Prospective Study of the Diagnostic Accuracy of the Simplify D-dimer Assay for Pulmonary Embolism in Emergency Department Patients: Conclusion
However, this may not occur in real practice. For example, in the present study, although the Canadian score < 2 produced a pretest probability of 2.9%, the actual measured posttest probability of PE was 1.2% (95% CI, 0.8 to 2.0%). In our population, the < 1.0% posttest probability objective was met only when a negative Simplify D-dimer result occurred when the physician’s unstructured pretest probability estimate was < 15%, which yielded a population with only a 2.7% prevalence of PE. With this combination, the actual measured posttest probability of PE was 0.7% (95% CI, 0.3 to 1.4%). We emphasize that these results are from one ED that appears to test for PE at a very low threshold.
Several factors could limit the external validity of this study. The upper limit of the 95% CI for the 0.7% post-test probability for PE after an unstructured estimate of PE < 15% plus a negative Simplify D-dimer result was 1.4%. Patients deemed low risk by clinicians were drawn from a cohort of urban ED patients with an overall 4.7% prevalence of PE. Researchers in Europe have found the prevalence of PE to be > 20% in patients referred from the ED; however, we submit that the threshold at which clinicians decide to order a d-dimer assay to rule out PE in the ED has decreased remarkably in the past 10 years, and continues to drop in the United States and also in Canada.” further
The accuracy and reliability of unstructured estimation of pretest probability assessment has been questioned. We previously found good interobserver agreement (Cohen к = 0.60) for the unstructured method of pretest probability assessment, and that this finding was independent of training level. These results have not yet been reproduced at other EDs, and the interobserver variation for the interpretation of the Simplify D-dimer assay used at the point of care in an ED has not been examined. Our ongoing research efforts may make our practitioners likely to remember the components of published decision rules, which might enhance their unstructured estimates. Fewer than one half of the patients received pulmonary vascular imaging as part of the criterion standard diagnosis. We use what we believe to be very thorough, explicit, and meticulous follow-up methodology. It remains possible that a patient who had PE could have been discharged after a false-negative d-dimer result did not seek additional medical care and reported better health at the 90-day follow-up. This effect could have resulted in an underestimate of false-negative Simplify D-dimer results. However, we also considered a patient with DVT diagnosed 89 days after study enrollment as tantamount to the diagnosis of acute PE on the day of enrollment. This effect may have resulted in an overestimate of false-negative Simplify D-dimer results. When interpreted together with the data from Hogg et al, and systematic analyses, we believe that our data allow the definitive conclusion that the Simplify D-dimer assay has a higher negative likelihood ratio than the quantitative d-dimer at a 500 ng/mL cutoff. We also conclude that the Simplify D-dimer assay must be used in a very-low-risk population to produce a posttest probability < 1.0%. Future work will directly compare the diagnostic accuracy and operational performance of the point-of-care Simplify D-dimer assay to the laboratory-based quantitative d-dimer.