Protected-Specimen Brush Technique in the Diagnosis of Ventilator-Associated Pneumonia

Protected-Specimen Brush Technique in the Diagnosis of Ventilator-Associated Pneumonia* Robert P. Baughman, MD, FCCP

(CHEST 2000; 117:203S–206S) Abbreviations: LR ⫽ likelihood ratio; PSB ⫽ protected-specimen brush; VAP ⫽ ventilator-associated pneumonia

brush (PSB) technique has T hebeenprotected-specimen used for almost 20 years in diagnosing pneumo-

nia,1 and the technique has been refined for VAP. Of the many studies, those included in our analysis met certain criteria. The 42 studies reviewed were identified through a MEDLINE search and presented primary data of bronchoscopy samples from patients receiving mechanical ventilation. All studies had rigorous clinical criteria for the diagnosis of pneumonia, including response to antibiotics, that were independent of the results of PSB or BAL procedures. The review methodology for this section is described in Dr. Steven H. Woolf’s section on the search of the literature. Independent nonbronchoscopic diagnostic criteria were required to ensure that there were no derivative data.108 Five studies compared PSB sampling to the use of specimens obtained immediately after autopsy as the final criteria for diagnosis. The remaining 13 studies used clinical assessment, response to antibiotics, and histologic data, including eventual autopsy information, when available. Of the 846 patients evaluated, pneumonia was diagnosed in 287. The quality of the PSB technique used in making the diagnosis was not described in detail for most of the studies in Tables 14 and 15. A few studies have examined assessing the quality of bronchial brushings. The issue of reproducibility was addressed by Marquette et al,27 who obtained five specimens from one area of the lung. In 25% of the patients, the variability in one or more of the bronchial specimens was ⬎ 103 cfu/mL, and in the remainder it was ⬍ 103 cfu/mL, which was the cutoff point. This suggests that 25% of the time a single PSB determination would have the possibility of being a false positive or a false negative. Another study, in patients who had localized pneumonia and were not receiving antibiotics, compared brush samples from the affected lobe and an unaffected lobe. The bacterial concentration in specimens from the affected area was at least 100-fold higher than in the unaffected lobe.26 Assessing the quality of the PSB sample involves examining a cell sample and looking directly at a brush sample. The problem is that two brush samples have to be obtained, since the slide could contaminate the culture results. Using the PSB technique allows for assessment of the quality of the sample, because the presence of squamous epithelial cells indicates that the sample is unaccept*From the University of Cincinnati Medical Center, Cincinnati, OH.

able. In one study, this was looked for and was found in some cases, but the number of unacceptable specimens was not given.109 In two studies, Gram’s stain was found to be sensitive and fairly specific in the identification of bacteria. It provides information within the first 24 h of the procedure, which is useful.109,110 As noted above, 5 of the 18 studies used autopsy data from VAP patients as a “gold standard.” Obviously, this does not provide information about patients who did not die. The other studies did not have histologic confirmation of the final diagnosis in all patients. None of the studies included a PSB procedure on every patient in the ICU. Entry criteria for all patients required either a clinical suspicion of pneumonia or availability of autopsy studies. One study18 had autopsy confirmation of the diagnosis, but the bronchoscopy had been performed up to 3 days before death. Among other factors contributing to variability are the skill of the health-care worker performing the procedure. It is not clear how much skill is required, but suctioning through the catheter before brushing may interfere with the results.111 Few studies specify that the secretions were cleared out with a separate bronchoscope before the PSB sample was taken.26 Also, where and what to brush is poorly defined.85 Other problems include the selection of patients to undergo the procedure and the lack of standard protocols for therapy, such as the selection of prior antibiotics and the use of empirical therapy pending the results of the cultures. Because clinical outcome is often used in assessing the PSB technique, the use of antibiotics pending the culture result can influence the study outcome.

Qualifications for Inclusion in the Tables Studies The 18 studies covered were published from 1984 to 1995 and used semiquantitative culture results for the PSB sample. All studies used the technique described for patients not receiving ventilator support.1 A standard method of bronchoscopy and handling of specimens was published in 1992,112–114 but was not significantly different from the methodology used in the studies.

How Were Patients Enrolled? All studies apparently enrolled patients prospectively, although two studies24,25 do not make this clear. Although the studies were prospective, none included all ICU patients. Most studies also were limited by the ability to perform the procedure. Some studies stated that patients were enrolled only during daytime.99 Most others probably followed this practice.

