Impact of BAL Data on the Therapy and Outcome of Ventilator-Associated Pneumonia

Impact of BAL Data on the Therapy and Outcome of Ventilator-Associated Pneumonia* Carlos M. Luna, MD, FCCP; Patricia Vujacich, MD; Michael S. Niederman, MD, FCCP; Carlos Vay , PhD; Carlos Gherardi, MD; Josue Matera, PhD; and Enrique C. Jolly, MD

Study objective: To define the impact ofBAL data on the selection of antibiotics and the outcomes of patients with ventilator-associated pneumonia (VAP). Design: Prospective observation and bronchoscopy with BAL, performed within 24 h of establishing a clinical diagnosis of a new episode of hospital-acquired VAP or progression of a prior episode of nosocomial pneumonia (NP). Setting: A 15-bed medical and surgical ICU. Patients: One hundred thirty-two patients hospitalized for more than 72 h, who were mechanically ventilated and had a new or progressive lung infiltrate plus at least two of the following three clinical criteria for VAP: abnormal temperature (>38°C or <35°C), abnormal leukocyte count (> 10,000/mm3 or <3,000/mm 3 ), purulent bronchial secretions. Interventions: Bronchoscopy with BAL within 24 h of establishing a clinical diagnosis of VAP or progression of an infiltrate due to prior VAPor NP. All patients received antibiotics, 107 prior to bronchoscopy and 25 immediately after bronchoscopy. Results: Sixty-five of the 132 patients were BAL positive (BAL[ + ]), satisfying a microbiologic definition ofVAP ( > 104 cfu/mL), while 67 were BAL negative (BAL[- ]). The BAL( +)patients had no differences in mortality, prior antibiotic use, and demographic features when compared with the BAL(-) patients. More of the BAL( +) patients (38/65) satisfied all three clinical criteria of VAP than did BAL(-) patients (24/67) (p<0.05). A total of 50 BAL( +) patients received antibiotic therapy prior to bronchoscopy, and when this prior therapy was adequate (n=16), as defined by the results of BAL, then mortality was 38%, while if prior therapy was inadequate (n=34), mortality was 91% (p<0.001), and if no therapy was given (n= 15), mortality was 60%. When therapy changes were made after bronchoscopy, more patients (n=42) received adequate therapy, but mortality in this group was comparable to mortality among those who continued to receive inadequate therapy (n=23). A total of 46 of the 65 BAL(+) patients died, with 23 of these deaths occurring during the 48 h after the bronchoscopy, before BAL results were known. When BAL data became available, 37 of the 42 surviving patients received adequate therapy, but their mortality was comparable to the patients who continued to receive inadequate therapy. Conclusions: Patients with a strong clinical suspicion ofVAP have a high mortality rate, regardless of whether BAL cultures confirm the clinical diagnosis ofVAP. When adequate antibiotic therapy is initiated very early (ie, before performing bronchoscopy), mortality rate is reduced if this empiric therapy is adequate, compared to when this therapy is inadequate or no therapy is given. If adequate therapy is delayed until bronchoscopy is performed or until BAL results are known, mortality is higher than if it had been given at the time of first establishing a clinical diagnosis of VAP. When patients were changed from inadequate antibiotic therapy to adequate therapy, based on the results of BAL, mortality was comparable to those who continued to receive inadequate therapy. Thus, even if bronchoscopy can accurately define the microbial etiology of VAP, this information becomes available too late to influence survival. (CHEST 1997; 111:676-85) Key words: antibiotic th erapy; broncboalveolar lavage; diagnosis; nosocomial pneumonia; pneumonia; ventilatorassociated pneumonia Abbreviations: BAL(-)=BAL negative; BAL(+)=BAL positive; EPI=extrapulmonary infection; NP=nosocomial pneumonia; PSB=protected specimen brush; VAP=ventilator-associated pneumonia

*From the Department of Medicine, Pulmonary Diseases Division and Critical Care Division, and from the Clinical Analysis Laboratory, Microbiology Division, Hospital de Clfnicas "Jose de San Martin," University of Buenos Aires, Argentina (Drs. Luna, Vujacich, Vay, Gherardi, Matera, and Jolly), and the Division of

676

Pulmonary and Critical Care Medicine, Winthrop-University Hospital, Mineola, NY (Dr. Niederman ). Manuscript received January 2, 1996; revision accepted September 24. Reprint requests: Dr. Niederman, 222 Station Plaza N, Suite 400, Mineola, NY 11501

