CTPA – Is This Test Just a Little Too Good?

1 Comment

CTPA rmOur ability to detect PEs has increased over time, but with hypersensitive testing we risk overdiagnosis, which can cause more harm than good

Our ability to detect PEs has increased over time, but with hypersensitive testing we risk overdiagnosis, which can cause more harm than good



I never thought I would be writing about a test that was too good.  But it is clear that CT pulmonary angiograms (CTPAs) fit the bill. But first some background.  

Perhaps 20 years ago clinicians were told that we were missing lots of PEs and that we should be much more attuned to the various and subtle presentations of this potential killer.


At the time not only was our threshold for considering the PE diagnosis too high, but the test that we had to make the diagnosis (V/Q scans), was thought to be too insensitive and not accessible enough – particularly during nights and weekends.

Within recent years we’ve had three diagnostic components that have driven up testing for PEs – risk scores (Wells, Geneva, etc), high-quality d-dimer tests and ready access to CT pulmonary angiograms.  For some of us more “senior” clinicians the hardest of the three to perform is calculation of the risk scores. Checking boxes to order d-dimers and CTPAs is, unfortunately, way too easy compared to risk scoring.

So now it seems that emergency providers are aware that patients who are short of breath or have pleuritic chest pain or have a variety of other symptoms may have a PE – and that it’s bad to miss the diagnosis. As a result, the number of CTPAs performed in the country’s EDs over the last decade has soared (an HMO survey noted a 14-fold increase in the ordering of CTPAs between 2001 and 2008 and a 52% decrease in V/Q scanning). And, unfortunately, the literature repeatedly documents that clinicians often are not justified in obtaining them.

Most authorities stress that it is imperative to risk stratify patients regarding the likelihood that a patient has a PE before ordering a d-dimer or CTPA.  This can be done via clinical gestalt or use of the various validated risk stratifying instruments.  


The advantage of gestalt is that there are no risk score equations to remember.  The disadvantage is that physicians vary in their abilities. Studies have shown, on average, that physicians are good at assessing risk for a PE.  But some are excellent, some are average and some are poor.  Using validated instruments helps eliminate the likelihood of poor clinician judgment.

Despite being very straight forward – low risk patients get a d-dimer and if it is negative the work-up is over (or low-risk patients who are PERC-negative have a PE ruled out) and all others get imaging – studies have demonstrated that a remarkable number of physicians order imaging without stratifying and without ordering a d-dimer in low risk patients or use the PERC rules.  

Here’s a recent paper from the EMA database indicating, in a very large study, that clinicians followed the diagnostic approach noted above in only about half of the cases and inappropriate CTPAs were done in about 80% of patients not meeting criteria.  There are easily four other papers in the EMA database demonstrating substantial deviation from established guidelines with unindicated ordering of CTPAs.

Adams, D.M., et al, Am J Med 126(1):36, January 2013
BACKGROUND: The PIOPED II investigators and various professional societies have recommended initial assessment of the pretest probability in patients with possible PE, D-dimer testing when the probability is low, and avoidance of CT pulmonary angiography (CTPA) when the pretest probability is low and D-dimer results are negative.

METHODS: The authors, from Intermountain Medical Center in Utah, performed an implicit chart review of 3,500 consecutive patients who had a CTPA performed for possible PE in two urban EDs. The pretest probability was based on attempted application of the Revised Geneva Score (with PE considered unlikely with a score of 10 or lower).

RESULTS: PE was diagnosed in 9.7% of the patients overall. Performance of CTPA was concordant with PIOPED II recommendations in 45.5%. Of the patients for whom CTPA was not concordant with recommendations, 83% involved a patient clinically unlikely to have PE who had no D-dimer testing, while the remainder involved a patient with both an unlikely pretest probability and a negative D-dimer. PE was diagnosed in 37.1% of patients with a likely pretest probability and 10.0% felt to be clinically unlikely but who had a positive D-dimer, but only 8.4% felt to be clinically unlikely and in whom D-dimer testing was not done, and 0.9% with both an unlikely pretest probability and a negative D-dimer test.

CONCLUSIONS: These findings are consistent with a high rate of nonadherence to recommendations for performance of CTPA for possible PE. 31 references (greg.elliott@imail.org for reprints)
Copyright 2014 by Emergency Medical Abstracts – All Rights Reserved 5/14 – #27

Then there is the “too good” problem.  A CTPA is ordered and it reveals a subsegmental pulmonary embolus (a small embolus out at the periphery).  Seems treatment would be straightforward – anticoagulation.

