D-dimer thresholds that depend on pretest probability of PE had highest efficiency, but highest predicted failure rate, too

Ruling out acute pulmonary embolism (PE) in various subgroups of patients using diagnostic strategies—including the Wells and Revised Geneva scores and the YEARS algorithm, combined with fixed and adapted D-dimer thresholds—are all acceptably safe. Combining these various strategies with pretest probability-dependent D-dimer thresholds had the highest efficiency, and the highest predicted failure rates as well, according to results from a large systematic review and meta-analysis.

Researchers concluded that applied adapted D-dimer thresholds to rule out suspected PE is useful, and published their results in the Annals of Internal Medicine.

“Currently recommended diagnostic strategies for suspected acute pulmonary embolism (PE) consist of a standardized assessment of the clinical pretest probability using a validated clinical decision rule (CDR) and D-dimer testing. The combination of a nonhigh clinical probability and a normal D-dimer test result safely rules out acute PE, allowing clinicians to refrain from performing imaging test,” wrote Milou A.M. Stals, MD, of Leiden University Medical Center, Leiden, the Netherlands, and colleagues.

The use of D-dimer thresholds dependent on age or clinical pretest probability has reduced the number of patients who require imaging to rule out PE from roughly 70% with the use of the fixed D-dimer threshold of 500 μg/L, to 40%-50%, they explained.

“Nevertheless, although the overall safety and efficiency of these strategies have been demonstrated in large management studies, it is also recognized that CDRs and D-dimer tests in general may be less safe and less efficient in specific patient subgroups, such as patients with renal insufficiency, patients with cancer, and elderly patients or inpatients. Thus, the preferred diagnostic strategy may be different for certain subgroups. Yet, how different CDR/D-dimer test combinations perform in relevant patient subgroups is unknown, as individual studies were often too small to perform reliable subgroup analyses,” wrote Stals and colleagues.

They conducted this systematic review and meta-analysis to assess the safety and efficiency of ruling out acute PE in patients using a combination of the Wells and revised Geneva scores with fixed and adapted D-dimer thresholds and the YEARS algorithm. After searching MEDLINE from January 1995 until January 2021, Stals and fellow researchers identified 16 studies that assessed at least one of these diagnostic strategies, which included individual-patient data from 20,553 patients.

Researchers defined safety as the diagnostic failure rate, which was VTE predicted incidence at three months after PE exclusion without baseline imaging; and efficiency as the proportion of patients classified by a given strategy as “PE considered excluded” without imaging.

Predicted failure rates were highest for those diagnostic strategies that used adapted D-dimer thresholds, ranging from 2%-4%. Overall, they were as follows:

  • Wells rule with a fixed D-dimer threshold of 500 μg/L: 0.36% (95% CI: 0.20%-0.63%).
  • Revised Geneva score with a fixed D-dimer threshold of 500 μg/L: 0.58% (95% CI: 0.37%-0.90%).
  • Wells rule with an age-adjusted D-dimer threshold: 0.76% (95% CI: 0.52%-1.1%).
  • Revised Geneva score with an age-adjusted D-dimer threshold: 1.1% (95% CI: 0.80%-1.5%).
  • YEARS algorithm with D-dimer threshold dependent on pretest probability: 1.8% (95% CI: 1.4%-2.4%).
  • Wells rule with D-dimer threshold dependent on pretest probability: 2.8% (95% CI: 2.3%-3.5%).

Predicted overall efficiency was highest in those diagnostic strategies that applied the D-dimer threshold dependent on pretest probability, and were highest with the Wells rule (47% (95% CI: 42%-52%), followed by the revised Geneva score (44%; 95% CI: 39%-50%), and the YEARS algorithm (41%; 95% CI: 36%-47%).

When combined with a fixed D-dimer threshold of 500 ug/L, the Wells rule (26%; 95% CI: 22%-31%) or the revised Geneva score (30%; 95% CI: 26%-36%) were the least efficient strategies. When using an age-adjusted D-dimer threshold, the corresponding predicted efficiencies of these two methods were 32% (95% CI: 27%-37%) and 37% (95% CI: 32%-41%), respectively.

Stals and colleagues also found that the efficiency of all strategies was highest in patients who were younger than 40 years (47%-68%), and lowest in those 80 years or older (6.0%-23%) and patients with cancer (9.6%-26%). However, with the application of pretest probability-dependent D-dimer thresholds, efficiency was considerably improved.

“Diagnosis of venous thromboembolism (VTE)—including pulmonary embolism (PE) and deep venous thrombosis—has been revolutionized over the last 2 decades by the development, validation, and dissemination of clinical decision rules that couple readily available clinical data with D-dimer levels to identify low-risk patients who do not require advanced imaging. But with multiple diagnostic strategies available, clinicians may be uncertain about which approach is best. One important question is whether to apply a fixed D-dimer cutoff (generally 500 μg/L) or to use a higher threshold in selected patients. Clearly, increasing the D-dimer cutoff will lower the number of patients who require radiographic imaging (improved specificity), but this comes with a risk for missing PE (lower sensitivity). Is this risk worth taking?” wrote Daniel J. Brotman, MD, of Johns Hopkins University, Baltimore, in his accompanying editorial.

“These results affirm what we already know about D-dimer testing for suspected VTE: Conditions that predispose to thrombosis through activated hemostasis—such as advanced age, cancer, inflammation, prolonged hospitalization, and trauma—drive D-dimer levels higher independent of the presence or absence of radiographically apparent thrombosis,” wrote Brotman, adding that patients with these conditions are at higher risk for VTE and not likely to have normal D-dimer levels.

Brotman also noted that although increasing cutoff levels for D-dimer will save some patients from unnecessary imaging, increasing the threshold also increases risks of diagnostic errors. Nevertheless, he concluded, these results are important.

“D-dimer may therefore serve as a useful risk-stratification biomarker regardless of whether PE is present: It is likely that patients who have marginally elevated D-dimer levels, regardless of the threshold used, have a better prognosis than those with extreme elevations, even when a small PE is missed. With these considerations and caveats in mind—all of which favor using higher D-dimer cutoffs in selected patients—the important work by Stals and colleagues offers reassurance that modifying D-dimer thresholds according to age or pretest probability is safe enough for widespread practice, even in high-risk groups,” he concluded.

Study limitations include differences between studies in scoring predictor items and D-dimer assays, possible differential verification bias including the classification of fatal events and subsegmental PEs, which may have caused overestimation of predicted failure rates for adapted D-dimer thresholds.

  1. Researchers compared the safety and efficiency of the Wells and Revised Geneva scores and the YEARS algorithm, coupled with fixed and adapted D-dimer thresholds, and found that all strategies had acceptable safety for diagnosis of pulmonary embolism.

  2. The efficiency of all strategies varied greatly according to combination and patient subgroup, and was highest in patients less than 40 years old and lowest in those 80 years and older and those with cancer.

Liz Meszaros, Deputy Managing Editor, BreakingMED™

This study was funded by the Dutch Research Council.

Stals reported no disclosures.

Brotman reported receiving investigator-initiated grant funding for a study, paid to Johns Hopkins, from Bristol Myers Squibb.

Cat ID: 195

Topic ID: 89,195,580,730,309,914,192,195,925