Clinical Investigations in Critical Care

H ospital-acquired

pneumonia complicates the course of mechanical ventilation in as many as 20% of all patients and serves as the hospitalacquired infection most likely to lead to the death of critically ill patients_! Current management strategies emphasize early diagnosis and therapy, with antibiotics focused on the most likely etiologic organisms, usually enteric Gram-negative bacteria and Staphylococcus au reus. Controversy remains about the most accurate way to diagnose this infection and the appropriate timing of antibiotic therapy in relation to the clinical suspicion of pneumonia. In some studies, antibiotic therapy itself has b een identified as a risk factor for both infection with highly virulent organisms and an increased mortality rate from YAP.2 •3 Concern that the prior use of antibiotics could enhance the mortality of YAP, coupled with an awareness that the clinical diagnosis of this infection is sensitive, but not specific, has led some investigators to suggest that antibiotic therapy decisions be guided by the results of bronchoscopically directed sampling of lower respiratory tract secretions .4 •5 Using this approach, antibiotics are administered only when the diagnosis of pneumonia is confirmed microbiologically by bronchoscopic methods.4 •5 An alternative approach is to use early and aggressive empiric antibiotic the rapy once aclinical diagnosis of YAP is made, not basing any decision making on the results of bronchoscopic sampling, because of a concern that the microbiologic diagnosis of YAP is not sensitive enough to recognize all infected patients.6 While controversy about the most accurate way to diagnose YAP continues, no investigator has shown that outcome, particularly mortality, in the setting of suspected YAP, is improved if antibiotics are prescribed based on a bronchoscopic diagnosis rather than empirically, once aclinical diagnosis of YAP has been made. However, some previous studies have shown that antibiotic therapy can have a favorable impact on the outcome of YAP, with lower mortality rates b eing observed if appropriate therapy, rather than inappropriate therapy, is prescribed. 7 ·8 Thus, it might be possible to evaluate the utility of bronchoscopy in the management of YAP from a therapeutic perspective. If bronchoscopy is valuable for the management of YAP, then the data obtained from quantitative bronchoscopic cultures could be used to select the most appropriate antibiotic and this should in turn improve survival in patients with suspected YAP. Alternatively, e ven if bronchoscopy can provide accurate microbiologic data, it may do so too late in the course of the illness to impact survival, because it may be necessary to choose adequate

empmc therapy from the start, even b efore the result of bronchoscopic cultures can be known. To address these questions, we performed bronchoscopy with quantitative BAL in all patients with a clinical suspicion of YAP, within 24 h of the clinical diagnosis of YAP or within 24 h of progression of a previously diagnosed episode of nosocomial pneumonia (NP) or YAP. We looked at the adequacy of antibiotic therapy, using bronchoscopic culture data to define the etiology of YAP, in all patients who satisfied both a clinical and a microbiologic definition of pneumonia (> 104 cfu/mL). We examined the adequacy of therapy at three time points : prior to BAL (empiric therapy), immediately after BAL (empiric therapy), and after the results of BAL were known. We examined whether the data obtained by BAL were used to modify therapy and whether such changes l ed to a favorable outcome. We confirmed the previous obse rvation that adequate antibiotic therapy can improve survival in patients with YAP, but only if given early in the course of illness, at a time when microbiologic data were not available. When BAL results were known and when they documented that the initial empiric therapy was inadequate, this information did not help patients, because the institution of adequate therapy at this time was associated with as high a mortality as when inadequate therapy was continued. Thus, even if BAL can provide accurate microbiologic information, our data suggest that it has no beneficial role in the selection of antibiotics, and that the outcome in YAP can be improved only if initial e mpiric therapy is accurate and timely.

MATERIALS AND METHODS

The study was conducted from April 1, 1992 to November 30, 1995 in the C1itical Care Unit of the "Clinicas Jose de San Martin" University Hospital, Buenos Aires, Argentina. This is a 15-bed m edical (noncoronary) and surgical unit at a 500-bed teaching hospital that serves as both a r eferral cente r and a primary care hospital. During the study period, 2,588 patients were admitted. Patients intubated and receiving mechanical ventilation who developed a n ewor progressive infiltrate on chest radiograph after b eing in hospital for more than 72 h were eligible for the study if th ey fulfilled at least two of the follo\ving criteria: (1) temperature >38°C or < 35°C; (2) leukocytes count > l0,000/mm 3 or < 3,000/mm 3 ; and (3) purulent bronchial secretions. P atients \\~th previous NP were included if they developed a newor progressive infiltrate along vvith a clinical deterioration, and if th ey fulfill ed the other clinical crite1ia of VAP. Severely neutropeni c .( < 500 neutrophils ) and AIDS patients were not eligible for the study. The follo~~ng information was recorded prospectively: age; sex; dates of admission and discharge from the ICU; prior trauma or surge1y; presence of COPD or other pulmonary disease; cardiac disease; heart failure; shock; diabetes; renal failure; non pharmacologic consciousness depression or coma; duration of tracheal intubation and mechanical ventilation p•ior to the CHEST I 111 I 3 I MARCH, 1997

677

development of VAP; use of corticosteroids; use of antacids, including H 2 blockers; and other underlying diseases. Shock was defined as a systolic BP ::;go mm Hg, requirement of vasopressors for more than 4 h, or urinary output less than 20 mUh for more than 4 hours; cardiac failure was defined by a pulmonary artery occlusion pressure 2o18 mm Hg or clinically (third heart sound gallop, neck vein distention and/or suggestive interstitial edema at the chest radiograph in patients with obvious severe cardiac disease) in the few patients in whom that pressure was not available; consciousness was evaluated by the Glasgow Coma Scale, and postoperative respiratory failure was diagnosed when patients were mechanically ventilated after surgery. All the patients who received antimicrobials for more than 24 h during the 10 days preceding the episode of VAP (mean, 119 h; range, 24 to 360 h) or those who were receiving antimicrobials for presumed VAP during the 24 h before the bronchoscopy were defined as receiving prior antibiotics. A total of 132 patients were studied, with 107 receiving prior antibiotics, while 25 were not receiving antibiotics prior to bronchoscopy. Prior antibiotics were given to 54 patients for NP, while in the other 53 patients, the indication was another condition (community-acquired pneumonia in 16 and extrapulmonary infection in 37) (Table 1). Mean duration of antibiotic therapy was 8.1 ::+::5.8 days, including prior use of antibiotics. At the onset of the study, 107 patients had bee n receiving prior antibiotics for a mean of 5.1::+::3.0 days. Mean duration of mechanical ventilation at the onset of the study was 7.8::+::14.4 days (5.8::+::5.7 days excluding three patients who were ventilated for more than 70 days) and from the time of bronchoscopy to the time of death in patients who died, 6.6::+::9.1 days. Fiberoptic bronchoscopy examination was performed in each patient within 24 h after the development of a new inHltrate or

