Therefore, we hypothesized that the combination of PetCO₂ and clinical probability with positive D-dimer test could improve diagnostic accuracy in the prehospital setting in patients with suspected PE. We wanted to determine if the combination of two level clinical probability assessment and PetCO₂ measurement could confirm or exclude PE in patients with an abnormal D-dimer test.

The aim of this study is to determine whether PetCO₂ improves sensitivity for exclusion of PE in unlikely patients with abnormal D-dimer results, and confirms PE with high specificity in PE likely patients in the prehospital setting.

Between October 2004 and December 2008, 131 patients with suspected PE and a positive D-dimer test were enrolled. Thirty-one patients were excluded because of anticoagulant treatment for more than 48 hours before inclusion (n = 10), inability to participate (n = 12) and receiving mechanical ventilation (n = 9). Recruitment, exclusion and subsequent grouping of all patients are shown in Figure 1. The baseline clinical and demographic variables of the study populations are displayed in Table 2. For the identification of the final diagnosis of PE, we examined 37 variables (Table 2) and 11 variables remained statistically significant after analysis. Variables from univariate analysis (with P < 0.05) were included into a model of multivariable analysis with logistic regression for identification of potential predictor variables of a final diagnosis of PE. Finally, after multivariable logistic regression analysis six variables were defined as independent predictor variables: PaCO₂, partial pressure of arterial oxygen (PaO₂), D-dimer, PetCO₂, cyanosis, previous deep vein thrombosis and/or PE (Table 3).

PE was diagnosed during the initial diagnostic work up by a positive spiral CT in 78 patients, a high probability ventilation/perfusion scan in 20 patients and pulmonary angiography in 2 patients. These patients were considered to have PE for the analysis of the diagnostic accuracy of PetCO₂. Thus, PE was confirmed in 41 patients (41%) and excluded in the remaining 59 patients (59%) on the basis of the results of initial diagnostic work up. Among 31 patients who met exclusion criteria for PE, no one had confirmed diagnosis of PE. Five of 41 patients with PE died, because of recurrent PE during hospitalization. Two of 59 patients without PE died: 1 of septic shock and 1 of cardiac arrest.

End-tidal carbon dioxide

Receiver operating characteristics analysis selected 28 mmHg as the optimal cut-off for PetCO₂ (Figure 2). The AUROC curve for PetCO₂ is 0.929 (95% CI = 0.881 to 0.977).

Thirty-eight of the 41 patients (92.6%) with PE had abnormal PetCO₂ of less than 28 mmHg compared with 10 of the 59 patients (16.9%) without PE. A PetCO₂ above 28 mmHg excluded PE with a sensitivity of 92.6% (95% CI = 79 to 98%), a negative predictive value of 94.2% (95% CI = 83 to 99%), a specificity of 83% (95% CI = 71 to 91%) and a positive predictive value of 79.2% (95% CI = 65 to 89%).

Thirty-five patients had a low (PE unlikely) clinical probability and PetCO₂ above 28 mmHg. This combination excluded PE with a sensitivity of 100% (95% CI = 89 to 100%), a negative predictive value of 100% (95% CI = 88 to 100%). Twenty-eight patients had a high clinical probability (PE likely) and a PetCO₂ below 28 mmHg. This combination had a specificity of 93.2% (95% CI = 83 to 98%) and a sensitivity of 58.5% (95% CI = 42 to 73%) for PE (Table 4).

Combinations	PE	No PE
PE likely OR PetCO2 < 28 mmHg (suggests PE possible)	41	24
PE unlikely AND PetCO2 > 28 mmHg (Rule out PE)	0	35
PE likely AND PetCO2 < 28 mmHg (suggests PE possible)	24	4
PE unlikely OR PetCO2 > 28 mmHg (Rule out PE)	17	55

PE = pulmonary embolism; PetCO₂ = partial pressure of end-tidal carbon dioxide.

In our study, we have demonstrated that the combination of PetCO₂ of more than 28 mmHg and low clinical probability (PE unlikely) is a potentially safe method for excluding PE in patients with suspected PE and positive D-dimer test in the prehospital setting. The results also suggest that the measurement of PetCO₂ alone has a lower negative predictive value (94%; 95% CI = 83 to 99%) than the previously mentioned combination of tests (100%; 95% CI = 89 to 100%).

In our study we found that the combination of high clinical probability (PE likely) and a PetCO₂ of less than 28 mmHg had 93.2% specificity (95% CI = 83 to 98%) for the confirmation of PE.

