Ventilator-Associated Pneumonia and PaO2/FIO2 Diagnostic Accuracy: Changing the Paradigm?

Background: Ventilator-associated pneumonia (VAP) is associated to longer stay and poor outcomes. Lacking definitive diagnostic criteria, worsening gas exchange assessed by PaO2/FIO2 ≤ 240 in mmHg has been proposed as one of the diagnostic criteria for VAP. We aim to assess the adequacy of PaO2/FIO2 ≤ 240 to diagnose VAP. Methods: Prospective observational study in 255 consecutive patients with suspected VAP, clustered according to PaO2/FIO2 ≤ 240 vs. > 240 at pneumonia onset. The primary analysis was the association between PaO2/FIO2 ≤ 240 and quantitative microbiologic confirmation of pneumonia, the most reliable diagnostic gold-standard. Results: Mean PaO2/FIO2 at VAP onset was 195 ± 82; 171 (67%) cases had PaO2/FIO2 ≤ 240. Patients with PaO2/FIO2 ≤ 240 had a lower APACHE-II score at ICU admission; however, at pneumonia onset they had higher CPIS, SOFA score, acute respiratory distress syndrome criteria and incidence of shock, and less microbiological confirmation of pneumonia (117, 69% vs. 71, 85%, p = 0.008), compared to patients with PaO2/FIO2 > 240. In multivariate logistic regression, PaO2/FIO2 ≤ 240 was independently associated with less microbiological confirmation (adjusted odds-ratio 0.37, 95% confidence interval 0.15–0.89, p = 0.027). The association between PaO2/FIO2 and microbiological confirmation of VAP was poor, with an area under the ROC curve 0.645. Initial non-response to treatment and length of stay were similar between both groups, while hospital mortality was higher in patients with PaO2/FIO2 ≤ 240. Conclusion: Adding PaO2/FIO2 ratio ≤ 240 to the clinical and radiographic criteria does not help in the diagnosis of VAP. PaO2/FIO2 ratio > 240 does not exclude this infection. Using this threshold may underestimate the incidence of VAP.


Introduction
Ventilator-associated pneumonia (VAP) remains the most frequent intensive care unit (ICU)-acquired infection [1,2]. Its incidence ranges from 9% to 27% of ventilated patients depending on diagnostic criteria [3]. VAP is associated with an increased length of stay, mortality and health care costs [4,5].
The diagnosis of VAP may be complex. Therefore, multiple efforts have been directed to overcome these difficulties and to identify the most appropriate diagnostic strategies [6], including integrative scoring systems that take into account multiple clinical data and microbiological findings.
The Clinical Pulmonary Infection Score (CPIS) score was developed to objectively diagnose VAP [6]. The score combines six variables: body temperature, white blood cell count, quantity and purulence of tracheal secretions, chest radiograph, bacterial growth in tracheal secretions and oxygenation (PaO 2 /F I O 2 ≤ 240 in mmHg). However, the reported poor sensitivity and specificity of CPIS score in most studies preclude its use as an accurate diagnostic strategy [7,8]. Similarly, the Centers for Disease Control (CDC) and Prevention have proposed the inclusion of a decline in oxygenation, assessed by a PaO 2 /F I O 2 ratio ≤ 240, in the diagnostic criteria for VAP [9]. Current American guidelines define pneumonia as the presence of a new or progressive radiological lung infiltrate plus the clinical evidence that the infiltrate is of an infectious origin and a decline in oxygenation [2]. Indeed, the most reliable evidence of an infectious origin in patients suspected of having VAP combines these criteria with the microbiological identification of the potentially-pathogenic microorganisms (PPM) that cause pneumonia [10].
To our knowledge, no studies have comprehensively evaluated the relationship between microbiological diagnosis in populations with clinical suspicion of VAP and better or worse oxygenation, as assessed by the PaO 2 /F I O 2 ratio.
We hypothesized that patients with worse oxygenation would have more frequent microbiological confirmation of VAP. In order to assess the adequacy of PaO 2 /F I O 2 ≤ 240 to help in the diagnosis of VAP, we compared the characteristics and outcomes of a real-life ICU population with suspected VAP and PaO 2 /F I O 2 ≤ or > 240, with special emphasis on microbiological confirmation.