Table !-Clinical Features of All Patients* Results of BAL Culture Clinical Characteristics

Positive (n=65)

60::+::21 Age, yr 41/24 Sex, M/F 38.5::+::1.1 Temperature, °C 15,006::+::7,202 Blood leukocytes/mm 3 188::+::91 PaO/Fio2 Bacteremia 18 27 Postoperative respiratmy failure Consciousness altered 37 29 Shock 27 Preexisting pulmonary disease 23 Preexisting cardiac disease Cardiac failure 8 14 Malignant neoplasm Mortality, % 7l 50 P1ior antibiotic 1 26 (6) For VAP, 1 54 (10) 8 (0) For CAP, 16 (1) 16 (1) For EPI, 37 (2) No. of criteria of pneumonia Only 2 27 38 All3

Negative (n=67) 60::+::19 40/27 38.6::+::1.0 14,148::+::5,390 202::+::99 5§ 29 38 25 32 23 6 15 64 57 28 (4) 8 (1) 21 (1) 43 11 2411

*CAP=community-acquired pneumonia; Fio 2 =fraction of inspired oxygen. 'Numbers in parentheses indicate S24 h. 1VAP or NP. §p=0.005. llp<0.05. 678

progression of a prior stable infiltrate. In the 25 patients not receiving prior antibiotics, therapy was started immediately after performing BAL. In 64 of the 107 patients receiving prior antibiotics, a change in therapy was done immediately after the BAL, while in the other 43, therapy either did not change at all or changed only after having the BAL culture results available. To perform the BAL, the bronchoscope was advanced and wedged into a distal airway in the area involved by the infiltrate, as seen on chest radiograph, or into a subsegment of the middle lobe or lingula if diffuse inflltrates were noted. After wedging, six 20-mL aliquots of sterile, nonbacteriostatic saline solution (0.9% NaCI) were instilled through the channel of the bronchoscope and subsequently aspirated by hand. The first aliquot instilled was discarded, the remaining aliquots were saved, with the total amount of saline solution instilled through the bronchoscope (after the first discarded aliquot) being 100 mL. Recovered BAL fluid was processed for semiquantitative culture for aerobic bacteria. Using a quantitative loop, 0.01 mL of the BAL fluid was plated onto each of four culture media (sheep blood agar, EMB Levine Agar, mannitol salt agar, and chocolate blood agar). After inoculation, each plate was incubated at 35°C in a C0 2 -enriched environment. Estimates of the number of bacteria originally in the fluid were made by colony counts and were expressed as colony-forming units per milliliter of fluid. Since other authors found that more than 104 cfulmL correlated with the final diagnosis of VAP with a sensitivity between 80% and 100% and a specificity between 69% and 100%, 9 - 11 we chose "more than 104 cfulmL" of at least one species as the cutoff point for bacteriologic confirmation of the diagnosis of pneumonia. In patients with BAL-confirmed pneumonia (n=65), antibiotic susceptibility was tested; if the organisms present at a concentration >than 104 cfulmL were demonstrated to be sensitive to the agent prescribed, the patient was defined as being treated with adequate antibiotics and if they were resistant, the patient was defined as being treated with inadequate antibiotics. Sensitivity patterns in the antibiogram, and not the response of the organisms to the therapy or the use of combination therapy instead of monotherapy, were used to define the adequacy of therapy. Antibiotic therapy was evaluated at three different time points: pre-BAL=before performing BAL; post-BAL=evaluating the empiric therapy received immediately after performing BAL; and postresult=evaluating the therapy choices that were guided by the antimicrobial sensitivity data. Each patient could be redefined at each time point as receiving adequate or inadequate antibiotics. The selection of antibiotic therapy was left to the discretion of the attending physician and the medical staff. Although the preferred approach of our group is to first perform the BAL and then start antibiotic therapy empirically, at the moment of performing BAL, 107 of 132 patients were receiving antibiotics or had received them for at least 24 h during the preceding 10 days. All of the patients defined as receiving adequate therapy at the pre-BAL time point were receiving this therapy at the time of performing the bronchoscopy except for two patients who had received antibiotics for 360 h and 120 h, and then the therapy was suspended for 24 h and 96 h, respectively, prior to performing bronchoscopy. Blood cultures were collected within 24 h of BAL in all patients. Bacteremia was present when two or more of each set of two or three blood cultures yielded a microorganism. In two cases, only one of hvo bottles was positive but the bacteria were also isolated from BAL and this was considered evidence of bacteremic VAP. All patients were monitored until their discharge from the hospital. Subsequent changes in their clinical course, chest radiograph, and modifications in antibiotic therapy were recorded in all the cases. Clinical Investigations in Critical Care

Statistical Analysis Data are expressed as means:!::SD. To assess if there was a difference between the clinical picture of the patients with positive and negative BAL specimens, clinical variables and mortality in both groups of patients were compared using Student's t test or Fisher's Exact Test. In addition, the same variables were compared in patients with BAL positive (BAL[ +]) pneumonia taking into account whether they were receiving prior antimicrobial therapy. Also, the frequency of finding different etiologic microorganisms was compared between patients receiving prior antibiotics and patients not receiving p1ior antibiotics and the relation between the etiologic agent and the a dequacy of the pre-BAL antibiotic therapy was related to mortality, using Fisher's Exact Test.

