Utility of Gram stain in the clinical management of suspected ventilator-associated pneumonia Secondary analysis of a multicenter randomized trial

Transcription

1 Journal of Critical Care (2008) 23, Utility of Gram stain in the clinical management of suspected ventilator-associated pneumonia Secondary analysis of a multicenter randomized trial M. Albert MD a, J.O. Friedrich MD, DPhil b, N.K.J. Adhikari MDCM, MSc b, A.G. Day MSc d, C. Verdant MD a, Daren K. Heyland MD, MSc c,, for the Canadian Critical Care Trials Group a Division of Critical Care, Department of Medicine, Université de Montréal, Montréal, Canada H3T 1J4 b Interdepartmental Division of Critical Care, University of Toronto, Toronto, Canada M5B 1W8 c Department of Critical Care, Queens University, Kingston, Canada K7L 2V7 d Clinical Research Centre, Kingston General Hospital, Kingston, Canada K7L 2V7 Keywords: Pneumonia; Ventilator-associated pneumonia; Gram stain; Diagnosis accuracy Abstract Purpose: Gram stains of endotracheal aspirates (EA) and bronchoalveolar lavages (BAL) may guide empiric antibiotic therapy in critically ill patients with suspected ventilator-associated pneumonia (VAP). Previous studies differ regarding the ability of the Gram stain to predict final culture results. The aim of the present study was to evaluate the relationship between EA or BAL Gram stains and final culture results in intensive care unit patients with a suspected VAP. Material and Methods: We retrospectively analyzed data from the Canadian multicenter VAP study to correlate EA or BAL Gram stain and final culture results. We categorized Gram stains as Gram positive (GP) and Gram negative (GN) if any GP or GN organisms respectively were seen on staining. Cultures were considered positive if they yielded pathogenic organisms on final results. Results: Seven hundred forty patients were enrolled in the study; 35 did not have a Gram stain done leaving 350 BALs and 355 EAs from 705 patients. Pooling BAL and EA results, we found the overall agreement between Gram stain class and pathogenic bacteria culture results to be poor (κ = 0.36; 95% CI, ). Among specimens with Gram stains showing no organisms, 99 (30%) of 331 grew pathogenic organisms. Among specimens with Gram stains showing no GN organisms, 113 (25%) of 452 grew pathogenic GN organisms. Among specimens with Gram stains showing no GP organisms, 45 (11%) of 428 grew pathogenic GP organisms. Supported by grants from the Canadian Institutes of Health Research and Physicians' Services, Inc, of Ontario, and by unrestricted grants from AstraZeneca and Bayer. Corresponding author. Tel.: x3339; fax: address: dkh2@queensu.ca (D.K. Heyland) /$ see front matter 2008 Published by Elsevier Inc. doi: /j.jcrc

2 Utility of Gram stain in the clinical management of suspected VAP 75 Conclusions: Gram stains performed for clinically suspected VAP poorly predict the final culture result and thus have a limited role in guiding initial empiric antibiotic therapy in such patients Published by Elsevier Inc. 1. Introduction Nosocomial infections such as ventilator-associated pneumonia (VAP) complicate the course of many critical care patients. Ventilator-associated pneumonia has been associated with increased length of stay, duration of mechanical ventilation, mortality, and cost of hospitalization [1]. Despite multiple investigations, the optimal diagnostic and treatment approaches remain uncertain. Clinicians treating patients with suspected VAP must decide whether to treat early with antimicrobial therapy, a strategy shown to improve outcomes in some observational studies [2,3], or wait until final culture results and therefore avoid the welldocumented risks of unnecessary antibiotic exposure [4]. Early empiric treatment forces the clinician to select antibiotics without the final culture and therefore often leads to Gram stains of endotracheal aspirates (EA) and bronchoalveolar lavages (BAL) being used to guide initial empiric antibiotic therapy for VAP. However, studies differ regarding the reliability of the Gram stain to predict the final