Study Population
The study was conducted in six medical and surgical ICUs, overall comprising of 45 beds, at an 800-bed university hospital. Data were prospectively collected from 2007 to 2017. Investigators made daily rounds of each ICU. Patients older than 18 years, mechanically-ventilated for 48 h or more, with suspected VAP, were consecutively included into the study and only the first episode was analyzed. We excluded patients with severe immune-suppression [11]. The institution's Internal Review Board approved the study (Comitè Ètic d'Investigació Clínica, registry number 2009/5427) and written informed consent was obtained from patients or their next-of-kin.
The microbiologic evaluation included the collection of at least one lower respiratory airways sample: tracheobronchial aspirates (TBAS) and/or bronchoscopic [14] or blind bronchoalveolar lavage (BAL) [15] if possible, within the first 24 hours of inclusion [16]. The same sampling method was performed on the third day if clinically indicated. Blood cultures and cultures from the pleural fluid were also collected if clinically justified.
Microbiologic confirmation of pneumonia was defined by the presence of at least one PPM [17] in the respiratory samples above pre-defined values (BAL > 10 4 and TBAS > 10 5 colony-forming units/mL, respectively, or any value if the patient was under antibiotic treatment), in pleural fluid or in blood cultures if an alternative cause of bacteremia was ruled out [18,19]. Microbiologic identification and susceptibility testing were performed by standard methods [20].
The initial empiric antimicrobial treatment was administered in all patients according to the local adaptation of guidelines [1,21], and subsequently revised according to the microbiologic results.
The empiric antimicrobial treatment was considered appropriate when the isolated pathogens were susceptible in vitro to at least one of the antimicrobials administrated at adequate dose [22].
The initial response to treatment was evaluated after 72 hours of antimicrobial treatment. In patients with initial non-response to treatment [23,24], respiratory samples and blood cultures were obtained again, and the empiric antimicrobial treatment was revised.

Assessment of the Systemic Inflammatory Response
We evaluated the serum levels of interleukin (IL)-6, IL-8, IL-10, tumour necrosis factor-alpha (TNF-alpha), C-reactive protein (CRP), Procalcitonin (PCT), and mid-regional pro-adrenomedullin (MR-proADM) within the first 24 hours and the third day after the diagnosis of pneumonia. All methods for this analysis have been described in detail elsewhere [25,26].

Data Collection
All relevant data were collected at admission and at onset of pneumonia from the medical records and bedside flow charts. Patients were followed until hospital discharge, death or up to 90-days after the diagnosis of pneumonia. Septic shock [27] and acute respiratory distress syndrome (ARDS) [28] were defined according to the previously described criteria.

Outcomes Variables
The rate of quantitative microbiological confirmation of patients with PaO 2 /F I O 2 ≤ 240 was compared with that of patients with PaO 2 /F I O 2 > 240. The lowest value of PaO 2 /FiO 2 on the day of VAP diagnosis was considered for patients' group allocation. Secondary outcomes were length of stay and mortality, and 90-day survival after VAP diagnosis.

Statistical Analysis
Categorical and continuous data are presented as number (percentage) and as mean ± SD (or median, inter-quartile range), respectively. Categorical variables were compared with the Chi-square or Fisher's exact tests. Quantitative continuous variables were compared using the unpaired Student's t-test or the Mann-Whitney test for normally and not normally distributed variables, respectively. The Kaplan-Meier curves were used to compare survival in the two groups.
Univariate and multivariate logistic regression analyses were performed to assess the association of PaO 2 /F I O 2 ≤ 240 or > 240 with positive microbiology. Variables that showed a significant result univariately (p < 0.10) were included in the corresponding multivariate logistic regression backward stepwise model. Adjusted odds-ratio (OR) and 95% confidence intervals (CI) were calculated.
In addition, the association between PaO 2 /F I O 2 at VAP onset as continuous variable and quantitative microbiological confirmation was assessed using receiver-operator-characteristics (ROC) curve analysis. The area under the ROC curve (AUC) with 95% confidence interval (CI), and the optimal cut-off value of PaO 2 /F I O 2 with sensitivity and specificity were calculated.

