Note: The cases discussed in this post happened before COVID-19. Given that D-Dimer may be elevated in patients with COVID-19, the pretest probability of PE in your patient population may be elevated at this time. Thus, the assumptions derived from Kohn et al’s paper highlighted in this post may not be applicable in the current environment [1].
Ah, the D-dimer. A little alliteration with the power to trigger even the “most chill” attending when ordered inappropriately by a well-intentioned intern. I’ll never forget the laughter during rounds the morning after I ordered a D-dimer on a patient with increasing oxygen requirements and tachypnea while alone on an overnight shift in the medical ICU. The D-dimer was, predictably, elevated, as it generally is in critically ill patients. We ordered a CTPA and found no clots but a whopping pleural effusion. I became wary of D-dimers.
Generation of D-dimer from degradation of cross-linked fibrin [2].
Recently in the ED I had a case of chest pain. The patient had diabetes and hypertension and a history of prostate cancer. His HEART score was 6 as he had an elevated troponin, new T wave inversions on EKG, age, and a few risk factors [3]. His Wells score was 2.5 points (tachycardia and malignancy), making him moderate risk with a pretest probability for PE (prevalence of PE in the moderate risk group) of 16.2% [4]. A bedside echo showed a normal ejection fraction and no pericardial effusion. Unfortunately, cardiac windows were limited and I was unable to get a good parasternal short or apical four chamber to assess for right heart strain [5].
Given the moderate risk for PE by Wells Score, I ordered a D-Dimer. As I clicked on the order, I remembered my embarrassing moment in the MICU. If this man was having acute coronary syndrome, could the D-dimer be elevated anyway? But if it was REALLY elevated, would that mean he is more likely to have a PE? Should we treat the D-dimer as a dichotomous variable – i.e. positive or negative – or should we be thinking of it as a continuous variable such that a higher D-dimer value gives more credence for PE than a lower D-dimer value even if both are considered positive?
What does the literature say about D-dimer cut-offs? The idea of adjusting D-dimer cut-offs to assess PE risk is not new. There are a few landmark studies that sought to reduce ED imaging done to diagnose PE by altering the D-dimer cut-off either based on age or risk. The idea of adjusting the D-dimer cut-off based on age originated in a 2010 study by Douma et al [6]. The ADJUST-PE study is the most recent validation study on age-adjusted D-dimers [7]. In 2017, the YEARS algorithm was developed to combine a clinical probability tool with adjusted D-dimer cutoffs and it has been validated in pregnancy [8,9]. In 2019, the PEGeD study was published and, similar to YEARS, it combines a clinical decision tool for risk assessment with an adjusted D-dimer threshold [10]. PEGeD has not been validated, so stay tuned for that development in the literature [11]. Check out these studies in the references to take a deep dive into D-dimers.
How can we apply likelihood ratios to the concept of D-dimer intervals to assess post-test probability of PE? Kohn et al’s paper applies another useful clinical tool to the concept of the adjusted D-dimer cut-off: the likelihood ratio [1,12]. In their paper, Kohn et al used pooled data from five PE management studies to estimate interval likelihood ratios (iLR) for eight D-dimer intervals with boundaries of 250, 500, 750, 1000, 1500, 2000, and 5000 ng/mL [1]. The conventional D-dimer cutoff of 500 ng/mL was used in all but one of the studies, which used an age-adjusted threshold. 500 ng/mL is the “conventional cutoff” if your lab reports Fibrinogen Equivalent Units (FEU’s). If your lab reports D-dimer Units (DDU), your cut-off is closer to 250 ng/mL. For more on FEU’s vs. DDU’s check out this post [11]. Using the previously validated Wells score to determine a pretest probability, a clinician can apply Kohn’s likelihood ratios for a particular D-dimer interval to a Fagan nomogram to find the post-test probability of PE [14,15]. An example: my chest pain patient had a calculated Wells score of 2.5 points which places him in the moderate risk group with a PE prevalence of 16%. His D-dimer was 3000, which falls in the 2500-4999 interval associated with an approximate iLR of 4. Plugging in 16% and LR of 4 into a Fagan nomogram shows that the probability of PE is between 40-50%. Applying a Fagan Nomogram to Kohn et al’s approach: pretest-probability (Wells Score) on the left, iLR in the center and probability of PE on the right [15]. Kohn et al’s data table for iLR’s and their associated D-dimer intervals [1]. Of note, Kohn et al used logistic regression to develop a fitted estimate of the iLR for each interval [1]. The purpose of this additional analysis was to fit the data to a model in which the iLRs increase by a constant ratio from one interval to the next. The authors used PE as the outcome and D-dimer interval as the predictor to calculate an estimated between-interval ratio of 2.0, so the iLR values increase by a ratio of 2 as you move up from one interval to the next. While this approach resulted in clean numbers for plugging into nomograms/calculations for determining the post-test probability of PE, it also resulted in a limitation of the study. For two of the D-dimer intervals (D-dimer between 250-500 and D-dimer > 5000), the fitted iLR estimate is higher than the point iLR estimate, so using the fitted iLR instead of the actual iLR will yield a higher post test probability of PE for either of those two intervals. Kohn et al’s approach views D-dimer as an ordinal rather than dichotomous data in PE management. Using likelihood ratios provides a numerical data point to support or refute the need for further testing in a moderate or low-risk patient. Kohn et al found that the likelihood ratio of 0.98 (essentially 1) for the D-dimer interval of 1000-1500 means that a PE for someone with a D-dimer within this range is neither more or less likely to have a PE than their pretest probability [1]. Of course, if their pretest probability is already high, you should not even be ordering the D-dimer. But if a pretest probability is somewhere between 0.33 and 33%, only a D-dimer > 1500 will increase the post-test probability for PE, and for any D-dimer value less than 1000, the D-dimer will decrease the post-test probability. 
