The Selection of Instrumental Variables
We screened instrumental variables for eGFR and UACR, identifying 26,159 independent SNPs strongly associated with eGFR and 3,063 with UACR, all achieving locus-wide significance (p < 1×10−8). After excluding SNPs in linkage disequilibrium (LD), 195 eGFR-related SNPs and 58 UACR-related SNPs remained. Each SNP had an F-value > 10, ensuring robust instrumental variables.
Causal Effect of eGFR and UACR on osteoporosis
After adjusting for confounders, we conducted MR analysis using 107 SNPs associated with eGFR to assess their potential causal relationship with osteoporosis (Figure 2, Figure 3). Both IVW and MR-Egger regression indicated a potential causal effect of eGFR on the risk of osteoporosis: IVW results (OR=1.022, 95% CI=1.009-1.035, P<0.001), MR-Egger regression results (OR=1.023, 95% CI=1.002-1.045, P=0.034). Additionally, WME analysis (OR=1.015, 95% CI=0.998-1.033, P=0.089) and WM analysis (OR=1.015, 95% CI=0.998-1.033, P=0.096) showed consistent causal effect directions without statistical significance. Cochran's Q tests indicated heterogeneity (IVW: Q=149.732, P=0.003; MR-Egger: Q=149.687, P=0.003). The intercept of MR-Egger regression showed no significant difference (P=0.859), suggesting no significant horizontal pleiotropy in our analysis. No outliers were observed by MR-PRESSO analysis. "Leave-one-out" sensitivity analysis did not find a causal relationship between eGFR and osteoporosis driven by potentially driving SNPs (Figure 4).
Similarly, we selected 27 SNPs to analyze the potential causal effect of UACR on osteoporosis.(Figure 2) Results from four causal estimation models (IVW: OR=1.004, 95% CI=0.994-1.014, P=0.394; MR-Egger: OR=1.001, 95% CI=0.967-1.033, P=0.987; WME: OR=1.005, 95% CI=0.994-1.016, P=0.405; WM: OR=1.003, 95% CI=0.983-1.025, P=0.747) all indicate that an increase in UACR may potentially elevate the risk of osteoporosis, but no causal relationship is observed. The MR-Egger intercept P-value indicates no directional pleiotropy (P=0.78), while Cochran's Q tests show heterogeneity (IVW: Q=52.096, P=0.002; MR-Egger: Q=51.930, P=0.001).
Overall, our results indicate an association and potential causal relationship between CKD and OP. However, the causal relationship may be influenced by potential confounding factors.
Causal Effect of eGFR and UACR on fracture
Using Phenoscanner V2 to identify and exclude confounding SNPs affecting bone health, such as obesity, T2D, thyroid issues, etc. For eGFR and UACR, all four causal estimation models show no causal relationship. (Figure 2) We selected 122 SNPs to explore the causal effect of eGFR on fractures. In the IVW model, the odds ratio (OR) was 1.005, 95% confidence interval (CI) = 0.994-1.016, P = 0.350; MR-Egger model: OR = 1.002, 95% CI = 0.984-1.021, P = 0.835; WME model: OR = 1.005, 95% CI = 0.994-1.016, P = 0.216; WM model: OR = 1.010, 95% CI = 0.989-1.032, P = 0.359. Although the direction of the results is consistent across all four models, none of the causal effects are statistically significant. No significant heterogeneity (P = 0.664) or outliers were observed (IVW: Q = 111.316, P = 0.725; MR-Egger: Q = 111.126, P = 0.707). MR-PRESSO analysis did not identify any outliers (P = 0.332).
We also selected 27 SNPs to investigate the causal effect of UACR on fractures (Figure 2). In the IVW model, the OR was 0.994, 95% CI = 0.987-1.002, P = 0.119; MR-Egger model: OR = 0.988, 95% CI = 0.964-1.012, P = 0.330; WME model: OR = 0.993, 95% CI = 0.983-1.003, P = 0.154; WM model: OR = 0.988, 95% CI = 0.971-1.005, P = 0.180. While the direction of the results is consistent across all four models, none of the causal effects are statistically significant. No significant heterogeneity (P = 0.588) or outliers were observed (IVW: Q = 20.35, P = 0.744; MR-Egger: Q = 20.05, P = 0.775). MR-PRESSO analysis did not identify any outliers (P = 0.090).