Patients' Characteristics
We prospectively identified 264 consecutive patients with suspected VAP during the study period; we excluded nine cases because PaO 2 /FiO 2 at onset of pneumonia was not available. Therefore, we included 255; PaO 2 /FiO 2 was ≤240 in 171 (67%) patients, and higher than this value in 84 (33%) patients ( Figure 1).   Compared to patients with PaO 2 /FIO 2 > 240, those with PaO 2 /FIO 2 ≤ 240 had a lower APACHE-II score at ICU admission. However, at onset of pneumonia, patients with PaO 2 /FIO 2 ≤ 240 had a higher SOFA score and CPIS, in both cases due to the lower values of PaO 2 /FIO 2 , presented ARDS criteria and shock more frequently, and had lower serum levels of sodium. Reasons for ICU admission were slightly different between groups (p = 0.051), with higher proportions of multiple trauma and septic shock among patients with PaO 2 /FIO 2 > 240, and higher proportions of postoperative care and hypoxemic respiratory failure in patients with PaO 2 /FIO 2 ≤ 240. The evolution of PaO2/FiO 2 at pneumonia onset and at day 3 in both groups is shown in Figure 2.
The most frequent pathogens identified were Pseudomonas aeruginosa, Gram-negative enteric bacteria and Staphylococcus aureus, with no significant differences in the relative proportion of any pathogen among patients with microbiological diagnosis of ICUAP.

Systemic Inflammatory Response
The serum levels of all inflammatory biomarkers at pneumonia onset and at day 3 are shown in Table 4. Patients with PaO 2 /FIO 2 ≤ 240 had higher serum levels of Procalcitonin at pneumonia onset, without any significant difference in the remaining biomarkers. MR-proADM: mid-regional pro-adrenomedullin.

Empiric Antibiotic Treatment and Outcome Variables
Initial non-response to treatment, length of stay and ICU mortality were similar between both groups (Table 5). Hospital mortality was higher, and the ventilator-free days were lower, in patients with PaO 2 /F I O 2 ≤ 240, while 90-day survival tended to be lower in this group (p = 0.070, Figure 3).

