INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During the last 20 yr, Legionella pneumophila has become widely recognized as a cause of community-acquired pneumonia (CAP). Recent studies regarding severe community-acquired pneumonia have shown that L. pneumophila is the second most common cause of admission to an ICU not far beyond pneumococcal pneumonia, which is the leading cause (1, 2). Since the frequency of community-acquired pneumonia (CAP) caused by Streptococcus pneumoniae is at least 5-fold greater than L. pneumophila pneumonia (3, 4), the latter seems more likely to result in severe pneumonia. More recently, cases of L. pneumophila pneumonia acquired in the hospital have been described both in endemic and outbreak forms (5). In this setting, Legionella pneumonia can also adopt severe presentations. Despite the fact that the epidemiology of CAP and nosocomial Legionella infection is well known, there are no specific reports concerning severe cases of L. pneumophila pneumonia admitted to intensive care units. Importantly, the empiric treatment of both nosocomial and CAP not always includes specific antibiotics against Legionella, and the current mortality rate of this type of lung infection is particularly high in patients who receive inappropriate empiric antibiotic therapy (6). We performed a prospective study upon a population with a reliable diagnosis of L. pneumophila pneumonia that required ICU admission in order to determine the prognostic factors, clinical, radiological and outcome variables of nosocomial and community-acquired cases of L. pneumophila pneumonia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Population

Between January 1983 and December 1991, all patients admitted consecutively to three ICUs of the three participant hospitals with the diagnosis of Legionella pneumonia (n = 84) were studied. Patients were admitted to the ICU because of severe acute respiratory failure and/or shock. Fifty-one patients had community-acquired Legionella pneumonia and 33 nosocomial Legionella pneumonia. Among the 33 patient with nosocomial acquired Legionella pneumonia, nine (27%) acquired the infection in the ICU.

Diagnosis of Pneumonia

For the purpose of the study, the inclusion criteria for pneumonia were the presence of new chest radiographic infiltrates on admission, and, at least, one of the major or two of the minor criteria given heretofore (7). Major criteria were cough, sputum production, or history of fever ( 37.8° C); minor criteria were dyspnea, pleuritic chest pain, pulmonary consolidation on examination, and leukocyte count > 12,000/ ml. In mechanically ventilated patients the clinical criteria for defining pneumonia were: presence of new chest radiographic infiltrates for > 48 h, leukocytosis or leukopenia, fever, and purulent secretions (8). These patients were followed up subsequently and noninfectious origin of infiltrates were discarded. Nosocomial Legionella pneumonia was defined when the acquisition was considered after 72 h of hospital admission. If a period of 15 d did not elapse since last hospital discharge, pneumonia was considered nosocomial as well. ICU acquired pneumonia was defined as the episode acquired after 72 h from admission to an ICU.

Data Collection

The following variables were recorded in all cases: age, gender, APACHE II score at admission, PA_O2/FI_O2 ratio on admission, presence of prior antibiotic treatment, underlying illness, immunosuppression, neoplasia, smoking habit, alcohol intake, presence of septic shock, presence of pleural effusion, need for mechanical ventilation, chest radiographic involvement on admission and discharge, whether the pneumonia was nosocomial or community acquired, serum creatinine, sodium, potassium, and urea levels, leukocyte count, differential white cell count, whether the diagnosis of Legionella pneumonia was direct or not, presence of other infections, radiographic progression of pneumonia or improvement, adequacy of initial ICU antibiotic treatment against Legionella, early specific treatment, and crude mortality rate.

Definitions

Chronic obstructive pulmonary disease (COPD) was diagnosed combining clinical criteria and the coexistence of an obstructive ventilatory impairment assessed either before or after ICU stay. Immunosuppression was considered present if patients were receiving chemotherapy and/or corticosteroid-equivalent prednisone dose  20 mg/day for more than one month. Patients were considered smokers if they had smoked more than 10 pack-years. Alcohol intake was considered present when the ingestion was  80 g/day. Chest radiographs were read independently at least by two of the authors and also by the attending physician. The following features were considered: unilateral or bilateral involvement, abscess formation and pleural effusion. Radiographic progression was defined as an increase in the size of the opacity by 50% or greater within 48 h of admission (2).

Improving pneumonia was assessed through chest radiographic improvement during admission, the latter defined as a decrease of the opacities in > 50% in relation to the admission radiograph. Early specific empiric treatment was defined as starting antibiotic treatment which covers Legionella (e.g., erythromycin or rifampicin) as soon as the diagnosis of pneumonia was established.

