Introduction
As far as dangerous causes of chest pain go, acute aortic syndromes (aortic dissection, intramural thrombus, and penetrating aortic ulcer) can present a real diagnostic dilemma. They’re rare (with a prevalence of around 2-4 per 100,000 individuals), seem to present in a myriad of ways, and are invariably life-threatening. It has been estimated that as many as 39% of aortic dissections may be missed on initial evaluation (1). Does that make you clench your butt a little? Me too.
So, little wonder that there is interest in a standardized approach to diagnosing aortic dissection (AD). A diagnostic algorithm has been shown to successfully standardize the approach to pulmonary embolism (PE), wherein patients are first risk stratified using the Wells’ score prior to application of either the pulmonary embolism rule-out criteria (PERC), D-dimer, or CT angiography for low, low/medium, and high risk patients, respectively. This strategy reduces the number of CT scans by ~30%, decreases length of stay, and limits radiation and contrast exposure. Could a similar strategy improve our ability to safely rule out aortic dissection in low risk patients?
Study 1 - Nazerian et al, 2014 The first study I want to highlight is from 2014 (2). This study is based on a prospective registry of patients that were recruited over five years in the EDs of two regional cardiothoracic surgery referral centers in Italy. Each ER had volumes of ~100,000 patients/year over the study period. Patients were only included if they fulfilled ALL of three criteria: (1) presenting with chest, back, or abdominal pain, syncope, or a perfusion deficit (CNS, mesenteric, myocardial, or limb ischemia), (2) an alternative diagnosis could not be established by the treating physician after initial medical evaluation, and (3) the treating physician’s clinical suspicion of AD was high enough to request an urgent aortic imaging exam. 1,455 patients were included in the registry, and 1,035 patients were enrolled for data analysis. Patients that did not get a D-dimer as part of their work-up were excluded. The authors performed a retrospective chart review to calculate an Aortic Dissection Detection (ADD) risk score (an externally validated risk stratification tool) and used 500ng/mL as their D-dimer cutoff. Reproduced from MDcalc.com In this study, using CT angiogram (CTA) as the reference standard, aortic dissection was diagnosed in 22.5% of patients. 22.5%!! The prevalence of aortic dissection in patients with an ADD risk score (ADD-rs) of 0 was 6%. Clearly, this is not a study of “all-comers”, as evidenced by the pretty stringent eligibility criteria. In patients with an ADD-rs of 0, a D-dimer less than 500ng/ml was 100% sensitive and had a negative likelihood ratio of 0. In patients with a retrospectively defined ADD-rs of 1 (who had a AD prevalence of 26%), D-dimer was below the 500ng/ml threshold in two of 133 cases of aortic dissection (a miss rate of 1.6%). Taken together, when applied to patients with an ADD-rs ≤ 1, D-dimer sensitivity was 98.7% (95% CI of 95.3%-99.8%) and negative likelihood ratio was 0.04 (95% CI, 0.01 to 0.15), with aortic dissection missed by D-dimer at a rate of 0.8% (95% CI, 0.1 to 2.6). This study had some serious limitations. While patients were recruited prospectively, ADD risk scoring was retrospective. Additionally, a significant portion of eligible patients was not enrolled, as a D-dimer was not obtained during their work-up, making this study prone to selection bias (although in post-hoc analysis, characteristics of excluded and enrolled patients were similar). Still, using D-dimer to rule-out AD had excellent operating characteristics among patients with an ADD-rs of 0, and applying D-dimer to patients with ADD-rs ≤ 1 yielded a miss rate of 0.8%. At the lower limit of sensitivity within the 95% confidence interval of 95.3%, the miss rate does increase to 2.6%, a rate that some clinicians may not tolerate for such a potentially catastrophic disease.
Study 2 - Asha et al, 2015 Study 1 was incorporated into a systematic review and meta-analysis (3) of five studies. Four of the included studies were small, prospective, cross-sectional analytical, with only the previously discussed Nazerian et al study being a retrospective analysis of prospectively recruited patients. Importantly, the 2014 Nazerian et al study contributed by far the largest number of cases (1035 cases analyzed and the next largest study had 224). The pooled sensitivity for ruling out AD with D-dimer was 98% (95% CI, 96% to 99%) and the negative likelihood ratio was 0.05 (95% CI, 0.02 to 0.09). Remarkably low study heterogeneity was observed, with I-square values of 0.0% for the pooled results.
Study 3 - Nazerian et al, 2018 More recently in 2018, Nazerian et al (4) published a prospective multicenter study (ADvISED). This study consecutively recruited outpatients presenting to EDs in six hospitals in four countries. To be eligible, patients were required to be >18 years old presenting to the ED with at least one of the following: chest pain, abdominal pain, back pain, syncope, or signs/symptoms of perfusion deficit within the last 14 days. Patients were included in the study only if acute aortic syndrome (AAS) was considered in the differential diagnosis by the attending physician (defined as a provider-determined need to rule out AAS). Acute aortic syndrome was diagnosed if conclusive imaging, surgical findings, angiography, or autopsy identified a type A or B AD, intramural aortic hematoma, aortic rupture, or penetrating aortic ulcer. The study analyzed 1850 patients with an overall AAS prevalence of 13%, the majority of cases (~75%) being either type A or B AD. This study found that in patients with an ADD-rs ≤ 1, D-dimer was falsely negative in only about one of every 300 patients (corresponding to a miss rate of only 0.3% (95% CI, 0.1% to 1%). When applied to patients with an ADD-rs of 0 or 1, the authors calculated a sensitivity of 98.8% (95% CI, 96.4 to 99.7%) and a negative likelihood ratio of 0.02 (95% CI, 0.01 to 0.07) for D-dimer, while reducing the number of CT scans by 60% (assuming all patients would have undergone imaging). Applying the ADViSED data to a hypothetical patient population with an AAS prevalence of 5% and ADD-rs ≤ 1, a negative d-dimer yields a post-test probability of AAS of ~0.1%, and reduces the number of CTAs by ~55%. At the low end of the confidence interval, a sensitivity of 96.4% still yields a post-test probability of ~0.3%.
