D-dimer is a peptide product of fibrinolysis and clinical biomarker of diseases involving activated coagulation, including venous thromboembolism and ischemic cardiovascular disease. Circulating D-dimer levels are heritable and heterogeneous across populations, with, for example, higher levels observed in African (AFR) ancestry relative to European (EUR) ancestry populations. To date, our understanding of genetic contributors to D-dimer variation has been limited to the European ancestry population. Here, we performed the largest, most diverse genome-wide analysis of circulating D-dimer [total n=46,031: 36,688 EUR, 7,397 AFR, 1,321 Hispanic, 654 East Asian, 7 American Indian/Alaska Native, 54 unknown]. We performed both single variant and gene-based aggregate rare variant tests in whole genome sequences (WGS) from 14,334 participants within the Trans-Omics for Precision Medicine (TOPMed) program. For single variant tests, we then meta-analyzed TOPMed-imputed genotype results from 31,697 participants within the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium using a fixed-effects, p-value based method. Single-variant analysis results revealed 6 known genetic loci associated (P<5E-9) with D-dimer, including the AFR-driven HBB sickle-cell locus, and 1 new signal on chromosome 20. Notably, the lead variant at this signal is a common (EAF=10%) intergenic variant in near-perfect correlation (r2>0.99) with rs867186, a missense variant in PROC, and several lead GWAS variants for coagulation factors and thrombotic disease. Gene-based aggregate tests implicate FGL1 missense variants (P=7.46E-06) in D-dimer regulation. Together, these loci provide new targets for functional work to disentangle mechanisms regulating fibrin production and fibrinolysis.