This study, performed in cardiac patients on VA ECMO shows several important findings: 1) consolidation of lower lobes was present in 94% of patients with acute respiratory failure 2) diffuse pulmonary edema was present on top of consolidation in 30% of patients with pneumonia and 15% of patients without pneumonia 3) Bedside chest radiography was not discriminating for pneumonia 4) The presence of an intrapulmonary shunt within consolidations, detected by color Doppler ultrasound, was highly suggestive of pneumonia 5) the diagnostic performance of the Lung Ultrasound Score sCPIS, where color Doppler replaces bedside chest radiography, was higher than the conventional sCPIS.
Components of sCPIS and LUS-sCPIS: Diagnostic Value in Patients on VA ECMO
HAP diagnosis is known to be difficult in postoperative cardiac surgery patients.[25] We hypothesized that the criteria usually used to diagnose pneumonia may not be discriminating in VA ECMO patients, as the extracorporeal circulation generates an inflammatory response, and the interpretation of chest radiography is complicated by the frequent presence of cardiogenic pulmonary edema.
As expected, clinical criteria as body temperature and leucocyte count were comparable in both groups. The level of vasopressor support was not discriminating. The difficulty of establishing the diagnosis using the conventional approach resulted in a high rate of inappropriate probabilistic antibiotic therapy in the non-pneumonia group (46%). Taking into consideration specificities of surgical cardiac patients, and specifically patients on VA ECMO, is a prerequisite to improve diagnostic performance at the bedside. Dureau et al.[14] showed that the LUS-sCPIS is much more performant than the sCPIS for the early diagnosis of postoperative pneumonia in cardiac patients. Intra-operative cardiopulmonary bypass-induced inflammatory reaction is a confounding factor jeopardizing the diagnostic value of bio-clinical and radiological parameters of the conventional sCPIS. Passive atelectasis of dependent lung regions is a common postoperative feature, resulting from diaphragm paralysis. The presence of an intrapulmonary shunt within the atelectatic areas (consolidations) is likely characteristic of an inflammatory/infectious process, a mechanism that likely explains why the LUS-sCPIS was more discriminating for pneumonia diagnosis. Intrapulmonary shunt has been reported in community-acquired pneumonia[17,26], ventilator-associated pneumonia[19], postoperative pneumonia[16], obstructive pneumonia caused by endobronchial tumor[17,27], and acute respiratory distress syndrome[18,28]
Few studies have specifically addressed pneumonia in VA ECMO patients[5] and our work is the first to investigate the LUS diagnostic performance. Lung ultrasound is easy to learn.[29] In acute respiratory failure, it provides an immediate and accurate assessment of the patient's respiratory status and can guide a rapid clinical decision. The presence of a pulmonary blood flow within a consolidation is significantly associated with the diagnosis of pneumonia. Incorporating the intrapulmonary shunt into the LUS-sCPIS yielded an AUC of 0.77, compared with 0.65 for the conventional sCPIS.[23] Confirming a previous statement13, the dynamic air bronchogram was significantly associated with the diagnosis of pneumonia with a positive predictive value of 95% and a specificity of 96% %. As fibroaspiration was not systematically performed before lung ultrasound examination, the true incidence of dynamic air bronchogram may have been underestimated [30], explaining the low sensitivity of 46%. The presence of juxta pleural condensation were not discriminating, contrasting with previous studies performed in patients with ventilator-associated pneumonia.[13,31] In comparison, only the abundant and purulent character of the tracheal secretions was discriminating, against none of the radiological criteria.
The poor performance of the chest radiography is easily explained by the high incidence of bilateral posterobasal consolidations (94% in our study) in mechanically ventilated patients following cardiac surgery.[32] As previously reported19, bilateral radiological infiltrate and coalescent B2 lines due to hydrostatic pulmonary edema were present in 30% of our patients on ECMO VA [33], complicating the chest radiography interpretation. These results support the concept that combining ultrasound and clinical criteria is more discriminating than combining radiological and clinical criteria for the diagnosis of pneumonia in patients on VA ECMO.
Limits
The first limitation of our study is its monocentric nature, with a number of patients limited transplantations (26%). This recruitment bias, linked to the specific characteristics of our center (first heart transplant center in France), may explain the high rate of pneumonia among included patients (63%). Cardiac transplantation and the postoperative context of cardiac surgery are indeed well-known risk factors for HAP.[34,35] The exact prevalence of pneumonia in our VA ECMO patients is difficult to estimate, as not all consecutive patients admitted to our department were included. Second, the lack of a gold standard for the diagnosis of HAP, resulting in heterogenous diagnostic criteria used in most studies, as well as the nonconsecutive recruitment of our patients depending on the presence of a LUS experts, may have led to biases. Despite these elements, the rate of pneumonia found in our study remains comparable to the highest incidences reported in the literature, estimated between 60% and 70%.[4,5,36] to 70. Most of them were cardiac surgery patients (71%), including a large proportion of heart
The sensitivity of color Doppler intrapulmonary shunt to detect pneumonia was only 66%, compared with 92% in the study by Dureau et al.14 The AUC of LUS-sCPIS was also relatively low (0.77). Physiologically, it is expected that a shunt within a consolidation will be detected in pneumonia because of the inhibition of hypoxic vasoconstriction due to local inflammation.[26] In VA ECMO patients, Doppler visualization of an intrapulmonary shunt could be compromised by the modification of transpulmonary blood flow related to the shunting of the cardiac chambers resulting from ECMO. Other parameters may also modify the intrapulmonary shunt, such as vasoactive support or mechanical ventilation settings.
Clinical implication
Any delay in diagnosis and antibiotic treatment of pneumonia in VA ECMO patients, who are already weakened by their critical condition, worsens the prognosis. Conversely, overexposure to unwarranted antibiotic treatment in uninfected patients favors the emergence of bacterial resistance.
Our study shows that in presence of acute respiratory failure complicating VA ECMO, the detection of an intrapulmonary shunt and/or a dynamic air bronchogram within a consolidation should alert the clinician to the possible diagnosis of pneumonia. The LUS-sCPIS score, which was previously described in cardiac surgery patients [14], is also more accurate than sCPIS in diagnosing pneumonia. Both ultrasound findings should incite to perform microbiological sampling and could help to identify patients who should receive empirical antibiotic therapy.