The nonpathogenic role of hyponatremia in the onset of osteoporosis: A Mendelian randomization study Authors: Zhang Hongxing, Peng Jiafeng, Gao Minglei, Li Danyang, Junchen Zhu Published: 2024 ABSTRACT The relationship between hyponatremia and osteoporosis is controversial, and it remains unclear if there is a causal link between the two. This study employed a 2-sample Mendelian randomization (MR) analysis to investigate the potential causal relationship between hyponatremia and osteoporosis. The instrumental variables were derived from genome-wide association studies conducted in European populations. These included hyponatremia (n = 465,348) as the exposure factor, with genetic summary data for bone mineral density (BMD) at the forearm (n = 8134), femoral neck (n = 32,735), lumbar spine (n = 28,498), and heel (n = 265,627) as outcomes. The inverse variance weighted method did not identify any causal effect of hyponatremia on BMD. Additionally, other methods, such as MR-Egger, weighted median, simple mode, and weighted mode, also did not show evidence of a causal relationship between hyponatremia and BMD. Pleiotropy and heterogeneity analyses indicated that the MR findings were robust. There is no significant causal relationship between hyponatremia and osteoporosis. The previously observed associations may be due to confounding factors. It is unlikely that hyponatremia is a causal factor for osteoporosis. 1. Introduction Osteoporosis is a common skeletal disorder characterized by reduced bone mass, deterioration of bone microarchitecture, and an increased risk of fragility fractures.[1] According to the World Health Organization, the number of osteoporosis patients in European Union countries alone is projected to rise from 3.5 million in 2010 to 4.5 million in 2025, representing a 28% increase.[2] The burden of osteoporosis is greater than that of almost all cancer types, except for lung cancer.[3] Dual-energy X-ray absorptiometry (DXA) is commonly used to diagnose osteoporosis, with measurements taken at the lumbar spine, femoral neck, forearm, and whole body. The overall prognosis for osteoporosis is poor; for instance, women diagnosed with osteoporosis at the age of 50 have approximately a 45% chance of experiencing a hip, spine, forearm, or proximal humerus fracture in the next 10 years. Each standard deviation decrease in bone mineral density (BMD) increases the risk of fractures by 1.5 to 3 times.[4] Hyponatremia is defined as a serum sodium concentration below 135 mmol/L and is one of the most common electrolyte imbalances in clinical practice. Hyponatremia has a high incidence rate and poses significant health risks, negatively impacting various systems, including the circulatory, nervous, and skeletal systems. Studies have found that heart failure patients may develop hyponatremia due to the activation of the neurohormonal system, leading to a decrease in sodium levels.[5] This hyponatremia, in turn, can exacerbate the progression of heart failure by further activating the neurohormonal system and impairing hemodynamics.[6] Other research indicates that a sustained decrease in serum sodium concentration can cause gait disturbances, accelerate the loss of contextual fear memory, impair cognition in novel object recognition tests, and damage long-term potentiation in hippocampal CA3-CA1 synapses.[7] In terms of the skeletal system, systematic reviews and meta-analyses have identified a statistical association between hyponatremia and osteoporosis[8]; however, higher-level evidence regarding whether a causal relationship exists between the two is still lacking. To better understand the relationship between hyponatremia and osteoporosis, epidemiological studies are needed. However, observational associations in traditional epidemiological research are often susceptible to bias and confounding, potentially leading to misleading conclusions.[9] To address this issue, we employed Mendelian randomization (MR), a genetic epidemiological method that uses genetic variations as instrumental variables to assess the effect of exposure on outcomes, thereby minimizing confounding factors and the influence of reverse causation.[10,11] For example, Zhao et al[12] used the MR method to reveal the association between cardiac and brain structure and function. In contrast, Zheng et al[13] employed MR to demonstrate that human herpesvirus is not associated with amyotrophic lateral sclerosis. In this study, we applied the MR approach to explore the potential causal relationship between hyponatremia and osteoporosis. 2. Methods 2.1. Data sources In this study, the exposure data for hyponatremia were retrieved from the GWAS Catalog (https://www.ebi.ac.uk/gwas/) using the search term “GCST90043666.” Researchers utilized a method based on generalized linear mixed models to analyze the relevant GWAS data of 465,348 European individuals.