Nearly every strategy addressing the diagnosis of pulmonary embolism (PE) revolves around the D-dimer test. These crosslink fragments resulting from the cleaving of fibrin mesh by plasmin have doomed many an unsuspecting soul to computed tomography pulmonary angiograms (CTPAs). The oft-lamented primary challenge associated with dependence upon D-dimer is its lack of specificity. The list of underlying conditions resulting in elevated levels of circulating D-dimer is extensive and includes:
- Increasing age
- African-American ethnicity
- Postoperative states
- Autoimmune and connective tissue disorders
- Smoking and illicit drug use
The first item on the list, increasing age, has been recently addressed by the reasonable approach of age-adjusting the D-dimer for every year over the age of 50.1 However, the other items on the list are simply part of the smorgasbord of collateral damage in our quixotic quest to identify every last PE.
What if there were a better way? What if we could make D-dimer great again? The answer is as unremarkable as it is obvious: Just like age adjustment, simply increase the commonly used dichotomous cutoff. Fewer D-dimer results above the testing threshold will result in fewer CTPAs and, by association, reduced harms and costs from unnecessary imaging. This is hardly a new concept—it’s more that the gradually enlarging body of evidence supporting such a strategy is novel. Clinician researchers from both sides of the Atlantic have called for a doubling of the D-dimer threshold over the last five years, with caveats.
The core concept underpinning this strategy stems from observations relating to the use of D-dimer as a continuous variable rather than a dichotomous cutoff. A recent study evaluated the use of interval likelihood ratios (the probability of a result in that interval for a disease-positive patient divided by the probability of a result in that same interval for a disease-negative patient) for D-dimer and found that even values in the interval between 750 and 1,000 ng/mL decreased the likelihood that a PE was present.2 Rather than using the typical 500 ng/mL basis as a cutoff to obviate the indication for further investigation, this implies there is more nuance to the testing threshold.
The most important cognitive consideration is to remember the goal of emergency department evaluation is not to rule out PE. Even the imprecisely named Pulmonary Embolism Rule-Out Criteria (PERC) do not strictly rule out PE. The various algorithms and decision instruments are built around the idea of reducing the posttest likelihood of PE to a point at which the harms of diagnosis and overdiagnosis outweigh the mortality benefit. Models for PERC were developed out of a test threshold of approximately 1.8 percent, meaning the acceptable miss rate for PE based on their assumptions is nearly 1 in 50. The difficulty with such models, however, is their dependence on estimates for prevention of morbidity and mortality. Unfortunately, the foundational evidence for these estimates can be traced back to a parallel group analysis dating to 1960 that compared 35 hospitalized patients with acute right heart failure and pulmonary infarction.3 It is a fantastic leap to generalize submassive PE to the segmental and subsegmental disease commonplace in modern practice, but these and other antiquated observational data are our source for estimates of harm from missed diagnoses.
Consideration of these outdated harm estimates forms the basis of the argument made by Kline et al in their proposal to double the D-dimer threshold.4 In a prospective, observational evaluation of 678 patients evaluated for PE, the overall miss rate in a “PE unlikely” population of Wells ≤4 or Revised Geneva Score ≤6 was 3.8 percent at a cutoff of 500 ng/mL. Increasing this threshold to 1,000 ng/mL increased the miss rate to 5.4 percent in this cohort, but 10 of the 11 missed PEs were subsegmental and none had concomitant deep venous thrombosis. Considering the current American College of Chest Physicians guidelines suggest clinical surveillance over anticoagulation for properly selected patients with subsegmental PE, the clinical importance of these “false negatives” is questionable.
While the argument made by Kline and colleagues is solely theoretical, at least two other articles describe putting this concept into practice. The Kaiser Permanente Hawaii region describes increasing the D-dimer threshold from 400 ng/mL to 1,000 ng/mL as part of an institutional effort to reduce unnecessary testing.5 Its retrospective review of patients evaluated for PE found zero missed from a cohort of 182 low- and intermediate-risk patients. Subsequently, after practice change, ongoing quality assessment did not detect any missed PEs in a similar cohort of 47 patients using 1,000 ng/mL as a threshold.
