Pulmonary embolism (PE) is a common condition with high untreated mortality rate, however less than 35% of patients suspected of PE have the disease. Arterial blood gas (ABG) analysis has not been shown to safely exclude PE when used alone. The purpose of this study was to evaluate the diagnostic value of ABG in combination with D-dimer in excluding PE.
A retrospective analysis was performed using data from a double-blind, randomized controlled trial comparing bedside diagnostic tests to ventilation/perfusion scanning in the exclusion of suspected PE. Validation of the statistically significant findings was attempted using a second database of patients with suspected PE.
Of the 399 participants, 57 were diagnosed with PE after initial investigations. ABG samples were taken in the initial assessment of 69.7% of subjects. In the diagnosis of PE, normal arterial carbon dioxide tension (PaCO2) of >36mm Hg or normal age adjusted alveolar-arterial oxygen gradient (P(A-a)O2) (2.5 + 0.21 X age) alone had sensitivities of only 54% and 90% respectively and negative predictive values (NPV) of 91% and 95% respectively. The sensitivity and NPV increased to 100% when each was combined with a negative D-dimer, however less than 30% of patients could be excluded. When a negative D-dimer was combined with either a normal paCO2 or normal P(A-a)O2, PE was excluded in 38% of patients while maintaining a sensitivity and NPV of 100%. In the validation set of 246 patients, a normal paCO2 or P(A-a)O2 with a negative D-dimer had a sensitivity of 89.2% and NPV of 91.3%.
In the derivation study, normal PaCO2 or normal P(A-a)O2 in combination with a negative D-dimer appeared useful in objectively excluding PE without diagnostic imaging. However, this clinical prediction rule did not validate in a second set of patients with suspected PE.
ABG data, alone or in combination, appears not to have a role in excluding PE without diagnostic imaging.
Diagnostic utility of various combinations of ABG and D-dimer in the exclusion of PE.VariableSensitivity (95% CI)Negative Predictive Value (95% CI)LR-True Negative Proportion (95% CI)PaCO2<3654.3% (36.7-71.2)90.6% (85.2-94.5)0.7255.4% (49.3-61.3)Abnormal P(A-a)O289.7% (72.7-97.8)94.5% (84.9-98.9)0.3924.0% (18.4-30.2)Positive D-dimer89.3% (78.1-96.0)96.3% (92.1-98.6)0.2340.1% (35.1-45.1)PaCO2 <36 or positive D-dimer100% (91.8-100)100% (96.2-100)028.7% (23.5-34.6)Abnormal P(A-a)O2 or positive D-dimer100% (90.2-100)100% (92.2-100)017.6% (12.7-23.5)Positive D-dimer or negative D-dimer with both paCO2<36 and abnormal P(A-a)O2 -in derivation population in validation population100% (90.2-100) 89.2% (74.6-97.0)100% (96.3-100) 91.3% (79.2-97.6)0 0.2637.6% (31.1-44.6) 30.0% (22.6-38.3)
Definition of abbreviations: P(A-a)O2 = alveolar-arterial oxygen gradient; PaCO2 = arterial carbon dioxide tension; 95% CI = 95% confidence interval; LR-= likelihood ratio of a negative test (1-sensitivity/specificity); True negative proportion=number of patients correctly excluded/number of patients tested.
Tara Keays, University grant monies University of Ottawa Medical Associates; Grant monies (from sources other than industry) Heart and Stroke Foundation of Ontario.