[14] As osteoporosis is diagnosed based on bone density measurements, we selected BMD from 4 body sites as the phenotypes. BMD measured by DXA at the forearm, lumbar spine, and femoral neck is commonly used for diagnosing osteoporosis, and we obtained their genetic data from the Genetic Factors for Osteoporosis Consortium (http://www.gefos.org/).[15] Additionally, quantitative ultrasound-estimated heel BMD is closely correlated with DXA-measured BMD.[16] Outcome data for heel BMD of 265,627 European individuals were retrieved from the online platform of the Integrative Epidemiology Unit Open GWAS Project (https://gwas.mrcieu.ac.uk/) using the search code “ukb-b-8875.” There was no evidence suggesting a causal relationship between myopia and osteoporosis; therefore, we selected genetic data from 460,536 European individuals from the Integrative Epidemiology Unit platform as the negative control group, using the search code “ukb-b-6353” (Table 1). Table 1Sources of genetic data.CategoriesSample sizeSourceSearch formulaHyponatremia465,348 https://www.ebi.ac.uk/gwas/ GCST90043666FA BMD8134 http://www.gefos.org/ FN BMD32,735 http://www.gefos.org/ LS BMD28,498 http://www.gefos.org/ HE BMD265,627 https://gwas.mrcieu.ac.uk/ ukb-b-8875Myopia460,536 https://gwas.mrcieu.ac.uk/ ukb-b-6353BMD = bone mineral density, FA = forearm, FN = femoral neck, HE = heel, LS = lumbar spine. Sources of genetic data. BMD = bone mineral density, FA = forearm, FN = femoral neck, HE = heel, LS = lumbar spine. 2.2. Selection of instrumental variables Instrumental variables must meet the 3 fundamental assumptions of MR analysis: they should be strongly associated with the exposure; they should be independent of any known confounding factors; they should only influence the outcome through exposure (Fig. 1). The selection criteria for instrumental variables are as follows: we select independent (r2 < 0.001) SNPs that are strongly associated with the exposure (P < 5 × 10−8) as instrumental variables. However, this criterion results in a small number of SNPs, so we adjust the threshold to 1 × 10−5 for SNP selection.[17,18] Subsequently, we identify SNPs associated with the outcome. If there are no corresponding SNPs in the outcome, we may consider using highly correlated proxy SNPs (r2 > 0.8). To ensure that the corresponding alleles of the exposure and outcome are aligned, we strive to coordinate the effects of these SNPs as much as possible. Then, we removed variants with F values (calculated as F = β2exposure/SE2exposure) <10 and SNPs with incompatible alleles.[19] The 3 basic assumptions of MR analysis. BMD = bone mineral density, MR = Mendelian randomization. 2.3. MR analysis This study utilized the TwoSampleMR package in R software (version 4.3.2).[20] MR analysis includes 5 methods: inverse variance weighted (IVW), weighted median, MR-Egger, simple mode, and weighted mode. Among them, the IVW method is the primary approach for estimating the causal effect between exposure and outcome. It calculates the ratio of the magnitude of the SNP-outcome association to the magnitude of the SNP-exposure association. We primarily used the inverse variance weighted method to assess causality between exposure and outcome.[21,22] When there is no pleiotropy (i.e., when the MR-Egger intercept P value > .05), the IVW analysis method is the most reliable.[23] The remaining 4 analyses ensure the robustness of the results. For IVW analysis with a P value <.05, we conducted MR-PRESSO analysis to remove potential pleiotropic instrumental variables. 2.4. Sensitivity analysis To further assess the robustness of our MR analysis, we conducted the following sensitivity analyses. First, we performed pleiotropy analysis. Pleiotropy analysis detects whether instrumental variables influence the outcome through factors other than the exposure. Generally, pleiotropy is assessed using the MR-Egger intercept. If the MR-Egger intercept is <0.05, it indicates the presence of pleiotropy, and the independence and exclusion assumptions are violated. Second, we conducted heterogeneity analysis. Heterogeneity analysis investigates whether genetic data from different analysis platforms and populations may affect the heterogeneity of the MR analysis. It evaluates the heterogeneity between instrumental variables using Cochran Q statistic. 2.5. Negative control To further ensure the reliability of our MR analysis, we used myopia as a negative control group, as there is no evidence indicating a causal relationship between myopia and hyponatremia. 