Is it safe to rule out acute pulmonary embolism with stand-alone D-dimer testing below 750 µg/L?
BEEM Bottom Line
Why is this study important?
Emergency department (ED) testing for pulmonary embolism (PE) can be complex and time consuming. This study evaluated the diagnostic utility of D-dimer testing alone, without clinical probability assessment (for example, the Wells score).
Which, if any, threats to validity are most likely to have an impact on the results and how?
This is a retrospective analysis of prospective studies with different assays and patient populations resulting in a variation in PE prevalence between studies.
Many patients with a D-dimer between the study cut-point and 750 µg/L underwent CT scanning and some patients were diagnosed with a small clinically unimportant PE; thereby, reducing the reported negative predictive value of the D-dimer test. Missing D-dimers were replaced by multiple imputation in 14% of cases, mostly in the high Wells score group, which was also the smallest group.
How do the key results compare with the current evidence?
The findings agree with prior research that shows D-dimer can exclude PE in non-high risk patients.[1] This analysis suggests that a higher cut-point of 750 µg/L could be used in low clinical probability patients. Currently, there is not enough evidence to raise the D-dimer threshold for all patients suspected for PE.
How should this study impact the care of ED patients?
Emergency physicians should continue to use clinical probability assessment (for example, the Wells rule) in combination with D-dimer to investigate patients for PE. We await further prospective validation for using a higher D-dimer cut-point in low clinical probability patients.
Study Summary
Pubmed ID
Study Reference
van Es N, van der Hulle T, Büller HR, et al. Is stand-alone D-dimer testing safe to rule out acute pulmonary embolism? J Thromb Haemost. 2017 Feb;15(2):323–328.
Study Design
Systematic review and meta-analysis with post-hoc analysis of data from 6 studies.
Population
Included: Prospective observational studies of hemodynamically stable adults with suspected acute PE evaluated in hospital. The studies had to report the Wells score by Wells ≤ 4 and Wells > 4. Patients with a non-high Wells score and normal D-dimer were followed for 3 months.
Excluded: Not described explicitly.
Intervention
Retrospective application of D-dimer cut-point 750 µg/L.
Comparison
None.
Outcomes
Primary: Sensitivity, specificity and negative predictive value of D-dimer cut-point of 750 µg/L.
Secondary: Sensitivity, specificity and negative predictive value in patients with a low, moderate, and high clinical pretest probability according to the Wells score; patients with active cancer; previous venous thromboembolism (VTE); inpatients.
Key Results
N = 7,268 patients.
| Outcome | Number | Measure (95%CI) | QUADAS | Heterogeneity |
|---|---|---|---|---|
| All Patients | ||||
| PE Incidence | 7,268 | 23% (17 to 30) | Low risk of bias | High |
| NPV | 97.2% (94.9 to 98.4) | |||
| LR– | 0.099 (0.07 to 0.13) | |||
| Sensitivity | 94.5% (91.5 to 96.4) | |||
| Specificity | 55.8% (47.6 to 63.7) | |||
| Low Probability | ||||
| PE Incidence | 2,852 | 10.1% (7.3 to 13.7) | Low risk of bias | High |
| NPV | 99.2% (98.6 to 99.5) | |||
| Sensitivity | 94.5% (90.7 to 96.8) | |||
| Specificity | 65.3% (56.9 to 72.9) | |||
| Moderate Probability | ||||
| PE Incidence | 4,061 | 28.8% (22.2 to 36.3) | Low risk of bias | High |
| NPV | 95.5% (92.0 to 97.5) | |||
| Sensitivity | 94.2% (90.2 to 96.7) | |||
| Specificity | 48.8% (38.4 to 59.2) | |||
| High Probability | ||||
| PE Incidence | 293 | 58.7% (52.8 to 64.2) | Low risk of bias | High |
| NPV | 79.3% (53.0 to 92.8) | |||
| Sensitivity | 96.3% (89.8 to 98.7) | |||
| Specificity | 20.9% (11.3 to 35.3) | |||
CI = Confidence Interval; LR– = Negative Likelihood Ratio; NPV= Negative Predictive Value; N = number of patients; QUADAS: quality assessment of diagnostic accuracy studies. 62 patients missing in study.
BEEM Critique
Risk of Bias Assessment
| Appraisers (A) | ||||
|---|---|---|---|---|
| A1 | A2 | A3 | ||
| 1 | The research question is sensible and answerable. | Maybe | Maybe | Maybe |
| 2 | The search for studies included all languages, databases, abstracts, bibliographies, and expert contact. | No | No | No |
| 3 | The search for studies was unbiased and reproducible. | Yes | Yes | Yes |
| 4 | The selection of studies was unbiased and reproducible. | Yes | Yes | Yes |
| 5 | The data abstraction was unbiased (e.g., conducted independently by 2 researchers). | Yes | Yes | Yes |
| 6 | The assessment of the quality of the primary studies was unbiased and reproducible. | Yes | Yes | Yes |
| 7 | The quality of the primary studies is high. | Yes | Yes | Yes |
| 8 | The methods used to combine the included primary studies are reported and valid. | Yes | Yes | Yes |
| 9 | The outcomes are clinically relevant. | Yes | Maybe | Yes |
| 10 | The statistical heterogeneity of the primary outcome is low (< 25%). | No | No | No |
Funding & Conflicts of Interest
Funding
None stated.
Conflicts of Interest
None stated.
Potential Threats to Validity
Chance
None identified.
Selection Bias
The authors combined results of a limited number of studies from different settings.
Measurement Bias
Many patients with a D-dimer < 750 µg/L underwent imaging, which identified clots of minimal clinical significance. The selected studies used different D-dimer assays, and in 1 study, different cut-points.
Analysis Bias
This was a post-hoc analysis with high heterogeneity in the incidence of PE
Confounding
None identified.
Footnotes
Contributors
Authors
Appraisers
Kanters D; Worster A; De Wit K & Sharif S.
Competing Interest Disclosure
Dr. de Wit - No conflicts of interest (ICMJE)
Dr. Sharif - No conflicts of interest (ICMJE)