Prior antibiotics was considered present if the patients received antibiotics during the preceding 48 h before their inclusion into the study. Empiric antibiotic treatment was initiated according to the attending physician's choice. For the purpose of the study, antibiotic treatment was considered adequate if the empiric drugs chosen followed the present recommendations of the literature, were changed according to the antibiotic susceptibility reports and the timing, route, dosage, and duration of therapy were considered correct (2). Delay of treatment was arbitrarily considered when there was a deferral of more than 72 hours in starting adequate antibiotic treatment against Legionella.

Septic shock was defined as a systemic inflammatory response to infection (presence of two or more of the following: temperature > 38° C, heart rate > 90 beats/min, respiratory rate > 20/min or Pa_CO2 < 32 mm Hg, and leukocyte count > 12,000/mm³ or > 10% band forms), in addition to hypotension (systolic blood pressure < 90 mm Hg or a reduction of > 40 mm Hg from baseline in the absence of other causes). Other causes of shock, such as hypovolemia or cardiogenic shock, were not considered (9).

Renal failure was defined as a serum creatinine level greater than 2 mg/dl and/or blood urea nitrogen of more than 30 mg/dl. Empyema was defined as the presence of pus in the pleural cavity as a consequence of infection, or the isolation of bacteria from pleural fluid. Direct diagnosis was considered if the diagnosis Legionella pneumonia was established as the first cause of pulmonary infiltrates.

Etiologic Testing

Two serial blood cultures and serologies were performed in all patients. Sera were collected on hospital admission, 3 w later, and 2 mo after discharge and tested for evidence of complement fixing antibody to influenza A and B, parainfluenza, adenovirus, Respiratory Syncytial Virus, IgG against Mycoplasma pneumoniae ( 1/512), Chlamydia psittaci ( 1/64), and Coxiella burnetti ( 1/64). Indirect microimmunofluorescence antibody technique was used for detecting IgG ( 1/ 512) and IgM ( 1/32) against C. pneumoniae. L. pneumophila (serogroups 1-6) was diagnosed using indirect microimmunofluorescence antibody technique IgG. Respiratory samples were obtained whenever possible from all patients.

Definite Diagnosis of Legionella Pneumonia

The diagnosis of L. pneumophila was considered as definite if the patient met one or more of the following criteria: (1) Isolation of L. pneumophila from a respiratory sample culture; (2) a 4-fold IgG titer rise with final titers for L. pneumophila IgG  1/128, and (3) Presence of a single IgG titer  1/256. L. pneumophila (serogroups 1-6) was diagnosed using indirect microimmunofluorescence antibody technique. A positive direct immunofluorescence test for Legionella was considered as a possible criterion for infection (> 5 cells/1,000×). In that case, the diagnosis of L. pneumophila was subsequently confirmed by culture on buffered charcoal yeast extract (BYCE-) or serology. Methods used to diagnose Legionella pneumonia are shown in Table 1. All the diagnostic procedures were carried out shortly after admission to the ICU or after diagnosis of pneumonia in patients previously admitted with other diagnoses. Sputum, protected specimen brush, transthoracic needle aspiration, pleural fluid and bronchoalveolar lavage samples were plated on the following media: blood agar, chocolate agar, Sabouraud agar, buffered charcoal yeast extract (BYCE-  ), thioglycolate broth, and Centers for Disease Control (CDC) medium for anaerobes. Laboratory diagnosis of pulmonary infection was done according to previous standardized reports (10).

Statistical Analysis

The primary end-point of the present study was to investigate the ICU mortality and prognosis factors. We also investigated the differences between nosocomial and community-acquired severe cases of L. pneumophila as regards clinical and radiographic features. For the comparison of means the nonparametric Mann-Whitney test was used for continuous variables when they departed from a normal distribution, otherwise, Student's t-test was used; and the chi-square (Fisher's exact test when needed) for discrete data. For the multivariate analysis, the unconditional logistic regression technique was applied to adjust for confounding variables for assessing prognostic factors. Common pitfalls associated with multivariable regression were avoided as described by Concato and coworkers (11). Continuous variables were categorized. In order to optimize the threshold that would discriminate two risk groups for these variables, the formula for threshold computation based on the median value (50th percentile) devised by Robert and colleagues was used (12). The relative risk for outcome (survival/death) was estimated for the following all variables: blood urea nitrogen (< 30 or  30 mg/dl), PA_O2/FI_O2 ratio < 130 or  130, serum Na levels (> 136 or  136 mEq/L), mechanical ventilation (yes/ no), septic shock (yes/no), liver diseases (yes/no), immunosuppression (yes/no), chronic renal failure (yes/no), age (< 50 or  50 years), COPD (yes/no), initial adequate antibiotic treatment (yes/no), improving pneumonia (yes/no), APACHE II score (< 15 or  15). Variables were selected according to their univariate analysis p value (< 0.10), or because they were deemed biologically plausible. Interactions were tested in the model; variables with association among each other were not included in the analysis. As for influential observations, no patients with outlier values in any variable were detected. All reported p values are two-tailed. The BMDP software package (BMDP Statistical Software; University of California Press, Berkley, CA) was used for the analysis.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