2.6. Informed agreement The acquisition of genetic data in this study was typically based on data that had been previously collected in prior research, and informed consent should have been obtained from the participants during the data collection process. This means that, at the time of the original data collection, participants were fully informed about the purpose of using their biological samples and related information, and they agreed to its use for research.[24,25] 3. Results First, we obtained 27 genetic instrumental variables from the GWAS of hyponatremia (Table 1, Supplemental Digital Content, http://links.lww.com/MD/O166). The selection of these instrumental variables was based on their significant associations with the phenotype of hyponatremia, laying the foundation for subsequent MR analysis. In the pleiotropy analysis, we employed MR-Egger regression and leave-one-out analysis to assess the possibility of horizontal pleiotropy. The nonsignificant intercept of MR-Egger regression (P > .05) suggests that our MR analysis results are less likely to be biased by horizontal pleiotropy (Table 2, Figures 1–4, Supplemental Digital Content, http://links.lww.com/MD/O167). Furthermore, we conducted heterogeneity analysis to examine whether there was significant heterogeneity among the selected instrumental variables. The results showed that there was no significant heterogeneity among the instrumental variables, indicating that our selected instrumental variables were statistically consistent and had similar effects on hyponatremia (Table 2). Negative control analysis further confirmed the rationality of our instrumental variable selection. The finding that hyponatremia is unrelated to myopia suggests that our instrumental variables are specific to hyponatremia rather than nonspecific factors affecting multiple phenotypes (Table 2, Supplemental Digital Content, http://links.lww.com/MD/O166). Table 2Mendelian randomization estimates for exposure on outcome.ExposureOutcomeNo. of IVsHeterogeneity testsDirectional horizontal pleiotropy testMR resultsMethodsCochran Q (p)MR-Egger intercept (p)Method β P HyponatremiaFN BMD23MR-Egger30.66 (0.08)0.75MR-Egger−8.87e-05.99Weighted median−3.47e-03.60Inverse variance weighted30.82 (0.10)IVW−3.17e-03.54Simple mode−2.87e-02.09Weighted mode1.91e-02.22HyponatremiaFA BMD23MR-Egger53.07 (1.34e-04)0.71MR-Egger0.02.49Weighted median4.02e-04.98Inverse variance weighted53.44 (1.97e-04)IVW0.01.44Simple mode−4.29e-03.89Weighted mode-2.71e-03.92HyponatremiaLS BMD23MR-Egger21.18 (0.45)0.81MR Egger−2.94e-03.78Weighted median−6.03e-04.94Inverse variance weighted21.34 (0.51)IVW−7.68e-04.88Simple mode−1.94e-02.21Weighted mode−2.58e-03.86HyponatremiaHE BMD21MR Egger29.82 (0.05)0.20MR Egger−2.89e-03.44Weighted median9.07e-04.69Inverse variance weighted32.59 (0.04)IVW1.27e-03.50Simple mode−4.99e-03.35Weighted mode−4.76e-03.31BMD = bone mineral density, FA = forearm, FN = femoral neck, HE = heel, IVs = instrumental variables, LS = lumbar spine. Mendelian randomization estimates for exposure on outcome. BMD = bone mineral density, FA = forearm, FN = femoral neck, HE = heel, IVs = instrumental variables, LS = lumbar spine. In MR analysis, IVW analysis at any site did not reveal a causal relationship between hyponatremia and BMD (Table 2), and when there is no evidence of pleiotropy, the IVW analysis is considered the most reliable method in MR analysis. Additionally, the other 4 MR methods also did not identify a causal relationship between hyponatremia and BMD. 4. Discussion In this study, we used 2-sample MR studies to determine whether there is an association between hyponatremia and osteoporosis. Despite utilizing a large amount of publicly available GWAS meta-analysis data, we were unable to demonstrate an association between the genetic instruments for hyponatremia and osteoporosis. Therefore, based on our analysis, there is no evidence to suggest a causal relationship between hyponatremia and osteoporosis. Previous studies have indicated a correlation between hyponatremia and osteoporosis. Verbalis et al[26] found in an animal model that sustained hyponatremia led to a significant decrease in BMD. Specifically, they observed a decrease of approximately 30% in BMD in rats exposed to hyponatremic conditions for 3 months. This finding suggests a potential direct impact of hyponatremia on bone health, although the exact mechanism remains to be investigated further. Additionally, analysis of National Health and Nutrition Examination Survey III data supported the association between hyponatremia and increased risk of osteoporosis. Using a linear regression model, the study revealed a correlation between hyponatremia and decreased T-scores in the hip and femoral neck among female patients. This may imply a relationship between hyponatremia and bone loss.