Effect of hospital-acquired ventilator-associated pneumonia on mortality of severe community-acquired pneumonia

Effect of Hospital-Acquired Ventilator-Associated Pneumonia on Mortality of Severe Community-Acquired Pneumonia Olivier

Leroy,

Jbrome

Guilley,

Hughes

Georges, Gilles

Philippe Choisy, Beaucaire

Benoit

Guery,

Serge

Alfandari,

and

Purpose: The purpose of this article is to evaluate, -two pairwise case-control studies, attributable mortality linked to hospital-acquired ventilator-associated pneumonia (HA-VAP) complicating the intensive care unit (ICU) stay of patients exhibiting severe community-acquired pneumonia (CAP). Materials and Methods: Over an II-year period, 498 patients with severe CAP were collected. Among them, 43 exhibited HA-VAP. In a first case-control study, these patients were matched with control on the basis of six confounding variables known to be general ICU prognosis factors. In a second case-control study, six variables specifically linked to CAP prognosis were used for matching. Results: In the two case-control studies, each case

patient was matched with one control patient. In the first analysis, success of matching was achieved in 198 of 258 (77%) variables used for matching. In thesecond analysis, matching was successful for 242 of 258 (94%) confounding variables used. Eighteen patients died, compared with, respectively, 6 (P = .003) and 7 (P = .Ol) controls. Attributable mortality of HA-VAP was similar in the two pairwise analyses, respectively, 28% (risk ratio = 3.0; 95% confidence interval, 1.32 to 6.82) and 26% (risk ratio = 2.57; 95% confidence interval, 1.2 to 5.52). Conclusion: When confounding factors were controlled, HA-VAP appeared to increase mortality of severe CAP requiring ICU admission. Copyright 0 1999 by W.B. Saunders Company

D

France. Received July 23, 1998. Accepted October 14, 1998. Address reprint requests to Olivier Leroy, MD, Service de R&animation Mkdicale et Maladies Infectieuses, Centre Hospitaliel; Rue du Pksident Coty, 59208, Tourcoing, France. Copyright 0 1999 by WB. Saunders Company 0883-9441/99/1401-0003$10.00/0

fully matched in the two populations, appears an interesting method at present. To date, only a few studies devoted to the mortality of HA-VAP have used such an analysis. MJ Variables chosen as confounding factors were current predictors of ICU mortality (age, severity of underlying illness, initial diagnosis, indication for ventilatory support, duration of mechanical ventilation). Including mainly surgical patients, only one of them demonstrated that HA-VAP was associated with an increased mortality.7 The prognostic impact of HA-VAP complicating the evolution of severe community-acquired pneumonia (CAP) is unknown. To evaluate the prognostic impact of such a complication and to determine its attributable mortality, according to an analysis using the case-control technique, the choice of the confounding variables used for matching appears as the most subtle point. The use of current predictors of ICU mortality is one of the possible choice. However, as suggested by reports studying the CAP prognosis,8-‘0 specific predictors of such a prognosis are different. Indeed, in one of our previous studieslo we identified, using a credit scoring technique, six ICU admission predictors of CAP prognosis including the following: (1) age > 40 years, (2) anticipated death < 5 years, (3) nonaspiration pneumonia, (4) chest radiograph involvement >1 lobe, (5) respiratory failure requiring mechanical ventilation, and (6) septic shock. Consequently, the goal of our study was to build two successive case-control analyses, the first one with

12

Journal

ESPITE THE USE of appropriate preventive measures, hospital-acquired ventilator-associated pneumonia (HA-VAP) remains a major problem in intubated patients receiving mechanical ventilation in intensive care units (ICU). The estimated prevalence of this infection in such patients ranges from 5%’ to more than 24%.* The observed mortality rates in patients acquiring HA-VAP varies from 30% to 70%.2-4 However, although these statistics clearly indicates that HA-VAP is a severe disease, its direct impact on mortality and morbidity remains, in recent literature, controversial. Indeed, numerous studies have demonstrated that severe underlying diseases predisposing to the development of HA-VAP also influence the prognosis? Thus, the question of whether HA-VAP or the host setting is the crucial factor responsible for death has not yet been clearly determined. To evaluate the respective responsibility of these various factors in increasing mortality and to draw firm conclusions about attributable mortality linked to HA-VAP, appropriate statistical analysis is required. Analysis based on a case-control study, in which confounding variables of prognosis are careFrom euses,

