Bronchoscopic BAL in the Diagnosis of Ventilator-Associated Pneumonia

Bronchoscopic BAL in the Diagnosis of VentilatorAssociated Pneumonia* Antoni Torres, MD, FCCP; and Mustafa El-Ebiary, MD

(CHEST 2000; 117:198S–202S)

and histologic parameters. Immediate postmortem studies have used histology studies, lung cultures, or both. These studies are of particular value. The calculated sensitivity and specificity vary among studies, depending mainly on prior antibiotic treatment, type of study population, and the reference test employed. Some studies do not clearly state how sensitivity and specificity were calculated. Few studies reported receiver operating characteristic curves.

Abbreviations: ICO ⫽ intracellular organism; PSB ⫽ protectedspecimen brush; VAP ⫽ ventilator-associated pneumonia

Studies

B

ronchoscopic BAL has been used since 1988 in the diagnosis of ventilator-associated pneumonia (VAP). Quantitative cultures retrieved by protected and nonprotected bronchoscopic BAL have been reported to be accurate for the diagnosis of this frequent complication of mechanical ventilation. This section covers BAL methods that utilize fiberoptic bronchoscopy. We reviewed 23 studies evaluating the accuracy of BAL methods using fiberoptic bronchoscopy to diagnose VAP in a total of 957 patients. The number of reported VAP episodes was 431. Most reported studies are case series, although five immediate postmortem studies have been described since 1995. Patient enrollment was prospective, but consecutive enrollment was used in only four studies. The duration of mechanical ventilation is described in most studies as lasting from 1 to 90 days. In most cases, except in the postmortem studies, investigators assessing the test results and reference tests were aware of the patients’ clinical histories. The techniques and bacteriologic methods used are well described, but in some studies the microbiological or BAL methods are referenced to prior publications. BAL technique is not completely standardized. The amount of liquid instilled varies roughly from 130 to 150 mL. The first aliquot usually is discarded, and the rest are pooled for quantitative bacterial analysis. Serial dilutions and calibrated loop methods have been used for quantitative cultures. Four studies have used protected bronchoscopic methods with balloon occlusion, and one study employed the mini-BAL technique via fiberoptic bronchoscopy. Only two studies assessed the quality of BAL specimens, and they did so by the detection of squamous epithelial cells. No study assessed the presence of ciliated bronchial cells to indicate contamination by upper airway secretions. In most studies, BAL was compared with either bronchoscopic procedures (usually the protected-specimen brush [PSB] technique) or nonbronchoscopic methods (usually endotracheal aspiration and other blinded techniques). Most studies also used quantitative bacterial cultures to distinguish colonization from infection. For that purpose, the most frequent cutoff point was 104 cfu/mL. Information regarding prior antibiotic administration is lacking from some studies, but 457 patients are known to have received such treatment. The reference test “gold standard” is not standardized and varies among studies. Most studies combine clinical, microbiological, *From the Servei de Pneumologia, Hospital Clı´nic, Barcelona, Spain. 198S

Qualifications We reviewed 23 studies published since 1988 regarding the use of bronchoscopic BAL for the diagnosis of VAP. All studies employed quantitative cultures on the BAL samples, and four studies used protected BAL methods with balloon occlusion.23,41,97,98 The remaining studies used the conventional BAL method. In five studies,18,19,41,97,99 the procedure was performed immediately after death in patients receiving mechanical ventilation.

How Were Patients Enrolled? The study population was recruited prospectively in all 23 studies. In 4 studies, enrollment was consecutive; in the remaining 19, patients were selected by the investigators. Some data in these studies were collected retrospectively.

Description of the Population The main descriptive parameter of the study populations was the duration of mechanical ventilation, which varied from 24 h93,100 to 90 days.101 The average period was 10 days. One study102 did not mention the duration. Another parameter was classification of the diagnosis as suspected, unsuspected, or postmortem. In 16 studies, the pneumonia episode was suspected on the patient’s inclusion into the study. In two of the five postmortem studies, pneumonia was not suspected before the patient’s death. In one study, the study population included suspected and nonsuspected pneumonia cases (Table 12).