RESULTS

During the 44-month period of this study, 132 episodes of suspected bacterial lung infection were evaluated by fiberoptic bronchoscopy and BAL; however, the BAL technique was able to confirm infection microbiologically (> 104 cfu/mL) in 62 individuals for a total of 65 episodes ( BAL[+] VAP) , 49.2% of the clinical pneumonias (Table 1). Three patients experienced more than one episode and their histories and outcomes for each pneumonia were treated as separate cases, the second episode in those patients happening at least 9 days after the first one. In one of these cases, the patient died after the second episode of VAP, and thus the patient was considered as alive for the first episode and as a mortality for the second. All 132 episodes had a clinical suspicion of VAP and all patients received antibiotic therapy for a mean of 4.1 days after performing BAL. Table 2 shows which types of

Table 2-Antimicrobials Used at Pre-BAL and PostBAL Times*

Results of BAL Culture in the Patients Studied Less than half of the patients (n =62) met all three clinical criteria (of the "at least 2" required for inclusion) for the diagnosis of bacterial pneumonia (Table 1). Cultures of BAL were negative (<104 cfu/mL) in 67 of 132 patients (50.8%). Among these patients with negative cultures, there was no growth in 30 patients and growth was between 10 1 and 104 cfu in 37 others. Most of the BAL( +)patients met all three criteria of bacterial pneumonia (38/65), while only 24/67 BAL negative (BAL[-]) patients met these three criteria (p<0.05). Demographic characteristics and some laboratory and clinical data that are markers of severity of illness are shown in Table 1. Besides the number of clinical criteria present, only the presence of bacteremia was significantly more common in BAL( +) than in BAL(-) patients (p<0.01). Bacteremia was present in 18 patients with BAL( +) VAP; in 13 cases, the organism was also obtained in the BAL culture. Bacteremia was also present in five patients with negative BAL culture, but the organism was isolated from BAL fluid (although in a count <10 4 cfu) in only one case. Mortality of the 132 episodes of suspected pneumonia was 67%. For the BAL( +) patients, mortality was 71% and for the BAL(-) patients, mortality was 64% (p=NS). There were no differences in age, sex, leukocytosis, oxygenation, use of ptior antibiotics (for treating the VAP or for other reason), or clinical diagnoses at ICU admission, when comparing BAL( +) and BAL(-) patients. Results in Patients With BAL( + ) Pneumonia

Pre-BAL

Monotherapy Combination AGS APS B+BI CEP IMID IMP VAN Others

antibiotics were used and if monotherapy or combination therapy was chosen prior to performing BAL and after BAL, distinguishing if therapy was given for the VAP or for another reason.

Overall

VAP

Other

Post-BAL

30 77 29 34 7 38 23 21 32 22

9 45 17 18 3 8t 7 14 21 9

21 32 12 16 4 30 16 7

13 119 36 40 4 27 19 63 1 861 20

ll

13

*AGS=aminoglycosides; APS= antipseudomonal agents;

B +BI = f3-lactam+f3-lactamase inhibitor; CEP =cephalosporins; IMID=imidazole; IMP = imipenem; VAN =vancomycin; Others=macrolides, antifungal agents; trimethoprim-sulfa; penicillins; quinolones. 1 p<0.00l =pre-BAL use of CEP for VAP vs other reason. 1p< 0.00l = use of IMP at pre- BAL time point vs post-BAL. 1p< 0.001 =use of VAN at pre-BAL time point vs post-BAL.

In a total of 50 of 65 episodes of BAL( +) pneumonia, patients received antibiotic treatment before fiberoptic bronchoscopy was performed: 16 for an extrapulmonary infection, eight for a previous community-acquired pneumonia, and 26 for presumed VAP or NP. Of the 26 patients receiving antibiotics for VAP, six were treated for < 24 h prior to BAL, and 20 were treated for > 24 h prior to BAL, but all had radiographic progression documented within 24 h prior to BAL being performed. There was no significant difference in mortality between the BAL( +) VAP patients who were receiving prior antibiotics and those who were not receiving prior antibiotics (37/50=74% vs 9/15=60%) (Table 3). Similarly, there were no significant differences b etween those BAL( +) patients receiving prior antibiCHEST I 111 I 3 I MARCH, 1997

679

Table 3-Features of BAL-Positive Patients in Relation to Prior Antibiotic Therapy Prior Antibioti c Th erapy Without (n=l5)

Clinical Characteristics Age, yr Sex, M/F Te mperature, oc Blood leukocytes/mm·3 PaOzfFI0 2 * Bacteremia Postoperative respiratory failure Consciousness altered Shock Preexistin g pulmonary disease Preexistin g cardiac disease Cardi ac f ailure Malignant neoplasm M01tality,% No. of criteria of pn eumonia Only 2 Al\ 3

With (n =50)

53 :+: 26 11/4 38.6:+: 0.5 14,080:+: 5,822 203:+:95 4 4 7 5 6 7 0 1 60

62:+: 20 30/20 38.4:+: 1.6 15,284:+: 7,545 185:+:90 13 23 30 24 21 16 8 13 74 21 29

6 9

*Fio 2 = fraction of inspired oxygen.

otics and those who were not (Table 3), with respect to age, sex, leukocytosis, oxygenation, bacteremia, clinical findings suggestive of severity of illnesses, or number of diagnostic clinical criteria of pneumonia present.