Service Universite’

de Rianimation Midicale et Maladies Infectide Lille, Centre Hospitaliel; Tourcoing,

of Critical

Care, Vol 14, No 1 (March),

1999: pp 12-19

EFFECT

OF HA-VAP

ON COMMUNITY-ACQUIRED

PNEUMONIA

the current ICU predictors of mortality, as confounding factors for matching,‘x7 the second one with specific predictors of CAP prognosis.” PATIENTS

AND METHODS

Study Design We performed two pairwise, retrospective case-control studies with 1:l matching. The study was conducted in the Service de Reanimation Mtdicale et Maladies Infectieuses (Hopital Chatiliez, Tourcoing, France) of Universite de Lille. From January 1987 to December 1997, we retrospectively (1987 to 1992) and prospectively (1993 to 1997) collected all patients exhibiting severe CAP. CAP was diagnosed according to Fine et al’s criteria.”

Collection of Data Variables recorded for patient evaluation have been previously described.8-10-‘* Briefly, main collected characteristics and used definitions were as followed: within 24 hours of ICU admission, patients underwent clinical, radiological, and biological tests. The underlying clinical conditions were classified according to the criteria proposed by McCabe and JacksonI Chronic respiratory insufficiency was assessed combining the usual clinical and radiological criteria and the coexistence of ventilatoty impairment assessed either before or after the ICU stay. Immunosuppression was defined as a leukocyte count less than l,OOO/ mm?, recent use of systemic corticosteroids, cytotoxic drugs, radiation treatment, or asplenia.” Vital sign abnormalities were assessed by simplified acute physiologic score (SAPS I),r4 and acute organ system failure (OSF) scoring system.‘s Shock was defined according to Bone et al’s criteria.r6 Neurological status and changes in mental status were stratified according to the Glasgow Coma Score.17 Aspiration pneumonia was diagnosed according to the usual criteria. 8~1’~‘2.‘8On the chest radiograph, the number of lobes involved and the unilateral versus bilateral involvement were recorded. Biological tests included measurement of hemoglobin level, total leukocyte count, total platelets count, creatinine and electrolytes levels, total serum protein, arterial blood gases, and Pao2/Fio2. A microbiological diagnosis, always attempted, was assigned using the usual criteria.8~1r~‘9 During the ICU stay, we collected information on the following therapeutic topics: supportive measures, such as hemodialysis or mechanical ventilation, duration of mechanical ventilation, use of systemic corticosteroids or inotropic drugs, and initial antimicrobial therapy. Adequacy and effectiveness of this initial CAP antimicrobial treatment were determined within 72 hours of the beginning of the treatment. Antimicrobial therapy was considered adequate if the causative pathogen, according to antibiotic susceptibility reports, was susceptible to this antimicrobial therapy or, when no precise microbiological diagnosis was made, the empirical drugs chosen accomplished the recommendations of the medical literature. The initial therapy was considered effective if the clinical situation improved and fever decreased within the first 72 hours of treatment. In all the other clinical situations, the initial therapy was considered a failure. This evaluation, based only on clinical response, should not be confused with the adequacy of therapy based on dose, route, and sensitivity patterns. The occurrence of complications was recorded. They were classified as either directly pneumoniarelated (secondary septic shock, acute respiratory distress syn-

13

drome, development of multiple organ failure, or HA-VAP) or no directly or nonspecific pneumonia-related (ICU-related complications and complications attributed only to underlying medical conditions). Multiple organ failure (MOF) was defined as the dysfunction of two or more of the seven evaluated organ systems. *’ Acute respiratory distress syndrome (ARDS) was defined according to usual criteria?’ All patients were monitored until their discharge from our ICU. As this ICU receives patients from all hospitals of a large geographic area (Region Nord-Pas de Calais) subsequent outcome after discharge from the ICU could not be determined.

Case Identification Patients exhibiting a late onset (25 days/KU admission) HAVAP were considered for potential case patients. The diagnostic criteria for the diagnosis of HA-VAP were modified from criteria established by the American College of Chest Physicians.” HAVAP was considered to be present when a new or progressive roentgenographic infiltrate, not present on ICU admission, developed after the fourth day following ICU admission in conjunction with one of the following criteria: a positive blood or pleural fluid culture or the presence of a positive quantitative culture of a sample of secretions from the lower respiratory tract (endotracheal aspiration 2 lo6 CFU/mL, protected brush catheter 2 lo3 CFU/mL, and bronchoalveolar lavage 2 lo4 CFU/mL); or two of the following criteria in the absence of an alternative explanation for the pulmonary infiltrates: new onset of production of purulent sputum, or temperature > 38.5”C, or leukocytes count > 10,000/mm3 or < 1 ,500/mm3. Positive pleural or blood cultures could not be related to initial CAP, or to another source than HA-VAP. Therefore, they were considered as significant when the same organism as that recovered from the sample of respiratory secretions was identified. To be significant for the diagnosis of HA-VAP, all bacteriological samples had to be obtained 25 days after ICU admission. If a patient had more than one episode of HA-VAP during the ICU stay, only the first episode was taken into account.