Were the Test Results and the Reference Standard Assessed by Investigators Who Were Unaware of the Results of the Other Investigators? The reference test used in most of the studies was not standardized, varying markedly among studies. It appears from the reports that in seven studies (29%) the test results and the reference standard were assessed blindly by investigators who were unaware of the results of the other investigations.15,18,20,41,97,102 In the remaining 13 studies, it appears that the investigators may have been aware of test results, reference standards, and the results of other investigations. In three studies, this could not be deduced.24,100,101

Were the Methods of Performing the Tests Described Adequately? Overall, the test methods were well described. Usually, they were of two types: bronchoscopic methods and

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

Table 12—Methodological Quality* Study/Year Chastre et al24/1988 Torres et al15/1989 Guerra et al102/1990 Pugin et al30/1991 Meduri et al101/1991 Meduri et al85/1992 Middleton et al93/1992 Torres et al23/1993 El-Ebiary et al13/1993 Sauaia et al14/1993 Torres et al41/1994 Sole´-Viola´n et al103/1994 Aubas et al104/1994 Rodriguez de Castro et al42/1994 Valle´s et al22/1994 Marquette et al18/1995 Timsit et al21/1995 Chastre et al99/1995 Papazian et al20/1995 Torres et al97/1996 Barreiro et al98/1996 Sole´-Viola´n et al19/1996 Luna et al33/1997

Enrollment Selected Selected Selected Selected Selected Selected Selected Consecutive Selected Selected Selected Selected Selected Selected Consecutive Selected Consecutive Selected Consecutive Selected Selected Selected Selected

MV Duration, d 8 ⫾ 6 (S) 7 ⫾ 5 (S) Unknown (S) 11 ⫾ 7 (S) ⫾ ⱖ 72 (S) 1–90† (S) 1–30‡ (S) 7 ⫾ 6 (NS) 7 ⫾ 7 (S, NS) ⫾ ⱖ 24 h (S) 9 ⫾ 7 (PM) 18 ⫾ 20 (S) 8 ⫾ 9 (S) 9 ⫾ 11 (NS) 9 ⫾ 7 (S) 11 ⫾ 8 (PM) 13 ⫾ 10 (S) 16 ⫾ 9 (PM) 23 ⫾ 27 (PM) 13 ⫾ 13 (PM) 11 ⫾ 10 (S) 4 ⫾ 3 (S) 8 ⫾ 14 (S)

Awareness of Results of Other Studies Unknown Yes Yes No Unknown Unknown No No No No Yes No No No No Yes No Yes Yes Yes No No No

Methodology Description Adequate Adequate Inadequate Adequate Adequate Inadequate Inadequate Adequate Adequate Adequate Adequate Inadequate Adequate Adequate Adequate Adequate Adequate Adequate Adequate Adequate Adequate Inadequate Adequate

microbiology

microbiology BAL

microbiology

BAL

*Values given as mean ⫾ SD. S ⫽ suspected VAP; NS ⫽ unsuspected VAP; PM ⫽ postmortem VAP. †Mean, 12. ‡Mean, 8.

microbiological methods. Five studies did not adequately describe the methodology;19,93,100 –103 three of those referenced microbiological methods to prior publications, and two referenced the BAL method to prior publications (Table 12).

Criteria Used to Assess the Quality of the Sample The best way to assess the quality of a BAL sample is to examine the cells of the fluid retrieved from preparations that had undergone centrifugation. The presence of ⬎ 1% squamous epithelial cells is evidence of oropharyngeal contamination. Ten studies did not mention the use of any criteria to assess sample quality. The remaining studies used cellular assessment, but only two looked for squamous epithelial cells.18,19 To our knowledge, none of the studies looked for ciliated bronchial cells as a marker of contamination.

Number of Patients and Episodes A total of 957 patients receiving mechanical ventilation were studied in the 23 series reported. The most patients in a single study was 102,13,98 and the fewest was 9.19 As shown in parentheses in Table 13, 431 episodes of VAP were studied. The criterion used for determining the number of episodes was the confirmation of VAP by clinical, microbiological, or histologic parameters. These parameters represent the reference standards. Many studies included control patients (ie, patients in whom pneumonia was not suspected). The selection of control patients was based on the absence of pulmonary infection or on the presence of a confirmed alternative diagnosis. Two studies prospectively included only control subjects.23,42 Some studies included patients who were suspected of having pneumonia but lacked definite criteria. These patients were not included in the 431 episodes we computed.

Intracellular Organism Detection

Study Design

Twelve studies looked for intracellular organisms (ICOs) as a possible marker for VAP. Ten of these reported ICO thresholds ranging from 2 to 25% to distinguish colonization from true infection. Of the 12 studies, eleven (92%) mentioned sensitivities and specificities. Sensitivities ranged from 37%18 to 100%,24 and specificities ranged from 89%99 to 100%.18,24,30,97,101,103 These studies suggest that ICO detection is a very specific marker for VAP.