Description of the Population The studies can be divided into two groups. The first includes patients who died while receiving mechanicalventilation. Bronchoscopy and autopsy were performed on these patients in the immediate postmortem period.20,41,99 There were no other selection criteria, so pneumonia was not necessarily suspected prior to autopsy. One study CHEST / 117 / 4 / APRIL, 2000 SUPPLEMENT

203S

Table 14 —Studies of Protected Specimen Brush* Investigators Blinded to Reference Standard

Adequate Methods of PSB

Criteria for Quality

Suspected VAP

No

Yes

Not described

Suspected VAP

Appears so

Yes

Not described

Suspected, 78; not suspected, 28

Yes

Yes

Not described

Postmortem

Yes

Yes

Not described

Postmortem

Yes

Yes

Not described

Postmortem on suspected VAP

Yes

Yes

Not described

Postmortem Postmortem

Yes Yes

Yes Yes

Not described Not described

Suspected VAP Suspected VAP

No No

Yes Yes

Not described Not described

Suspected VAP

No

Yes

Not described

Suspected VAP

Yes

Yes

Vent ⬎ 48 h, no change in ATB 15 mo study, vent ⬎ 72 h

Suspected VAP

No

Yes

Compared 2 brushes, 17–24% discordance Not described

Suspected VAP

No

Yes

Reason for Enrollment

Study/Year

Enrollment

Description of Population

Fagon et al /1988

Prospectively

Marquette et al35/1993

Prospectively

El-Ebiary et al13/1993

Not stated

Chastre et al29/1984

Prospectively

Papazian et al20/1995

Prospectively

Marquette et al18/1995

Prospectively

Torres et al41/1994 Chastre et al99/1995

Prospectively Prospectively

Lambert et al16/1989 Torres et al45/1988

Prospectively Unclear

Meduri et al100/1992

Prospectively

Timsit et al115/1993

Prospectively

All patients in surgical ICU who developed clinical criteria, were on vent ⬎ 72 h 3-mo period, 3 ICUs, vent time unknown Patients in respiratory ICU, on vent ⬎ 72 h, who either had suspicion for pneumonia (74) or bronchoscopy (28) Patients on vent ⬎ 1 h, not prolonged, brain-dead Vent ⬎ 72 h, nonimmunocompromised Vent ⬎ 1 h, suspected of pneumonia, had bronch, died within 3 d Vent ⬎ 72 h, died Patients died in daytime, Vent ⬎ 72 h, no change in ATBS for 72 h Vent ⬎ 72 h, ATBS ⬍ 24 h Vent ⬎ 24 h, nosocomial (17) and CAP (8) One yr study, vent ⬎ 24 h, no ATBs for 48 h Vent ⬎ 24 h, could tolerate bronch

Timsit et al28/1995

Prospectively

Sole´-Viola´n et al103/1994

Prospectively

Richard et al117/1998 Baughman et al26/1987

Prospectively Prospectively

10 mo, vent unknown time Vent ⬎ 48 h, not ATBS 48 h, localized infiltrate

Suspected VAP Suspected VAP

No No

Yes Yes

Sole´-Viola´n et al116/1993 Chastre et al24/1988

Prospectively Unclear

Vent ⬎ 48 h, 14 mo study Vent ⬎ 48 h, no antibiotics ⬎ 10 d

Suspected VAP Suspected VAP

No No

Yes Yes

40

Compared 2 brushes Not described Compared PSB in affected and nonaffected Not described Not described

*Bronch ⫽ bronchoscopy; vent ⫽ ventilation; ATB ⫽ antibiotic; CAP ⫽ community-acquired pneumonia. Values in parentheses are No. of patients.

compared the results in ventilator-assisted patients with and without clinically suspected pneumonia.13 The rest of the studies report results in patients studied because of the clinical suspicion of pneumonia. One study analyzed findings in patients who had bronchoscopy for possible pneumonia then underwent autopsy within 72 h.18

Were the Test Results and the Reference Standard Assessed by Investigators Unaware of the Results of the Other Investigators? In all patients studied during life, the results from the PSB procedure affected long-term treatment.13,16,18,21,26,40,45,85,103, 204S

Clinical response was one of the criteria for defining pneumonia in all studies except for the one that used autopsy.18 The four studies in which postmortem bronchoscopy and autopsy were performed20,29,41,99 did not have that bias. The fact that the results from the autopsy studies were similar to those from the clinical studies supports the results of the latter. 115–117

Were Methods of Performing the Tests Described Adequately? Overall, the authors provided sufficient details on methodology. As noted, the PSB technique has remained virtually unchanged for years. In the one study using two

Evidence-Based Assessment of Diagnostic Tests for Ventilator-Associated Pneumonia

Table 15—Studies of Protected Specimen Brush* Study Design Patient no.