culture results [5-7]. Many of these have been small, single-center studies using suboptimal diagnostic criteria for VAP. Some investigators have suggested increasing reliability by microscopic examination of the Gram stain for intracellular organisms in the clinical setting of pneumonia, whereas others could not demonstrate any utility of such information [8,9]. Consequently, important issues remain: firstly, whether withholding antibiotics based on absence of bacteria on Gram stain is a safe strategy when compared to the final culture results, and, secondly, whether the spectrum of initial antibiotic therapy can be restricted based on the initial Gram stain. Given the controversy surrounding the accuracy of Gram stains, our primary objective was to evaluate the utility of the BAL and EA Gram stain in decision making related to empiric antibiotic administration. As a secondary objective, we examined the relationship between the quantity of organisms seen on Gram stain and other clinically important outcomes. 2. Material and methods We retrospectively analyzed data from the Canadian multicenter randomized controlled VAP trial to examine the correlation between EA or BAL Gram stains and final culture results [10]. In this factorial RCT, mechanically ventilated adults in the intensive care unit (ICU) for 4 days or more who developed a clinical suspicion of pneumonia were randomized to undergo either bronchoscopy with quantitative cultures or standard nonquantitative EAs, and to initial empiric broad-spectrum combination antibiotics or monotherapy. A clinical suspicion of pneumonia was defined by the presence of new or persistent radiographic features suggestive of pneumonia with no other obvious cause and the presence of any 2 of the following: fever higher than 38 C, leukocytosis (N /L) or neutropenia (b /L), purulent endotracheal secretions, isolation of potentially pathogenic bacteria from the EA, and increasing oxygen requirements. These criteria had to be met within 48 hours of enrollment. The trial excluded immunocompromised patients (postorgan transplantation, human immunodeficiency virus infected, neutropenic [b /L absolute neutrophils], patients receiving corticosteroids N20 mg/d for 6 months) and patients infected or colonized with Pseudomonas species or methicillin-resistant Staphylococcus aureus (MRSA). Full selection criteria are described elsewhere [10]. Bronchoalveolar lavages and EA were performed in a standardized manner according to conventional techniques, and laboratory processing was standardized [10]. Patients were prepared for bronchoscopy with 100% FiO 2 and adequate sedation with or without paralysis. The sampling area was selected on the basis of the location of the new, persistent, or progressive infiltrate seen on chest radiograph. Suctioning of secretions in the endotracheal tube, trachea, and large airways was avoided when passing the bronchoscope into the lung. Lidocaine was not used because of its bacteriostatic properties. The tip of the bronchoscope was wedged into a subsegment of the lung, and 20 ml sterile saline solution was injected, aspirated, and discarded. A new trap was positioned and additional 20- to 60-mL aliquots were injected slowly and aspirated. The total amount of fluid injected was approximately 140 ml. For endotracheal suctioning, patients were prepared with 100% FiO 2, bag ventilation (if necessary), and adequate sedation. A sterile suction catheter and suction trap were then used. Three to five milliliters of sterile saline was instilled if an adequate specimen was not obtained. Both EA and BAL specimens were transported immediately to the laboratory. All samples were plated for culture within 1 hour. Gram stains were performed on all specimens using usual technique which involved primary staining with crystal violet followed by iodine, subsequent washing of the slides with 95% ethanol, and, finally, secondary staining with safranin. Cultures were performed using standard media. All organisms recovered were identified and their susceptibility was tested using standard techniques.

