100], p = 0.279), FEV1/FVC (OR [95% CI]: 0.949 [0.797–1.130], p = 0.554) and FEV1 (OR [95% CI]: 1.013 [0.824–1.246], p = 0.902) on OP were found using the IVW method (Fig. 3). The trend lines in Fig. 4 (four subplots) suggest that, despite using different MR methods (IVW, Weighted Median, and MR Egger), the potential causal relationship between lung function (including FVC, PEF, FEV1/FVC, and FEV1) and OP appears to be weak or non-significant. The findings of MR Egger and weighted median methods were consistent with those from the IVW method (Fig. 4).

Fig. 3
figure 3

MR analysis between the lung function (FVC, PEF, FEV1/FVC and FEV1) and OP. Three methods: IVW, MR Egger, and weighted median method.

Full size image
Fig. 4
figure 4

Scatter plot of MR analysis. (A) FVC and OP; (B) PEF and OP; (C) FEV1/FVC and OP; (D) FEV1 and OP.

Full size image

Cochran’s Q test revealed no heterogeneity (p > 0.05) in the MR analyses for FVC, PEF, FEV1, and OP. However, heterogeneity was observed in the MR analysis between FEV1/FVC and OP (p < 0.05) (Supplementary Table 6). MR-Egger regression indicated no horizontal pleiotropy (p > 0.05) in the MR analyses for FVC, PEF, FEV1/FVC, FEV1, and OP (Supplementary Table 6). The MR-PRESSO approach indicated that there was no horizontal pleiotropy (p > 0.05) in the MR analysis between FVC, PEF, FEV1, and OP. In the MR analysis between FEV1/FVC and OP, there was horizontal pleiotropy (p < 0.05) (Supplementary Table 6). The “leave-one-out” analysis demonstrated that the results remained consistent even after excluding each individual SNP, indicating robustness in the findings (Fig. 5). In addition, the MR-PRESSO approach showed that there were no outliers in the MR analysis between FVC, PEF, FEV1, and OP, and there were 5 outliers in the MR analysis between FEV1/FVC and OP (Supplementary Table 3). Besides, the radial MR approach suggested that there were 6, 11, 18, and 12 outliers in the MR analyses between FVC, PEF, FEV1/FVC, FEV1, and OP, respectively (Supplementary Tables 1–4). The power values for FVC, PEF, FEV1/FVC, FEV1, and OP were 0.59, 0.71, 0.89, and 0.69, respectively, which indicated that our MR results lacked sufficient statistical power (Supplementary Table 7).

Fig. 5
figure 5

“Leave one out” analysis. (A) FVC and OP; (B) PEF and OP; (C) FEV1/FVC and OP; (D) FEV1 and OP.

Full size image

MR analysis after removing outliers

Due to the presence of heterogeneity and horizontal pleiotropy in the analysis between FEV1/FVC and OP, along with the identification of five outliers by the MR-PRESSO approach, after the removal of these outliers, we conducted an MR analysis again. After the removal of these outliers detected by the MR-PRESSO approach, the IVW method showed no causality between FEV1/FVC and OP (OR [95% CI]: 0.960 [0.815–1.129], p = 0.619) (Supplementary Table 6). The Cochran’s Q test showed no heterogeneity (p > 0.05) in the MR analysis between FEV1/FVC and OP. The MR-Egger regression showed that there was no directional pleiotropy (p > 0.05) in the MR analysis between FEV1/FVC and OP (Supplementary Table 6). The MR-PRESSO approach indicated that there was no horizontal pleiotropy (p > 0.05) in the MR analysis between FEV1/FVC and OP (Supplementary Table 6). The “leave-one-out” analysis demonstrated that the results remained consistent even after excluding each SNP, indicating robustness in the finding.

After removing the outliers identified by the MR radial method, we performed a third round of MR analysis. The results of the IVW method suggested that there was no causal relationship between FVC (OR [95% CI]: 1.291 [0.923–1.629], p = 0.052), PEF (OR [95% CI]: 0.890 [0.721,1.098], p = 0.276), FEV1/FVC (OR [95% CI]: 1.011 [0.861–1.188], p = 0.894) and FEV1 (OR [95% CI]: 0.918 [0.744–1.133], p = 0.425) and OP (Fig. 6). No heterogeneity and pleiotropy was found, which suggested that our findings are very robust (Supplementary Table 6). The “leave-one-out” analysis showed that the results remained consistent after excluding each SNP, confirming the robustness of the findings.

Fig. 6
figure 6

MR analysis between the lung function (FVC, PEF, FEV1/FVC and FEV1) and OP after exclusion of the outliers detected by MR radial. Three methods: IVW, MR Egger, and weighted median method.

Full size image

Discussion

In our study, MR analysis was used to evaluate the causal relationship between lung function (FVC, PEF, FEV1/FVC and FEV1) and OP. The MR method carried out a causal analysis of the lung function (FVC, PEF, FEV1/FVC, and FEV1) and OP at the genetic level, circumventing the problem associated with the observational study. Our findings showed that lung function (FVC, PEF, FEV1/FVC, and FEV1) was not causally associated with OP at the genetic level. However, the possibility of associations between lung function and OP at non-genetic levels cannot be ruled out.