Selection of Controls and Matching A control was defined as a patient initially exhibiting severe CAP requiring mechanical ventilation for more than 4 days, but without later occurrence of any directly pneumonia-related complications. Each possible control patient was matched to one case patient on the basis of variables considered to be potential confounding factors. In a first case-control study called “ICU matching,” we used six variables previously used by several authors’,’ and known to be associated with the risk of death in ICU patients. Thus, each case patient was matched to one control patient on the basis of the following variables considered to be potential confounding factors: (1) age of the case and control patient had to be matched within 5 years, (2) case and control had to be of the same sex, (3) case and control had to have the same severity of underlying diseases according to the McCabe and Jackson classification, (4) SAPS I calculated within the first 24 hours of admission to the ICU had to be matched within 3 points, (5) case and control had to be matched for the presence or the absence of shock on admission, (6) duration of mechanical ventilation received by control and case (before the onset of HA-VAP) had to be matched within 2 days. However, as the study endpoint-mortality in the ICU-was closely related to underlying diseases, initial

14

LEROY

severity of CAP, and evolution during the ICU stay, the adequacy of the matching was controlled by comparing, for each casecontrol pair, several variables previously identified as prognosis factors of CAP.@ On admission, these factors were immunosuppression, nonaspiration CAP, bilateral or >2 lobes on chest radiograph involvement, bacteremia, OSF score > 2, leukocytes count < 3,500/mm3, creatinine level > 20 mg/L, and use of antimicrobial combination as initial treatment of CAP. During the ICU stay, the predictive factors were ineffective initial antimicrobial therapy instituted for CAP and the occurrence of nonspecific CAP-related complications. In a second analysis called “CAP matching,” we used six variables specifically linked to CAP prognosis and previously identified by our group in a multivariate analysis of CAP prognosis.l” All of them were available within a few hours following ICU admission. Therefore, each case patient was matched to one control patient on the basis of the following variables considered to be potential confounding factors: (1) age had to be in the same class (5 or > 40 years), (2) severity of underlying diseases according to the McCabe and Jackson classification must be similar (anticipated death < or 2 5 years), (3) case and control had to be matched for the presence or the absence of aspiration on admission, (4) involvement on chest radiograph had to be in the same class (1 or > 1 lobe), (5) case and control had to be matched for the presence or the absence of shock on admission, and finally, (6) as all control and case patients were initially mechanically ventilated, we used as the last variable the duration of exposure to risk; consequently, each potential control had a total length of ventilation equal (2 2 days) to the duration of ventilation received by the case before the onset of HA-VAP.

Statistical Analysis Statistical calculations were performed using the Statistical Analysis System software package (SAS version 6.1; SAS Institute, Cary, NC).” The characteristics of patients in each group were compared by chi-square test for categorical variables and Student’s t test for continuous variables. For the statistical analysis of the matched cohort study, chi-square and McNemar’s tests were used. For all statistical tests used, a P value less than .OS was considered as statistically significant. The attributable mortality due to HA-VAP was defined as the crude mortality of the controls subtracted from that of cases. The risk ratio and its 95% confidence interval (CI), defined as the ratio of mortality rate of cases to that of the controls, and used to measure the relative risk of HA-VAP in inducing death, was determined using Taylor’s method.

RESULTS

Incidence of HA-VAP During the 11-year period, 498 patients exhibiting severe CAP were collected. Acute respiratory failure leading to mechanical ventilation was present in 298 patients (60%). Of these, 184 patients were ventilated ~5 days (62%). HA-VAP was diagnosed in 43 of these patients. The mean interval from mechanical ventilation to identification of HA-VAP was 12.4 ? 6.1 days with a range of 5 to 35 days. Therefore, the incidence of HA-VAP