3 76 M. Albert et al. Table 1 Baseline characteristics of study patients Gram stain GP, GN, or both (n = 374) No organisms (n = 331) Total (N = 705) Age (y) 58.8 ± ± ± 17.9 Sex (female) 106 (28.3%) 112 (33.8%) 218 (30.9%) APACHE II 19.1 ± ± ± 6.3 Admission category Medical 226 (60.4%) 198 (59.8%) 424 (60.1%) Surgical 148 (39.6%) 133 (40.2%) 281 (39.9%) Primary admission diagnosis Cardiovascular 105 (28.1%) 65 (19.6%) 170 (24.1%) Trauma 105 (28.1%) 79 (23.9%) 184 (26.1%) Respiratory 52 (13.9%) 69 (20.8%) 121 (17.2%) Neurologic 51 (13.6%) 45 (13.6%) 96 (13.6%) Gastrointestinal 26 (7.0%) 33 (10.0%) 59 (8.4%) condition 24 (6.4%) 15 (4.5%) 39 (5.5%) Sepsis 7 (1.9%) 21 (6.3%) 28 (4.0%) Renal 4 (1.1%) 4 (1.2%) 8 (1.1%) PaO 2 /Fio ± ± ± 83 On vasopressors 80 (22.5%) 72 (20.6%) 152 (21.6%) Days in ICU before 8.1 ± ± ± 5.2 enrollment Antibiotics at randomization Not on any 187 (50.0%) 75 (22.7%) 262 (37.2%) within 3 d Yes but none 104 (27.8%) 137 (41.4%) 241 (34.2%) started within 3 d New antibiotics started within 3 d 83 (22.2%) 119 (36.0%) 202 (28.7%) All statistics are counts (percentages) or mean ± SD. P b.01 between GP, GN, or both and no organisms. The following data were collected on all study patients: age, sex, admission diagnosis, admission APACHE II (Acute Physiology and Chronic Health Evaluation II) score [11], comorbidities, and dates of hospital and ICU admission. Patients were monitored daily for signs and symptoms of infection (heart rate, respiratory rate, temperature, blood pressure, PaO 2 /FiO 2, secretions) and organ dysfunction. Gram stains, culture results (including sensitivity profile), antibiotics, duration of mechanical ventilation, and ICU stay were recorded. Patients were followed daily until death, ICU discharge, or for a maximum of 28 days. At the end of the study period, local site investigators unblinded to allocation reviewed all patient charts to determine whether the patient had VAP; the final clinical and microbiological outcome was determined according to standard definitions [10]. Methods center personnel reviewed all study charts to ensure completeness and consistency across sites. We categorized Gram stains as follows: Gram positive (GP) for stains showing only GP bacteria; Gram negative (GN) for stains showing only GN bacteria; both for stains showing GP and GN bacteria; and neither for stains showing neither GP nor GN bacteria. The amount of bacilli and cocci seen on Gram stain was classified as none, few, moderate, or numerous. We defined positive cultures by the presence of pathogenic bacteria (requiring treatment as defined by a priori consensus; see Appendix A) on EA or BAL cultures regardless of their concentration. Only the first enrollment culture per patient was used. The potential undertreatment rates were defined by the proportion of patients with positive cultures that were not predicted by the Gram stain (either GP or GN on final cultures in the absence of GP or GN on the Gram stain). Qualitative breakdown of the maximum amount of cocci or bacilli on microscopic examination was also done to explore possible relationships with clinical outcomes Statistical analysis Baseline characteristics were compared between patients with and without organisms identified on Gram staining at enrolment. Continuous data are reported as mean ± SD and categorical data as proportions or percentages. Differences were tested by the independent t test or the χ 2 test for continuous and categorical variables, respectively. Agreement between Gram staining and organisms cultured was measured by the simple κ. All agreement analyses were done using individual samples from the different diagnostic tests and pooled samples (BAL and EA combined). For BAL results, we assessed agreement on all positive cultures and on those growing more than 10 4 colony-forming unit (CFU)/mL. Agreement was measured for the specific classification of Gram stain (GP, GN, neither, or both). The association between the amount of bacilli and cocci on Gram stain and the selected outcome variables was tested by the Cochran-Mantel-Haenszel test for linear association. All analyses were performed using SAS version 9.1 (SAS, Inc, Cary, NC). 3. Results Between May 2000 and February 2005, 2535 patients were screened at 28 ICUs; 1168 were eligible and 740 patients were randomized. After withdrawal of consent (n = 1) and exclusion due to lack of gram stain (n = 34), we included 705 patients in this study. There were 350 samples collected by BAL and 355 by EA. Baseline characteristics of the patients are presented in Table 1. Comparing baseline characteristics of patients with no organisms on Gram stain and all other patients with either GN, GP, or both organisms, we observed that patients with no organisms on Gram stain had higher admission APACHE II scores (20.8 ± 6.6 vs 19.1 ± 5.9, P b.01), were more likely to be admitted with a respiratory condition (20.8% vs 13.9%, P =

