A-a Gradient (Alveolar-arterial Oxygen Gradient) Use: To quantify ABG results to estimate the average gradient across the alveolar membrane. An increased A-a gradient is seen in pulmonary embolism, pneumonias, ARDS, etc. Normals: Normal values increase with age, and smoking or occupational pneumoconiosis will also increase the A-a gradient. David Plummer, writing in Tintinalli's ACEP "Emergency Medicine: a Comprehensive Study Guide" 4E, page 195 "A normal AaO2 is under 10 in young healthy patients and increases with age. Consider values over 30 always abnormal." However, also consider that Hockberger, writing on page 372 of this same book, says "A normal A-a gradient = (age/4)+4." Also consider this table: Age Upper Limit of Normal ----------------------------- 20 17 30 21 40 24 50 27 60 31 70 34 80 38 [Morris A, Kamer R, Crapo R, et al. Clinical pulmonary function: a manual of uniform laboratory procedures, 2E. Salt Lake City: InterMountain Thoracic Society, 1984:45. Quoted in Emergency Medicine Reports, 12/14/92.] Equation: A-a gradient = (713 (FIO2) - 1.2 (pCO2)) - (pO2) For room air, A-a gradient is 150 - 1.2(pCO2) - pO2 Quick and dirty approximation: 145 - pCO2 - pO2 (Wilson and Barton, writing in Tintinalli's ACEP "Emergency Medicine: a Comprehensive Study Guide" 4E, page 113: "One can also estimate the alveolar oxygen in patients with a normal cardiac output breathing room air by subtracting the PaCO2 from 145. This is possible because PAO2 and PACO2 add up to about 145 when a patient breathes room air at sea level.") Derivation: (pAlveolarO2) = (Fractional Inspired O2) * (Atmospheric Pressure - partial pressure of water in saturated air) - (pAlveolar CO2/R) (where R is the respiratory coeffecient, the amount of CO2 produced per molecule of O2 used, or about .83, the reciprocal of which is 1.2) or: pAO2 = (FIO2(760-47)) - (pACO2/R) or, since rapid diffusion of CO2 across the alveolar membrane means that pACO2 and paCO2 are the same: pAO2 = 713(FIO2) - 1.2(paCO2) therefore: A-a O2 = (713(FIO2) - 1.2(pCO2)) - pO2 Some say it is best to measure A-a O2 on 100% O2 to remove effects of V/Q mismatch on the calculation; others disagree. >> Clearly, the literature is consistent in that at most 4-8% of PE's will >> have an A-a gradient of less than 10. John, IMHO, this is just not true. In a modern population of patients with angiographically proven PE, approximately 15% - 30% have a normal Aa gradient (see citations below). I personally reviewed every positive angiogram for 18 months at George Washington Medical Center, and found 15% with a perfectly normal blood gas, including a normal Aa gradient. There were many more cases during the same time period in which patients with a high clinical likelihood and an abnormal V/Q did not proceed to angiography, in defiance of common sense and in violation of the standard of care. If one includes all the cases where segmental or larger defects were seen on the V/Q and no other pathology was ever found to explain it, the Aa gradient was normal in 25% of the cases. Of course this doesn't even include all of the patients with PE's who never even got a V/Q scan. The literature is full of studies that suffer from enrollment bias: In older studies, a normal Aa gradient or a PO2 above 80 were considered 'proof' that the patient couldn't have a PE, thus the 'up to 8%' statement became a self-fulfilling prophecy. Anyway, the likelihood of a normal Aa gradient in a patient with a known PE is not important. As Harvey has pointed out, the important thing would be to know whether the Aa gradient can discriminate between patients WITH and those WITHOUT pulmonary embolism, in a population of patients at risk. The truth is, in a patient with signs and symptoms of PE, knowing the reults of a room air ABG does not appreciably change the prior clinical likelihood -- the test likelihood ratio is very close to unity for any set of criteria you can devise. As a matter of fact, for patients with signs and symptoms consistent with PE, the worse the Aa gradient (or the lower the PO2) the more likely the patient is to have something OTHER than a PE. Here is a (chopped up) brief excerpt from the 4th edition of the Rosen text (in press). ================= Arterial Blood Oxygen A circulatory obstruction due to pulmonary embolism may or may not produce abnormalities of pulmonary gas exchange. In the setting of hyperventilation or hypoventilation the PO2 may reflect the minute ventilation more than the underlying pulmonary function. As late as 1970 it was believed that pulmonary thromboembolism invariably produced an arterial PO2 below 80 torr, and patients with an arterial PO2 greater than 80 torr were commonly excluded from clinical trials of the era. We now know that many patients with pulmonary embolism are asymptomatic, and most have a normal arterial PO2. Even in a population of classically symptomatic patients with angiographically proven PE, at least 17 percent will have a normal PO2, and one in twenty has an arterial PO2 level above 100 Torr on room air.{Menzoian1979, UPET1973, USPET1974} The arterial blood gas PO2 has a _zero_predictive_value_ in a population of patients with a clinical suspicion of PE. To understand why this is so, one must recognize that this population includes some patients with PE along with others who have some other pulmonary pathology. The clinical diagnosis of PE is uncertain, and the incidence of PE in a typical population under suspicion will usually be less than 50%. Many of the patients have other pathologies that affect pulmonary gas exchange as much as or more than pulmonary embolism does. The end result is that if any reasonable level of PaO2 is chosen as a dividing line, the incidence of PE will be higher in the group with a PaO2 above the dividing line than in the group below the divider. This seems counter-intuitive, but its general truth may be demonstrated mathematically for any test finding with a Gaussian distribution. Clinical evidence that it is true for the PaO2 in patients with suspected pulmonary embolism has been available since the early work of Menzoian and Williams{Menzoian1979} (See table below.) Arterial oxygen levels simply have no predictive value in the evaluation of patients with suspected pulmonary embolism. 