ET AL

was 8.6% for all patients admitted into the ICU for CAP during the study period, 14.4% for ventilated patients, and 23.4% for patients ventilated 2.5 days. A potential causal pathogen was isolated in 36 patients (84%) from blood cultures (n = 2), protected brush (n = 6), and endotracheal aspiration (n = 2X). There were 40 microorganisms recovered. The most common causative organisms were Staphylococcus spp (n = 16), Pseudomonas sp (n = lo), and Acinetobacter sp (n = 10). A polymicrobial flora was identified in only four cases (9%). Seven patients were considered to have HA-VAP despite negative bacteriological proofs of infections. All were subsequently treated by antimicrobial therapy and were included in this study. Among the 184 patients mechanically ventilated 2.5 days, 85 were unharmed from any pneumonia-related complications and, thus, were possible control patients. Characteristics of the Study Population Main characteristics of the case patients and the two control group patients are summarized in Tables 1 and 2. In the “ICU matching,” study, cases and controls were comparable for age, underlying conditions, initial severity of CAP, initial biological and microbiological data, and most of therapies used during ICU stay (Tables 1 and 2). The only significant difference was the higher incidence of ineffective initial antimicrobial therapy in the case patients (51% v 21%). In the “CAP matching” study, the two group of patients were also similar for all characteristics. Patients only significantly differed on the incidence of bacteremia at ICU admission (26% v 9%), the use of hemodialysis (0% v 14%), and the effectiveness of initial CAP antimicrobial therapy (81% v 49%) that were more frequent in the control patients. Effective and Control of Matching In the two case-control studies, we were able to match each case patient exhibiting HA-VAP with one control patient. In the “ICU matching” study (Table 3), the effectiveness of matching was as follows: Age-matching of case-control pairs amounted to 42%. Sixty-seven percent of pairs were sex-matched. In 81%, SAPS I scores and severity of underlying disease according to the McCabe and Jackson classification were matched. Matching for presence or absence of septic shock on admission was performed in 98% of case-control pairs. Finally, 91% of control patients had a duration of

EFFECT

OF HA-VAP

ON COMMUNITY-ACQUIRED

Table

PNEUMONIA

1. Characteristics

15

of the Case and Control

Groups:

Data on ICU Admission Control Patients

Characteristics

Underlying

Case Patients

CAP Matching

conditions

Mean age (years) Sex M/F Anticipated

63.7 t

death

<5 years

COPD lmmunosuppression Aspiration Severity of initial SAPS I

Septic

Score

Chest radiograph >2 lobes

68.3 i

9.1

32/l 1

24 (56%)

26 (60%)

24 (56%) 21 (49%)

2 (5%) 4 (9%)

2 (5%) 5 (12%)

3 (7%) 5 (11%)

15.0 i 5.8 12.8 ? 3.9 1.4 t 0.7

14.1 i: 4.6 12.7 i 3.7 1.3 2 0.7

8 (19%)

9 (21%)

13.5 I! 4.6 12.4 2 3.7 1.3 i 0.8 8 (19%)

involvement

Bilateral Biological data (mean Pao2/Fio2 (mm Hg)

value

8 (19%)

5 (12%)

9 (21%)

14 (33%)

7 (16%)

12 (28%)

213 ? 102 46.5 t 20

229 + 110 49.7 + 17.7

229 2 107 50.4 + 17.6

? SD)

Hg)

Serum creatinine level (mg/L) Total serum protein (g/L) Leukocytes count (103/mm3)

15.2 t 8.8 60 2 10.2 14.0 -t 7.1

Platelets

267 i

count

Table

67.6 i: 10.6 3419 21 (49%)

shock

(mm

11.3

3419 21 (49%)

CAP

Glasgow Coma OSF score

Pace,

KU Matching

(103/mm3)

2. Microbiological

Data,

14.96 ? 7.7 59.6 ? 6.9 13.7 2 7.3

123

Main Therapies,

14.7 -t 7.3 60 -t 9 13.3 t 7.1

223 -t 103

and Complications

During

226 i

117

ICU Stay

Control Patients Case Patients

Microbiological Unknown

diagnosis

S. pneumoniae Staphylococcus M. catarrhalis

ICU Matching

CAP Matching

of CAP

spp.

