“…Across our two severity phenotypes (hospitalized vs. non-hospitalized cases; severe vs. non-hospitalized cases), we found no new associations at P<5x10 -8 ( Supplementary Figure 1), indicating that these analyses were underpowered for genome-wide discovery. To increase power, we combined results from our COVID-19 hospitalization phenotype (2,175 cases vs. 651,047 controls) with those from two published GWAS [19,21], for a combined sample size of 5,461 hospitalized cases and 661,632 controls with no record of SARS-CoV-2 infection. In this larger analysis of disease risk, seven loci reached genome-wide significance (Figure 2), including the four highlighted by the replication analysis above (LZTFL1, MHC, DPP9 and IFNAR2) and three novel associations ( Table 1): rs79833209 near CCNG1 (5q34; MAF=2%, OR=1.54, P=2x10 -8 ); rs4782327 in ACSF3 (16q24.3; MAF=22%, OR=1.17, P=8x10 -9 ); and rs12461764 in FPR1 (19q13.41; MAF=35%, OR=1.18, P=10 -8 ).…”