Although we found no causal link between lung function and OP, a recent investigation explored the correlation between lung function indices and BMD in roughly 6000 middle-aged and older individuals. The findings revealed several lung function indices positively linked with BMD34. Research conducted in Italy indicates that FVC, FEV1, and the FEV1/FVC ratio are correlated with bone fragility. Furthermore, individuals with compromised lung function may face an increased risk of vertebral fractures35. In addition, numerous studies investigating the association between lung function and BMD in patients with COPD have demonstrated a decline in BMD with worsening lung function impairment36,37. The association found in observational studies might be attributable to shared risk factors. Common risk factors for both diseases include smoking, decreased physical activity, vitamin D deficiency, and the use of glucocorticoids37,38. Additionally, the discrepancy between our findings and those of traditional observational studies may be due to the fact that participants in traditional observational studies might have a history of lung diseases such as bronchitis, COPD, or asthma. Declining lung function is a concomitant symptom of these lung diseases. These lung conditions themselves may be associated with OP. Furthermore, the inability of observational studies to effectively control for confounding factors could also be a reason for the differences in conclusions.

Smoking stands out as a paramount risk factor for lung function decline in patients with COPD39. Moreover, it serves as an independent risk factor for OP24. The mechanisms linking smoking and OP are not fully understood. However, it is proposed that tobacco smoke could impact BMD by the proinflammatory effects40. Physical activity plays a crucial role in maintaining bone health due to the influence of mechanical stress stimuli on bone metabolism and its beneficial effects on skeletal muscle. Additionally, physical activity contributes to bone-muscle crosstalk, further supporting skeletal health41. Vitamin D plays a regulatory role in bone metabolism, and its deficiency can affect calcium absorption in the intestines, leading to impaired bone mineralization. This can trigger secondary hyperparathyroidism, resulting in increased bone turnover, heightened bone loss, and elevated fracture risk, ultimately promoting the development of OP42. Among patients with COPD experiencing lung function impairment, vitamin D deficiency is often prevalent due to factors such as decreased physical activity, inadequate participation in outdoor activities, and limited exposure to sunlight43. Studies have found a positive correlation between lung function and serum vitamin D levels in patients with COPD. Vitamin D deficiency may promote production of inflammatory cytokines, thus contributing to COPD44. In the long-term management of COPD patients with impaired lung function, steroid therapy is often necessary to alleviate pulmonary symptoms45. However, glucocorticoids can affect osteoblasts and osteoclasts in OP patients and inhibit osteoblast function. Inhalation of glucocorticoid can also reduce osteoblast generation and disrupt bone homeostasis, leading to OP. When osteoblast function is suppressed, it impacts bone metabolism46.

This study presents several advantages. Firstly, our study harnessed Mendel’s principle of independent assortment, opting for IVs as the exposure in MR analysis. This choice greatly boosts the reliability of our results as we investigate the cause-and-effect connection between lung function and OP. Secondly, reverse causation is precluded as genes precede the onset of the disease. Thirdly, our study leveraged data from publicly available GWAS pooled investigations, benefiting from a substantial sample size that enhances the robustness of our analysis. Additionally, to make our results more reliable, we eliminated outliers unitized MR-PRESSO and Radial Regression approaches and gained a consistent outcome, improving the overall robustness of our results.

Nevertheless, our study has inherent limitations. Firstly, the population under investigation for lung function and OP predominantly comprises individuals of European descent. This potential ethnic homogeneity may limit the broader applicability of our findings to the entire population. Secondly, despite the exclusion of SNPs linked to potential confounding factors identified in prior research, we were unable to completely eliminate the influence of other unaccounted confounding variables on the overall estimates. Thirdly, while we hypothesize that the association observed between lung function and osteoporosis in previous studies may be attributable to shared risk factors, our research does not provide sufficient evidence to confirm the validity of this inference. Fourthly, our results show that the selected IVs only explain 2.54% of the variation in FVC, 3.31% of the variation in PEF, 5.46% of the variation in FEV1/FVC, and 3.21% of the variation in FEV1. Additionally, because our MR results lacked sufficient statistical power, the negative findings may have been due to insufficient statistical power. Therefore, the results we obtained require further validation in larger sample cohorts. Lastly, although FVC, FEV1, FEV1/FVC ratio, and PEF are used to measure lung function, they are limited in terms of their reliability and accuracy in measurement.

Conclusion

In conclusion, according to our findings, there is no causal connection between lung function and OP. We suggest that the association observed in clinical settings might be attributable to shared risk factors such as smoking, decreased physical activity, vitamin D deficiency, and the use of glucocorticoids. Due to the shared risk factors between the two, it is reasonable to prevent OP in patients with lung function impairment routinely. Proper management of patients with lung function impairment is essential to reduce the risk of developing OP. Future multidisciplinary cooperation may play a vital role in clinical practice and influence the prognosis of OP.