Background— Although genetic factors contribute to the onset of heart failure (HF), no large-scale genome-wide investigation of HF risk has been published to date. We have investigated the association of 2 478 304 single-nucleotide polymorphisms with incident HF by meta-analyzing data from 4 community-based prospective cohorts: the Atherosclerosis Risk in Communities Study, the Cardiovascular Health Study, the Framingham Heart Study, and the Rotterdam Study. Methods and Results— Eligible participants for these analyses were of European or African ancestry and free of clinical HF at baseline. Each study independently conducted genome-wide scans and imputed data to the ≈2.5 million single-nucleotide polymorphisms in HapMap. Within each study, Cox proportional hazards regression models provided age- and sex-adjusted estimates of the association between each variant and time to incident HF. Fixed-effect meta-analyses combined results for each single-nucleotide polymorphism from the 4 cohorts to produce an overall association estimate and P value. A genome-wide significance P value threshold was set a priori at 5.0×10 ⁻⁷ . During a mean follow-up of 11.5 years, 2526 incident HF events (12%) occurred in 20 926 European-ancestry participants. The meta-analysis identified a genome-wide significant locus at chromosomal position 15q22 (1.4×10 ⁻⁸ ), which was 58.8 kb from USP3 . Among 2895 African-ancestry participants, 466 incident HF events (16%) occurred during a mean follow-up of 13.7 years. One genome-wide significant locus was identified at 12q14 (6.7×10 ⁻⁸ ), which was 6.3 kb from LRIG3 . Conclusions— We identified 2 loci that were associated with incident HF and exceeded genome-wide significance. The findings merit replication in other community-based settings of incident HF.