Causal relationship between chronic obstructive pulmonary disease and BMD at different sites: A bidirectional Mendelian randomization study

Abstract

Observational studies have demonstrated a correlation between chronic obstructive pulmonary disease (COPD) and osteoporosis (OP). However, it is unclear whether there is genetic causality between COPD and bone mineral density (BMD) reduction at different sites. This study assessed the causal relationship between COPD and BMD in various anatomical locations. Data associated with COPD and BMD were obtained from published genome-wide association studies (GWAS). We selected single nucleotide polymorphisms (SNPs) that were strongly associated with COPD and BMD could serve as instrumental variables for the analysis. Inverse variance weighted, MR-Egger and weighted median were manipulated to evaluate causality. Subsequently, we conducted heterogeneity tests using Cochran Q test and tested for pleiotropy using the MR-Egger intercept. We performed leave-one-out sensitivity analysis to assess the robustness of the results. Additionally, we obtained more accurate causal genetic associations by removing any pleiotropic outlying SNPs and performed Mendelian randomization (MR) analysis with the remaining data. Our findings established that COPD was negatively associated with Heel-BMD (odds ratio[OR] = 0.978, 95% confidence interval [CI] = 0.966, 0.990, P = .0003) but not LS-BMD (OR = 0.981, 95% CI: 0.943, 1.020, P = .335), FA-BMD (OR = 0.984, 95% CI: 0.927, 1.046, P = .616), and FN-BMD (OR = 0.981, 95% CI: 0.950, 1.014, P = .249). In reverse MR analysis, the results showed no significant causal effect of BMD at different sites on COPD. The results were proved to be dependable and steady by sensitivity, heterogeneity, and pleiotropy analysis. We found that COPD increases the risk of decreased heel BMD, however, there is no evidence that the loss of BMD increases the risk of COPD.

1. Introduction

Chronic obstructive pulmonary disease (COPD) is a chronic pulmonary ailment characterized by persistent respiratory symptoms and restricted airflow.^[1] It has been found that COPD is a chronic inflammatory illness that can impact several systems throughout the body, such as bones, blood, and the cardiovascular system, and it poses a severe threat to the health of older adults.^[2] National cross-sectional research in China revealed that the incidence of COPD in Chinese adults aged 20 and above was 8.6%, while that in people aged >40 years was as high as 13.7%.^[3] The latest prediction from the World Health Organization on death rate and reasons for deaths shows that the incidence of COPD will continue to increase over the next 40 years due to the rising smoking rate in less developed nations and the elderly population in high-income countries. Predictions indicate that the number of people who die from COPD and related diseases each year will exceed 5.4 million by 2060.^[4]

Osteoporosis (OP) is a prevalent chronic condition characterized by decreased bone mineral density, weakened bone strength, and a heightened risk of fractures in fragile bones.^[5] OP is a frequent non-respiratory complication observed in individuals with COPD.^[6] The meta-analysis revealed that the incidence of OP among individuals with COPD was 38% (95% CI, 34%–43%), whereas it was 15% (95% CI, 6%–27%) in the control group.^[7] Furthermore, another study demonstrated that the proneness to OP in patients with severe COPD is 4 times higher than in healthy individuals.^[8] The incidence of comorbid OP in COPD patients is notably more significant than that in healthy people of the same age, which is likely related to the use of glucocorticoids, smoking, decreased lung function, and malnutrition in COPD patients.^[7] Moreover, OP can exacerbate COPD and vertebral compression fractures (VCFs) caused by OP are common complications that can lead to humpback deformity and further impair lung function.^[9]

Although an increasing number of studies have demonstrated a correlation between COPD and OP, causality is still under investigation. Randomized controlled trials (RCTs) are widely considered the most rigorous research design for inferring causality in epidemiology, however, the implementation of RCTs is often limited by ethical considerations.^[10] Furthermore, many factors, such as confounding variables and reverse causality, influence the association between exposure and outcome and the limitations of high time, workforce, and resource requirements.