Finally, the second article prospectively describing practice using a threshold of 1,000 ng/mL comprises a Dutch effort directly following the Kline analysis.6 These authors performed their own retrospective and observational prospective analyses to clarify which elements of risk stratification conveyed the greatest risk with relation to pretest likelihood of PE.7,8 Their resulting protocol, the YEARS algorithm, was then prospectively evaluated across 12 Dutch hospitals across almost two years. Their results were published in The Lancet in May.
Their algorithm dramatically simplifies the approach to risk stratification and testing. All patients considered for PE are tested using D-dimer. Patients with one of the three YEARS high-risk items—clinical signs of deep vein thrombosis, hemoptysis, and whether pulmonary embolism is the most likely diagnosis—could be excluded from further testing using the conventional cutoff of 500 ng/mL. Then patients without any of those high-risk features used a cutoff of 1,000 ng/mL.
In their cohort of 3,465 patients undergoing the protocol, rate of subsequent venous thromboembolism was nearly identical between those excluded from additional testing by variable D-dimer thresholds and those with the diagnosis excluded by CTPA. Of 1,633 non-anticoagulated patients in follow-up after a D-dimer below their specific threshold, only seven were ultimately diagnosed with PE in the following three months. Similarly, in follow-ups of 1,138 with negative CTPA results, eight patients were ultimately diagnosed. The face validity of nearly identical miss rates between imaging and non-imaging pathways lends substantial credibility to their algorithm.
There are a few oddities specific to the YEARS protocol precluding universal recommendation as is. Most important, D-dimer results were routinely available prior to risk stratification. This meant clinicians had access to the D-dimer result when assessing the patient as “PE is most likely diagnosis.” It is naive to expect foreknowledge of the D-dimer result did not influence their risk stratification, introducing potentially detrimental effects on internal validity and certainly on generalizability.
At the least, these prospective results, along with the retrospective and observational series, support revisiting our thresholds for D-dimer. In patients for whom PE is not likely, it is probably reasonable practice to consider D-dimer results below 1,000 ng/mL as properly further decreasing the likelihood of PE. Overtesting and overdiagnosis of PE have long been recognized as problematic, and applying even this simplest additional layer of Bayesian reasoning to testing for PE will help our patients by decreasing the costs and harms from testing without a disproportionate increase in harms from missed PE.
- Righini M, Van Es J, Den Exter PL, et al. Age-adjusted D-dimer cutoff levels to rule out pulmonary embolism: the ADJUST-PE study. JAMA. 2014;311(11):1117-1124.
- Kohn MA, Klok FA, van Es N. D-dimer interval likelihood ratios for pulmonary embolism. Acad Emerg Med. 2017;24(7):832-837.
- Barritt DW, Jordan SC. Anticoagulant drugs in the treatment of pulmonary embolism. A controlled trial. Lancet. 1960;1(7138):1309-1312.
- Kline JA, Hogg MM, Courtney DM, et al. D-dimer threshold increase with pretest probability unlikely for pulmonary embolism to decrease unnecessary computerized tomographic pulmonary angiography. J Thromb Haemost. 2012;10(4):572-581.
- Char S, Yoon HC. Improving appropriate use of pulmonary computed tomography angiography by increasing the serum D-dimer threshold and assessing clinical probability. Perm J. 2014;18(4):10-15.
- van der Hulle T, Cheung WY, Kooij S, et al. Simplified diagnostic management of suspected pulmonary embolism (the YEARS study): a prospective, multicentre, cohort study [published online ahead of print May 23, 2017]. Lancet.
- van der Hulle T, den Exter PL, Erkens PG, et al. Variable D-dimer thresholds for diagnosis of clinically suspected acute pulmonary embolism. J Thromb Haemost. 2013;11(11):1986-1992.
- van Es J, Beenen LF, Douma RA, et al. A simple decision rule including D-dimer to reduce the need for computed tomography scanning in patients with suspected pulmonary embolism. J Thromb Haemost. 2015;13(8):1428-1435.