Episodes

Def/Prob/ Pneu

147

147

34

PSB results compare to clinical

Marquette et al95/1993

52

52

22

Compare EA to PSB vs EA vs clin PSB to clinical

El-Ebiary et al13/1993

102

102

26

Compare clin vs EA, PSB, BAL PSB, BAL, EA vs clin

Chastre et al29/1984 Papazian et al20/1995

26 38

26 38

6 18

AUT vs PSB AUT vs clin

Marquette et al18/1995 Torres et al41/1994

28 30

28 30

9 18

Chastre et al99/1995

20

20

14

AUT vs PSB AUT vs PBS, BAL, EA, blinded BAL AUT vs PSB/BAL EA vs PSB vs AUT AUT vs PSB/BAL/ AUT vs PSB/EA/ needle FNA AUT vs PSB/BAL AUT vs PSB/BAL

Lambert et al16/1989

22

22

18

Clin/AUT vs BRO Clin vs EA/PSB

Torres et al45/1988

25

25

18

Meduri et al100/1992

25

25

9

Timsit et al115/1993

26

42

Timsit et al28/1995

161

Sole´-Viola´n et al103/1994

Study/Year Fagon et al40/1988

Overall Design

Intervention Comparisons PSB vs clin

Prior Antibiotics

Reference Standard

Clin ⫽ 10 Auto ⫽ 24 False positive ⫽ 4 37 (19 ⬍ 1 d) Clin No ⫽ 23 Clin Yes ⫽ 22 Uncertain ⫽ 7 100% Clin: PNE ⫽ 26 ⫹ 10 Pos AUT Uncertain ⫽ 48 Clin Neg ⫽ 28 14/26 AUT findings 21 within 3 d, AUT 10 with none Hist ⫹/⫺ Culture 13/28 Hist All 3 with new AUT Hist ATB 18/20, but no AUT changes for 3d 10 Clin 18: Aut 4: 18/22 felt to have PNE 4/25 Clin with AUT in 13 patients 90/147

Sensitivity, Specificity, % % LR 100

95

20

64

96

16

60

93

8.58

100 33

60 95

2.5 6.6

57 36

88 50

4.75 0.72

82

77

3.56

73

100

73

Clin PNE vs BRO Metras, PSB vs Clin vs blind catheter Clin vs Clin vs PSB vs None for 48 h PSB/protected PBAL BAL

66

100

66

33

100

33

15

Clin vs two PSB

67

94

11.16

161

13

Effect of prior ATB on BAL/ PSB results

74

88

6.16

33

33

16

PSB vs BAL

75

100

75

Richard et al117/1988

24

25

13

Clin vs PSB

77

100

77

Baughman et al26/1987

21

21

8

62

100

62

Sole´-Viola´n et al116/1993

45

45

25

64

100

64

Chastre et al24/1988

21

21

5

PSB two areas vs clin Clin vs PSB or BAL Clin vs PSB or BAL

9 PNE, 2 W hist 14 not PNE, 7 with hist 2 uncertain Clin vs two 8 in previous 15 certain PNE separate PSB 24 h, 17 no PNE 15 overall 10 uncertain Clin vs PSB/BAL 41%, but 13 definite divided type Probable 6 of ATB uncertain 7 excluded many PSB vs BAL vs clin 6/33 No info on diagnosis PNE Clin vs PSB/ 12 Clin at end of nonprotected hospital, blind to brush culture results PSB vs PSB vs Clin None for 48 h Clin with AUT PSB and BAL vs Clin PSB or BAL vs Clin

*AUT ⫽ autopsy; EA ⫽ endotracheal aspirate; BRO ⫽ bronchoscopy; PNE ⫽ pneumonia. See Table 14 for other abbreviations.

bronchoscopes (the first used to suction the airways, and the second used without suction) on all patients,85 the results were not significantly different from those of other studies. One study found that it did not matter which segment of the lung was brushed.118