As an alternative method, Mendelian randomization (MR) can effectively use the results of existing genome-wide association studies (GWAS) data, use genetic variation as instrumental variables (IVs), and explore the causal relationship between risk factors and outcomes.^[11] Based on Mendel law of inheritance, parental alleles are randomly assigned to offspring, which is equivalent to random grouping in RCT study. In theory, genetic variation is unaffected by common confounding factors, such as the postnatal environment, and genetic variation occurs earlier than exposure and outcome, eliminating reverse causality. Therefore, genetic variants as IVs to analyze causality have been gradually applied in epidemiological research. With the publication of large-scale GWAS data, numerous reliable genetic variations are available for MR studies, and many studies have used the MR method to explore the causal relationships between multiple traits.^[12]

While previous observational studies have reported associations between COPD and BMD, establishing a cause-and-effect relationship remains challenging due to the presence of confounding factors and reverse causation.^[13] To address these limitations, we propose a Mendelian randomization study design, which utilizes genetic variants as instrumental variables to assess the causal impact of COPD on BMD. By leveraging this robust and innovative approach, we aim to provide a more definitive understanding of whether COPD directly influences BMD at different skeletal sites, shedding light on potential mechanisms underlying this relationship. The findings of this study hold the promise of informing clinical decision-making and public health strategies, ultimately contributing to the development of targeted interventions to mitigate the impact of COPD on bone health and improve the quality of life for affected individuals.

2. Materials and Methods

2.1. Study design

A 2-sample MR study analyzed the data to evaluate the causal relationship between COPD and BMD in different parts of the body. In addition, heterogeneity tests and gene pleiotropy were conducted as quality controls to verify the reliability of the causality results. The MR analysis is based on 3 key assumptions^[14]: The IVs are strongly related the exposure factors. IVs are not associated with confounding factors affecting expose-outcome. The IVs affect the outcome only through exposure factors. The research design of the MR study is illustrated in Figure 1.

Figure 1.

Schematic representation of the MR study. MR = Mendelian randomization.

2.2. Data sources

Summary-level genetic data for femoral neck BMD (FN-BMD), lumbar spine BMD (LS-BMD), and forearm BMD (FA-BMD) were obtained from the genetic factors for osteoporosis consortium (http://www.gefos.org/), including 53,236 individuals and 32,735 individuals of European ancestry.^[15] Summary statistics for heel BMD (Heel-BMD) were obtained from the UK Biobank (http://www.nealelab.is/uk-biobank), including 265,627 individuals of European descent.^[16] Summary data for COPD retrieved from the FinnGen database(https://www.finngen.fi/en), containing 6915 COPD cases and 186,723 controls of European descent.^[17] Additional data details can be seen in Supplementary Table 1, http://links.lww.com/MD/K210.

2.3. Selection and validation of single nucleotide polymorphisms (SNPs)

In the forward MR analysis with COPD as exposure and BMD at various anatomical locations as the outcome, we extracted SNPs with significant genome-wide significance (P < 5 × 10⁻⁸) from the of COPD GWAS database to gain strong IVs. The significant SNPs’ linkage disequilibrium was set to r² = 0.001 and KB = 10,000. However, only 5 SNPs have been identified in COPD. Previous studies have shown that including multiple instrumental variables can enhance the precision and dependability of analytical outcomes^[18]; therefore, based on the correlation analysis results of relatively reliable thresholds in previous literature, we eased the criteria to 1 × 10^−[6] for screening COPD-related IVs.^[19] We then removed the palindromic and ambiguous SNPs and used PhenoScanner webs (http://www.phenoscanner.medschl.cam.ac.uk/ accessed on April 7, 2023) to analyze the selected SNPs and SNPs that were related to potential confounders (mineral and other dietary supplements: fish oil, basal metabolic rate and smoking) for BMD were removed.^[20,21]

In the backward MR analysis with BMD at different sites as exposure and COPD as the outcome, we extracted SNPs with significant genome-wide differences (P < 5 × 10⁻⁸) from the GWAS database of BMD to gain strong IVs. The significant SNPs’ linkage disequilibrium was set to r² = 0.001 and KB = 10,000. Similarly, we excluded palindromic and ambiguous SNPs and SNPs associated with possible confounding factors. Subsequently, the F-statistic was utilized to assess whether there was a weak IV bias (F-statistic < 10 was used to define a weak IV), and the IVs of F-statistic < 10 were removed from our analysis. The F-statistic was calculated utilizing the following formula^[20]:

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{F}=\frac{\mathrm{N}-\mathrm{K}-1}{\mathrm{K}}\times \frac{{\mathrm{R}}^2}{1-{\mathrm{R}}^2}}\end{array}}$$

R² is the proportion of variability in the exposure explained by the IVs, k represents the number of IVs used in the model, and n represents the sample size.