But not so fast.  Here is just the case where the finding of a small clot poses a big problem that would likely not have occurred if CTPAs were not so good at picking up PEs.  To put it another way, despite the 80% increase in the diagnosis of PEs since introduction of CTPA, there has been essentially no change in the absolute numbers of patients dying of a PE.  This must mean we are picking up a large number of PEs that are not associated with subsequent fatal PEs – and essentially the only diagnosis that has increased is the number of subsegmental PEs (identified in 15% of CTPA imaging).  To put this into perspective, when V/Q scans were the norm, only 1% of “high probability” V/Q scans was associated with subsegmental PEs.  The bottom line – we are identifying a lot more small clots of uncertain clinical significance.

Do these small clots need to be treated?  Is the risk of future embolization greater than the bleeding risk of treatment?  Here’s an analysis making the case that treatment may result in worse outcomes than non-treatment.  Specifically, in the largest case series of patients treated with anticoagulation for isolated subsegmental PE (91 patients), the risk of major bleeding far exceeded the risk of recurrent venous thromboembolism (5.3% vs. 0.7%) (Donato, AA.,  et al, Thromb Res, 2010, pg 766-770).

There is no doubt that radiologists clearly see great anatomic detail on a CTPA, but do we want them to?  Some have suggested we go back to doing more V/Q scans just because they don’t have the ability to pick up subsegmental clots very well – strange reasoning for sure, but none the less, an option.

Wiener, R.S., et al, Br Med J 347:f3368, July 2, 2013
The authors, coordinated at Boston University, comment on the overuse of CT pulmonary angiography (CTPA) and the overdiagnosis of pulmonary embolism (PE). Although it is traditionally taught that all PE must be identified and treated, the expanding use of CTPA for any patient with a remote possibility of PE has resulted in the increased identification of subsegmental PE of uncertain clinical significance. A survey of ED physicians in 2005 found that most considered CTPA to be the first-line imaging test for suspected PE, and there has been a 14-fold increase in the use of this imaging modality from 2001 to 2008. Isolated subsegmental PE are identified in 15% of CTPA imaging compared with only 1% of “high probability” V/Q scans. The authors feel that some PE need not be identified and treated. Despite a significant 80% increase in the identification of PE during the eight years after the introduction of CTPA, PE-related mortality has remained essentially unchanged. Complications related to anticoagulation have increased. In one large series of patients treated with anticoagulation for isolated subsegmental PE, the risk of major bleeding far exceeded the risk of recurrent venous thromboembolism (5.3% vs. 0.7%). Other potential disadvantages of increased use of CTPA include an increased risk of contrast nephrotoxicity as well as the risk associated with radiation exposure. The radiation dose of CTPA is 10-15mSv compared with 2.0-2.5mSv with V/Q scanning and 5mSv for invasive pulmonary angiography. The authors suggest that no imaging is required for patients with a low-probability Wells score (below 4) and a negative D-dimer test. They further suggest consideration of V/Q scanning as an alternative first-line imaging modality for stable patients in whom PE is a possibility and the chest x-ray is normal. 50 references (rwiener@bu.edu – no reprints)
Copyright 2014 by Emergency Medical Abstracts – All Rights Reserved 1/14 – #28

So we are in a quandary – CTPAs are a great test to pick up even small PEs, but perhaps in the majority of cases treating these small PEs results in more harm than good.  The problem – we just don’t know who is at risk and who is not.  Do small PEs foreshadow the development of large PEs?  And remember, the patients who get work-ups for PEs (often inappropriately) do have symptoms that someone thinks may represent a significant PE (significant enough to show up at the ED).

So, here is a great example of the growingly popular phenomena known as “over diagnosis” – making a diagnosis of a condition that we think is pathologic but in which treatment or downstream testing may result in more harm than good. Just one more example of a way that emergency physicians need to be armed with the facts and leaders in healthcare treatment policymaking.

Richard Bukata, MD
is the editor of Emergency Medical Abstracts


1 Comment

  1. sarah mcnamee on

    Hi there
    What is the article about the 80% increase in PE diagnosis but essentially no change in mortality?
    I’m interested to know the follow up period of this?

Leave A Reply