Take Away Points:
-Think of a D-dimer as an ordinal variable rather than a dichotomous one if PE is on your differential.
-D-dimer < 1000 decreases your pretest probability of PE. D-dimer > 1500 increases your pretest probability of PE.
-Use a Fagan nomogram to get an actual percentage of the post-test probability using the pretest probability (from Wells) and the iLR from Kohn et al’s paper [1].
Note: A bedside lung/cardiac US in patients with significant chest pain, shortness of breath, or hypoxemia is a great place to start gathering information. Aside from quickly identifying pleural/pericardial effusion and pulmonary congestion, bedside lung/cardiac US can detect right ventricular strain to help determine high pre-test probability and obviate D-dimer testing [5].
References 1. Kohn MA, Klok FA, van Es N. D-dimer Interval Likelihood Ratios for Pulmonary Embolism. Acad Emerg Med. 2017 Jul;24(7):832-837. doi: 10.1111/acem.13191. Epub 2017 Jun 14. Review. PubMed PMID: 28370759.
2. Siemens-healthineers.com. n.d. D-Dimer Testing. [online] Available at: <https://www.siemens-healthineers.com/point-of-care-testing/featured-topics-in-poct/cardiac-featured-topics/d-dimer-testing> [Accessed 1 June 2020].
3. Six AJ, Backus BE, Kelder JC. Chest pain in the emergency room: value of the HEART score. Neth Heart J. 2008;16(6):191‐196. doi:10.1007/BF03086144
4. Wells PS, Anderson DR, Rodger M, et al. Excluding pulmonary embolism at the bedside without diagnostic imaging: management of patients with suspected pulmonary embolism presenting to the emergency department by using a simple clinical model and d-dimer. Ann Intern Med. 2001;135(2):98‐107. doi:10.7326/0003-4819-135-2-200107170-00010
5. Jiang L, Ma Y, Zhao C, et al. Role of Transthoracic Lung Ultrasonography in the Diagnosis of Pulmonary Embolism: A Systematic Review and Meta-Analysis. PLoS One. 2015;10(6):e0129909. Published 2015 Jun 15. doi:10.1371/journal.pone.0129909
6. Douma RA, le Gal G, Söhne M, et al. Potential of an age adjusted D-dimer cut-off value to improve the exclusion of pulmonary embolism in older patients: a retrospective analysis of three large cohorts. BMJ. 2010;340:c1475. Published 2010 Mar 30. doi:10.1136/bmj.c1475
7. 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 [published correction appears in JAMA. 2014 Apr 23-30;311(16):1694]. JAMA. 2014;311(11):1117‐1124. doi:10.1001/jama.2014.2135
8. 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 correction appears in Lancet. 2017 Jul 15;390(10091):230]. Lancet. 2017;390(10091):289‐297. doi:10.1016/S0140-6736(17)30885-1
9. van der Pol LM, Tromeur C, Bistervels IM, et al. Pregnancy-Adapted YEARS Algorithm for Diagnosis of Suspected Pulmonary Embolism. N Engl J Med. 2019;380(12):1139‐1149. doi:10.1056/NEJMoa1813865
10. Kearon C, de Wit K, Parpia S, et al. Diagnosis of Pulmonary Embolism with d-Dimer Adjusted to Clinical Probability. N Engl J Med. 2019;381(22):2125‐2134. doi:10.1056/NEJMoa1909159
11. Berland, N. Working up PE in the ED: Negative Likelihood Ratios and Fagan Nomograms [Internet]. County EM Blog. 2020 [cited 2020 Jun 10]; Available from: http://blog.clinicalmonster.com/2020/02/28/working-up-pe-in-the-ed-negative-likelihood-ratios-and-fagan-nomograms/
12. Goldstein, I. Biostats/Bored=Review [Internet]. County EM Blog. Available from: http://blog.clinicalmonster.com/2015/12/09/biostatsboredreview/
13. Wolf SJ, McCubbin TR, Feldhaus KM, Faragher JP, Adcock DM. Prospective validation of Wells Criteria in the evaluation of patients with suspected pulmonary embolism. Ann Emerg Med. 2004;44(5):503‐510. doi:10.1016/j.annemergmed.2004.04.002
14. Fagan TJ. Letter: Nomogram for Bayes theorem. N Engl J Med. 1975;293(5):257. doi:10.1056/NEJM197507312930513
15. Schwartz A. Diagnostic Test Calculator [Internet]. 2002 [cited 2020 Jun 10]; Available from: http://araw.mede.uic.edu/cgi-bin/testcalc.pl.
16. Brown MD, Reeves MJ. Evidence-based emergency medicine/skills for evidence-based emergency care. Interval likelihood ratios: another advantage for the evidence-based diagnostician. Ann Emerg Med. 2003 Aug;42(2):292-7. doi: 10.1067/mem.2003.274. Review. PubMed PMID: 12883521.
From the Archives:
D-dimer for aortic dissection screening: is it ADvISEDable? A (relatively) brief lit review
Working up PE in the ED: Negative Likelihood Ratios and Fagan Nomograms
Andrea Greene
Latest posts by Andrea Greene (see all)
- Just Another Fast Track Lip Lac? - September 18, 2020
- Learning in a Low-Volume ED - June 26, 2020
- Toxic shock-like syndrome and COVID-19: A case report of multisystem inflammatory syndrome in children (MIS-C) - June 18, 2020
0 Comments