2.4. The primary MR analysis

To determine the potential causality between COPD and BMD at various anatomical locations, we ran a primary MR analysis using 3 methods: inverse variance weighted (IVW), MR-Egger, and weighted median.^[22,23] The IVW method is the most precise method for estimating causality when directional pleiotropy does not exist in the results (p for MR-Egger intercept P > .05) and is commonly used as the primary analytical method for estimating causal effects.^[24] The MR-Egger method is capable to identify and adjust for directional pleiotropy. However, it exhibits limited a statistical power.^[25] To use the weighted median method for estimating causality, a minimum of 50% of SNPs were considered valid instrumental variables.^[23] We ensured that the causal effects were uniform across all 3 methods and deemed them, significant by setting a significance threshold of P < .05.^[26,27]

2.5. The secondary MR analysis

To guarantee the precision of the MR findings in our study, we performed secondary MR analyses, including sensitivity, heterogeneity, and pleiotropy analyses, to account for potential biases stemming from variations and polymorphisms among individual SNPs. Specifically, we employed Cochran Q test to detect any variations in MR analysis, indicating potential heterogeneity.^[28] Afterwards, the “leave-one-out” method was employed to evaluate the causal genetic impacts of exceptional SNPs and ascertain whether the exclusion of those SNPs influenced MR estimations.^[11] To assess horizontal pleiotropy, we investigated the intercept coefficient in the MR-Egger regression. When the intercept coefficient significantly deviates from zero (P < .05), it suggests the existence of horizontal pleiotropy.^[22] In addition, the MR-PRESSO approach was used to assess directional pleiotropy and detect and adjust for potential anomalous data points.^[27] To enhance the representation of the IVs analysis, we opted for radial variants instead of funnel plot. Additionally, we utilized RadialMR imaging for automated identification of outliers.^[29] Furthermore, we utilized RadialMR and Outlier MR-PRESSO techniques to obtain more accurate causal genetic associations to remove any pleiotropic outlying SNPs and perform MR analysis with the remaining data.

3. Results

In the forward MR analysis, COPD constituted the exposure, and BMD at different sites were set as the outcome. There are 23 SNPs of COPD for FA-BMD, 22 SNPs of COPD for FN-BMD, 23 SNPs of COPD for LS-BMD, and 21 SNPs of COPD for Heel-BMD remained (Supplementary Tables 2–5, http://links.lww.com/MD/K211, http://links.lww.com/MD/K212, http://links.lww.com/MD/K213, http://links.lww.com/MD/K214) after removing SNPS related to palindromic, ambiguous and confounding factors. In the backward MR analysis, BMD at different sites constituted the exposure, and COPD was set as the outcome. 3 SNPs of FA-BMD for COPD, 14 SNPs of FN-BMD for COPD, 17 LS-BMD for COPD, and 298 SNPs of Heel-BMD for COPD remained (Supplementary Tables 7–10, http://links.lww.com/MD/K216, http://links.lww.com/MD/K217, http://links.lww.com/MD/K218, http://links.lww.com/MD/K219) after removing SNPs related to palindromic, ambiguous, and confounding factors. To ensure the reliability of our findings, we eliminated any outliers identified using the MR-PRESSO and RadialMR methods and obtained robust results (Supplementary Table 6, http://links.lww.com/MD/K215, 11, http://links.lww.com/MD/K220, Supplemental Figures 1–2, http://links.lww.com/MD/K221, http://links.lww.com/MD/K222). Our research found that the F-statistics for the IVs linked to exposure were all above 10, which suggests a low likelihood of bias in weak IVs.

3.1. Causal effects of COPD on BMD at different parts

The IVW method was employed as the main analytical method; only Heel-BMD was casually influenced by COPD (OR = 0.978, 95% CI = 0.966, 0.990, P = .0003), and the weighted median analysis provided further evidence in support of this result (OR = 0.978, 95% CI = 0.961, 0.995, P = .016). There was no causal association between COPD and the other 3 types of BMD, including COPD for FA-BMD (OR = 0.984, 95% CI: 0.927, 1.046, P = .616), COPD for FN-BMD (OR = 0.981, 95% CI: 0.950, 1.014, P = .249), and COPD for LS-BMD (OR = 0.981, 95% CI :0.943, 1.020, P = .335). MR Egger and weighted median analyses had the same effects as the IVW estimates. This study provided a comprehensive overview of the MR analysis outcomes derived from various causal effect assessment methods. Detailed estimates are available in Supplementary Table 6, http://links.lww.com/MD/K215, which were visualized using forest plots and scatter plots (Figs. 2 and 3A). Furthermore, the result did not indicate pleiotropy or heterogeneity (Supplementary Table 6, http://links.lww.com/MD/K215). The leave-one-out plot demonstrated that removing a single SNP from the genetic variants had a minimal effect on the outcome, as shown in Figure 3B.

Figure 2.