!00%

Although the 50 patients who received prior antibiotics did not have asignificantly different mortality than the 15 who did not receive prior antibiotics, there were mortality differences among these patients depending on the adequacy of pre-BAL antibiotic choices (Fig 1). Patients who received adequate antibiotics early (prior to the BAL) had a significantly lower (p < 0.001 ) mmtality rate of 38% (6/16) than the patients who r eceived inadequate antibiotics at this time and who had a mortality rate of 91 % (31134). The 16 patients receiving adequate therapy at the pre-BAL time were being treated for a n ew YAP (n=4 ), progression of prior YAP or NP (n=9 ), or extrapulmonary infection (EPI) (n=3). As previously mentioned (Table 3), mortality was 60% (9/15) for the group of patients who received no antibiotics prior to BAL. Adding these 15 patients to the 34 who received inadequate antibiotics prior to BAL, the mortality was 81.6% (40/49) , even though some patients subsequently had therapy modified to an adequate agent, either empirically or based on BAL culture results . The mortality rate of these patients was significantly higher than the mortality observed in patients receiving adequate antibiotics prior to BAL (p<0.005 ) (Fig 2). Antibiotic therapy was initiated immediately after performing BAL in the 15 patients not receiving previous antibiotics. For the 50 patients already

p·=·N-.s·.· -- ··· ··-····

80%

~

~

c:t::

0

~

SO%

40%

20%

pre-BAL (N:66)

NO-ATB ADEQUATE NOT-ADEQUATE

(9/15) 38% (6/16) 91% (31/34)

post-BAL (N:66)

post RESULT (N:42)

60%

71% 70%

(30/42) (16/23)

57% 40%

(21/37) (2/5)

I:2:J NO-ATB .ADEQUATE ts:J NOT-ADEQUATE FIGURE l. Mortality rates are plotted in re lation to the adeguacy of antibiotic therapy at three diffe rent

tim es (pre-BAL, post-BAL, and postresult). Statistical diffe re nces between adequate and inad equate the rapy are present only at the pre-BAL time, when mortality was lower for p ati ents receiving adequate the rapy.

680

Clinical Investigations in Critical Care

p< 0.005

100%

80%

20%

O%JL--------------

ADEQUATE NOT-ADEQUATE or NO-ATB

38% 81,6%

(6/16) (40/49)

.ADEQUATE ts::JNOT·ADEQUATE or NO·ATB

FIGURE 2. For the 65 patients with a positive BAL culture, the impact of the initial therapy, at the pre-BAL time, on the outcome was evaluated. Patients receiving adequate initial antibiotic therapy had a significantly lower mortality rate than patients receiving either inadequate antibiotic therapy or no antibiotics.

recervmg antibiotics, therapy was changed in the immediate post-BAL period in 33 cases while it was not changed in the other 17. Three of 16 patients receiving adequate antibiotics pre-BAL had a change to inadequate antibiotic therapy at the post-BAL time and two of these patients continued with this therapy, even after BAL data were known, because they were responding clinically and radiographically. With these changes, 42 of the 65 BAL( +) patients were receiving adequate therapy in the immediate post-BAL time. At the time that BAL results were known, antibiotic therapy was changed in 20 cases, not changed in 22, and 23 patients had already died. When BAL data were known, 37 of the 42 surviving patients were receiving adequate therapy. As shown in Figure 1, the prescription of adequate antibiotics, compared with inadequate therapy, reduced mortality only at the pre-BAL time. At the post-BAL time (empiric therapy) and at the postresult time (culture-guided therapy), even though a larger percentage of patients received adequate therapy, this did not reduce the mortality when compared with the outcomes observed in those receiving inadequate antibiotics. This lack of reduction in mortality applied even after excluding the 23 early mortalities, and examining only the remaining 42 patients evaluated when BAL results were known.

Thus, doing BAL and defining adequate therapy did not significantly reduce the mortality for patients who initially received inadequate therapy, and much of the potentially useful bacteriologic information became available very late, after 23 of the patients had died. Etiology and Outcome of BAL( +) V AP: Influence of Previous Antimicrobial Therapy

A total of 123 microorganisms were cultured from BAL in the 6.5 episodes of BAL( +) VAP, at a concentration> 10 cfu/mL (1.9 microorganisms per pneumonia). The most frequently isolated organisms were S aureus, Acinetobacter species, Klebsiella pneumoniae, and Pseudomonas aernginosa. They were involved in 49%, 49%, 26%, and 20% of these pneumonias, respectively. Twenty of the 32 isolated organisms of S aureus were identified as methicillin resistant. Acinetobacter species and/or S aureus (at least one of them) were involved in 74% of the episodes of pneumonia. There were no differences in mortality for any specific microorganism (Table 4). No specific pathogen (Table 4) was present significantly more frequently in the group of patients receiving prior antibiotics than in the group not receiving prior antibiotics (p>0.05) . CHEST I 111 I 3 I MARCH, 1997

681

Table 4-Etiology, Adequacy of Antibiotic Therapy, and Mortality in 65 Cases of BAL+ VAP* Prior Antibiotic Therapy Microorganism Gram-negative bacilli Acinetobacter sp K pneumoniae P aemginosa Proteus mirabilis Escherichia coli Neisseria sp Enterobacter cloacae Pseudorrwnas putida Citrohacter sp 1-Iaemophylus injluenzae Alcaligenes xiloxida Gram-positive cocci S aureus Streptococcus viridans Staphylococcus epidennidis Streptococcus agalactiae Corynebacterium sp Enterococcus faecium Fungi Candida sp Total

Mortality

NO-ATB 1 (n=15)

ATB (n=50)

Total (n=65)

