Gram staining of protected pulmonary specimens in the early diagnosis of ventilator-associated pneumonia

O. Mimoz*, A. Karim, J. X. Mazoit, A. Edouard, S. Leprince and P. Nordmann

Department of Anaesthesiology, Bicêtre Hospital, Assistance Publique-Hôpitaux de Paris, F-94275 Le Kremlin Bicêtre, France

Accepted for publication: 26 June, 2000


    Abstract
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
We evaluated prospectively the use of Gram staining of protected pulmonary specimens to allow the early diagnosis of ventilator-associated pneumonia (VAP), compared with the use of 60 bronchoscopic protected specimen brushes (PSB) and 126 blinded plugged telescopic catheters (PTC) obtained from 134 patients. Gram stains were from Cytospin slides; they were studied for the presence of microorganisms in 10 and 50 fields by two independent observers and classified according to their Gram stain morphology. Quantitative cultures were performed after serial dilution and plating on appropriate culture medium. A final diagnosis of VAP, based on a culture of >=103 c.f.u. ml–1, was established after 81 (44%) samplings. When 10 fields were analysed, a strong relationship was found between the presence of bacteria on Gram staining and the final diagnosis of VAP (for PSB and PTC respectively: sensitivity 74 and 81%, specificity 94 and 100%, positive predictive value 91 and 100%, negative predictive value 82 and 88%). The correlation was less when we compared the morphology of microorganisms observed on Gram staining with those of bacteria obtained from quantitative cultures (for PSB and PTC respectively: sensitivity 54 and 69%, specificity 86 and 89%, positive predictive value 72 and 78%, negative predictive value 74 and 84%). Increasing the number of fields read to 50 was associated with a slight decrease in specificity and positive predictive value of Gram staining, but with a small increase in its sensitivity and negative predictive value. The results obtained by the two observers were similar to each other for both numbers of fields analysed. Gram staining of protected pulmonary specimens performed on 10 fields predicted the presence of VAP and partially identified (using Gram stain morphology) the microorganisms growing at significant concentrations, and could help in the early choice of the treatment of VAP. Increasing the number of fields read or having the Gram stain analysed by two independent individuals did not improve the results.

Br J Anaesth 2000; 85: 735–9

Keywords: intensive care; infection, nosocomial; complications, ventilator-associated pneumonia


    Introduction
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Ventilator-associated pneumonia (VAP) is a serious complication in mechanical ventilation.1 Inaccurate or delayed antibiotic therapy can affect the prognosis.1 2 The best method of diagnosis for VAP is not clear. Protected specimen brushing (PSB) is a method with sensitivity of 90% and specificity of 75% if 103 colony-forming units (c.f.u.) ml–1 is taken as the threshold.3 Blinded plugged telescopic catheter (PTC) sampling may also be reliable with the same threshold, although it seems more sensitive and less specific than PSB.1 3 Whatever the method of sampling, a 24-h delay is needed for quantitative cultures, which limits early diagnosis. Gram staining of these protected pulmonary specimens could overcome this problem, but its diagnostic yield remains controversial.1 3

The present study was undertaken to evaluate prospectively whether the Gram stain of PSB or PTC predicts the results of quantitative cultures and can guide the choice the treatment of pneumonia. We also studied the possible improvement of the accuracy of Gram staining by analysing more fields or by having two independent observers analyse the stains.


    Patients and methods
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Study population
We studied 134 successive adult patients in the medical and surgical intensive care units of the University Hospital of Bicêtre who were suspected of having nosocomial pneumonia. All had new and persistent pulmonary infiltrates on chest radiography that were unrelated to cardiogenic causes, macroscopically purulent tracheal aspirates, and leucocytosis (>12 000 cells mm–3), and had been mechanically ventilated for at least 48 h. Although most (>90%) of the patients were febrile (>38.5°C), fever was not systematically used as a clinical criterion of pneumonia as some patients were given muscle relaxants.

One hundred and eight protected pulmonary specimens were taken from patients who had not received antimicrobial therapy during the previous 3 days, and 78 protected pulmonary specimens were taken from patients who had received antibiotics (for more than 72 h). No patient received local antimicrobial therapy or digestive decontamination with antibiotics.

Specimen collection
We obtained 126 PTC and 60 PSB samples after careful endotracheal suctioning and without topical anaesthesia.3 The choice of one of these two techniques was left to the physician in charge of the patient. During the procedures, 100% oxygen was given and the patients were sedated with midazolam; pancuronium was used to achieve neuromuscular block. Heart rate, arterial pressure and arterial oxygen saturation were monitored during the investigation.