4 Utility of Gram stain in the clinical management of suspected VAP 77 Table 2 Organisms present from enrollment specimens in study patients GP Pathogenic organisms Staphylococcus aureus 119 (16.9%) ecies 33 (4.7%) Enterococcus species 14 (2.0%) MRSA 12 (1.7%) GN Pathogenic organisms Hemophilus influenzae 98 (13.9%) ecies 64 (9.1%) ecies 56 (7.9%) Pseudomonas species 47 (6.7%) Escherichia coli 41 (5.8%) ecies 21 (3.0%) Acinetobacter species 14 (2.0%) Proteus species 14 (2.0%) Moraxella catarrhalis 13 (1.8%) Stenotrophomonas maltophilia 12 (1.7%) s Common or mixed flora 111 (15.7%) No growth 124 (17.6%) Candida species 112 (15.9%) Aspergillus species 8 (1.1%) Coagulase-negative Staphylococci 12 (1.7%) 46 (6.5%) Pathogenic and nonpathogenic ( s ) organisms were defined a priori..002), had worse initial PaO 2 /FiO 2 ratio (207 ± 80 vs 227 ± 85, P =.001), and were more likely to be on antibiotics before randomization (77.3% vs 50.0%, P b.001). Table 2 presents culture results for all pathogenic and nonpathogenic organisms. We found that 46% of patients had no pathogenic organism cultured, whereas multiple pathogenic organisms were identified in 20% of the cultures. The raw agreement between enrollment Gram staining and organism Gram typing was 55% by both BAL and EA. Furthermore, concordance between Gram stain findings and culture results, as measured by κ (95% CI), was similar for EA (κ = 0.37 [ ]) and BAL (κ = 0.35 [ ], P =.70). Hence, EA and BAL results were pooled and the concordance between the enrollment Gram stain and the culture results for pathogenic organisms is shown in Table 3. For Gram stains containing neither GP nor GN organisms, 99 (30%) of 331 grew pathogenic organisms in the eventual culture results. Among specimens with Gram stains showing no GN organisms, 113 (25%) of 452 grew pathogenic GN organisms. Among specimens with Gram stains showing no GP organisms, 45 (11%) of 428 grew pathogenic GP organisms. We also compared how the Gram stains correlated with quantitative BAL culture results, using more than 10 4 CFU/ ml pathogenic organisms as the threshold for a positive culture. For BAL Gram stains containing neither GP nor GN organisms, 46 (25%) of 187 grew pathogenic organisms in quantitative cultures; for Gram stains showing no GN organisms, 44 (18%) of 246 specimens grew pathogenic GN organisms; and for Gram stains showing no GP organisms, 25 (11%) of 230 grew pathogenic GP organisms. Table 4 summarizes sensitivity, specificity, positive (PPV) and negative predictive value (NPV), positive (LR+) and negative likelihood ratio (LR ), and κ correlation for overall pooled analysis and each diagnostic test separately. Overall, the absence of either a GP or GN organism seen on Gram stain had an NPV of 70%. For EAs, the absence of a GP organism yielded an NPV of 93%, whereas the absence of a GN organism yielded an NPV of 81%. For BALs, the NPV for GPs was 86% and for GNs was 70%. Overall, 144 (26%) of 549 samples with Gram stains showing either no bacteria or bacteria from a single Gram class grew pathogenic organisms from a different Gram class. This overall potential undertreatment rate was different according to prerandomization antibiotic status: 60 (36%) of 167 for patients not on antibiotics up to 3 days before enrollment, 39 (20%) of 207 for patients on antibiotics but not changed within 3 days before enrollment, and 45 (26%) of 175 for patients with new antibiotics started within 3 days of enrollment (P =.001). Qualitative breakdown of bacteria on microscopic examination demonstrated that the amount of bacteria tended to be greater in patients with positive final culture results (P b.001), with confirmed pneumonia (P =.002), and who received targeted therapy by study day 5 (P =.014). No significant correlation was found between amount of bacteria Table 3 Agreement between Gram stain and culture results for pathogenic organisms Gram stain result Pathogenic organism Total Neither GP GN Both Neither 232 (32.9) 25 (3.6) 60 (8.5) 14 (2.0) 331 (47.0) GP 43 (6.1) 39 (5.5) 16 (2.3) 23 (3.3) 121 (17.1) GN 7 (1.0) 1 (0.1) 84 (12.0) 5 (0.7) 97 (13.6) Both 42 (6.0) 16 (2.3) 64 (9.1) 34 (4.8) 156 (22.3) Total 324 (46.0) 81 (11.5) 224 (31.8) 76 (10.8) 705 (100.0) Values are shown as n (overall%). Pooled data for BALs and EAs.