The most common design was a case series of patients with suspected VAP, selected by the investigators. Some studies evaluated control patients, as defined above. Two studies investigated only control subjects.23,42 Five studies were performed immediately after death and included data from histologic examination or microbiological studies of lung tissue. The diagnostic values of BAL and other sampling methods were compared in histologic and nonhistologic studies. Comparisons included endotracheal CHEST / 117 / 4 / APRIL, 2000 SUPPLEMENT

199S

Table 13—No. Results*

Study/Year

No. of Patients (No. of VAP Episodes)

Study Design

Antibiotics†

Chastre et al24/1988

21 (8)

Case series, Only BAL

Torres et al15/1989 Guerra et al102/1990

41 (25) 60 (30)

Control subjects, EA vs BAL Case series‡ BAL only

Pugin et al30/1991

28 (40)

Unknown

Meduri et al101/1991

25 (9)

Control subjects, BAL vs blind BAL PSB vs PBAL

Meduri et al100/1992 Middleton et al93/1992

25 (9) 12 (12)

9/14 6/6

Torres et al23/1993

27 (0)

PSV vs PBAL Case series, EA, ACCU, cath, PSB, BAL Control subjects only, EA, PSB, BAL EA, PSB, BAL

El-Ebiary et al13/1993

102 (26)

0/21 25/16 39/21

0/25

30 (18)

30/0

Sole´-Viola´n et al103/ 1994 Aubas et al104/1994

33 (16)

EA, PSB, BAL, PLNA, postmortem biopsy Case series, PSB vs BAL

Unknown

72 (28)

Case series, only BAL

Unknown

Rodriguez de Castro et al42/1994 Valle´s et al22/1994

22 (0)

Control patients, PSB vs BAL

0 /22

41 (18)

Case series, PSB vs BAL

20 /22

Marquette et al18/1995

28 (19)

13 /15

Timsit et al21/1995

45 (122)

Chastre et al99/1995

20 (NR)

Papazian et al20/1995

38 (18)

Torres et al97/1996

25 (23)

Postmortem study, EA, PSB, BAL Random sample, PSB, BAL, PTC Postmortem study, PSB vs BAL Postmortem study, EA, PSB, BAL, mini-BAL Postmortem study, BAL vs PBAL Case series, PSB vs PBAL

Luna et al33/1997

132 (65)

2.2 n˜

64 (72)㛳 53 (53)

71 (71)㛳 100 (100)

None None

⬁

93 (93)¶

CC (100)¶

(73/100/⬁)

92 (97)

CC (92)

2% (85/100/⬁)

90 (100) 50 (42)

100 (100) CC (NR)

NR None

CC

NR

65 (65)

None

4.3

57 (57)

87 (87)

None

⬁

67 (67)

100 (100)

None

0.9

50 (50)

45 (45)

None

Histology/clinical/ ⬁ microbiology Clinical/ 4.9 microbiology Clinical CC

87 (87)

100 (100)

ⱖ 4% (62/100/⬁)

89 (89)#

CC (82)#

5% (39/98/19.5)

Histology/ microbiology Clinical/ microbiology Histology

Torres et al41/1994

Postmortem, organ donors, PSB vs BAL Case series, BAL

69 (69)

54/48 16/2

9 (7)

80 (NR)

Histology/clinical

EA vs BAL

Sole´-Viola´n et al19/1996

2.6

23/4

18 (6)

102 (35)

Sensitivity‡

Histology/clinical/ microbiology Histology/clinical Histology/clinical/ microbiology Clinical score

56 /66

Histology/clinical/ microbiology Histology

100 (100)

ⱖ 5% (67/96/16.7)

⬁

CC (47)

CC (100)

(37/100/⬁)

CC (77)

CC (77)

5% (59/98/2.95)

91 (91)

78 (78)

ⱖ 5% (91/89/8.2)

50 (58)

95 (95)

None

17 /8

Histology

79 /23

Histology/clinical/ microbiology Histology

107 /25

Microbiology

None

89 (89)

12 /26

0 /9

82 (82)**

25% (100/100/⬁)

⬁

Histology/clinical/ 3.3 microbiology Lung 4.1 microbiology Histology 11.6

18 /2

NR

Specificity‡

ICO§ (sensitivity/ specificity/ Likelihood ratio)