Blinded PTC samplings were performed by the previously described standardized technique.4 The PTC unit was composed of an outer catheter sealed by a polyethylene glycol plug and an inner catheter for sampling. After blind introduction of a PTC unit into the bronchial system, the inner catheter was advanced, extruding the plug, and three brief aspirations were then applied to the inner catheter with a 10-ml syringe connected to its proximal port. After the catheter had been retracted into the outer sheath, the entire unit was removed from the patient; the distal end of the outer catheter was dried and then cut off with sterile scissors, the inner catheter was advanced, and 1 ml of saline was flushed through its proximal port and collected into a sterile vial. Finally, the distal segment of the catheter was transected and collected into the same vial. For the PSB, a bronchoscope (Olympus BF P 20D; Olympus Optical Corporation of America, New Hyde Park, NY, USA) was introduced through an adapter (Bodai Suction Safe Y; Sontek Medical, Lexington, MA, USA) and advanced to the bronchial orifice, selected according to the radiographic position of the new pulmonary infiltrate. After brushing, the tip of the PSB was cut aseptically and dropped into a sterile glass vial containing 1 ml of saline, according to a standard technique.3 All the specimens were taken to the laboratory within 15 min of collection.

Microbiological processing
Microscopic examination
From the original suspension, 0.2-ml aliquots were dropped into a Cytospin and centrifuged at 300 g for 10 min. The slides were Gram-stained and independently examined twice (by A.K. and S.L.) at high magnification (x100) without knowledge of the quantitative culture results. The presence of microorganisms was looked for on 10 and 50 fields and any microorganisms found were classified according to their Gram stain morphology.

Quantitative cultures
Each vial was mechanically vortexed for 60 s. Two successive 1:100 saline dilutions were prepared. Aliquots (0.1 ml) of the original suspension and of each dilution were plated onto different media for quantitative culture and identification: fresh blood agar, fresh blood agar to which nalidixic acid, amphotericin B and colistin had been added, Drigalski agar, chocolate agar and anaerobic blood agar (BioMérieux, La Balme-les-Grottes, France). All samples were incubated for 48 h in an appropriate atmosphere and the microorganisms recovered were identified by standard methods and their number expressed as c.f.u ml–1. The cut-off value of protected pulmonary specimens for the diagnosis VAP was >=103 c.f.u. ml–1.1

Statistical analysis
Data are expressed as medians and ranges. To assess the degree of qualitative correlation between Gram staining and quantitative cultures, the results of Gram staining were divided into four categories. Gram staining was considered to be true-positive if each Gram stain morphotype present also grew at significant concentration (i.e. >=103 c.f.u. ml–1), true-negative if no organism was present on Gram staining and quantitative culture was either sterile or non-significantly positive (i.e. <103 c.f.u. ml–1), false-positive if a morphotype present on Gram staining was not grown in significant concentration, and false-negative if a microorganism grew at significant concentration but was not present on Gram staining. For mixed cultures, all the bacterial isolates grown at significant concentration had to be morphologically present on Gram staining. If not, the case was classified as false-negative. Similarly, the case was classified as false-positive if all the morphotypes present on Gram staining did not grow at significant concentration. Sensitivity, specificity and predictive value were calculated.5 The Spearman rank test was used to assess the relationship between the number of bacilli and/or cocci seen on Gram staining and the number of microorganisms subsequently growing on appropriate culture medium; P<0.05 was considered to be statistically significant. The {kappa} statistic was used to measure the agreement between the two observers.6 A {kappa} value of 0 indicates no agreement beyond chance, whereas a {kappa} value of 1 indicates perfect agreement.


    Results
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
We studied 99 men and 35 women aged 56 (16–95) yr who had been in hospital for 9 (3–233) days. They had a final stay in the intensive care unit (ICU) of 24 (4–356) days and ICU mortality of 24%. The primary reasons for ventilation were trauma (78 patients), postoperative respiratory failure (12 patients), exacerbation of chronic obstructive pulmonary disease (18 patients), severe sepsis (23 patients) and acute pancreatitis (three patients). No adverse event was observed during any procedure.

