Seven studies met the inclusion criteria (n=3,225 patients; 1888 first unprovoked venous thromboembolism), of which two were RCTs (n=1,109 patients; 550 first unprovoked venous thromboembolism) and five prospective cohort studies (n=2,116 patients; 1,338 first unprovoked venous thromboembolism); only patients with first unprovoked venous thromboembolism were included in the analysis. All studies used independent blinded outcome assessment and had at least 12 month follow-up. Four of the cohort studies reported consecutive recruitment of patients. Two studies reported loss to follow-up (less than 2%) and two reported funding source.
Of the 907 patients with a positive D-dimer test after stopping anticoagulation, 165 (18.2%) had a recurrent venous thromboembolism in 2,462 person-years follow-up, a risk of 8.9% per year (95% CI 5.8 to 11.9). Statistically significant heterogeneity was observed (p<0.001) thought to be due to variation in the location of the initial venous thromboembolism and the D-dimer cut-offs used for positive and negative results.
Of the 981 patients with a negative D-dimer test, 74 (7.5%) had a recurrent venous thromboembolism in 2,040 person years follow-up, a risk of 3.5% per year (95% CI 2.7 to 4.3). No statistically significant heterogeneity was observed (p=0.832).
The pooled incidence rate showed the risk for recurrent venous thromboembolism in patients with a positive D-dimer test was over twice that of patients with a negative test (2.2, 95% CI 1.65 to 2.94). Publication bias was thought to be present for the positive D-dimer outcome, but not the negative outcome.
A reanalysis was undertaken in 2010 using individual patient data (see Other Publications of Related Interest); the annualised risk for recurrent venous thromboembolism was 3.7 per 100 patient years (95% CI 3.2 to 4.3) with a negative D-dimer test, and 8.8 per 100 patient years (95% CI 6.2 to 11.3) with a positive D-dimer test. A univariate regression using the cut-offs for the D-dimer test reported in the source studies, gave a hazard ratio for recurrent venous thromboembolism of 2.59 (95% CI 1.90 to 3.32); the results were not significantly altered when re-analysed with cut-offs of 500 and 250 µg/L.