LR

Histology/clinical/ 12 microbiology Histology/clinical ⬁ Clinical features CC

Sauaia et al14/1993

Barreiro et al98/1996

Reference Standard

⬁ 1 9.6

CC (22)†† CC (45)‡‡ 87 (87)

1.7

86 (86)

50 (50)

None

NR

49 (NR)

CC (NR)

None

CC (100)†† 2% (58/100/⬁) CC (55)‡‡ 2% (75/98/37.5) CC (91) None

*EA ⫽ endotracheal aspirates; PLNA ⫽ percutaneous lung aspiration; PBAL ⫽ protected BAL; PTC ⫽ plugged telescoping catheter; LR ⫽ likelihood ratio; ICO ⫽ intracellular organisms; CC ⫽ cannot calculate; BI ⫽ bacterial index; SBI ⫽ simplified bacterial index; SEC ⫽ squamous epithelial cells; BC ⫽ ciliated bronchial cells; NR ⫽ not reported. †Values given as No. of patients receiving/not receiving antibiotics. ‡Values given as calculated figures/figures estimated by authors. Threshold, 104 cfu/mL. §Values given as threshold used (sensitivity/specificity/LR). 㛳Threshold, 103 cfu/mL. ¶BI, 5. #SBI, 4. **Threshold, 105 cfu/mL. ††PBAL. ‡‡BAL.

aspiration, PSB, balloon-tipped BAL, and mini-BAL. Only 3 studies investigated conventional BAL alone, and the remaining 20 studies compared one or more of the techniques mentioned above. The most common technique was PSB, which was compared in 15 studies. 200S

Antibiotics Sixteen studies included patients with and without prior antibiotic treatment, four studies included only patients not receiving antibiotic therapy,19,24,42,101 and one study

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

included only patients receiving antibiotic treatment.41 In three studies, the presence or absence of antibiotic therapy is unknown.30,103,104 More patients (544) received antibiotics than did not (375). Most studies stated the presence of antibiotics as a dichotomous variable.

Reference Standard Several reference standards have been used to differentiate between the presence and absence of pneumonia. Seventeen studies used histology studies as the reference standard. The histologic criterion was the presence of neutrophilic infiltration in the alveoli. Five of the 17 studies used immediate postmortem histology.18 –20,97,99 Eleven studies combined histologic and clinical or microbiological data. Only one study used postmortem lung cultures as the reference test.99 Seven additional studies used clinical data with or without microbiological reference tests (Table 13). Clinical reference standards varied among studies and included the following: the presence of radiographic cavitation; classical signs of pulmonary infection (chest radiographic infiltrates, fever, leukocytosis, and purulent secretions); and good response to antibiotics. All reference standards included ruling out alternative diagnoses that could explain the presence of fever and pulmonary infiltrates. In one study,30 the authors developed a clinical pulmonary infection score that included the following six points: body temperature, leukocyte count, characteristics of tracheal secretions, oxygenation, chest radiographic findings, and a semiquantitative score for cultures of tracheal aspiration. The threshold to distinguish infection from noninfection was six points. Other studies used microbiological standards as reference tests. Most studies included the results of blood and pleural fluid cultures, the isolation of definite pathogens, and the isolation of pathogens above threshold levels from respiratory sample cultures.

specifically for sensitivity. The remaining articles do not describe how sensitivity was calculated in relation to this issue. Importantly, the calculation of sensitivity depends on the cutoff points of quantitative cultures. Most manuscripts used 104 cfu/mL, one study used 103 cfu/mL,15 and another study used 105 cfu/mL.42 One study30 calculated sensitivity by using the bacterial index instead of a definite threshold of quantitative cultures. The bacterial index is the sum of the logarithms base 10 (log10) of the different isolated microorganisms. The cutoff point used was 5. Another study104 used a similar index (simplified bacterial index) with a cutoff point of 4. Several articles13,18,20,21 calculated sensitivity using different cutoff points.