Discussion
To the best of our knowledge, this is the first study investigating the relationship between quantitative microbiological confirmation and oxygenation in a prospective cohort of patients mechanically ventilated and suspected of having VAP. We found that PaO 2 /F I O 2 ≤ 240 was associated with less microbiological confirmation in this subgroup of patients. In addition, patients with worse oxygenation had higher hospital mortality.
The diagnosis of VAP yields on clinical criteria; however, due to their potential subjectivity, other parameters has been proposed in order to better define VAP episodes [29]. Over the last years, due to an excess of antibiotic treatment, the United States Critical Care Collaborative Societies and the Department of Health and Human Services have proposed the implementation of "ventilator-associated complications" and "ventilator-associated infections" [30]. However, the use of "more objective" criteria such as chest X-ray and oxygenation has important limitations. Chest X-rays may be difficult to interpret in ICU patients due to multiple non-infectious causes such as congestive heart failure, atelectasis, prior chronic lung disease, or ARDS.
Microbiological confirmation is probably the best matching to assess VAP diagnosis quality [10]. We hypothesized that patients with more impaired oxygenation would have more frequently etiologic diagnosis and therefore they would be considered as definitive VAP episodes. Surprisingly our findings demonstrated the opposite. We previously reported that negative microbiology in patients with both community-acquired and ICU-acquired pneumonia was associated with renal and cardiac co-morbidities [31,32]. These studies suggested that some of these cases might also represent, at least in part, fluid overload because of renal failure or congestive heart failure added to the underlying inflammatory process potentially mimicking pneumonia. These conditions often exhibit more severe deterioration of oxygenation, thus possibly explaining the association of lower PaO 2 /F I O 2 with less microbiological confirmation of VAP. However, lack of detection of a PPM would not preclude existence of an infection since many patients are previously treated with antibiotics. Therefore, until more specific diagnostic criteria are available, initiation of antimicrobial treatment should still be based on the clinical signs of infection together with radiographic criteria.
The use of oxygenation has been proposed in many diseases as a criterion for diagnosis. Indeed, in case of ARDS PaO 2 /F I O 2 ratio is both a diagnostic and stratification tool [28,33]. Therefore, since pneumonia is the most frequent cause of ARDS [33], PaO 2 /F I O 2 ratio has also been considered for pneumonia diagnosis too. Unfortunately, although PaO 2 /F I O 2 might be a good parameter of severity, it might lead clinicians to over treat patients. In many situations of daily clinical practice, the use of PaO 2 /F I O 2 ratio might trigger antibiotics start when they are probably not needed, and conversely patients with clear infiltrates in the X-rays but without major oxygenation derangements might preclude that the patient is not developing a VAP. In our population, 33% of suspected VAP episodes had a PaO 2 /F I O 2 >240. Therefore, the use of this threshold of PaO 2 /F I O 2 in the diagnosis probably would have underestimated the incidence of VAP. This has important implications, either in the development of antimicrobial resistance due to unnecessary treatment, and an increase in mortality due to inappropriate (or none) antibiotic treatment. Using PaO 2 /F I O 2 may be important to stratify patients' severity in controlled clinical trials, and probably for an earlier or more aggressive treatment, as we report higher hospital mortality for those patients with worse oxygenation. We think, however, that patients with higher PaO 2 /F I O 2 should not be excluded from clinical trials dealing with VAP.
One probable reason to point out oxygenation as a diagnosis marker is due to a wrong interpretation of CPIS as a clinical marker for VAP [6]. The CPIS can be a useful tool for severity identification when the patient has been diagnosed of VAP and can help to address patients' response to treatment. Luna et al. using PaO 2 /FIO 2 ratio could demonstrate that this was the best parameter for patients' stratification of treatment response and a marker of clinical cure [34]. Moreover, we have previously reported that lack of improvement of PaO 2 /FIO 2 was among the best predictors for adverse outcomes in ICU-acquired pneumonia [24]. This, however, faces against the use of PaO 2 /F I O 2 as a diagnostic criterion, especially in some subsets of patients such as those with multiple trauma, surgery, or major burns [8]. Indeed, we found in the present study that patients with postoperative VAP episodes presented more hypoxemia compared to patients with multiple trauma. On the other hand, the recent HAP/VAP International guidelines [35] made a weak recommendation for short antibiotic courses in patients with low CPIS (i.e., 6 or less). However, this does not seem to be related to diagnosis purposes.
Diagnosis of VAP including oxygenation has not been universally accepted by many authors. In a European cohort study conducted in 465 patients with VAP, worsening oxygenation was present in 3 out of 4 patients [36], whilst Martin-Loeches et al. found it in two-thirds of their patients with VAP [37]. Similar to severe CAP [38], oxygenation can be and probably should be taken as a severity criteria similar to septic shock development, but it should not substitute current diagnostic parameters widely used. Several studies performed over the last decade have found that prompt antibiotic therapy was the most important determinant for improved outcomes in a general sepsis population [39]. It does not seem reasonable to add a test that based on our results would delay and probably mislead VAP diagnosis.
The current study has important strengths. It includes a long cohort of patients prospectively collected and without significant loss of follow-up due to a careful surveillance with a dedicated research team. In all the patients, blood gases were taken at the moment of diagnosis based on local protocols. Because of inaccuracy to assess the actual FiO 2 in non-intubated patients, we did not include cases of non-ventilator ICU-acquired pneumonia [40]. Moreover, all patients had at least one lower respiratory tract sample processed for culture at onset of pneumonia, and the frequency and type of previous antibiotic treatment is reported, without differences related to severity of oxygenation in both variables. An additional value for this study is the acquisition of inflammatory markers at the onset of pneumonia and after three days of antibiotic treatment. However, we have to acknowledge some limitations. The most important would be the single centre analysis, which might limit reproducibility of the results. However, since the same research team worked on the project during the inclusion period, the study appears reliable in terms of diagnosis and homogeneity. Second, we did not collect blood gas data at ICU admission and therefore cannot assess the evolution of oxygenation before patients were suspected of having VAP.

Conclusions
In conclusion, we have found that adding PaO 2 /F I O 2 ratio ≤ 240 to the clinical and radiographic criteria does not help in the diagnosis of VAP. PaO 2 /F I O 2 ratio > 240 does not exclude this infection. Using this threshold may underestimate the incidence of VAP. Changing this paradigm might have beneficial effects on the early antibiotic treatment for VAP.