120 patients with suspected pulmonary embolism (54 positive for PE){Menzoian1979} Room air ABG results correlated with angiography results PO2 selected Incidence of PE in patients Incidence of PE in patients BELOW selected PO2 ABOVE selected PO2 ======== ============ ================== 80 Torr 45/101 (44.6%) 9/19 (47.4%) 70 Torr 39/87 (44.8%) 15/33 (45.5%) 65 Torr 29/69 (42%) 25/51 (49%) The Alveolar-arterial oxygen gradient The PaO2 is very sensitive to the minute ventilatory volume. If the patient takes one or two extra breaths per minute the PaO2 may be normal even when pulmonary gas exchange is significantly impaired. When the minute volume rises without an increase in cellular respiration, the PCO2 will drop proportionately, and for this reason the alveolar-arterial oxygen gradient may be used as an encompassing and more reliable measure of impaired gas exchange. The A-a gradient will never be zero because pulmonary gas exchange is not perfect, but in the absence of impaired pulmonary gas exchange it should be less than 10 torr in a young person and less than 20 torr in old age.{Skorodin1984} One useful rule of thumb is that the A-a gradient should be no greater than 10 plus one-tenth of the patient's age. The A-a gradient is a better measure of gas exchange than the PO2 alone, but it is still nonspecific and insensitive: small pulmonary emboli do not impair pulmonary gas exchange, and nearly every type of pulmonary pathology can affect the A-a gradient. A large pulmonary embolism often causes an increased A-a gradient, but a perfectly normal ABG (normal PO2, normal PCO2, normal pH, and normal A-a gradient for age) will be seen in up to 23 percent of patients in whom pulmonary embolism is diagnosed.{Overton1988, Stein1995} The A-a gradient is, no doubt, normal in many more patients with small emboli in whom the diagnosis is never made. As with hypoxemia, most patients with an abnormal A-a gradient have some pulmonary pathology _other_ than PE as the cause for their impaired gas exchange. Craig Feied, MD FACEP cfeied@ncemi.org I ran across the article that Jon mentioned in his review of A-a gradients in PE. Stein's 1995 article does explictity address the issue of the most appropriate formula to use in calculating the A-a gradient. I've reproduced the abstract below for those interested. What conclusions can one draw from this? Only that a normal A-a gradient, no matter how calculated, cannot be used to reliably exclude PE with sufficient accuracy to obviate the need for further studies. If your clinical suspician is high enough further diagnostic testing beggining with V/Q scan and, if indeterminate, venous doppler studies or pulmonary angiography is indicated. I don't think that anyone would accept missing 10% of all PEs because it was 'excluded' based upon a normal A-a gradient. Certainly we would not accept any algorithm that routienly sends home 10% of acute MIs. H. Louzon MD -------------------------------------- On 25 Jan 96 at 15:30, Harvey Louzon wrote: [snip] > What conclusions can one draw from this? Only that a normal A-a > gradient, no matter how calculated, cannot be used to reliably exclude PE > with sufficient accuracy to obviate the need for further studies. If > your clinical suspician is high enough further diagnostic testing > beggining with V/Q scan and, if indeterminate, venous doppler studies or > pulmonary angiography is indicated. I don't think that anyone would > accept missing 10% of all PEs because it was 'excluded' based upon a > normal A-a gradient. Certainly we would not accept any algorithm that > routienly sends home 10% of acute MIs. Hmmm. I have some problems with this statement. But having read Orwell, I know how to deal with it! Assume I have a patient with some chest pain and some tachycardia; not a lot of eaither, though, and not much in the way of risks for PE. So I'm not _really_ thinkng about PE; I just get an ABG and calculate an Aa gradient on general principles. After seeing the result, I wasn't thinking about PE. I wasn't, I wasn't, I wasn't, nobody saw me do it, and you can't prove anything. (Is this Orwell or The Simpsons?) But seriously now, I think there are situations where one has a slight suspicion of PE (1:10,000-1:100,000), and the (slight) information obtained from an Aa gradient serves to diminsh that suspicion even more, to the point where one is comfortable with the risk of sending the patient out without additional testing (1:1,000,000, assuming a normal Aa gradient makes a PE just 10 times less likely). The alternative is to order a V/Q every time you think about PE. I think an Aa gradient is a poor but still marginally useful test in such situations. What we truly need is a prospective study where patients are prospectively stratified into - low clinical suspicion - moderate clinical suspicion - high clinical suspicion And each patient gets an ABG, a venous doppler of the legs, and a V/Q scan. And then we can look at the utility of studies for real, and develop an appropriate algorithm for evaluating for possible PE. This _must_ be a prospective study, and carefully designed, so we can capture those patients that might otherwise be sent home without any formal record of considering a PE but nonetheless had an ABG done as part of a chest pain workup (you didn't see me, nobody can prove anything . . .) -Keith Conover, M.D. See Also