Forest plots of Mendelian randomization analyses of the causal effects of COPD on BMD at various anatomical locations. BMD = bone mineral density, COPD = chronic obstructive pulmonary disease.

Figure 3.

Scatter plots and Leave-one-out plot of Mendelian randomization analyses of the causal effects of COPD on BMD at various anatomical locations. A Scatter plots; B Leave-one-out plot. BMD = bone mineral density, COPD = chronic obstructive pulmonary disease.

3.2. Causal effects of BMD at different parts on COPD

Our study found no evidence of a causal relationship between the 4 types of BMD and COPD using the IVW analysis, including COPD for FA-BMD (OR = 0.840, 95% CI: 0.704, 1.003, P = .054), COPD for FN-BMD (OR = 1.160, 95% CI: 0.954, 1.412, P = .688), COPD for LS-BMD (OR = 0.688, 95% CI: 0.860, 1.209, P = .820), and COPD for HE-BMD (OR = 1.015, 95% CI: 0.932, 1.105, P = .730) (Supplementary Table 11, http://links.lww.com/MD/K220), which were presented in the form of forest plots and scatter plots (Figs. 4 and 5A). We also obtained similar results using the MR-Egger and weighted median methods. The absence of heterogeneity was confirmed using Cochran Q test (P > .05). The MR-Egger regression also indicated that our MR results were not affected by horizontal pleiotropy (intercept P > .05), and the MR-PRESSO approach suggested that our results were not influenced by directional pleiotropy (Global Test P > .05), detailed results of heterogeneity and pleiotropy analysis are available in Supplementary Table 11, http://links.lww.com/MD/K220. Furthermore, the leave-one-out analysis confirmed the robustness of our findings (Fig. 5B).

Figure 4.

Forest plots of Mendelian randomization analyses of the causal effects of BMD at various anatomical locations on COPD. BMD = bone mineral density, COPD = chronic obstructive pulmonary disease.

Figure 5.

Scatter plots and Leave-one-out plot of Mendelian randomization analyses of the causal effects of BMD at various anatomical locations on COPD. A Scatter plots; B Leave-one-out plot. BMD = bone mineral density, COPD = chronic obstructive pulmonary disease.

4. Discussion

In this bidirectional MR analysis, we employed GWAS summary statistics to explore the causal association between COPD and BMD at different sites. In the forward analysis, our study results demonstrated no causal relationship between COPD and BMD at different sites, except for Heel-BMD, and the IVW and weighted median method suggested a negative causal relationship between COPD and Heel-BMD. It is worth noting that the COPD GWAS datasets used in this study consisted of individuals of European ancestry, whereas the lumbar spine, forearm, and hip-neck BMD GWAS datasets included individuals of mixed ancestry. This difference in population stratification could have introduced bias into our results. In the backward analysis, after using the 3 MR methods, no evidence was found to suggest a causal association between the 3 types of BMD (FA-BMD, LS-BMD, and Heel-BMD) and COPD. Additionally, the MR-Egger method showed a positive causal association between FN-BMD and COPD. However, the inference of a causal relationship cannot be verified by employing supplementary methods, implying that the existing evidence is inadequate and that the conclusions drawn should be approached with caution.

Numerous risk factors, including glucocorticoid administration, inadequate vitamin D levels, compromised lung function and inflammation, are associated with reduced BMD in individuals with COPD.^[30] Glucocorticoids are extensively employed in the management of COPD.^[31] Glucocorticoid-induced OP represents the most common form of secondary OP.^[32] On the one hand, glucocorticoids directly inhibit osteoblast proliferation and differentiation, as well as suppress the production of osteoblast-specific markers (osteocalcin, bone-specific alkaline phosphatase) through the insulin-like growth factor I (IGF-I) and Wnt signaling pathways.^[32,33] On the other hand, glucocorticoids activate Caspase3 through various apoptotic signaling pathways, promoting apoptosis.^[34] Furthermore, glucocorticoids exert their effects by modulating the growth factors in the bone microenvironment; for instance, IGF-I enhances bone formation and type I collagen synthesis, while reducing collagen degradation and osteoblast apoptosis, whereas glucocorticoids inhibit IGF-I transcription.^[35] Vitamin D undergoes hepatic and renal conversion to the biologically active form, 1,25-(OH)₂D₃. Serum levels of 1,25-(OH)₂D₃ are directly related to bone mineral density and inversely proportional to PTH levels.^[36] PTH plays a crucial role in maintaining calcium homeostasis, promoting collagen matrix degradation for bone mineralization, leading to decreased BMD, OP, and increased risk of fractures.^[37] In COPD patients, vitamin D deficiency is frequently observed as a result of reduced physical activity, insufficient engagement in outdoor activities, and limited exposure to sunlight.^[38] Airway inflammation constitutes a critical pathological factor contributing to the persistent and progressive exacerbation of COPD. Moreover, the systemic inflammatory response also plays a pivotal role in the advancement of OP among COPD patients.^[39] On the one hand, pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α), interleukin (IL)-6, IL-1β, IL-11, and IL-17, govern bone resorption by promoting osteoclast activity.^[40] On the other hand, these cytokines directly impede bone formation. The Wnt/β-catenin signaling pathway plays a pivotal role in skeletal development, significantly influencing the differentiation of mesenchymal stem cells into mature osteoblasts. TNF-α, acting as a key inducer of DKK-1 (a Wnt antagonist), can suppress osteoblast differentiation.^[40,41] According to findings from the Xinxiang Rural Cohort Study conducted in China, a decrease in BMD of 0.1 g/cm² was associated with a reduction of 53.0 mL in forced vital capacity (FVC) and 33.5 mL in forced expiratory volume in 1 second (FEV1) after adjusting for possible confounding variables using linear regression analysis, which suggests that a decline in BMD is linked to impaired lung function.^[42]

