Utility of the Alveolar-arterial Oxygen Gradient in Diagnosing Pulmonary
Embolism
Clinical Question
How likely is a patient (in whom the possibility of a PE is being
entertained) to have a pulmonary embolus if the A-a gradient is normal?
Clinical Bottom Line
- In a patient who you are considering the diagnosis of pulmonary embolism,
the presence of a normal A-a gradient nearly rules out the likelihood of having
a PE (high negative predictive value of a negative test, >90% -- especially
in the absence of any previous history of DVT or PE -- or a negative LR of
0.10.)
- However, the converse is not true. The positive predictive value of an
abnormal A-a gradient is <33%, and does not rule in a PE.
The Evidence
Derivation Set: All Patients
|
|
Pulmonary Embolism |
|
Predictive Value |
Likelihood Ratio |
|
|
present |
absent |
|
|
| A-a gradient |
abnormal |
104 |
364 |
0.22 |
1.13 |
|
normal |
5 |
67 |
0.93 |
0.30 |
|
|
109 |
431 |
540 |
|
Derivation Set: Patients without previous DVT or PE
|
|
Pulmonary Embolism |
|
Predictive Value |
Likelihood Ratio |
|
|
present |
absent |
|
|
| A-a gradient |
abnormal |
82 |
307 |
0.21 |
1.17 |
|
normal |
1 |
56 |
0.98 |
0.08 |
|
|
83 |
363 |
446 |
|
Validation Set: Patients without previous DVT or PE
|
|
Pulmonary Embolism |
|
Predictive Value |
Likelihood Ratio |
|
|
present |
absent |
|
|
| A-a gradient |
abnormal |
48 |
282 |
0.15 |
1.16 |
|
normal |
1 |
53 |
0.98 |
0.13 |
|
|
49 |
335 |
384 |
|
Combined Derivation & Validation Sets: Patients without previous
DVT or PE
|
|
Pulmonary Embolism |
|
Predictive Value |
Likelihood Ratio |
|
|
present |
absent |
|
|
| A-a gradient |
abnormal |
130 |
589 |
0.18 |
1.17 |
|
normal |
2 |
109 |
0.98 |
0.10 |
|
|
132 |
698 |
830 |
|
Comments
- This was a well-intended study, however several flaws exist.
- There was no independent comparison of the test with the "gold
standard", which in this case should have been the pulmonary angiogram.
The presence or absence of a PE as recorded in the chart was accepted as
correct. This was addressed fully in the comment section of the paper.
- There may be an element of incorporation or verification bias present as
the authors do not clearly state whether or not the results of the ABG
influenced the decision to perform the gold standard (which in this case was the
V/Q scan.)
- They did create a derivation set of patient data, and confirm it against a
validation set. The data were then combined and compared to each individually.
There does not appear to be any significant difference except when comparing to
the initial derivation sample which included all patients.
- Univariate analysis of the subgroups that showed increased risk of PE for a
given risk factor (immobility, stroke, pulmonary disease, or previous DVT/PE)
revealed good discrimination only for patients who did or did not have a
previous history of DVT or PE.
- Despite the flaws in this paper, patients with a moderate to high suspicion
of PE clinically should have an ABG performed. If it is normal, the LR for
this (negative) test is 0.30 (considering all patients, not just those with no
prior history of PE or DVT). For example, if your pre-test probability is 50 %
of having a PE, the post-test probability is ~ 20%. If your pre-test
probability is close to the prevalence reported in this study (~20%), then the
post-test probability may be 5% or less. This has potential implications for
cost savings and reducing the number of patients subjected to an invasive
procedure like the VQ scan.
APPRAISED BY: Patrick Sousa, MD
DATE: September 4, 1996
McFarlane MJ and Imperiale TF. Use of the Alveolar-Arterial Oxygen
Gradient in the Diagnosis of Pulmonary Embolism. Am J Med, 1994;96:57-62.