Specificity The situation regarding the specificity of BAL is similar to that of sensitivity. Three studies14,33,93 included only patients with VAP, so they did not report specificity values. The mean (⫾ SD) calculated for specificity in the 23 studies considered was 82 ⫾ 19%. The two studies that investigated only specificity for BAL reported values of 65% and 82%, respectively.23,42 In eight studies, specificity was reported by the authors but could not be calculated.18,21,30,93,97,98,101,104 In the remaining studies, the calculated and reported specificities were the same. Variability in specificity and in sensitivity is explained by differences in prior antibiotic treatment, the type of study population, and the manner in which pneumonia was confirmed (the reference test). The way specificity was calculated was not standardized among the studies, although formulas apparently were applied uniformly. Specificity can be calculated on the basis of the number of patients or the number of microorganisms, and we found only one article that specifically addressed this concern.23 Several articles13,18,20,21,23 calculate sensitivity using different cutoff points. Receiver operating characteristic curves were reported in only four studies.13,18,21,104

Risks Sensitivity Sensitivity values are reported in all but two studies,23,42 which included control patients (ie, patients without pneumonia). The BAL method to detect VAP had a variable sensitivity (22% in the study by Torres et al97 to 100% in the study by Meduri et al100). The calculated mean (⫾ SD) sensitivity in the 23 studies was 73 ⫾ 18%. The variability depends on prior antibiotic treatment, type of study population, and the reference test used. In six studies, the calculated sensitivity differed from that reported by the authors.15,20,93,100 –102 In three studies, sensitivity could not be calculated.18,21,97 In some studies, sensitivity was calculated for patients with and without prior antibiotic treatment. Obviously, such treatment decreased sensitivity. The method of calculating sensitivity was not standardized among studies, although formulas were applied uniformly. Sensitivity can be calculated by taking into account the number of patients or the number of microorganisms, and only one study23 made this distinction in computing the diagnostic yield parameters, although it did not do so

BAL is not without risk. The risks are of the following two types: those inherent to the use of the fiberoptic bronchoscope (which this report does not address); and those inherent to the instillation of fluid during bronchoscopy. The most important and most common consequences involve blood-gas exchange and include important decreases in oxygenation and slight increases in CO2 values. In addition, hemodynamic parameters can be altered after the BAL procedure. Alterations in blood-gas exchange depend on the type of lavage used (conventional BAL or protected BAL) and the amount of liquid instilled, which ranges from 50 mL in mini-BAL to 150 mL in conventional BAL. Recent articles have examined the effects of BAL on gas exchange. Steinberg and coworkers105 did not find significant changes in oxygenation, mean arterial pressure, heart rate, peak inspiratory pressure, or static thoracic compliance after BAL in 110 patients with ARDS. Papazian and colleagues20 found a significant decrease in Pao2 after BAL and a moderate drop in Paco2. Guerra and BaughCHEST / 117 / 4 / APRIL, 2000 SUPPLEMENT

201S

man102 observed a median decrease in Pao2 of 8 mm Hg (range, 63 to 29 mm Hg), which could be treated by increasing the fraction of inspired oxygen. However, the authors concluded that BAL using fiberoptic bronchoscopy is well tolerated in critically ill patients who are receiving mechanical ventilation. Our group examined the effects on oxygenation of protected BAL and mini-BAL in patients with VAP and in control subjects. We found important reductions in the Pao2/fraction of inspired oxygen ratio 5 and 24 h after the procedure (a 20% drop from baseline to postbronchoscopy values). The decrease was independent of the type of BAL used. Paco2 values showed a minor, transient increase (average, 7 mm Hg).106 Montravers et al25 observed similar effects on arterial oxygenation 3 and 5 h after BAL. These authors did not find important alterations in hemodynamic parameters (arterial pressure, heart rate, and cardiac index) after BAL. From these studies, it seems that the major side effect of BAL is the postprocedural reduction in arterial oxygenation in VAP patients. The patient may recover from the oxygenation impairment after several hours or may not recover completely. The variables involved in this response need detailed study. They include type of lavage, amount of liquid instilled, prior alteration in blood-gas exchange, and severity of the VAP.

202S

One study described sepsis-like effects after BAL in VAP patients.107 This response is characterized by fever and by a decrease in mean arterial pressure and arterial oxygenation, and it seems related to the level of endotoxins in the BAL fluid. The reaction may be a bacterial migration-like effect from the alveoli to the systemic circulation during BAL. Further investigation is warranted.

Conclusions • The sensitivity of quantitative BAL fluid cultures ranges from 42 to 93%, with a mean of 73%. The clinical implication is that BAL cultures are not diagnostic for pneumonia in almost one fourth of cases. • The specificity of quantitative BAL fluid cultures ranges from 45 to 100%, with a mean of 82%. This means that the diagnosis is incorrect (ie, false-positive result) in about 20% of cases. • Detecting intracellular organisms in BAL cultures is a quick, specific test and has a high positive predictive value. • Only two studies assessed the quality of BAL samples as measured by the presence of squamous epithelial cells.

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