19 (44%)

13 (30%)

10 (23%)

12 (40%) 3 (10%)

16 (41%) 4 (10%) 6 (15%)

16 (36%) 10 (22%) 5 (11%)

4 (10%) 0

5 (11%) 1 (2%)

H. infloenzae E. co/i

2 (7%) 8 (27%) 0

Klebsiella-Enterobacter-Serratia

1 (3%)

2 (5%)

2 (4%)

0

1 (3%)

1 (3%) 3 (10%) 4 (9%)

3 (8%) 3 (8%) 8 (19%)

2 (4%) 1 (2%) 3 (6%) 11 (26%)*

24 (56%) 13 (30%)

18 (42%) 6 (14%)

17 (40%) 7 (16%)

Proteus spp Chlamydia spp Miscellaneous Positive blood

cultures

Therapies lnotropic support Corticosteroids Dialysis Antimicrobial

ot

6 (14%)

0"

therapy

Monotherapy Adequate Effective Non CAP-related

4 40 21 15

complications

(9%) (93%) (49%) (35%)

*P c.05,

comparison

case patients

versus

CAP matching

control

tP c.05,

comparison

case patients

versus

ICU matching

control

8 39 34 12 patients. patients.

(19%) (91%) (79%)t (28%)

9 (21%) 39 (91%) 35 (81%)" 11 (26%)

LEROY

16

Table

3. ICU Matching Control

Study:

Effectiveness

and

Table

4. CAP Matching

Characteristics

age (25 years) sax SAPS I (-t 3 points)

Same Same

McCabe class incidence of septic

Same

duration

(i

Characteristics

n/n

%

18143

42 67

Same Same

ai

factors

Same Same Same

29143 35143 35143 shock

of exposure

to risk

Other prognostic factors lmmunosuppression (yes or no) Aspiration (yes or no) Chest radiograph involvement 5 or >2 lobes uni or bilateral 5 or >2 5 or >20 mg/L count 2 or <3,500/mm3

Bacteremia (yes or no) Initial antimicrobial therapy Monotherapy or combination Effective or ineffective Non-CAP-specific

complications

Confounding factors Same age class (5 or >40

81

42143

98

Same

39143

91

Same duration (2 2 days)

41143 34143

35 73

32143 28143 36143

74

39143 40143 31143

65 a4 91 33 72

33143

77

I a/43* 24143

42 56

(yes or no) “P < .05

years)

McCabe class incidence of aspiration

Same chest radiograph (I or >I lobe)

2 days)

OSF score Creatinine Leukocytes

Effectiveness

of Matching Proportion of Cases Matched to Controls

Proportion of Cases Matched to Controls

Confounding

Study:

of Matching

ET AL

incidence

involvement

of septic

shock

of exposure

to risk

n/n

%

43143 36143

100 a4

42143 40143

98 33

43143 38143

100 aa

of septic shock. Only 7 pairs differed on the underlying conditions class. One pair exhibited difference about the incidence of aspiration pneumonia. Chest radiograph involvement was similar in 40 of 43 (93%) pairs. Finally, 88% of control patients had a duration of mechanical ventilation at least as long as the case patients before the onset of HAVAP t 2 days. The mean duration of mechanical ventilation in control patients was 12.5 + 6.2 days, compared with a mean interval of occurrence of HA-VAP set at 12.4 + 6.1 days in the case patients. Mortality

mechanical ventilation at least as long as those of the case patients before the onset of HA-VAP + 2 days. For case patients, the mean interval from mechanical ventilation to identification of HA-VAP was 12.4 + 6.1 days. For control patients, the mean duration of mechanical ventilation was similar, 12.4 li: 6.5 days. Overall, success of matching was achieved in 198 of 258 (77%) variables used for matching. Because the study endpoint, mortality of CAP in the ICU, was also related to factors other than the confounding factors used, we controlled the matching, for each case-control pair, by comparing 11 prognostic factors of CAP evaluating the severity of disease in the two groups of patients (Table 3). For 10 of these variables, there was no significant difference between case and control pairs. The only significant difference observed was about the efficacy of the initial CAP antimicrobial therapy. Only 18 of 43 (42%) of the pairs were matched. In the “CAP matching” study (Table 4), matching was successful for 242 of 258 (94%) confounding variables used: all pairs were similar for the age class and the occurrence (or the absence)

Eighteen cases (42%) died, compared with, respectively, 6 (14%) and 7 (16%) of the control patients in the “ICU matching” and “CAP matching” studies (Table 5). In the first study, 27 matched pairs had a concordant outcome (23 lived and 4 died). Sixteen pairs had a discordant outcome and, in 14 of the pairs, the case died (P = .003). In this analysis, the estimated attributable mortality linked to HA-VAP complicating severe CAP was 28%, and the estimated risk ratio was 3.0 (95% CI, 1.32 to 6.82). In the “CAP matching” analysis, 24 pairs

