Diagnosing PE by Blood Gas

Motivation: On an early morning at 5:30 am last winter, I was huddling with my surgery team waiting for our chief to arrive and start rounding.  In the haze of early morning stupor, I heard one of the residents mention that overnight someone was short of breath.  He had called one of his fellow surgical residents, who had advised him to draw a blood gas to rule out pulmonary embolism (PE).  One of the other members of the team was about to say something when the chief arrived, and we cut short our talk to begin marching to the patient rooms.  But, I wondered how sensitive is an ABG with its characteristic signs of hypoxemia and hypocapnia in ruling out PE?

Paper: Diagnostic Value of Arterial Blood Gas Measurement in Suspected Pulmonary Embolism.  Rodger, M., et. al. Am. J. Respir. Crit. Care Med. (2000) 162: 2105-2108.  http://ajrccm.atsjournals.org/cgi/content/full/162/6/2105

Methods: For a 30 month period, consecutive patients (inpatient and outpatient) suspected of PE were subjected to a blood gas and D-dimer test.  Referring clinicians were then asked to provide a clinical gestalt of the likelihood of PE.  Subsequently, all patients underwent V/Q scanning.  For low probability V/Q scan with high clinical probability of PE or high probability V/Q scan with low clinical probability, patients were referred for pulmonary angiogram at the clinician's discretion.  Inclusion criteria for patients were essentially greater than 18 years of age with capacity for informed consent and physiological capability to undergo pulmonary angiogram if necessary.

Results: 
Patient Selection: In total, 246 patients were considered for PE.  In 49 patients, PE was accepted as the final diagnosis and ruled out in 163 other patients.  In 34 patients, the final diagnosis was unclassified.  These patients were not referred for angiogram by their clinicians but had either low probability V/Q scan with high pre-test likelihood or high probability V/Q scan with low pre-test likelihood.  The patients deemed unclassified were removed from further analysis.

Blood Gas Measurements: The clinical variable PaO2<80 mm Hg was present in 57.9% of patients with PE and 46.6% of patients without PE.  The clinical variable PaCO2<36 mm Hg was present in 44.4% of patients with PE and 39.7% of patients without PE.  Other clinical rule results are as follows:

• Abnormal (A-a) gradient- Sensitivity: 84.2%, Specificity: 27.4%, PPV: 27.4%, NPV: 84.2%
• D-dimer positive - Sensitivity: 83.0%, Specificity: 57.6%, PPV: 39%, NPV: 91.2%
• PaO2<80 mm Hg or D-dimer positive - Sensitivity: 91.9%, Specificity: 32.4%, PPV: 32.4%, NPV: 91.9%
• PaO2<80 mm Hg or D-dimer positive or Respiratory Rate>20 breaths/min - Sensitivity: 96.9%, Specificity: 21.3%, PPV: 30.7%, NPV: 95.0%
• PaCO2<36 mm Hg or Abnormal (A-a) O2 gradient - Sensitivity: 91.9%, Specificity: 14.7%, PPV: 25.6%, NPV: 85.0%

Discussion: To rule out PE, we ideally want clinical criteria with high sensitivity.  Individually, hypoxemia and hypocapnia have really poor sensitivities at 57.9% and 44.4%!  Calculating an abnormal alveolar-arterial gradient is a more sensitive indicator for PE, but the negative predictive value (84.2%) is still unacceptably low.  While evaluating the results, I was surprised to see that a D-dimer test was no more sensitive than an abnormal A-a gradient.  The D-dimer has the advantage of specificity rather than sensitivity over the A-a gradient.  If I were stuck on an island without radiology, I would choose the rule of PaO2<80 or D-dimer positive criteria to start ruling out PE though we would miss about 8% of pulmonary embolism cases.  With an estimated untreated mortality of 30%, such an approach would still be dangerous.  This study shows that at present, radiology remains an integral part of ruling out PE.

This study has some limitations.  Of the 246 patients originally included in the study, the diagnosis remained indeterminate in 34 patients.  Often these patients are the most troubling because they have a mismatch in the clinical probability and probability provided by V/Q scan.  A very interesting question is whether blood gas results could be used to further stratify risk status in these patients with intermediate risk status.  Pulmonary angiogram studies or at least further clinical follow-up history would be useful.