Eighty-one (44%) protected pulmonary specimens (54 PTC and 27 PSB) had a positive culture, including 17 specimens with two (11 PTC and six PSB) or three (one PSB) microorganisms with different Gram stain morphologies. Microorganisms isolated at significant concentration were 24 Staphylococcus aureus, 1 isolate of coagulase-negative staphylococci, 7 Streptococcus spp., 10 Streptococcus pneumoniae, 1 Neisseria spp., 9 Escherichia coli, 14 Haemophilus influenzae, 2 Proteus spp., 1 Klebsiella spp., 4 Enterobacter spp., 3 Serratia spp., 19 Pseudomonas aeruginosa, 1 Acinetobacter spp., 1 Stenotrophomonas maltophilia, 2 Branhamella catarrhalis, and 1 Bacteroides spp. Ten patients (12%) died rapidly and the remaining 71 recovered from their bacterial pneumonia while receiving appropriate antibiotic therapy for organisms cultured at a significant growth in protected pulmonary specimens. The causes of death were overwhelming sepsis (n=2), multiple organ failure of unknown origin (n=5), brain death (n=2) and refractory hypoxaemia (n=1). Two (6%) of the 31 patients with a non-significant positive protected pulmonary specimen developed bacterial pneumonia (defined as a protected pulmonary specimen culture significantly positive and a clinical outcome consistent with pneumonia while receiving appropriate antibiotic therapy) caused by the same organisms 4 and 5 days later. None of the patients with a sterile protected pulmonary specimen culture later developed bacterial pneumonia during the 5 days after their inclusion in the study.

When 10 fields were analysed, a strong relationship was found between the presence of bacteria on Gram staining and the final diagnosis of VAP (Table 1). The ability of PTC Gram staining to predict VAP seemed better than that of PSB Gram staining, although the difference was not significant. Increasing the number of fields read to 50 was associated with a slight decrease in specificity and in the positive predictive value of the Gram stain, but with a small increase in its sensitivity and negative predictive value (Table 2). Whatever the number of fields read, the agreement between the two observers was excellent ({kappa}=0.88 and 0.82 for 10 and 50 fields respectively).


View this table:
[in this window]
[in a new window]
 
Table 1 Performance of Gram staining in predicting ventilator-associated pneumonia when 10 fields were analysed by two independent observers. Values in parentheses are 95% confidence intervals
 

View this table:
[in this window]
[in a new window]
 
Table 2 Performance of Gram staining in predicting ventilator-associated pneumonia when 50 fields were analysed by two independent observers. Values in parentheses are 95% confidence intervals
 
The ability of Gram staining to predict which microorganisms subsequently grew at significant concentration when 10 fields were read is shown in Table 3. Again, the performance of PTC Gram staining was not significantly better than that of PSB Gram staining. Increasing the number of fields observed to 50 gave a slight decrease in specificity and in the positive predictive value of the Gram stain and a small increase in its sensitivity and negative predictive value (Table 4). The agreement between the two observers was also excellent ({kappa}=0.87 and 0.85 when 10 and 50 fields were studied respectively).


View this table:
[in this window]
[in a new window]
 
Table 3 Performance of Gram staining in predicting the morphology of bacteria growing at significant concentration when 10 fields were analysed by two independent observers. Values in parentheses are 95% confidence intervals
 

View this table:
[in this window]
[in a new window]
 
Table 4 Performance of Gram staining in predicting the morphology of bacteria growing at significant concentration when 50 fields were analysed by two independent observers. Values in parentheses are 95% confidence intervals
 
Whatever the number of fields studied or whoever performed the Gram stain analysis, a strong correlation (P<0.001) was found between the number of bacilli and/or cocci observed on Gram staining and the number of bacteria obtained in quantitative cultures.


    Discussion
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
The diagnosis of bacterial pneumonia in severely ill, mechanically ventilated patients is difficult.7 Diagnostic strategies for pneumonia range from a clinical approach to the use of invasive sampling. None of these strategies is without problems. The clinical approach may be too sensitive, and patients may be treated for pneumonia when a non-pulmonary infection or a non-infectious pulmonary process is present. Invasive sampling obviates the need for broad-spectrum antibiotic coverage for all patients with suspected pneumonia, reduces the risk of emergence of resistant microorganisms in the ICU and reduces the cost of antibiotic treatment. However, these techniques are expensive and need a specialized laboratory and specific clinical skills. Moreover, these techniques may be inaccurate in patients receiving antibiotics at the time of testing and may not be sensitive enough to diagnose all cases of ventilator-associated pneumonia, especially early infections.8 No study has formally shown that the outcome of patients with ventilator-associated pneumonia can be improved if invasive methods are used in place of clinical diagnosis.