5 78 M. Albert et al. Table 4 Endotracheal and BAL Gram stain diagnostic performance Sensitivity Specificity PPV NPV LR+ LR κ (95% CI) Overall GP + GN 74% 72% 75% 70% ( ) EA GP 81% 65% 36% 93% ( ) BAL GP 64% 76% 47% 86% ( ) EA + BAL GP 71% 70% 40% 89% ( ) EA GN 73% 79% 71% 81% ( ) BAL GN 53% 88% 78% 70% ( ) EA + BAL GN 62% 84% 74% 75% ( ) For the overall GP + GN row: sensitivity, specificity, PPV, NPV, LR+, and LR- were done pooling all Gram stain types together (ie, GP, GN, and GN + GP), but the κ did require Gram stain types to match. Overall κ was similar when quantitative (N10 4 CFU/mL) BAL culture results are used: 0.29 (95% CI ). and 28-day mortality (P =.061) or number of days off all antibiotics (P =.23). 4. Discussion We performed a secondary analysis of data from a multicenter randomized trial of diagnostic strategies and empiric antibiotics for VAP to examine the agreement of Gram stains with final culture results. Our study is the largest published to date on the topic. Our principal finding is that the Gram stain poorly predicts the final culture results. In particular, a high proportion (25%-30%) of specimens with no organism seen on Gram stain end up growing pathogenic bacteria. In our data set, these results did not differ whether the samples were obtained using EA or all BAL cultures or quantitative cultures from BAL. Our findings highlight the potential risk of withholding antibiotic therapy for patients with suspected VAP based on the absence of organism on initial Gram stain results. According to our results, the only possible utility to the Gram stain would be the high NPV of EA Gram stains showing no GP bacteria. In our study, this result was highly predictive (93%) of a final culture result with no GP organism. It is important to note that we excluded patients from this trial if they had a known history of colonization or infection with MRSA. Based on this high NPV, we believe it is reasonable to withhold empiric coverage for GP bacteria if the initial Gram stain from the EA does not demonstrate any such bacteria if there is no prior history of infection or colonization of MRSA. However, the NPV for a GP Gram stain is substantially lower for BAL specimens (86%). In addition, if patients are in ICUs where the overall prevalence of GP bacteria is high or patients are known to be colonized or previously infected with MRSA, clinicians may still consider empiric or specific GP coverage. Qualitative breakdown of bacteria on microscopic examination did correlate with positive final culture results and final diagnosis of pneumonia. However, given the overall low predictive value of the Gram stain, its qualitative assessment is unlikely to be of any clinical utility as a guide for initial antibiotic therapy in suspected VAP. Diagnosis of VAP in the ICU remains a challenge despite numerous studies of diagnostic techniques and treatment algorithms. Retrospective observational data show increased mortality and morbidity from VAP when initial antibiotic therapy does not treat subsequently cultured bacteria [2]. Many approaches have been suggested to decrease the risk of inadequate empiric coverage: using local resistance patterns to guide antibiotic selection, invasive techniques for diagnosis, and clinical protocols based on the initial specimen Gram stain [1,12-14]. However, the latter approach assumes that Gram stains of EA and BAL reliably predict final culture results. Our results challenge this assumption. Many studies have examined the value of Gram stains in a clinical setting of VAP. Nevertheless, most of them were either retrospective, with small sample size or single center [5-7,12,15-21]. The largest study prospectively included 232 patients with suspected VAP using invasive BAL techniques [15]. The authors' conclusion was that Gram stain of BAL correlated poorly with quantitative cultures and was therefore not reliable for dictating empiric therapy. Interestingly, a recent publication examining an ICU trauma population found similar results [16]. In a retrospective analysis of 155 patients with VAP over 5 years in one