Table

5. Mortality

in Case

and Control

ICU Matching Control Patients Survival Survival 23

Patients

Death 2

Total 25

4 6

18 43

Death 4

Total 25

3 7

18 43

Case Patients Death

Total

14 37 CAP Matching Control Patients Survival Survival 21

Case Patients Death Total

15 36

EFFECT

OF HA-VAP

ON COMMUNITY-ACQUIRED

PNEUMONIA

had a concordant outcome (21 lived and 3 died). Nineteen pairs had a discordant outcome (P = .Ol). In 15 pairs, cases died and controls survived. The estimated attributable mortality due to HA-VAP and the risk ratio were, respectively, 26% and 2.57 (95% CI, 1.2 to 5.52). DISCUSSION

The incidence of HA-VAP ranges, in the literature, from 5% to up 24%.‘s2 In the present work, nosocomial pneumonia occurred in 8.6% of patients admitted into the ICU for severe CAP. The incidence increased to 14.4% for patients mechanically ventilated generally and to 23.4% for patients ventilated 25 days. Despite such a worldwide high incidence and numerous studies devoted to this infectious disease, the relationship between HA-VAP and in-hospital mortality still remains unclear. Factors predisposing the development of HA-VAP and influencing the outcome of HA-VAP are so numerous that their control is required to study the attributable mortality linked to HA-VAP. Case-control analysis appears, at the moment, as one of the best statistical techniques to reach this goal. In previous studies using this statistical analysis to assess the attributable mortality linked to HA-VAP,‘g6s7variables currently used for matching were age, gender, severity scores, underlying diseases, initial diagnosis, indication for ventilatory support, and duration of mechanical ventilation. In this study population, all patients were already matched for diagnosis (severe CAP) and indication for ventilatory support (acute respiratory failure due to CAP). Therefore, we performed a first case-control analysis, using as confounding variables, age, gender, severity of underlying conditions, severity of present disease (ie, septic shock, SAPS I), and duration of mechanical ventilation. Success of matching was achieved in 198 of 258 (77%) variables used as confounding factors. However, such factors were general ICU prognosis factors and not CAP-specific prognosis factors. Consequently, even if case and control patients had similar characteristics on admission, we compared, to independently verify the adequacy of the matching for the severity of initial CAP, cases and controls adjusting on 11 previously identified prognosis factors of CAP. In contrast to the other studies that compared the mean of each variable in the two groups, 1,7the comparison was performed in our study for each case-control pair. This control of matching revealed only one significant differ-

17

ence between the two groups, which was the rate of effective initial CAP antimicrobial therapy. Such data emphasized, to our mind, the adequacy and the efficacy of our matching and suggested that ineffective initial CAP antimicrobial therapy was the only significant variable associated with the occurrence of HA-VAP. In our second analysis, to avoid bias caused by the confounding factors chosen, we carried out a study matching case and control patients on the basis of specific CAP predictors of outcome, collected on ICU admission, rather than on the basis of general ICU prognosis factors. Success of matching was achieved in 242 of 258 variables used. In the current literature, HA-VAP mortality ranges from 30% to 70%.2-4 In our study, the rate was 42%. In our series, independently of the confounding factors used, a significant difference in outcome was noted between patients developing HA-VAP and the matched control patients. Moreover, the mortality attributable to HA-VAP was similar in our two case-control studies, respectively, 28% (risk ratio = 3.0) in the “ICU matching” study and 26% (risk ratio = 2.57) in the “CAP matching” study. Such findings, observed after controlling for the other determinants of outcome, suggest strongly that the occurrence of HA-VAP increases the mortality of severe CAP. In the recent literature, among the studies using a matched cohort analysis, attributable mortality linked to HA-VAP varies widely from one study to another. Leu et alz4did not find any significant difference in mortality between case and control patients, whereas Craig and Connellyz5 demonstrated a higher rate of mortality in HA-VAP. However, in these two latter studies, both ventilated and nonventilated patients were included. Moreover, in pairs in which the case and control patients were both mechanically ventilated, the mortality was not reported. Fagon et al7 in a study featuring 48 pairs of patients, a mortality rate of 54.2% was reported in patients developing HA-VAP, whereas the rate in patients without HA-VAP development was 27.1% Therefore, the attributable mortality was 27.1% and the death risk ratio was 2.0. On the other hand, in a matched cohort study including 85 pairs of patients carefully matched for the usual confounding factors, Papazian et al’ found no higher mortality linked to HA-VAP. Mortality rate was similar in both case (40%) and control (38.8%) patients. However, in the subgroup of patients with an initial medical problem, they observed an in-

