Best Evidence Topics
  • Send this BET as an Email
  • Make a Comment on this BET

Everything in graduation: Arterial/end-tidal CO2 gradient and the diagnosis of pulmonary embolism.

Three Part Question

In [adult patients with suspected pulmonary embolism] does [an elevated gradient between the arterial and end-tidal CO2 values] increase the likelihood of [a positive diagnosis of pulmonary embolism]?

Clinical Scenario

A 66-year-old man presents to the Emergency Department (ED) acutely short of breath, on a background of 10 days of fever, cough and pleuritic chest pain. He tells you that he had a positive COVID-19 swab in the community 5 days ago. His oxygen saturations are initially 85% on 10L oxygen by face mask and he looks tired. He is rapidly intubated and stabilised in the department, but his oxygen requirements remain at 60% despite optimised ventilation. You are unsure whether he has a concurrent pulmonary embolism (PE). A colleague says that looking at the difference between the arterial partial pressure of carbon dioxide (PaCO2) and end-tidal carbon dioxide (ETCO2) can sometimes be helpful. They suggest that an increased alveolar dead space fraction (AVDSf), calculated by PaCO2-ETCO2/PaCO2, can increase the likelihood of PE. You resolve to consult the literature to see if there is anything in this recommendation.

Search Strategy

Using the Healthcare Databases Advanced Search, Embase and Medline were both searched using the following term:

((Pulmonary embol* OR pulmonary thromboembol*) AND (end-tidal OR capnogra* OR alveolar dead space OR dead space fraction)).ti,ab

The search was limited to human studies.

The Cochrane library was also searched using the terms ‘pulmonary embol*’ or ‘carbon dioxide’.

Embase returned 183 results, Medline returned 169. All results were reviewed by title and then abstract, if appropriate. Studies that looked solely at ETCO2 without PaCO2 were not included. 16 results from Embase and 16 from Medline were relevant to the three-part question. 12 were duplicates.

Search Outcome

20 results were reviewed in regard to the quality of evidence. 1 result was a meta-analysis of 9 studies relevant to the three-part question. Of the remaining results, 2 additional studies published subsequent to the meta-analysis were included. All other studies were excluded, based on: limited analysis of the AVDSf in isolation, study methodology, or a low level of evidence such as a case series. References for all included studies were reviewed, however no additional papers were found. No relevant Cochrane reviews were identified.

The 3 identified studies providing the best level of evidence are reviewed in table 1.

Relevant Paper(s)

Author, date and country Patient group Study type (level of evidence) Outcomes Key results Study Weaknesses
Manara et al.
Systematic review and pooled meta-analysis of 14 observational studies between 1997 and 2010 looking at AVDSf. 2991 patients with suspected PE were included. 1781 had AVDSf calculated and compared to a reference standard diagnosis.Level 1b. Pooled analysis of cross-sectional studies with risk of bias assessment.Pooled sensitivity0.73 (95% CI 0.69-0.78). Significant heterogeneity between included studies (I2 = 31.3%) and half the included studies within the systematic review were found to be at high risk of bias. Many studies used ETCO2 readings in awake and spontaneously ventilating patients thus limiting generalisability to those patients requiring mechanical ventilation. In addition, ETCO2 was measured using different types of equipment and different methods.
Pooled specificity 0.61 (95% CI 0.58-0.64).
Pooled positive likelihood ratio2.76 (95% CI 1.62-4.7).
Pooled negative likelihood ratio0.38 (95% CI 0.31-0.48).
Yüksel et al.
159 patients presenting to the Emergency Department (ED) with symptoms suggestive of PE. All patients had AVDSf calculated and compared to a reference standard diagnosis.Level 3b. Single-centre observational study. Sensitivity0.68 (95% CI 0.50 to 0.82).Single-centre, observational study with subjective inclusion and exclusion criteria. Patients lost to follow up were excluded from the results. A high proportion (a third) of included patients had an underlying pulmonary malignancy. Minimal numbers of patients required mechanical ventilation.
Specificity0.74% (95% CI 0.65 to 0.81).
AVDSf by diagnosisAVDSf was significantly higher in patients with PE than those without (0.48 vs 0.35, p<0.001).
AUC0.74 (95% CI 0.65 to 0.82).
Yücel et al.
100 ED patients presenting with symptoms of PE who had a positive d-dimer or a high clinical risk prediction score. All patients had AVDSf calculated and compared to a reference standard diagnosis.Level 3b. Single-centre observational study.Sensitivity0.81 (95% CI NR) Single-centre, observational study. Only patients at high risk for PE were included Main analysis and discussion focused on AVDSf in combination with risk-stratification tools although patients with a negative d-dimer were excluded. 50% of patients were excluded, implying potential convenience sample or selection bias.
Specificity0.63 (95% CI NR)
AVDSf by diagnosis AVDSf was significantly higher in patients with PE than those without (0.217 vs 0.098).
AUC0.734 (95% CI NR)


