Genetically Predicted COPD and Osteoporosis: Bidirectional Mendelian Randomisation with Robust Sensitivity Analyses and Colocalisation

Background: Chronic obstructive pulmonary disease (COPD) and osteoporosis are highly comorbid, but the direction and sharing of genetic liability remain unresolved. Methods: We integrated the Global Biobank Meta-analysis (GBMI) COPD summary statistics (58,559 European-ancestry cases/937,358 controls), the ReGGen East Asian bone-mineral-density meta-analysis (365,403 individuals) and FinnGen R12 (osteoporosis 10,461 cases; pathological fractures 2429 cases). Instruments were selected at p < 5 × 10−8 and pruned for linkage disequilibrium (1000 Genomes reference panels, r2 < 0.001 within 10 Mb); per-instrument F statistics, multivariable MR (MVMR) adjusting for liver-disease liability, and PI*Z-excluded (rs28929474) sensitivity MR were pre-specified. Colocalisation used the Bayesian approximate Bayes factor method in its casecontrol form with prior sensitivity p12 in {1 × 10−6, 1 × 10−5, 1 × 10−4}. Results: Genetically predicted COPD showed directionally consistent positive associations with osteoporosis (fixed-effect IVW OR = 1.226, 95% CI 1.095–1.373, p = 4.2 × 10−4; weighted-median p = 0.016; multivariable IVW adjusting for liver disease p = 0.040; PIZ-excluded OR = 1.160, p = 0.013). Under multiplicative random effects the estimate attenuated to borderline non-significance (OR = 1.169, 95% CI 0.997–1.371, p = 0.055; Cochran Q = 46.6, p = 0.008), and the fracture association was nominal and did not survive Bonferroni correction for five primary tests (p = 0.042, alpha = 0.01) (OR = 1.390, p = 0.042). The reverse directions were null (BMD-to-COPD OR = 0.930, p = 0.127; osteoporosis-to-COPD within FinnGen OR = 1.031, p = 0.503). Colocalisation supported no region reaching PP.H4 > 0.5 at any prior setting; at the SERPINA1 locus the posterior favoured H3 (PP.H3 = 0.72), i.e., distinct causal variants rather than one shared variant, and PI*Z is therefore proposed as a candidate, direction-consistent locus rather than a proven shared causal variant. Lung co-expression modules associated with FEV1% predicted (21/32 modules BH-significant; strongest module beta = −11.06 per SD, p = 5.4 × 10−21), while bone MSC differential expression and peripheral-blood single-cell (Supplementary, exploratory) analyses were exploratory and hypothesis-generating only. Conclusions: The analyses provide weak, directionally consistent, genetically supported evidence linking COPD liability to osteoporosis risk, sensitive to between-instrument heterogeneity, and identify no statistically shared causal locus. Clinical translation should be framed as hypothesis-generating; bone-health surveillance in COPD and monitoring of SERPINA1-carrier cohorts are future research directions rather than established recommendations. Robust sensitivity estimators (weighted median, weighted mode, MR-Egger, MR-PRESSO with 1000 bootstrap distributions, MR-RAPS, contamination mixture) were pre-specified and are reported in full. At the SERPINA1 locus, a targeted conditional analysis conditioned the bone-side (M13_OST) statistics on rs28929474 using LD from the 1000 Genomes high-coverage EUR panel (n = 503) and repeated the colocalisation on the conditioned statistics. Five pre-specified primary tests were corrected at alpha = 0.05/5 = 0.01 (Bonferroni): F1 remained significant (p = 4.2 × 10−4), whereas the fracture association (F2, p = 0.042; Cochran Q p = 0.997, no heterogeneity) did not survive correction and is interpreted as nominal. All outlier-robust estimators remained directionally consistent for F1: MR-RAPS OR = 1.211 (p = 0.0017), contamination mixture OR = 1.294 (95% CI 1.039–1.565, p = 0.040), weighted median OR = 1.297 (p = 0.0034), and the MR-PRESSO outlier-corrected estimate OR = 1.190 (95% CI 1.060–1.337, p = 0.0064; global heterogeneity p = 0.006).