Some studies [21–23] have evaluated the diagnostic accuracy of capnography in patients with suspected PE. The multicenter study by Kline and colleagues [21] calculated sensitivity as 67.2% (95% CI = 55.0 to 77.5%) and specificity as 76.3% (95% CI = 71.2 to 85.6%). Rodger and colleagues [22] calculated sensitivity as 79.5% (95% CI = 63.5 to 90.7%) and a specificity of 70.3% (95% CI = 61.2 to 78%). A negative predictive value varied from 90.7% [22] to 91.9% [21]. Hogg and colleagues [23] calculated sensitivity as 100% (95% CI = 84.5 to 100%) but a low specificity of 22.7% (95% CI = 18.8 to 27.2%). The combination of a normal alveolar dead space fraction and normal D-dimer concentration excluded PE with a sensitivity ranging from 90.5% to 98.4% [21–23]. Sanchez and colleagues [24] combined alveolar dead space fraction and clinical probability assessment in patients with a positive D-dimer. The combination of a normal alveolar dead space fraction and a low clinical probability excluded PE with a sensitivity of 99.1% (95% CI = 94.9 to 100%) and a negative predictive value of 97.8% (95% CI = 88.2 to 99.9%). Our study shows similar results as this study, the difference being that we combined PetCO₂ and clinical probability assessment in patient with a positive D-dimer.

Our study suggests that a simple method of nasal measurement of PetCO₂ in combination with clinical evaluation can safely exclude PE without blood gas analysis and calculations of PaCO₂ - PetCO₂ gradient (unpractical for diagnostics in the field).

What impact do these results have on patient care, and what patient benefit is derived from the out-of-hospital study? The primary goal of our observational, prospective study was to find out the diagnostic rule of PetCO₂ in patients with suspected PE and abnormal D-dimer results. The study showed that these results could be useful in the emergency department. Corwin and colleagues [26] and Hirai and colleagues [27] reported that emergency physicians did not use D-dimer effectively to determine the need for CT or angiography in the evaluation of acute PE. The use of quantitative D-dimer in combination with PetCO₂ in the prehospital setting would decrease unnecessary imaging and irradiation, costs and time for patients seen in the admission department. The results from the field can help in diagnostic decisions. The prehospital emergency physician can organize direct transport from the field to pulmonary vascular imaging. Squizzato and colleagues [28] showed in a meta-analysis that 928 patients with symptomatic PE were treated completely as outpatients or discharged early.

In previous recommendations for the diagnosis of PE [25], bedside echocardiography was recommended for high-risk PE (presence of shock or persistent hypotension). Mansencal and colleagues [29] found that echocardiography (using a portable ultrasound device) is a reliable method for screening patients with suspected PE, especially in patients with dyspnea or with high clinical probability. The prehospital point-of-care ultrasound is reality [30], and its findings can help in thrombolytic therapy in the field. Thrombolytic therapy is the first-line treatment in patients with high-risk PE presenting with cardiogenic shock and/or persistent arterial hypotension because it rapidly exerts beneficial effects on hemodynamic parameters.

We used the prediction rule described by Wells and colleagues. There is no clear difference in diagnostic performance between using another clinical prediction rules, but the score by Wells and colleagues has been more extensively validated.

However, this study has some limitations. Firstly, prehospital emergency physicians and hospital physicans at admission were not blinded to the values of PetCO₂ and the D-dimer test as assessed by the rapid quantitative test because capnometry represents the routine procedure in our prehospital management.

Secondly, the study was realized in a single emergency center with a relatively small sample size. Thirdly, all patients would have had the reference standard (CT, pulmonary angiography or a normal ventilation/perfusion scan result). Finally, severe additional factors may have an impact of the reliability of PetCO₂ (some patient hyperventilated, had periodic breathing or had ventilation/perfusion mismatch).

We conclude that implementing D-dimer and quantitative capnometry into standard daily prehospital care of acute dyspnea exerted a beneficial effort on the PE diagnosis and the eventually need for immediately treatment. However, the diagnostic accuracy of these methods should be confirmed in a larger multicenter study in the field (in combination with the point-of-care prehospital ultrasound) for determination of whether these findings can be safely incorporated.

In conclusion according to the results of our study, PetCO₂ measurement has been demonstrated as a useful adjunct to standard clinical evaluation and identification of PE in patients with positive D-dimer tests in the prehospital setting. The combination of clinical assessment and PetCO₂ measurement in patients with low clinical probability (PE unlikely) has better diagnostic value than D-dimer in combination with clinical assessment. The combination of PetCO₂ concentrations and high clinical probability (PE likely) is a potentially safe method for confirmation of PE in patients with suspected PE and positive D-dimer test in prehospital setting.