Study 4 - Tsutsumi et al, 2020 The last study to discuss is a recent systematic review and meta-analysis (5) in the European Heart Journal of Cardiovascular care. In this study, the investigators calculated pooled negative likelihood ratios from the studies analyzed for a variety of rule-out criteria (either using ADD-rs alone or in conjunction with D-dimer). Both of the Nazerian et al papers were included in their analysis, as well as seven other studies (all of which were retrospective). The table below applies the pooled negative likelihood ratio data that Tsutsumi et al derived, and applies it to two different hypothetical AAS prevalences. Reproduced from Tsutsumi et al (5) Based on their results, the authors suggest that in a patient population with a low AD prevalence (~5%) and an ADD-rs of 0 alone could be used to rule out aortic dissection (if we are willing to accept a miss rate of 0.8%). For patients with an ADD-rs ≤ 1 and a negative D-dimer (again in a patient population with low AAS prevalence), the miss rate in this meta-analysis was 0.1%. In the above table, Tsutsumi et al also choose to apply the various rule-out criteria to a hypothetical patient population with a 20% prevalence of AAS. They suggest that this prevalence would be representative of those found in EDs of cardiovascular or tertiary care centers (a prevalence in line with the original Nazerian et al study data (2), and obtained a failure rate of only 0.5% in using a rule-out strategy of combined ADD-RS ≤ 1 with a negative D-dimer. However, it is important to note that the true prevalence of AAS is likely far lower, as the study populations only included patients already under investigation for AAS.
Index test
Hypothetical prevalence (%)
Pooled negative likelihood ratio
Failure rate (%)
Missed cases (per 1000 patients)
ADD-RS = 0 with D-dimer
5
0.01 (0.00- 0.07)
0.05
0-1
(4 studies)
20
0.2
2
ADD-RS ≤ 1 with D-dimer
5
0.02 (0.01-0.06)
0.1
1
(4 studies)
20
0.5
5
Conclusions: ADvISE me!!
There are a few major take-home points here from the studies that we have discussed. One is that while the 2018 Nazerien et al study was large, prospective, and well-executed, this data has not been prospectively reproduced in any additional patient populations (externally validated). So, of course, any application of this algorithm to patients outside of the study group would be premature. Additionally, about half of the study population did not undergo conclusive diagnostic testing for AD. These patients were followed for 14 days, as this time course was felt to likely capture patients with undiagnosed acute aortic syndromes who developed major adverse events as a result. However, this was an assumption by the authors and not a pre-validated follow-up period.
Finally, patients in the majority of these studies were only eligible for enrollment if the treating physician already had a clinical suspicion for an AAS. In this way, this study (as well as the preceding papers we’ve discussed) only applies to patients who are presenting with “typical” symptoms of AD. If one were to inappropriately apply this diagnostic algorithm more broadly, i.e. to all chest pain patients, the approach will actually result in a significant INCREASE in the rate of CT scans. This occurs because as a rare disease the base rate of AAS is low, and many, many more false positives than true positives will be expected upon screening for AAS with D-dimer.
Thinking about this another way, pulmonary embolism occurs in 112 cases per 100,000 individuals, whereas AD occurs in 2-4 per 100,000 individuals(6). This is a 50-fold difference in prevalence. Since pulmonary embolism is relatively much more prevalent than AD, a diagnostic strategy wherein we are able to reduce the number of CT scans by 30% in each group would clearly have a much larger overall impact when trying to diagnose PE. I think the question we need to ask is, “What are we really gaining by reducing CT imaging by 50% in a very small subset of patients?” While the adverse consequences of radiation exposure and incidental findings leading to increased downstream testing will often pale in comparison to the consequences of missing an acute aortic syndrome, these complications are significant in young healthy patients who may have a very low pretest probability of AAS.
So ultimately, there is a large amount of data, including at least one well-executed prospective trial, that suggests a diagnostic strategy using ADD-rs combined with D-dimer to rule-out AAS in a low-risk segment of patients already suspected of having an AAS could safely exclude the diagnosis while significantly reducing the need for advanced imaging. Importantly, these results have thus far only been applied to patients in which there is already a suspicion for AAS, and as such should not be erroneously applied to all comers with chest pain. Lastly, this diagnostic strategy will require external validation prior to its adoption.
(Editor’s note: A 2020 meta-analysis in Acad Emerg Med may be more useful Tsutsumi et al since it only pooled data from studies that included d-dimer testing. Systematic Review of Aortic Dissection Detection Risk Score Plus D-dimer for Diagnostic Rule-out Of Suspected Acute Aortic Syndromes. PMID: 32187432 DOI: 10.1111/acem.13969)
trevorc
1 Comment
Eden · May 16, 2020 at 11:37 pm
I would argue that the ADD-rs is NOT all that helpful. what they should have done is compare it to usual care and see if it changes anything. have any of us ever suspected AD in a patient without any of the high risk features? we very rarely consider AAS in any patient with a score of 0 or 1 even without using the ADD-rs. is there really any added benefit to adding d-dimer to our clinical gestalt for something that is uncommon? Also, when you look at the d-dimer negative group who had AAS, every single one of them had something that would make a doctor consider scanning them anyway. unless this thing is proven to have benefit over clinical gestalt, you don’t know if it would really significantly reduce need for advanced imaging because they didn’t specifically look at that. the one area where this may be helpful is in inexperienced/indecisive clinicians who need assistance with MDM.