As far as we know, this is the first MR analysis to explore the causal relationship between COPD and BMD at different sites. This research presents several benefits. Firstly, the data were attained from the most extensive sample size in GWAS studies. Secondly, to make our results more robust, we eliminated aberrant data points using multiple approaches, resulting in consistently positive outcome. Furthermore, we achieved identical results using different MR techniques, thus enhancing the overall stability of our findings. Finally, ethical hazards were not involved in our study.

Nevertheless, this study also had several limitations. Firstly, the population of COPD and BMD are main European descent, which may constrain the representativeness of our findings to the entire population. Secondly, our results indicated that individuals with COPD were prone to reduce Heel-BMD. Meanwhile, it also showed that COPD had no significant causal effect on LS-BMD, FA-BMD, or FN-BMD, which warranted scrupulous consideration. This result can be attributed to the following reason. The population of COPD was the European ancestry, whereas the LS-BMD, FA-BMD, and FN-BMD were Mixed ancestry; population stratification could distort the results. The lack of GWAS datasets for the LS-BMD, FA-BMD, and FN-BMD in European ancestry prevented us from correcting bias in the results. To achieve more precise results, additional MR studies with larger sample sizes or RCTs are warranted in the future.

In conclusion, our findings established that COPD is negatively associated with Heel-BMD. In addition, none of the 4 types of BMD were positively or negatively associated with COPD. Therefore, physicians should strengthen the monitoring of BMD in COPD patients and guide them in adopting correct and healthy habits to reduce the risk of OP. This approach will facilitate prompt detection of COPD patients with concurrent OP. Prompt preventive measures against OP, coupled with appropriate management of preexisting OP, may elevate the quality of life and daily functioning of individuals with COPD, preserve pulmonary function, and eventually lead to a more favorable prognosis.

Acknowledgments

We are grateful to the Genetic factors for Osteoporosis Consortium (GEFOS) consortium, the UK Biobank, the FinnGen database, and all participants in our study.

Author contributions

Conceptualization: Rui Jiang, Zheng Zhang.

Formal analysis: Zheng Zhang.

Validation: Feng Luo.

Writing – original draft: Rui Jiang.

Writing – review & editing: Shuanglin Mou.

Supplementary Material

medi-102-e35495-s001.xlsx ^{(10.9KB, xlsx)}

medi-102-e35495-s002.xlsx ^{(13.4KB, xlsx)}

medi-102-e35495-s003.xlsx ^{(13.3KB, xlsx)}

medi-102-e35495-s004.xlsx ^{(13.5KB, xlsx)}

medi-102-e35495-s005.xlsx ^{(13.7KB, xlsx)}

medi-102-e35495-s006.xlsx ^{(10KB, xlsx)}

medi-102-e35495-s007.xlsx ^{(12.7KB, xlsx)}

medi-102-e35495-s008.xlsx ^{(13.2KB, xlsx)}

medi-102-e35495-s009.xlsx ^{(57.5KB, xlsx)}

medi-102-e35495-s010.xlsx ^{(12.9KB, xlsx)}

medi-102-e35495-s011.xlsx ^{(13KB, xlsx)}

medi-102-e35495-s012.docx ^{(43.3KB, docx)}

medi-102-e35495-s013.docx ^{(48.5KB, docx)}