When invasive sampling is used, bacterial growth in the specimens is quantified and the presence of pneumonia and the identity of the causative pathogen(s) are defined by the recovery of bacteria at a concentration above a predetermined threshold (>=103 c.f.u. ml–1).7 Results are delayed until the quantitative cultures are available, and initial therapy is therefore empirical. Previous studies found that Gram staining of protected specimens had low sensitivity, prompting several authors to use microscopic analysis of bronchoalveolar lavage material along with quantitative cultures of protected specimens.9 10 The use of two invasive techniques in combination increases the cost, carries an additional risk for the patient and increases the difficulty and the work-load of the laboratory.

The present study shows that Gram staining of protected pulmonary specimens correctly predicts the presence of VAP and may partially identify (using Gram stain morphology) the microorganisms subsequently growing at significant concentrations. Our results support previous studies in which Gram staining was performed on slides prepared directly from the brushes. Teague et al. used two brushes, one for Gram staining and the other for quantitative culture, and found a marked correlation between Gram staining and the presence of bacteria at a concentration of >=103 c.f.u. ml–1.11 Similar results were reported by Pollock et al. when the brush was smeared directly onto a glass slide before the brush was placed in the holding medium.12 In contrast, early studies using Gram staining on centrifuged specimens reported low sensitivity for the method.4 9 13 More recently, Marquette et al. reported 85% sensitivity and 94% specificity for cytocentrifuged Gram staining.14 The authors said they obtained good results because they first screened for bacteria using May-Grünwald Giemsa-stained slides at high magnification before performing (if these stains were positive) Gram staining on slides to determine the precise morphology of the bacteria detected. We believe that our results may be accounted for by several factors: cytocentrifuged preparations were made with a high, calibrated volume for the initial dilution (200 µl, i.e. 20% of the total dilution volume), and Acinetobacter spp., which may be detected either as Gram-negative coccobacilli or as Gram-positive cocci depending on its state of development, were rarely isolated. However, increasing the number of fields read or having the Gram staining analysed by two independent individuals were both time-consuming and did not usefully improve accuracy. Our results agree partially with those obtained previously when bronchoalveolar lavage was used to diagnose bacterial pneumonia. In a study reported recently in this journal, Gram staining of bronchoalveolar lavage was found to be useful for the rapid diagnosis of ventilator-associated pneumonia, but was not reliable for the early adaptation of empirical therapy.15 By contrast, the role of Gram staining of tracheal aspirates was not useful in the diagnosis of ventilator-associated pneumonia or in guiding empirical therapy.16

Our study has several limitations. First, the effects on patient outcome and antibiotic management in relation to the data collected were not evaluated because Gram staining was generally performed many days after sampling and without knowledge of the results of quantitative cultures. Secondly, the performance of Gram staining of protected pulmonary specimens was not compared with results obtained by bronchoalveolar lavage or tracheal aspirates. However, it was not our policy to use such sampling methods routinely to investigate patients with suspected nosocomial pneumonia. The use of both invasive techniques in combination is expensive and time-consuming and carries an additional risk for the patient. Finally, PSB and PTC were considered as reference methods of diagnosing VAP. Until recently, lung histology has been considered the gold standard in this situation. However, when 39 open lung biopsies obtained from dead patients were reviewed independently by four pathologists, the prevalence of histological VAP, as determined by each of the pathologists, varied from 18 to 38%.17 Histological features of VAP were found in seven of nine organ donors who had no clinical evidence of pulmonary infection and who were not receiving antibiotic therapy.18 Finally, studies have found discrepancies between histology and bacteriological cultures. Neither the bacterial densities from the four quantitative airway cultures nor the bacterial density from the quantitative culture of lung parenchyma separated the histological pneumonia and non-pneumonia groups accurately.19

In summary, Gram staining of protected pulmonary specimens performed on 10 fields predicted the presence of VAP and partially identified (using Gram stain morphology) the microorganisms growing at significant concentrations. Further studies on antibiotic management and patient outcome, based on this method, are warranted.