12 4

65 28 16 9 4 2 2 1

77 32 17 13 4 3 2 2

4 0 0 1 0 0 0 1 13 9 2 0 1 1 0 0 0 25

1 1 0 29 23 3 2 0 0 4 4 98

42 32 5 2

4 4 123

ADEQ ATB

o/o

NO-AD ATB

%

78 76 58 50 100

1/5 1/4 416 l/3

20 25 67 33

24/27 12/13 3/7 1/1

s9t 92§ 43 100

21/32 2/5 2/2 1/1 1/ 1 1/1

66 40

1/7 1/3

14 33

20/25 1/2

56t 50

3/4

75

0/1

0

3/3

General

%

25/32 13/17 7/13 2/4 3/3 2/2 112 0/1 1/1 0/l 1/ l

100

*NO-ATB=patients not receiving prior antibiotics; ATB=patients receiving prior antibiotics; Mortality=No. of cases dead/cases; ADEQ ATB=patients receiving prior adequate antibiotics; NO-AD ATB=patients not recei\~ng prior adequate antibiotics (NO ATB+ INADEQUATE ATB). 1p=NS; p value compares frequency of isolation of the different microorganisms in the group receiving antibiotics \vith the group not receiving antibiotics. tp< 0.005. \p<0. 01; p value compares mortality related to different microorganisms in cases with adequate vs no-adequate (inadequate+ NO-ATB) pre-BAL antibiotics.

The use of adequate pre-BAL antibiotics for infections with Acinetobacter species, K pneumonia, and S aureus (Table 4) led to a lower mortality rate than the use of inadequate antibiotics for these organisms.

DISCUSSION

Bacterial infection of the lung is a common complication of mechanical ventilation and most researchers have adopted or modified the clinical definition of pneumonia used by Johanson et alt 2 (radiographic appearance of a new or progressive pulmonary infiltrate; fever; leukocytosis; and purulent tracheobronchial secretions). However, several studies have demonstrated that the same clinical picture can occur without bacterial pneumonia, making this definition sensitive, but not specific. Controversy has therefore developed over whether a clinical definition of pneumonia should be used to guide decisions about the initiation or continuation of antibiotic therapy. Some investigators have proposed 682

that antibiotic decisions be guided by microbiologic confirmation of the clinical diagnosis of pneumonia, relying on quantitative cultures obtained from bronchoscopic techniques, including BAL and the protected specimen brush (PSB), because these methods have been reported to accurately confirm or exclude the diagnosis of pneumonia.l3-16 While emotions run high in this area, no investigator has demonstrated that bronchoscopic data can favorably influence the outcome of patients with established NP. Therefore, we examined the initial and subsequent choices of antibiotics, and the impact of BAL data (collected within 24 h of the clinical diagnosis of a new or progressive pneumonia), on the course and outcomes in 132 episodes of clinically suspected YAP. In this study, pneumonia was confirmed microbiologically (> 104 cfu/mL on BAL) in 65 episodes, and the bacteriology of BAL cultures and antimicrobial susceptibility patterns were used to determine if antibiotic therapy was adequate or inadequate. These designations were made for each episode at three different times: pre-BAL, when the Clinical Investigations in Critical Care

pneumonia was first recognized (early empiric therapy); post-BAL when the bronchoscopy was completed (empiric therapy); and postresult, when BAL data were known (culture-guided therapy). At each successive time point, alarger percentage of patients received adequate therapy, but the use of adequate th erapy was only significant (p<0.001) , compared with inadequate therapy, if it was given at the pre-BAL time (Fig 1). If adequate therapy was given at the post-BAL or postresult time, mortality was higher than if adequate therapy had been given earlier. In addition, at these later times, mortality was comparable whether patients received adequate or inadequate therapy. We also found that among the 46 patients who died, 23 died within the first 48 h after bronchoscopy, a time before BAL data were available to guide therapy. Thus, definitive bacteriologic data often became available too late to be clinically useful, and there was no evidence that the information provided by BAL could be used to reduce mortality in these seriously ill patients. Adequate antibiotic therapy did have benefit for some patients, but this benefit was most clear if therapy was initiated empirically at the earliest possible time after the clinical recognition of infection. When the 65 BAL( +) patients were compared with the 67 BAL(-) patients (Table 1), mortality rates were comparable and the two groups differed only with the finding that significantly more BAL( + ) patients had all three clinical criteria of pneumonia present and more were bacteremic. Thus, when VAP was confirmed by BAL culture, patients were more likely to have three clinical criteria present than to have only two such criteria, while the opposite situation was found in patients whose pneumonia was not confirmed by BAL (p<0.05 ). These data suggest some value in carefully observing the clinical features of VAP, even though some investigators have suggested that these clinical features are overly sensitive for the diagnosis of VAP and thus of little value.l 0·13 · 16 ·li However, others have found , as we did, that the clinical definition of pneumonia does correlate with BAL data, reporting a higher frequency of clinical pneumonia in BAL( + ) patients than in BAL(- ) patients. 18 One limitation to the use of bronchoscopy in critically ill patients is the finding that prior antibiotic therapy can lead to fals e-negative microbiologic data, even though some studies have collected quantitative cultures in patients receiving antibiotics. 11 · 16 In the present study, BAL was done regardless of whether the patient was receiving antibiotics and the overall clinical picture in the 65 BAL( +) patients was similar regardless of whether the patient was receiving antibiotic therapy (Table 3). In fact, 50 of these 65 patients with microbiologically confirmed pneu-