In conclusion, after excluding confounding factors affecting bone health, there is an association between CKD and fractures, but the causal relationship is not statistically significant.
Causal Effect of eGFR and UACR on TB-BMD OF different ages
We assessed the causal impact of eGFR and UACR on total body bone mineral density (TB-BMD) across different age groups. In the primary IVW analysis, both eGFR and UACR showed no causal effects on TB-BMD in different age groups.
For eGFR, we selected 119 and 112 SNPs to investigate the potential causal effects of eGFR on TB-BMD in the age groups of 30-45 and 45-60, respectively (Table1). IVW results were as follows: TB-BMD OF 30-45: β=0.128, 95% CI=-0.536 to 0.792, P=0.706; TB-BMD OF 45-60: β=1.005, 95% CI=0.994 to 1.016, P=0.927. For UACR, we selected 34 and 32 SNPs to explore the potential causal effects of UACR on TB-BMD in the age groups of 30-45 and 45-60, respectively (Table2). IVW analysis results were as follows: TB-BMD OF 30-45: β=0.239, 95% CI=-0.155 to 0.634, P=0.234; TB-BMD OF 45-60: β=0.143, 95% CI=-0.129 to 0.415, P=0.303.
Table1: Mendelian Randomization (MR) Analysis Summary of the relationship between estimated glomerular filtration rate (eGFR) and bone mineral density (BMD) at different sites.
Exposure: eGFR
|
Outcome
|
IVW
|
MR Egger
|
Weighted median
|
Weighted mode
|
nSNP
|
β
|
pval
|
nSNP
|
β
|
pval
|
nSNP
|
β
|
pval
|
nSNP
|
β
|
pval
|
TB-BMD
(age45-60)
|
112
|
1.005
|
0.927
|
112
|
1.002
|
0.835
|
112
|
0.139
|
0.763
|
112
|
0.116
|
0.805
|
TB-BMD
(age35-40)
|
119
|
0.128
|
0.706
|
119
|
0.774
|
0.185
|
119
|
0.334
|
0.521
|
119
|
0.261
|
0.628
|
HE-BMD
|
113
|
-0.139
|
0.340
|
113
|
0.009
|
0.970
|
113
|
-0.155
|
0.077
|
113
|
-0.124
|
0.133
|
LS-BMD
|
102
|
-0.112
|
0.659
|
102
|
0.214
|
0.642
|
102
|
0.154
|
0.690
|
102
|
-0.129
|
0.791
|
FA-BMD
|
104
|
0.003
|
0.583
|
104
|
0.004
|
0.710
|
104
|
0.012
|
0.216
|
104
|
0.008
|
0.420
|
Table2: Mendelian Randomization (MR) Analysis Summary of the relationship between urinary albumin-to-creatinine ratio (UACR) and bone mineral density (BMD) at different sites.