ICU, the authors were unable to show a clinically significant relationship between Gram stain and culture results. However, they did find that the absence of GP on Gram stains had an NPV of 83%, similar to our results. The specificity of Gram stain for final cultures showing GP and GN bacteria was only 49% and 59%, respectively. studies contradict our findings. For example, Fagon et al [14] evaluated the effect of BAL vs EA to diagnose VAP, without standardizing antibiotic selection, in a multicenter RCT that enrolled 413 ICU patients. Both strategies used direct examination of either invasive or noninvasive specimens to select one of the predetermined treatments. If the Gram stain examination was negative, antibiotics were withheld except if severe sepsis criteria were present. They found that the invasive approach was associated with a decrease in mortality at 14 days and improved secondary outcomes such as earlier attenuation of organ dysfunction and less antibiotic use. Only 3 (17%) of 18 nonseverely septic

6 Utility of Gram stain in the clinical management of suspected VAP 79 patients with negative Gram stains in the noninvasive group and 7 (7%) of 97 such patients in the invasive group who initially had antibiotics withheld subsequently had positive cultures requiring antibiotics. These proportions appear lower than in our study. The Fagon et al study states that the mortality of all initially untreated patients was similar; however, it does not provide separate outcomes data for those with subsequent positive cultures. Even if these data were reported, only 10 patients fell into this category, limiting any attempt to reliably determine whether the delay in starting antibiotics contributed to significantly higher mortality. It may be that patients with no bacteria on Gram stain but subsequent positive cultures represent a lower severity of bacterial pneumonia and are at lower risk of delayed antibiotic therapy. The Fagon et al trial appears to have too few patients to determine this, and the Canadian VAP study [10] is unable to provide additional insight as all patients received broad spectrum antibiotics regardless of Gram stain results. Recently, Veinstein et al [13] attempted to validate a diagnosis and treatment algorithm based on direct examination of specimens of EAs and protected telescopic catheter in suspected VAP. In this observational study of 76 patients, clinical severity criteria and Gram stains were used to empirically treat patients with a positive Gram stain and to suspend treatment until cultures were available if the Gram stain was negative. They found that early management was appropriate in 80% of the patients and that 83% received appropriate antimicrobial regimen. In other words, 17% of patients did not receive adequate initial antibiotics. Similarly to our results, they found that among 21 patients with negative Gram stain of EA and protected telescopic catheter, 3 (14%) had bacteria growing on final culture results. Clinicians using this algorithm could potentially withhold antibiotics for many days in 14% of the patients that finally had microbiologic evidence of pneumonia. This approach may be unwarranted given the potential increase in morbidity and mortality associated with withholding appropriate antibiotics [2,22,23]. Our study is the largest published to date on the topic and involves multiple academic and community ICUs with many different personnel collecting and analyzing the ETT and BAL samples. This greatly increases the generalizability of the results. It also has the advantage of comparing the validity of Gram stains from both invasive and noninvasive techniques. Another strength of our study is that patients needed to have clinical and radiological suspicion of pneumonia to be included in this study, which we believe is similar to what drives suspicion an empiric treatment of pneumonia in clinical practice. Finally, standardized protocols for microbiological procedures were used, and all data were acquired prospectively. Our study has some limitations. Because patients with MRSA colonization were