LEROY

18

creased mortality rate when HA-VAP occurred (55%), compared with the control group (41%). Nevertheless, this difference did not reach statistical significance, possibly because of the small number of patients (22 pairs) in this group. In these two previous studies, it must be emphasized that the population studied included a large number of surgical patients, respectively 60% and 75%. In a specific study of intubated trauma patients, based on a retrospective case-control analysis, Baker et al6 demonstrated the same mortality rate (24%) whether HA-VAP occurred or not. Finally, in a study based on a multivariate analysis of the prognosis, Kollef et alz6 showed that the occurrence of HA-VAP did not increase mortality in patients ventilated for more than 5 days. One possible explanation of the discordant results observed in these well-conducted studies was the inclusion of patients exhibiting different initial illnesses whose final outcome varied largely even if initial presentation or initial severity indices were similar. Consequently, studies focusing on patients exhibiting similar initial disease seem inevitable to draw firm conclusions about the attributable mortality linked to HAVAP occurring in such patients. Those considerations were the basis of our study, which included a homogeneous population of patients exhibiting the same initial illness, in this case severe CAP requiring ICU admission and mechanical ventilation. However, some limitations of our study must be addressed. First, we chose to pair cases and controls on the basis of severity of illness on admission and on the basis of the duration of exposure to risk. Initial microbiological diagnosis of CAP and daily evaluation of variations in illness severity and therapeutic activity between admission and the day of the diagnosis of HA-VAP were not taken into account for matching. The comparison of cases and controls about such data exhibited only a few significant differences. One of the most important was the higher incidence of ineffective initial CAP antimicrobial therapy in patients exhibiting HA-VAP. This finding suggests that ineffective initial CAP

ET AL

antimicrobial therapy is one of the factors associated with the occurrence of HA-VAP. Second, the criteria to diagnose HA-VAP used the combination of usual clinical, radiological, and microbiological data. In 7 out of our 43 studied patients, no precise microbiological diagnosis was performed. In 28 out of 36 cases with a precise microbiological documentation, the causative organism was isolated from quantitative culture of specimens obtained by endotracheal aspiration. Despite controversies that may surround our diagnostic tools, some recent results suggest their appropriateness. The specificity and sensitivity of quantitative cultures of specimens obtained by endotracheal aspiration appear, in recent publications, similar to those obtained by other techniques, such as protected specimen brush or bronchoalveolar lavage.27,28 Moreover, some studies29,30founding a weak relationship between lung histopathology and microbiological biopsy cultures suggest that none of the available quantitative cultures had a reliable positive predictive value for histological pneumonia. Finally, the small number of patients collected in our series could represent another limitation. In comparison with previous studies featuring from 29 to 85 pairs,‘%‘js7 we matched only 43 pairs. However, in contrast to these studies, which included without distinction a general population of patients admitted to the medical-surgical ICU, we studied a selective population of patients, in this case, patients exhibiting severe CAP requiring mechanical ventilation. In conclusion, our report suggests that, when confounding factors are controlled, the occurrence of HA-VAP causes increased mortality of severe CAP. Consequently, prevention and, perhaps, early diagnosis and adequate treatment of HA-VAP in patients exhibiting severe CAP appear as major therapeutic goals to improve the prognosis of such patients. ACKNOWLEDGMENT The writers thank Terence writing of this manuscript.

Moran

for his collaboration

in the

REFERENCES 1. Papazian L, Bregeon F, T&ion X, et al: Effect of ventilatar-associated pneumonia on mortality and morbidity. Am .I Respir Crit Care Med 154:91-97, 1996 2. Torres A, Aznar R, Gate11 JM, et al: Incidence, risk, and prognosis factors of nosocomial pneumonia in mechanically ventilated patients. dm Rev Respir Dis 142523-528, 1990

3. Fagon JY, Chastre .I, Domart Y, et al: Nosocomial pneumonia in patients receiving continuous mechanical ventilation. Am Rev Respir Dis 139:877-884, 1989 4. Kollef MH: Ventilator-associated pneumonia: A multivarate analysis. JAMA 270:1965-1970, 1993 5. American Thoracic Society: Hospital-acquired pneumonia