monia were receiving antibiotic therapy at the time that BAL was performed. The high rate of positive BAL results in these patients may be a reflection of the fact that only 16 individuals were receiving adequate therapy. Previous studies have suggested that prior antimicrobial therapy can cause a change in the distribution of infecting pathogens , le ading to infection with organisms that are associated with a high mortality rate. 2,17,19 In the present study, we did not observe significant changes in the distribution of any specific pathogen as a result of prior use of antibiotics (Table 4). In addition, mortality in the BAL( +) patients was not higher for patients receiving antibiotics compared with those not receiving antibiotics, even when taking into account the time that antibiotics were given (>24 h pre-BAL vs ::::;24 h pre-BAL) and the reason that antibiotics were prescribed (VAP vs other causes) (Tables 1 and 3). Although we observed only a small, nonsignificant difference (71 % vs 64%) in mortality between BAL( + ) and BAL(- ) patients, previous studies have reported that the mortality of patients with microbiologically confirmed VAP (using BAL or PSB ) is higher than the mortality of patients with clinical pneumonia but negative BAL or PSB cultures.l 1·20 Our findings may reflect ahigh rate of false (-) BAL results , possibly as a consequence of the fact that 57 of the 67 BAL(-) patients received prior antibiotic therapy. It is difficult to comment on the rate of false-negative results in the BAL(- ) patients, although some of these patients probably were treated with adequate therapy. For example, one patient had two of two blood cultures positive for Acinetobacter species but was BAL(- ), with < 104 cfu/mL for Acinetobacter species, while receiving adequate antibiotic therapy for that microorganism. It is possible that other BAL(-) patients also had false-negative results as a result of antibiotic therapy or had negative BAL data as a reflection of the inherent limitations of bronchoscopic cultures. Alternatively, both BAL( +) and BAL(-) patients could have comparable, and high, mortality rates because of a lack of attributable mortality of VAP in this population of individuals with a high degree of severe illness. The greatest impact of prior antibiotic therapy on mortality was related to whether that therapy was adequate for the organisms that were subsequently identified b y BAL. The following were the most common bacteria in our study: S aureus (most of the m m ethicillin resistant) ; Acinetobacter species; K pneumoniae, and P aeruginosa. Although we did not find overall differences in mortality related to specific etiologic pathogens, we did observe that the use of adequate therapy led to an improved outcome for CHEST I 111 I 3 I MARCH, 1997

683

certain pathogens (Table 4). For the group as a whole, as well as with certain organisms, when adequate therapy was given prior to performing BAL (at the time that pneumonia was clinically recognized), mortality was lower than if adequate therapy or no therapy had been given (Fig 2). However, our data did have a trend toward the finding that when pneumonia was first recognized, the use of inadequate empiric therapy may have been worse than the use of no therapy (Fig 1). Even when therapy was later modified to reflect bronchoscopic data, mortality did not fall when compared with continued inadequate therapy (Fig 1). These data fit well with the findings of Montravers et al 2l who showed that microbiologic evaluation using a PBS after 3 days of antibiotic therapy can define a bacteriologic failure, but that the recognition of such a microbiologic nonresponse did not lead to modifications in therapy that could improve outcome. Others have also suggested that the use of adequate antibiotic therapy cannot affect outcome in VAP,22,23 but they did not stratify the timing of adequate therapy in the same way that we did, looking specifically at early vs late time points. Our study has a number of limitations that are a reflection of the fact that the data, although collected prospectively, were observational, and patients were not managed with a specific therapeutic protocol. Other limitations include the following: (1) there was a high mortality of both BAL( +) and BAL(-) patients, which might be related to the fact that the patients in this study were predominantly old and seriously ill, although comparable mortality rates have been reported by other investigators, particularly in patients receiving prior antibiotics; 24 (2) most of the patients were receiving antibiotics prior to BAL so that the adequacy of therapy of BAL(-) patients could not be assessed; and (3) many patients receiving prior antibiotics were being treated for previous pneumonia or for EPI. Paradoxically, 12 of 16 patients getting adequate therapy at the pre-BAL time point had developed a new or progressive infiltrate, with microorganisms susceptible to the antibiotics used despite >24 h of adequate therapy. Future studies are needed to confirm our findings and to determine if they can be generalized to other critical care settings. In our study, many patients were infected with potentially resistant Gram-negative bacteria, and our observations may not apply to a population with more easily treated organisms. In addition, our initial antibiotic therapy was often inadequate, reflecting a need to modify our empiric approach, although it was impossible to comment on whether our initial therapy choices in BAL(-) patients were correct. In addition, we had a surprisingly high early mortality rate in our patients, which may 684

have been multifactorial and not entirely related to the presence of pneumonia. Quantitative bronchoscopic cultures of BAL and PSB may permit more precise etiologic diagnosis of pneumonia and must be considered a reference standard in confirming or ruling out infection in cettain clinical investigations or epidemiologic surveys ofVAP. Nevertheless, some authors believe that invasive diagnostic methods do not currently represent an accepted standard for routinely treating patients suspected of having VAP. 25 In our study, BAL permitted us to know the etiology of YAP, but this knowledge did not help reduce mortality or lead to an improvement in outcome for any patient group. Thus, it may be necessary to use appropriate empiric therapy from the time pneumonia is first recognized, and if this therapy is adequate, outcome may improve. Our data suggest that if initial therapy is incorrect, the results of BAL cultures cannot be relied on to modify therapy and to lead to a favorable outcome.