excluded from the study, caution should be taken before extrapolating these results to ICU with high prevalence of such bacteria. Also, we did not collect other microscopic examination findings on the specimens we sampled that could have been correlated to final culture results, such as intracellular microorganism quantification. However, a recent study by Sirvent et al [8] did not show any advantage in using percentage of cells containing intracellular organisms to further classify patients with clinical suspicion of VAP. Another limitation of our study is that many patients were already on antibiotics at the point of enrollment into this study. This could have affected our results by decreasing the number of positive cultures even if the Gram stain did initially show some amount of bacteria. However, this situation likely reflects routine clinical practice and supports the generalizability of our results. Unexpectedly, our subgroup analysis demonstrated a higher potential undertreatment rate in patients not receiving any antibiotic at randomization. Finally, we did not specifically examine the agreement and potential undertreatment rate in the subgroup of patients with suspected VAP and organ dysfunction. 5. Conclusion Our study demonstrates that EA and BAL Gram stains performed for clinically suspected VAP poorly predict the antibiotic treatment required according to final culture results. The only possible exception would be for GP bacteria: the absence of GP bacteria on Gram stain of an EA excluded a GP culture in 93% (NPV) of patients in a cohort of patients not known to be previously colonized or infected with MRSA. wise, no clinically reliable correlation was found using the Gram stain as a diagnostic tool. These data suggest that Gram stain has a very limited role in guiding initial empiric antibiotic therapy in patients with suspected VAP. Acknowledgments We thank the investigators, staff, and members of the steering committee of the Canadian Critical Care Trials Group who participated in this study; see the citation for a complete list of participants [10]. Appendix A Consensus classification of pathogenic and nonpathogenic organisms Organism name Organism class Pathogenic GN organisms Acinetobacter baumannii Acinetobacter sp Acinetobacter sp Acinetobacter sp Acinetobacter xylosoxidans Acinetobacter sp (continued on next page)

7 80 M. Albert et al. Appendix A (continued) Organism name Aerogenes sp Aeromonas hydrophilia Bordetella not pertussis Burkholderia cepacia Capnocytophaga sp Citrobacter braakii Citrobacter Freundii Citrobacter sp Eikenella sp Enterobacter aerogenes Enterobacter cloacae Enterobactera asburiae Escherichia coli Flavimonas oryzihabitans Hafnia alvei Hemophilus influenza Hemophilus parainfluenza Klebsiella azaenae Klebsiella oxytoca Klebsiella pneumonia Leclercia adecarboxylata Moraxella catarrhalis Morganella sp Neisseria meningitides Neisseria sp Pasteurella multocidia Proteus mirabilis Proteus sp Pseudomonas sp Serratia liquefaciens Serratia marcescens Serratia odorifera Sphingomonas sp Stenotrophomonas maltophilia Pathogenic GP organisms Enterococcus sp Staphylococcus aureus Staphylococcus aureus methicillin resistant Streptoccocus agalactiae Streptococcus beta haemolytic group B Streptococcus beta haemolytic group F Streptococcus pneumoniae Streptococcus viridans Nonpathogenic organisms Aspergillus sp Bacteroides sp Candida sp Common or mixed flora Corynebacteria sp Organism class Aeromonas sp Bordetella sp Burkholderia sp Citrobacter sp Citrobacter sp Citrobacter sp Eikenella sp Escherichia sp Flavimonas sp Hafnia sp Hemophilus sp Hemophilus sp Moraxella sp Morganella sp Neisseria sp Neisseria sp Pasteurella sp Proteus sp Proteus sp Pseudomonas sp Sphingomonas sp Stenotrophomonas sp Enterococcus sp Staphylococcus sp Staphylococcus sp Aspergillus sp Bacteroides sp Candida sp Appendix A (continued) Organism name Diptheroids sp Prevotella sp Scopulariopsis sp Staphylococcus coagulase negative References Organism class Prevotella sp Scopulariopsis sp Staphylococcus sp [1] Shorr AF, Kollef MH. Ventilator-associated