Re: “How to Avoid Missing an Aortic Dissection”

In a recent article, the authors have provided a thoughtful review of aortic dissection (AOD) cases while promoting the aortic dissection detection risk score (ADD-RS) to improve the ED diagnosis of these patients.¹  Unfortunately, the ADD-RS was not constructed using appropriate methodology for creating clinical decision rules (CDRs) and should not be relied upon.^2,3 The ADD-RS was derived by cardiologists via a consensus and retrospectively tested on patients entered into an AOD registry.^4,5 

Relying on a chart review of clinical features, potentially obtained after hospital admission, when the final diagnosis is known is unreliable and inappropriate for creation of a CDR. Importantly, that “validation study” did not follow standard chart review methodology and did not follow basic reporting guidelines for observational studies (STROBE). There was no blinding to final diagnoses, no prospective evaluation of individual criteria, no assessment of interrater reliability for individual features, no statistical evaluation of the contribution of each criterion to a final model, no purposeful model building, no inclusion of patients without AOD, and no comparison of this rule to clinical judgement. A prior meta-analysis of the ADD-RS included eight retrospective chart reviews and one prospective study with no study being registered, no study following STROBE guidelines with comprised populations already suspected of having AOD, and most patients already selected to undergo advanced aorta imaging (CT or transesophageal echocardiography).⁶ 

Importantly, flaws in patient selection and data collection within these studies would tend to over inflate the sensitivity of the ADD-RS and D-dimer in diagnosing AOD potentially missing AOD cases scored as low or indeterminate risk. Looking at individual ADD-RS features, trained emergency physicians are already aware of classic AOD high risk historical features (connective tissue disease, aortic valve disease, known aneurysm), high risk complaints (abrupt onset, tearing/ripping, severe), and examination findings (pulse deficit, new aortic regurgitation murmur, neurologic deficit, shock). Many of these “high risks” may be absent from ED medical records or unknown to patients (connective tissue disease, valvular heart disease), are subjective (onset, pain descriptors, and severity) while the interrater reliability of physical examination and subjective historical features is unknown. In particular, the suggestion that abrupt chest pain requiring morphine should be a red flag is anecdotal and not evidence based while the term “abrupt” can be interpreted in multiple different manners by patients and providers (i.e., potentially poor interrater reliability and thus useless for CDRs). While the ADD-RS might be useful to remind physicians of selected known risk factors for AOD, current studies only show this to be a potential predictor in those already suspected of having AOD. It is unknown if this score (with/without D-dimer) can accurately predict AOD in a general ED population that includes patients not initially suspected of AOD. As such, this score cannot and should not be promoted as a proven predictor of AOD in the general ED population of patients with chest pain, back pain, syncope, weakness, or the multitude of other potential presentations.