Introduction Currently the same threshold value is used to identify a positive D-dimer result for all patients presenting to our ambulatory clinic with suspected pulmonary emboli (PE). It has been suggested that adjusting the threshold value according to the pre-test probability would exclude PE in more patients than using the same cut-off point regardless of clinical probability.
Methods Data from 362 consecutive patients presenting to the ambulatory PE clinic was collected. A pre-test probability of PE was recorded for all patients and those with a high pre-test probability had radiological investigations. Patients with a low or intermediate pre-test probability had a latex agglutination D-dimer test. If this result was =0.5 μg/ml they had further investigations, otherwise they were discharged. The diagnosis of PE was made if a VQ scan showed ventilation/perfusion mismatch or CTPA report demonstrated PE. Receiver operating characteristic curve analysis was performed separately for patients with low and intermediate probability and the optimum cut-off value to exclude PE determined. Sensitivity, specificity, negative predictive value and positive predictive value for different cut-off points were determined.
Results 362 patients were included in the analysis, 207 (57%) had low, 129 (36%) intermediate and 26 (7%) high pre-test probability. Prevalence of PE was 2% in the low probability group, 14% in the intermediate probability group and 42% in the high probability group. No patients with a D-dimer of <0.5 μg/ml who were discharged without further tests have re-presented with similar symptoms. In the low pre-test probability group, a cut-off point of 1.07 improved the specificity from 64% to 89% while maintaining a sensitivity of 100% and negative predictive value of 100%. Analysis in patients in the intermediate risk group suggested that a cut-off of 0.5 μg/ml was appropriate. By adjusting the D-dimer threshold to >1.0 μg/ml in the low probability group, a further 53 patients could have been discharged home without need for radiological investigation.
Conclusion The diagnostic accuracy of D-dimer testing may be improved in patients with a low pre-test probability by adjusting the cut-off threshold.