ATB ⫽ antibiotics;

22 (49%)

Clin with AUT

None for 10 d

Clin with AUT (3 equivocal cases)

Metras ⫽ Metras

catheter;

100

hist ⫽ histology;

100 100 clin ⫽ clinical;

Criteria Used to Assess the Quality of the Sample One article discussed the examination of the bush sample for the presence of cells.109 Technically, this may not be feasible, because dilution of the sample relies on CHEST / 117 / 4 / APRIL, 2000 SUPPLEMENT

205S

the vigorous agitation of the brush in a solution immediately after sampling.1,85 The Gram’s stain may provide an early guide to therapy.25,109

specificity of 95%. The PSB may be more specific than sensitive.

Study Design

Likelihood Ratio

The most common study design selected patients suspected of having VAP who could undergo bronchoscopy. One study was limited to patients with localized infiltrates, so that an unaffected lobe could be analyzed.26 One study included patients who had no clinical suspicion of pneumonia.13 One postmortem study examined bronchoscopy results obtained within 3 days before death.18 The remaining postmortem studies compared only the specimens obtained after death.

Table 14 also includes the calculated likelihood ratios (LRs), based on sensitivity and specificity. Since several studies reported the PSB procedure to be highly specific, the likelihood that a positive finding would mean pneumonia was quite high. The median LR was 16. Only one study reported an LR of ⬍ 1,41 and several studies reported ratios of ⬎ 50 based on a 100% specificity.16,24,26,100,103,116,117 In the one study reporting an LR of ⬍ 1, all patients were receiving antibiotics and only three patients had recently changed antibiotics.41 Since histology studies were the criteria for diagnosis, it is not clear whether a negative result for a PSB culture and a positive result of a histologic examination may indicate a treated, but not resolved, case of pneumonia.

Antibiotics Antibiotic therapy is a significant variable in the assessment of a diagnostic technique. In three studies, patients were not receiving any antibiotics.24,26,100 Others studies attempted to study only patients who had not undergone a recent change in antibiotic therapy.41,99 In the rest of the studies, patients took antibiotics for varying periods. Since most patients with VAP are already taking antibiotics at the time of bronchoscopy, this remains an unresolved dilemma.

Reference Standard Table 14 discusses the reference standards for each study. The postmortem studies used histology reports and culture results.18,20,29,41,99 In the clinical studies, the most common criteria were response to therapy or the identification of an alternative diagnosis. In many studies, some patients died and underwent postmortem examination. The autopsy results did not change the clinical impression to any degree.

Sensitivity The sensitivity of the studies varied from about 33 to 36%20,41 to ⬎ 95%.29,40 This wide range cannot be explained by diagnostic criteria alone. The lowest20 and highest29 sensitivities were in autopsy studies. As noted above, the recommended cutoff of 103 cfu/mL85 is arbitrary. In one series looking at the reproducibility of results for the brush technique, 25% of the results were discordant.27 The median sensitivity of brush sampling for all studies was 67%.

Specificity The calculated specificity, shown in Table 14, ranged from 50%41 to 100%,16,24,26,45,100,103,116,117 with a median

206S

Comparison to Other Techniques Fourteen studies were direct comparisons between PSB and other invasive diagnostic techniques, usually BAL.13,18,20,24,28,41,93,99,100,103,116 None of the studies comparing the PSB and BAL techniques convincingly showed a benefit for one over the other, although BAL was generally more sensitive and PSB more specific. It is not clear that these tests are complimentary. The other studies compared PSB with relatively less invasive techniques, such as aspiration samples,13,20,118 blind BAL,20,45 and nonprotected brush.117 The comparisons showed little difference between the techniques.

Risks The reports are strangely silent regarding complications from the PSB procedure. Bronchoscopy in the patient receiving mechanical ventilation is associated with significant risks, but it is not clear that the brush technique adds to the risk.

Conclusion Despite the wide array of studies and the use of high-quality diagnostic techniques in several studies to determine the cause of pneumonia, sensitivity ranged from 33 to 100% and specificity ranged from 50 to 100%. Overall, PSB appears more specific than sensitive in diagnosing VAP. In a patient with suspected VAP and a positive result of testing a PSB sample, the LR of VAP appears to be much ⬎ 1.

Evidence-Based Assessment of Diagnostic Tests for Ventilator-Associated Pneumonia