[26] Another historical cohort study analyzed postfracture follow-up data from 5160 female patients and found a significant correlation between hyponatremia and decreased BMD in the hip and femoral neck, although no association was found with lumbar spine BMD. This suggests a potentially more significant impact of hyponatremia on specific skeletal sites.[27] Usala et al[28] further emphasized the effect of hyponatremia in diabetic patients. Their case-control study revealed an independent association between hyponatremia and increased risk of fragility fractures, suggesting that hyponatremia may be an additional risk factor for osteoporosis and fragility fractures in diabetic patients.[28] However, evidence regarding the direct impact of hyponatremia on BMD remains lacking. For instance, a retrospective study suggested that the association between hyponatremia and osteoporosis may not be causal but may instead indicate sodium levels as a marker of disease severity in elderly patients. This suggests that hyponatremia may reflect a variety of health conditions in elderly patients rather than directly causing osteoporosis.[29] Additionally, research on elevated levels of circulating arginine vasopressin in hyponatremic patients has provided another potential explanation. While the increase in arginine vasopressin levels may be related to bone loss, its potential role in the relationship between hyponatremia and osteoporosis cannot be ruled out.[30] In investigating the contentious views and potential confounding factors regarding the relationship between hyponatremia and osteoporosis, this study employed MR to elucidate the causal relationship between the two. To achieve this, we utilized the largest-scale GWAS data to date on hyponatremia and BMD, selecting multiple genetic variants significantly associated with both conditions. Through this approach, we enhanced our ability to detect causal effects. In contrast to traditional small-scale studies based on individual-level data, our study utilized data from large-scale GWAS databases to more precisely evaluate the effect size between hyponatremia and osteoporosis. The advantage of this method lies in its ability to mitigate common confounding factors found in traditional epidemiological research, such as age, lifestyle, and environmental factors, thereby providing a more reliable pathway for causal inference. Following the MR analysis, we did not observe a direct causal link between hyponatremia and osteoporosis. This finding challenges previous research conclusions and suggests the presence of potentially overlooked confounding variables. For instance, age is a well-known significant risk factor for osteoporosis, while sodium levels may reflect the severity of diseases in elderly patients. In light of this, future studies should pay closer attention to these potential confounders to better understand the direct pathogenic factors of osteoporosis. This endeavor will aid in the development of more precise preventive and therapeutic strategies, thereby addressing osteoporosis as a public health concern more effectively. However, our study has some limitations. First, the research findings are based on genetic data from European populations. This implies that our conclusions may be constrained by the specific genetic background and environmental factors of these populations. Genetic differences exist among different populations, which may affect the validity and generalizability of genetic instruments. Therefore, our findings may not be applicable to other racial or regional populations, particularly those with significantly different genetic backgrounds than European populations. Second, the genetic summary data for hyponatremia are derived from case-control GWAS studies with potential biases. This imbalance may lead to sample selection bias, thereby affecting the statistical power and stability of the results. In case-control studies, if the selection criteria for cases and controls are inconsistent, or if there are other forms of systematic error, the study results may be biased. Additionally, if the distribution of genetic variation between cases and controls is uneven, biases may be introduced in MR analysis, whether directional or nondirectional. In conclusion, our study suggests that there may not be a causal relationship between hyponatremia and osteoporosis, and the previously observed associations may be confounded by underlying factors, which warrants further investigation. Author contributions Conceptualization: Zhang Hongxing. Writing—original draft: Zhang Hongxing. Methodology: Peng Jiafeng, Gao Minglei. Validation: Peng Jiafeng. Software: Gao Minglei. Project administration: Li Danyang. Writing—review & editing: Li Danyang. Supervision: Junchen Zhu. Supplementary Material