Plasma D-dimer concentration rises more than 100-fold during acute deep vein thrombosis, but there are no prospective data concerning D-dimer as a risk factor for incident venous thrombosis in a general population. Incident venous thrombosis was ascertained in 2 prospective observational studies, the Atherosclerosis Risk in Communities Study and the Cardiovascular Health Study. Of 21 690 participants enrolled between 1987 and 1993, after 8 years of follow-up, D-dimer was measured using baseline stored plasma of 307 participants who developed venous thrombosis and 616 who did not. Relative to the first quintile of the distribution of D-dimer, the age-adjusted odds ratios for future venous thrombosis for the second to fifth quintiles of D-dimer were 1.6, 2.3, 2.3, and 4.2, respectively (P for trend <.0001). Following added adjustment for sex, race, body mass index, factor V Leiden, prothrombin 20210A, and elevated factor VIII coagulant activity (factor VIII:c), these odds ratios were 1.5, 2.1, 1.9, and 3.0, respectively (P for trend <.0001). Among those with idiopathic thrombosis or secondary thrombosis unrelated to cancer, the adjusted fifth quintile odds ratios were 3.5 and 4.8, respectively. By contrast, D-dimer in the fifth versus first quintile was not related to occurrence of cancer-associated thrombosis (odds ratio, 1.1). Odds ratios for elevated D-dimer were consistently elevated in subgroups defined by age, sex, race, duration of follow-up, and thrombosis type (deep vein thrombosis or pulmonary embolus). D-dimer is strongly and positively related to the occurrence of future venous thrombosis.