REFERENCES 1 Niederman MS, Craven DE, Fein AM, et a!. Pneumonia in the critically ill hospitalized patient. Chest 1990; 97:170-81 2 Rello J, Ausina V, Ricart M, et a!. Impact of previous antimicrobial therapy on the etiology and outcome of ventilator-associated pneumonia. Chest 1993; 104:1230-35 3 Fagon JY, Chastre J, Dormat Y, et al. Nosocomial pneumonia in intubated patients undergoing mechanical ventilation (prospective analysis of 52 episodes with the use of a protected specimen brush and quantitative culture techniques). Am Rev Respir Dis 1989; 139:877-84 4 Griffin JJ, Meduri CU. New approaches in the diagnosis of nosocomial pneumonia. Med Clin North Am 1994; 78:1091122 5 Chastre J, Fagon JY. Invasive diagnostic testing should be routinely used to manage ventilated patients \vith suspected pneumonia. Am J Respir Crit Care Med 1994; 150:570-74 6 Niederman MS, Torres A , Summer W. Invasive diagnostic testing is not needed routinely to manage patients suspected of having ventilator acquired pneumonia. Am J Respir Crit Care Med 1994; 150:565-69 7 Celis R, Torres A, Gatell J, et a!. Nosocomial pneumonia, a multivariate analysis of risk and prognosis. Chest 1988; 93:318-24 8 Torres A, Aznar R, Gatell JM, et a!. Incidence, risk, and prognosis factors of nosocomial pneumonia in mechanically ventilated patients. Am Rev Respir Dis 1990; 142:523-28 9 Meduri GU, Beals DH , Maijub AG, et a!. Protected bronchoalveolar lavage: a n ewbronchoscopic technique to retlieve uncontaminated distal ai1way secretions. Am Rev Respir Dis 1991; 143:855-64 lO Chastre J, Fagon JY, Soler P, et a!. Diagnosis of nosocomial bacterial pneumonia in intubated patients undergoing ventilation: comparison of the usefulness of the bronchoalveolar lavage and the protected specimen brush. Am J Med 1988; 85:499-506 ll Guerra LF, Baughman RP. Use ofbronchioalveolar lavage to diagnose bacterial pneumonia in mechanically ventilated patients. Crit Care Med 1990; 18:169-73 Clinical Investigations in Critical Care

12 Johanson WG Jr, Seidenfeld JJ, Gomez P, eta!. Bacteriologic diagnosis of nosocomial pneumonia followin g prolonged mechanical ventilation . Am Rev Respir Dis 1988; 137:259-64 13 Chastre J, Viau F , Brun P, et a!. Prospective evaluation of protected specimen brush for th e diagnosis of pulmona1y infections in ventilated patients . Am Rev Respir Dis 1984; 130:924-29 14 Torres A, Puig de Bellacasa J, Xaubet A , et a!. Diagnostic value of quantitative cultures of bronchoalveolar lavage and telescoping plugged catheters in mechanically ventilated patients with bacterial pneumonia. Am Rev Respir Dis 1989; 140:306-10 15 Thorpe JE, Baughman RP, Frame PT, et a!. Bronchioalveolar lavage for diagnosing acute bacte1ial pneumonia. J Infect Dis 1987; 22:855-69 16 Fagon JY, Chastre J, Hance AJ, et a!. Detection of nosocomial lung infection in ventilated patients: use of a protected speci men brush and quantitative culture techniques in 147 patients. Am Rev Respir Dis 1988; 138:110-16 17 Meduri GU, Wunderink RG , Leepe r KV, et al. Management of bacterial pneumonia in ventilated patients: protected bronchoalveolar lavage as a diagnostic tool. Chest 1992; 101:500-06 18 Pugin J, Auckenthaler R, Mili N, et a!. Diagnosis of venti lator associated pneumonia by bacteriologic analysis of broncho-

19

20

21

22

23 24

25

scopic and non-bronchoscopic "blind" bronchoalveolar lavage fluid . Am Rev Respir Dis 1991; 143:1121-29 Wunderink RG. Mortality and ventilator associated pneumonia: th e bes t antibiotics may be the least antibiotics. Chest 1993; 104:993-94 Fagon JY, Chastre J, Han ce AJ , et a!. Nosocomial pneumonia in ventilated patients: a cohort study evaluating attributable mortality and hospital stay. Am J M ed 1993; 94:281-99 Montravers P, Fagon JY, Chastre J, et a!. Follow-up protected specimen brushes to assess treatment in nosocomial pneumonia. Am Rev Respir Dis 1993; 147:38-44 Seidenfield JJ, Pohl DF, Bell RC, et al. Incidence, site and outcome of infections in patients with adult respiratory distress syndrome. Am Rev Respir Dis 1986; 134: 12-6 Stevens RM , Tenes D , Skillman JJ, et al. Pneumonia in an intensive care unit. Arch Intern Med 1962; 110:847-64 Fagon JY, Chastre J, Dormat Y, et al. Nosocomial pneumonia in patients receiving mechanical ventilation: prospective analysis of 52 episodes with use of a protected specimen brush and quantitative culture techniques. Am Rev Respir Dis 1989; 139:877-84 Timsit JF, Misset B ,Francoual S, eta!. Is protected specimen brush a reproducible method t o diagnose ICU-acquired pneumonia? Chest 1993; 104:104-08

.-\ M E RI C A N

PHY

S

CO l l E G E

I

C

I

O F

AN

S

Mechanical

~ntilation

The First Annual Symposium and Workshop for Critical Care Providers June 26-28, 1997 San Diego, California

FOR INFORMATION CALL: 1-800-343-ACCP or 847- 498-1400

CHEST /111 /3/ MARCH, 1997

685