Background: Whether statin continuation during sepsis should differ by infecting pathogen remains unresolved. Drug-target Mendelian randomization (MR) can proxy lifelong LDL-lowering to estimate causal effects on infection outcomes, but the statistical power of this design for pathogen-specific endpoints has not been formally evaluated. Methods: We conducted a pre-registered two-sample drug-target MR study using cis-variant instruments for three LDL-lowering targets - PCSK9 (6 SNPs), HMGCR (3 SNPs), and NPC1L1 (6 SNPs) - extracted from the GLGC 2021 European-ancestry LDL-C GWAS (N approximately 1.6 million). Outcomes comprised six pathogen-stratified infection endpoints and one negative control (forearm fracture) from FinnGen Release 12. Primary analysis used inverse-variance weighted MR with Benjamini-Hochberg correction across 21 tests. Sensitivity analyses included MR-PRESSO, radial MR, leave-one-out, and Bayesian colocalization (coloc.abf). Power calculations used the Burgess (2014) formula to determine minimum detectable odds ratios and required GWAS sample sizes. Results: All 21 analyses were substantially underpowered: minimum detectable odds ratios ranged from 1.78 to 106.8 at 80% power (Bonferroni-corrected alpha). No target-outcome pair reached significance after multiple testing correction (all FDR q > 0.56). Sensitivity analyses detected no horizontal pleiotropy (zero MR-PRESSO outliers) and the negative control was null (OR 0.94-1.13). Colocalization found no evidence of shared causal variants (all H4 < 0.07). Achieving adequate power for even the best-powered pair (PCSK9 x septicaemia, OR 0.85) would require an outcome GWAS approximately 20-fold larger than FinnGen R12. For rare outcomes such as pneumococcal pneumonia, required sample sizes exceeded 85 million. Conclusions: Current pathogen-specific infection GWASs are insufficient to power drug-target MR for the statin-in-sepsis question. We provide a quantitative roadmap of the outcome GWAS sample sizes required, identifying this as a priority gap for large-scale biobank harmonisation efforts. Until this gap is closed, the question of whether statin management in sepsis should be pathogen-stratified cannot be resolved by genetic epidemiology.
Hayden Farquhar (2026) studied this question.