EFFECT

OF HA-VAP

ON COMMUNITY-ACQUIRED

PNEUMONIA

in adults: Diagnosis, assessment of severity, initial antimicrobial therapy, and preventative strategies. Am J Respir Crit Care Med 153:1711-1725, 1995 6. Baker AM, Meredith JW, Haponik EF: Pneumonia in intubated trauma patients. Am J Respir Crit Care Med 153:343349, 1996 7. Fagon JY, Chastre J, Hance AJ, et al: Nosocomial pneumonia in ventilated patients: A cohort study evaluating attributable mortality and hospital stay. Am J Med 94:281-288, 1993 8. Leroy 0, Santri C, Beuscart C, et al: A five-year study of severe community-acquired pneumonia with emphasis on prognosis in patients admitted to an intensive care unit. Int Care Med 21:24-31, 1995 9. Leroy 0, Georges H; Beuscart C, et al: Severe communityacquired pneumonia in ICUs: Prospective validation of a prognostic score. Int Care Med 22: 1307-13 14, 1996 10. Leroy 0, Devos P, Bosquet C, et al: Simplified prognostic score for patients with severe community-acquired pneumonia, in Program and Abstracts of the 37th Interscience Conference on Antimicrobial Agents and Chemotherapy. Toronto, American Society for Microbiology, Washington, DC, K143, 1997 (abstr) Il. Fine MJ, Smith DN, Singer DE: Hospitalization decision in patients with community-acquired pneumonia: A prospective cohort study. Am J Med 89:713-721, 1990 12. Leroy 0, Vandenbussche C, Coffnier C; et al: Community-acquired aspiration pneumonia in intensive care unit: Epidemiological and Prognosis data. Am J Respir Crit Care Med 156:1922-1929, 1997 13. McCabe WR, Jackson CG: Gram-negative bacteremia: Etiology and ecology. Arch Intern Med 110:847-855, 1962 14. Le Gall JR, Loirat P, Alperovitch A: Simplified acute physiologic score for intensive care patient. Crit Care Med 12:975-977, 1984 15. Knaus WA, Draper EA, Wagner DP, et al: Prognosis in acute organ system failure. Ann Surg 202:685-693, 1985 16. Bone RC, Fischer CJ, Clemmer TP, et al: The methylprednisolone severe sepsis study group. Sepsis syndrome: A valid clinical entity. Crit Care Med 17:389-393, 1989 17. Teasdale G, Jennet B: Assessment of coma and impaired consciousness. Lancet ii:81-84, 1974

19

18. Mier L, Dreyfuss D, Darchy B, et al: Is penicillin G an adequate initial treatment for aspiration pneumonia? Int Care Med 19:279-284, 1993 19. Marrie TJ, Durant H, Yates L: Community-acquired pneumonia requiring hospitalization: 5-year prospective study. Rev Infect Dis 11:586-599, 1989 20. Tran DP, Groenevald ABJ, Van Der Meulen J, et al: Age, chronic disease, sepsis, organ system failure, and mortality in a medical intensive care unit. Crit Care Med l&474-479, 1990 21. Bernard GR, Artigas A, Brigham KL, et al: The American-European consensus conference on ARDS. Am J Respir Crit Care Med 149:818-824, 1994 22. Pingleton SK, Fagon JY, Leeper KV: Patient selection for clinical investigation of ventilator-associated pneumonia: Criteria for evaluating diagnostic techniques. Chest 102:553S5563, 1992 (suppl) 23. SAS Institute: SAS/STAT guide for personal computers, version 6.1 edition, SAS Institute Inc, Cary, NC, 1995 24. Leu HS, Kaiser DL, Mori M, et al: Hospital-acquired pneumonia: Attributable mortality and morbidity. Am J Epidemiol 129:1258-1267, 1989 25. Craig CP, Connelly S: Effect of intensive care unit nosocomial pneumonia on duration of stay and mortality. Am J Infect Control 12:233-238, 1984 26. Kollef MH, Silver P, Murphy DM, et al: The effect of lateonset ventilator-associated pneumonia in determining patient mortality. Chest 108:1655-1662, 1995 27. Marquette CH, Copin MC, Wallet F, et al: Diagnostic tests for pneumonia in ventilated patients: Prospective evaluation of diagnostic accuracy using histology as a diagnostic gold standard. Am J Respir Crit Care Med 151:1878-1888, 1995 28. Torres A, El-Ebiary M, Padro L, et al: Validation of different techniques for the diagnosis of ventilator-associated pneumonia: Comparison with immediate post mortem pulmonary biopsy. Am J Respir Crit Care Med 149:324-331, 1994 29. Fabregas N, Torres A, El-Ebiary M, et al: Histopathologic and microbiologic aspects of ventilator-associated pneumonia. Anesthesiology 84:760-771, 1996 30. Kirtland SH, Corley DE, Winterbauer RH, et al: The diagnosis of ventilator-associated pneumonia. Chest 112:445457, 1997