Go Ahead, Age-Adjust the D-Dimer

If you include D-dimer as part of your practice to exclude pulmonary embolism, you’re probably already aware background levels gradually increase with age.  This results in further deterioration of the already poor specificity.  Happily, this publication in JAMA demonstrates it’s probably safe to gradually increase your cut-off level for D-dimer with age without sacrificing sensitivity.

These authors enrolled 3,324 patients in a cohort with ultimate prevalence for PE of 19.0%.  Of these, 2,898 fell into the “pulmonary embolism unlikely” group and underwent D-dimer testing.  817 met the traditional D-dimer <500 μg/L cut-off and no CTA was performed.  An additional 337 patients met their “age-adjusted cutoff” – “10 x patient age” for those aged greater than 50 years.  Only one patient with a negative D-dimer met their final adjudicated outcome of eventual venous thromboembolism.  The authors, therefore, conclude this strategy of age-adjusted D-dimer cut-off is safe.

However, a handful of patients were lost to follow-up, and the confidence interval for 3-month VTE, considering the sample size, ranges from 0.1-1.7% in their age-adjusted cohort.  Additionally, 7 patients died and 7 patients were evaluated for suspected VTE during the follow-up period.  Only 1 was judged by their 3 experts to have been due to VTE, but this consensus measure for their primary outcome could have profound effects on their ultimate conclusion.  Six different D-dimer assays were also used across their multi-center study, which hinders external generalizability.

But, in the end, it’s probably reasonable.  The harms and costs secondary to testing and treating false-positives and small VTE still likely outweigh any additional misses through this strategy.  Considering this comes from JAMA, it’s likely defensible in the face of peer-review to begin using age-adjusted cut-offs immediately.

“Age-Adjusted D-Dimer Cutoff Levels to Rule Out Pulmonary Embolism”
https://jama.jamanetwork.com/article.aspx?articleid=1841967

6 thoughts on “Go Ahead, Age-Adjust the D-Dimer”

  1. Ryan – great post and this could potentially have a significant impact on PE CT scanning. My only issue is something you brought up which is the variable assays. My shop, and many others based on crowdsourcing, have the more sensitive assay where the cutoff is 250 instead of 500. Seems reasonable to change it to "5 X age" as the upper limit but would love some endorsement of this strategy from experts in PE. Any thoughts?

  2. It sounds reasonable to me – but, of course, the ultimate judge of reasonableness in this situation will be your patient. My D-dimer, for example, is 400, rather than 500. I would likewise scale the age cut-off … and would then have to discuss the implications of the test results in the context of my knowledge translation.

  3. They validated age x 10. Can you use age x 13? Perhaps. You wouldn't be able to measure much a difference in outcomes, but your sensitivity would be slightly weakened.

  4. 'm a 3rd year resident in Winnipeg, Canada. I have been looking into the different assays – ours is the Hemosil D-Dimer HS with a cutoff of <230. I found this article from 2008 ;
    Performance of the automated and rapid HemosIL D-Dimer HS on the ACL TOP
    analyzer, by Salvagno et al.
    It is a lab analysis comparing the HS assay to the VIDAS (which was one of the assays used in the ADJUST-PE trial and has the <500 cutoff), and found good correlation between the two using the equation 1.3 x VIDAS – 384. This would mean that instead of using 800 as a cutoff for an 80 year old, you could use 656 etc.
    Does this make sense? I haven't seen any reasoning given when others suggest using Age x 5 or some other lower number for the assays with a lower cutoff. Let me know what you think,

    Rachel Kroeker

  5. I'm not an expert on all the different assays in use, but – assuming they have the same collinear response to serum levels – I'd imagine some sort of transposition of age-adjustment would be appropriate. There was an article I peer-reviewed for a journal a few months back that described just about the same application as you're proposing (I am not sure if it was accepted, due to other flaws).

Comments are closed.