The overall conclusion that emerges from these studies is one that emergency physicians will be familiar with; no test is sufficiently sensitive or specific to be used in isolation when evaluating patients with suspected PE. Despite the AVDSf lacking diagnostic utility, studies in both awake and intubated patients show that AVDSf will generally be increased in patients with PE, compared to those without. To further support this, case studies have suggested that patients with a massive PE exhibit a reduction in their AVDSf following thrombolysis (4). Unfortunately, many other pathologies can also cause a rise in AVDSf, such as pulmonary hypertension, acute respiratory distress syndrome (ARDS) and a low cardiac output state (5). As such, it remains unclear as to how we can best use this test to help patients. Many studies on the topic have highlighted the benefit of combining AVDSf with risk-stratification tools to exclude PE. Perhaps most notable was a randomised-controlled trial combining AVDSf with a negative d-dimer result and Well’s score (6). This study, along with others included in the meta-analysis described above, concluded that the primary role of AVDSf may be in patients with a low pre-test probability of PE. Unfortunately, there is significant heterogeneity in the studies to date, the specific cut-offs for the AVDSf that would be used in such a ‘rule-out’ protocol remain unclear, and this evidence is not currently generalisable to critically ill patients who are likely to have a higher pre-test probability. In theory, calculating the AVDSf using the ETCO2 and PaCO2 is an appealing risk stratification tool for PE. It utilises easily accessible data in patients who are invasively ventilated and can help identify the likelihood of concurrent pathology in patients with respiratory infection. In patients with COVID-19 this is a particularly difficult issue. As a result many centres are opting to use higher doses of anticoagulation as a prophylactic measure, with the potential associated harms, rather than pursue further investigation to establish if a concurrent PE is present (7). Given there is evidence to support using AVDSf to help exclude PE in patients presenting to the ED, further research on AVDSf in COVID-19 patients would be of benefit.

Editor Comment


Clinical Bottom Line

There is good quality (level 1b) evidence to support the use of AVDSf to exclude PE in patients with a low pre-test probability, when combined with established risk stratification methods.. There is no good evidence to support AVDSf as a ‘rule-in’ test for PE and current evidence may not be generalisable to critically ill patients.


  1. Manara A, D'Hoore W, Thys F. Capnography as a Diagnostic Tool for Pulmonary Embolism: A Meta-analysis. Annals of emergency medicine. 2013;62.
  2. Yuksel M, Pekdemir M, Yilmaz S, Yaka E, Kartal A. Diagnostic accuracy of noninvasive end-tidal carbon dioxide measurement in emergency department patients with suspected pulmonary embolism. Turkish Journal of Medical Sciences 2016;46:84-90
  3. Yücel Z, Aksu N, Camkurt M. The combined use of end-tidal carbon dioxide and alveolar dead space fraction values in the diagnosis of pulmonary embolism. Pulmonology 2020;26.
  4. Moreira MM, Terzi RGG, Carvalho CHN, de Oliveira Neto AF, Pereira MC, Paschoal IA Alveolar dead space and capnographic variables before and after thrombolysis in patients with acute pulmonary embolism. Vasc Health Risk Manag. 2009;5(1):9-12.
  5. Robertson HT. Dead space: the physiology of wasted ventilation. uropean Respiratory Journal. 2015;45(6):1704-16.
  6. Rodger MA, Bredeson CN, Jones G, Rasuli P, Raymond F, Clement AM, et al The Bedside Investigation of Pulmonary Embolism Diagnosis Study: A Double-blind Randomized Controlled Trial Comparing Combinations of 3 Bedside Tests vs Ventilation-Perfusion Scan ... Archives of Internal Medicine. 2006;166(2):181-7.
  7. Al-Samkari H, Karp Leaf RS, Dzik WH, Carlson JCT, Fogerty AE, Waheed A, et al. COVID-19 and coagulation: bleeding and thrombotic manifestations of SARS-CoV-2 infection. Blood 2020;136(4):489-500.