    Footnotes
 
* Corresponding author: Department of Anaesthesiology, Paul Brousse Hospital, Assistance Publique-Hôpitaux de Paris, F-94804 Villejuif, France Back


    References
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
1 Anonymous hospital-acquired pneumonia in adults: diagnosis, assessment of severity, initial antimicrobial therapy, and preventive strategies. A consensus statement. Am J Respir Crit Care Med 1995; 153: 1711–25[ISI][Medline]

2 Luna CM, Vujacich P, Niederman MS, Vay C, Gherardi C, Matera J, et al. Impact of bronchoalveolar lavage data on the therapy and outcome of ventilator-associated pneumonia. Chest 1997; 111: 676–85[Abstract/Free Full Text]

3 International Consensus Conference—clinical investigation of ventilator-associated pneumonia. Chest 1991; 102 (Suppl. 1): 551–84[Abstract]

4 Pham LH, Brun-Buisson C, Legrand P, et al. Diagnostic of nosocomial pneumonia in mechanically ventilated patients. Comparison of a plugged telescoping catheter with the protected specimen brush. Am Rev Respir Dis 1991; 143: 1055–61[ISI][Medline]

5 Youden WJ. Index for rating diagnostic tests. Cancer 1950; 3: 32–5[ISI]

6 Posner KL, Sampson PD, Caplan RA, Ward RJ, Cheney FW. Measuring interrater reliability among multiple raters: an example of methods for nominal data. Stat Med 1990; 9: 1103–15

7 Chastre J, Fagon JY, Trouillet JL. Diagnosis and treatment of nosocomial pneumonia in patients in intensive care units. Clin Infect Dis 1995; 21 (Suppl. 3): 226–37

8 Nierderman MS, Torres A, Summer W. Invasive diagnostic testing is not needed routinely to manage suspected ventilator-associated pneumonia. Am Rev Respir Dis 1994; 150: 565–9

9 Higuchi JG, Coalson JJ, Johanson WG. Bacteriologic diagnosis of nosocomial pneumonia in primates. Am Rev Respir Dis 1982; 125: 53–7[ISI][Medline]

10 Meduri GU, Beals DH, Maijub AG, Baselski V. Protected bronchoalveolar lavage: a new bronchoscopic technique to retrieve uncontaminated distal airway secretions. Am Rev Respir Dis 1991; 143: 855–64[ISI][Medline]

11 Teague RB, Wallace RG, Awe RJ. The use of quantitative sterile brush culture and Gram stain analysis in the diagnosis of lower respiratory tract infection. Chest 1982; 81: 556–62[Abstract]

12 Pollock HM, Hawkins EL, Bonner JR, Sparkman T, Bass JB. Diagnosis of bacterial pulmonary infections with quantitative protected catheter cultures obtained during bronchoscopy. J Clin Microbiol 1983; 17: 255–9[ISI][Medline]

13 Bavoux E, Thaller F, Baillet A, Bure A, Loirat P. Bacterial pneumonia in intensive care units: evaluation of direct Gram stain examination of protected brush compared to quantitative culture [abstract]. Program and Abstracts of the 32nd Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), Anaheim, 1992. A1122.

14 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: 105–13[Abstract/Free Full Text]

15 Allaouchiche B, Jaumain H, Chassard D, Boulétreau P. Gram stain of bronchoalveolar lavage fluid in the early diagnosis of ventilator-associated pneumonia. Br J Anaesth 1999; 83: 845–9[Abstract/Free Full Text]

16 Namias N, Harvill S, Ball S, et al. A reappraisal of the role of Gram’s stains of tracheal aspirates in guiding antibiotic selection in the surgical intensive care unit. J Trauma 1998; 44: 102–5[ISI][Medline]

17 Corley DE, Kirtland SH, Winterbauer RH, et al. Reproducibility of the histologic diagnosis of pneumonia among a panel of four pathologists: analysis of a gold standard. Chest 1997; 112: 458–65[Abstract/Free Full Text]

18 SoleViolan J, deCastro FR, Rey A, et al. Comparison of bronchoscopic diagnostic techniques with histological findings in brain dead organ donors without suspected pneumonia. Thorax 1996; 51: 929–31

19 Kirtland SH, Corley DE, Winterbauer RH, et al. The diagnosis of ventilator-suspected pneumonia: a comparison of histologic, microbiologic, and clinical criteria. Chest 1997; 112: 445–57[Abstract/Free Full Text]





This Article
Abstract
Full Text (PDF)
E-Letters: Submit a response to the article
Alert me when this article is cited
Alert me when E-letters are posted
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Search for citing articles in:
ISI Web of Science (3)
Disclaimer
Request Permissions
Google Scholar
Articles by Mimoz, O.
Articles by Nordmann, P.
PubMed
PubMed Citation
Articles by Mimoz, O.
Articles by Nordmann, P.