pneumonia: insights from recent clinical trials. Chest 2005;128:583S-91S. [2] Iregui M, Ward S, Sherman G, et al. Clinical importance of delays in the initiation of appropriate antibiotic treatment for ventilatorassociated pneumonia. Chest 2002;122: [3] Kollef M. Appropriate empirical antibacterial therapy for nosocomial infections: getting it right the first time. Drugs 2003;63: [4] Neuhauser MM, Weinstein RA, Rydman R, et al. Antibiotic resistance among gram-negative bacilli in US intensive care units: implications for fluoroquinolone use. JAMA 2003;289: [5] Allaouchiche B, Jaumain H, Chassard D, et al. Gram stain of bronchoalveolar lavage fluid in the early diagnosis of ventilatorassociated pneumonia. Br J Anaesth 1999;83: [6] Duflo F, Allaouchiche B, Debon R, et al. An evaluation of the Gram stain in protected bronchoalveolar lavage fluid for the early diagnosis of ventilator-associated pneumonia. Anesth Analg 2001;92: [7] Blot F, Raynard B, Chachaty E, et al. Value of gram stain examination of lower respiratory tract secretions for early diagnosis of nosocomial pneumonia. Am J Respir Crit Care Med 2000;162: [8] Sirvent JM, Vidaur L, Gonzalez S, et al. Microscopic examination of intracellular organisms in protected bronchoalveolar mini-lavage fluid for the diagnosis of ventilator-associated pneumonia. Chest 2003;123: [9] Allaouchiche B, Jaumain H, Dumontet C, et al. Early diagnosis of ventilator-associated pneumonia. Is it possible to define a cutoff value of infected cells in BAL fluid? Chest 1996;110: [10] Heyland D, Dodek P, Muscedere J, et al. A randomized trial of diagnostic techniques for ventilator-associated pneumonia. N Engl J Med 2006;355: [11] Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985;13: [12] Brun-Buisson C, Fartoukh M, Lechapt E, et al. Contribution of blinded, protected quantitative specimens to the diagnostic and therapeutic management of ventilator-associated pneumonia. Chest 2005;128: [13] Veinstein A, Brun-Buisson C, Derrode N, et al. Validation of an algorithm based on direct examination of specimens in suspected ventilator-associated pneumonia. Intensive Care Med [14] Fagon JY, Chastre J, Wolff M, et al. Invasive and noninvasive strategies for management of suspected ventilator-associated pneumonia. A randomized trial. Ann Intern Med 2000;132: [15] Croce MA, Fabian TC, Waddle-Smith L, et al. Utility of Gram's stain and efficacy of quantitative cultures for posttraumatic pneumonia: a prospective study. Ann Surg 1998;227:743-51, discussion [16] Davis KA, Eckert MJ, Reed Jr RL, et al. Ventilator-associated pneumonia in injured patients: do you trust your Gram's stain? J Trauma 2005;58:462-6, discussion

8 Utility of Gram stain in the clinical management of suspected VAP 81 [17] Marquette CH, Wallet F, Neviere R, et al. Diagnostic value of direct examination of the protected specimen brush in ventilator-associated pneumonia. Eur Respir J 1994;7: [18] Marquette CH, Copin MC, Wallet F, et al. Diagnostic tests for pneumonia in ventilated patients: prospective evaluation of diagnostic accuracy using histology as a diagnostic gold standard. Am J Respir Crit Care Med 1995;151: [19] Prekates A, Nanas S, Argyropoulou A, et al. The diagnostic value of gram stain of bronchoalveolar lavage samples in patients with suspected ventilator-associated pneumonia. Scand J Infect Dis 1998;30:43-7. [20] Mimoz O, Karim A, Mazoit JX, et al. Gram staining of protected pulmonary specimens in the early diagnosis of ventilator-associated pneumonia. Br J Anaesth 2000;85: [21] Wahl WL, Ahrns KS, Brandt MM, et al. Bronchoalveolar lavage in diagnosis of ventilator-associated pneumonia in patients with burns. J Burn Care Rehabil 2005;26: [22] Steinberg MAM, Parkes R, Ferguson ND, Rizoli SB, McLeod RS, Marshall JC, Canadian Critical Care Trials Group. Empiric antimicrobial therapy for suspected ICU-acquired infection: The AATIC pilot study. Intensive Care Med 2006;32:113. [23] Baker AMMJ, Chang M, Dunagan D, Smith A, Haponik E. Bronchoscopically guided management of ventilator-associated pneumonia in trauma patients. J Bronchol 2003;10:7-16.