Exposure: UACR
|
Outcome
|
IVW
|
MR Egger
|
Weighted median
|
Weighted mode
|
nSNP
|
β
|
pval
|
nSNP
|
β
|
pval
|
nSNP
|
β
|
pval
|
nSNP
|
β
|
pval
|
TB-BMD
(age45-60)
|
34
|
0.143
|
0.303
|
34
|
-0.219
|
0.644
|
34
|
0.081
|
0.690
|
34
|
0.075
|
0.820
|
TB-BMD
(age35-40)
|
32
|
0.239
|
0.234
|
32
|
-0.225
|
0.751
|
32
|
0.449
|
0.121
|
32
|
0.665
|
0.250
|
HE-BMD
|
27
|
-0.158
|
0.043
|
27
|
-0.503
|
0.060
|
27
|
-0.120
|
0.037
|
27
|
-0.129
|
0.056
|
LS-BMD
|
23
|
-0.211
|
0.109
|
23
|
-0.035
|
0.942
|
23
|
-0.186
|
0.307
|
23
|
-0.110
|
0.749
|
FA-BMD
|
35
|
-0.065
|
0.339
|
35
|
-0.730
|
0.376
|
35
|
-0.065
|
0.822
|
35
|
0.265
|
0.574
|
FN-BMD
|
31
|
-0.052
|
0.401
|
31
|
0.062
|
0.885
|
31
|
-0.052
|
0.729
|
31
|
0.148
|
0.645
|
No significant pleiotropy was found (MR-Egger, eGFR for TB-BMD OF 30-45: P=0.174; TB-BMD OF 45-60: P=0.239; UACR for TB-BMD OF 30-45: P=0.496; TB-BMD OF 45-60: P=0.250) and no heterogeneity (IVW Cochran Q, eGFR for TB-BMD OF 30-45: P=0.124; eGFR for TB-BMD OF 45-60: P<0.001; UACR for TB-BMD OF 30-45: P=0.388; UACR for TB-BMD OF 45-60: P=0.507). No outliers were observed by MR-PRESSO analysis (eGFR for TB-BMD OF 30-45: P=0.707; eGFR for TB-BMD OF 45-60: P=0.142; UACR for TB-BMD OF 30-45: P=0.243; UACR for TB-BMD OF 45-60: P=0.305).
Causal Effect of eGFR and UACR on HE-BMD
eGFR showed null association with HE-BMD in four causal estimation models (IVW:β=-0.139,95%CI=-0.424 to 0.146,P=0.340; MR-Egger:β=0.009,95%CI=-0.457 to 0.0.475,P=0.970; WME:β=-0.155, 95% CI=-0.327 to 0.017,P= 0.077; WM:β=-0.124, 95% CI :-0.283 to 0.036,P=0.133) (Table1)
In contrast, UACR demonstrated a causal effect in both IVW and WME analyses (IVW: β=-0.158, 95% CI= -0.312 to 0.005, P= 0.043; WME: β=-0.12, 95% CI= -0.327 to 0.017, P= 0.037). The other two analytical methods did not reveal a significant causal effect, but the direction of the association was consistent (MR-Egger: β=-0.503, 95% CI= -1.003 to 0.003, P=0.060; WM: β=-0.129, 95% CI= -0.333 to 0.075, P=0.056). (Table2)
All findings indicated no significant pleiotropy but revealed the presence of heterogeneity. Notably, 12 outliers were identified in the eGFR dataset, and 2 outliers were observed in the UACR dataset through MR-PRESSO analysis (eGFR: P=0.143; UACR: P=0.230).
Causal Effect of eGFR and UACR on FA-BMD
Consistently, there was also no relationship between eGFR or UACR and FA-BMD in the results of IVW (eGFR: β=-0.003, 95% CI=-0.008 to 0.015, P=0.583; UACR: β=-0.065, 95% CI=-0.633 to 0.503, P= 0.339) or three other models. All results showed no significant pleiotropy or heterogeneity. No outliers were observed by MR-PRESSO analysis (eGFR : P=0.565; UACR: P=0.346)(Table1, Table2)
Causal Effect of eGFR and UACR on LS-BMD
The IVW results indicate no causal relationship between eGFR or UACR and LS-BMD: eGFR: β= -0.112, 95% CI=-0.607 to 0.384, P = 0.659; UACR: β= -0.211, 95% CI=-0.469 to 0.047, P = 0.109. No significant pleiotropy was found (MR Egger, eGFR: P = 0.397; UACR: P = 0.250). Heterogeneity tests (IVW Cochran Q, eGFR: P < 0.01; UACR: P = 0.479) were conducted, revealing significant heterogeneity for eGFR. MR-PRESSO analysis identified one outlier for eGFR, while no outliers were observed for UACR (eGFR: P = 0.788; UACR: P = 0.121). (Table1, Table2)
Causal Effect of UACR on FN-BMD
The IVW results indicate no causal relationship between UACR and FN-BMD: UACR: β = -0.052, 95% CI = -0.343 to 0.240, P = 0.109. No significant pleiotropy was found (MR Egger, P = 0.147), and there was no evidence of heterogeneity (IVW Cochran Q, P = 0.401). MR-PRESSO analysis did not observe any outliers (P = 0.408). (Table2)