Expanded Testing for Pulmonary Embolism Leads CT Scan Overuse

In the vast ocean of medicine, few diagnostic dilemmas descend so quickly into madness as does pulmonary embolism (PE). In the classical teaching, PE remains one of a handful of life-threatening diagnoses considered in the context of chest pain or shortness of breath. The proliferation of advanced imaging technology has also dramatically eased evaluation for PE, leading to an explosion of testing. Sadly, the cumulative effect of such expanded testing appears to be a pervasive preponderance of negative studies and low-yield, but costly, utilization.

And, frankly, it’s even worse than we’ve acknowledged.

The vast majority of PEs are diagnosed using one test, the computed tomography (CT) pulmonary angiogram. This test gained widespread acceptance with the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) studies, demonstrating adequate sensitivity for PE compared with conventional angiography.¹ Sensitivity is a valuable test attribute for a disease believed to have a high case-fatality rate. However, as technology has improved, CT has begun detecting smaller and smaller clots. By assigning the same clinical significance across the disease severity spectrum, it becomes unclear whether this improved sensitivity benefits our patients and whether our test specificity is adequate for our current strategy.

The problem is twofold, and two specialties are complicit in this predicament: radiology and emergency medicine. In radiology, the subsegmental PE is the culprit. As vessel size decreases, the quality of opacification and contrast capture diminishes. This results in consistent ambiguity regarding the presence of a flow-limiting lesion.

For example, a group of authors in Pennsylvania reviewed 415 images from their institution judged diagnostic for PE, focusing mostly on segmental and subsegmental PE.² Using five radiologists, four of whom were subspecialty trained in thoracic radiology, each image was individually re-reviewed. Based on their sample of 192 images read initially as segmental PE, a majority of authors could not agree on a positive finding in 5.7 percent of cases. For subsegmental PE, at least one reviewer dissented in 60 percent of cases. When compared with the original community radiologist’s official read, the consensus was a false-positive rate of 3.6 percent for segmental PE and 15 percent for subsegmental PE.

A second radiology department, this time in Ireland, reviewed 174 CTs reported positive for PE.³ Three subspecialty-trained thoracic radiologists subsequently reviewed each of the studies read initially by one of 15 general radiologists. In this study, 45 (25.9 percent) cases were judged erroneously reported positive, including 26.8 percent segmental and 59.4 percent subsegmental. The authors reported the most common causes of diagnostic error were technical image-acquisition artifacts underappreciated by the general radiologists.

The authors offer a few specific suggestions relevant to radiologists to improve image quality and account for technical issues, but their primary complaint was simply this: we scanned too many patients who did not have a PE. After subtracting the patients with false-positives, yield in this study was 129 of 937, or 13.7 percent, falling at the low end of most published performance characteristics. This prompted another recommendation: the best way to improve yields is to refer patients for scan only when they have a higher pretest likelihood of disease.

Referring appropriate patients for CT is, unfortunately, something we do terribly in the United States. A comparison of populations of patients evaluated for PE in several observational studies, with 3,174 patients in Europe and 7,940 patients in the United States, showed patients were reliably higher risk in European populations.⁴ Interestingly, this was most exaggerated in the clinical gestalt of treating clinicians: in Europe only a third of patients were thought to be low risk, while in the United States these totaled nearly two-thirds. The net effect in this study was an overall yield for PE of 28.1 percent in Europe compared with 7.1 percent in the United States. The PEs diagnosed in the United States were also generally less severe as stratified by the Pulmonary Embolism Severity Index, and PE-related deaths were likewise lower. The concise summary: we’re performing astounding numbers of negative CTs and finding less significant disease, and it’s almost certain our already-low numbers of positive results are further diluted by false-positives.

Using a validated diagnostic strategy, grounded in sound risk assessment, can reduce excessive testing. The fantastic Ali Raja, MD, leads a team that recently published updated American College of Physicians clinical guidelines for the evaluation of patients with suspected acute PE.⁵ These guidelines include most of the same strategies espoused in ACEP’s prior guideline but now updated to include age-adjusted D-dimer.⁶ The age-adjusted threshold, age × 10 ng/mL added to the generic 500 ng/mL in patients older than age 50, has been validated in multiple studies. Most recently, a review of a large cohort of Kaiser Permanente patients revealed a small handful of additional missed PEs, but the corresponding decrease in radiation exposure and contrast-induced nephropathy provided a net benefit.⁷ These authors did not account for the likelihood of false-positive CTs in patients with low pretest probability, and it is reasonable to suggest their study overstates the excess misses while understating the harm reduction.

Looking further at how our emergency medicine experts view the evaluation for PE, Jeff Kline offers a comprehensive summary of risk factors and diagnostic considerations.^8,9 In his view, nonspecific cardiopulmonary symptoms are not sufficient in isolation to reasonably consider the possibility of PE. Patients must have physiologic manifestations of PE absent an alternative cause, paired with the presence of at least one known risk factor for PE. Risk-stratification into low, intermediate, or high risk can be performed by gestalt, Wells score, or revised Geneva score. Low-risk patients who meet the PE rule-out criteria fulfill an unfavorable risk-to-benefit ratio, and testing should be avoided. For otherwise low- and intermediate-risk patients, quantitative D-dimer testing is recommended.

Better yet, Kline has also proposed dynamically adjusting the D-dimer cutoff level based on the pretest probability.¹⁰ In a review of 126 patients diagnosed with PE, there were 11 patients for whom the pretest likelihood of PE was low and who had D-dimer levels less than 1000 ng/mL. All but one was subsegmental, representing less than 5 percent of the pulmonary vascular tree, and none had concomitant deep venous thrombosis. Accounting for the risks of anticoagulation, the increasing prevalence of false-positive CT pulmonary arteriography, and the risks of contrast-induced nephropathy, it may yet prove reasonable to forgo CT in this subset of patients. However, until better evidence becomes available, such a strategy should be approached via shared decision making, balancing the risks of small, undiagnosed PE against those associated with anticoagulation.

As vessel size decreases, the quality of opacification and contrast capture diminishes. This results in consistent ambiguity regarding the presence of a flow-limiting lesion.

There is no question that widespread use of CT has provided substantial benefit to patients and the health care system. However, its ubiquity and ease of use is leading to unintended consequences, particularly in overdiagnosis paired with substantial risks of unnecessary treatment. Every effort should be made to reduce use of CT in those with low pretest likelihood of PE, and small, subsegmental PE should be viewed with suspicion in the context of individual patient factors. We must continue to refine and reflect upon our routine evaluation of cardiopulmonary complaints, lest our pursuit of this white whale slip into madness.

Dr. Radecki is assistant professor of emergency medicine at The University of Texas Medical School at Houston. He blogs at Emergency Medicine Literature of Note and can be found on Twitter @emlitofnote.