L. pneumophila pneumonia was detected in 84 of 3,283 patients admitted to all three participating ICUs consecutively. This represented 2.6% of the total number of admissions. Pneumonia was deemed to be community-acquired in 51 (61%) and hospital acquired (or nosocomial) in 33 (39%) patients. The main causes of ICU admission were because of severe acute respiratory failure (n = 65, 77%), and septic shock (n = 19, 23%). The mean ± SD age of the study population was 59 ± 14 years. Male gender was more frequent (63 versus 21 females). Mean APACHE II score at admission for the whole population was 16 ± 7. The mean (± SD) length of ICU stay was 18 ± 22 days. Fifty-four (64%) patients needed mechanical ventilatory support. Their mean ± SD PA_O2/FI_O2 ratio was 156 ± 86.

Comparison Between Community-Acquired and Nosocomial Legionella Pneumonia

Table 2 shows the main characteristics and the underlying diseases in patients with both types of Legionella pneumonia. Importantly, patients with nosocomial legionellosis were more likely to have underlying chronic lung disease (64%) and other processes such as compared with the community-acquired group (41%) (p = 0.044). By the same token, those patients with underlying chronic renal failure, cardiac diseases, and alcoholism were more liable to acquire Legionella pneumonia in the hospital (21%, 39%, and 54% versus 4%, 10%, and 18%, respectively) (p < 0.05 for all comparisons). Patients with nosocomial Legionella pneumonia suffered more commonly, although not statistically significantly (p = 0.059), from septic shock (33%) as compared with the community-acquired group (16%). A higher proportion (57%) of community-acquired pneumonia showed bilateral extension of the chest X-ray infiltrates as compared with the nosocomial group (36%). No significant differences were found between hospital and community-acquired patients with Legionella pneumonia as regards age, gender, APACHE II scores, PA_O2/FI_O2, prior antibiotics, diabetes mellitus, liver diseases, immunosuppression, neoplastic disease, smoking habit, pleural effusion, presence of mechanical ventilation, duration of mechanical ventilation, unilateral chest radiographic involvement on admission and discharge, bilateral chest radiographic involvement on discharge, and crude mortality rate.

Radiographic Findings

Chest radiographs on admission showed unilateral involvement in 40 patients (n = 20, 39% in the community-acquired and n = 20, 61% in the nosocomial; p = 0.055). On discharge, however, these rates became closer (21% in the community-acquired and 27% in the nosocomial; p = 0.54). On the other hand, bilateral chest X-ray involvement was more common in the community-acquired cases (n = 29, 57% and n = 21, 41%, on admission and discharge, respectively) as compared with nosocomial cases (12, 36% and 14, 42%, on admission and discharge, respectively), p = 0.066 and p = 0.9, respectively. Pleural effusion was radiographically evident on admission in 20 cases (13 community-acquired (25%), and seven nosocomial (21%), p = 0.65. Empyema was only detected in two patients; one from each group. Cavitation was not observed in any case.

Antiobiotic Treatment

Fifty-nine (70%) patients received prior antibiotics for their pneumonia before ICU admission. Fourteen (24%) of them died. The most frequent drugs given were: second and third generation cephalosporins together with aminoglycosides. Seventy-four (83%) patients received specific treatment with erythromycin; 19 (26%) of them died. However, patients who did not receive adequate antibiotic treatment against Legionella had poor prognosis compared with those receiving adequate treatment (six of 10, 60% versus 19 out of 74, 26%; p = 0.025). Among these 74 patients, in 10 (12%) cases the treatment was delayed (average 7 d) until the confirmation of diagnosis. No differences were found when comparing patients' outcome in relation to the delay of specific treatment (18, 28% versus 1, 10%; p = 0.22, respectively). Antibiotics given to all patients during ICU admission are shown in Table 3.

Etiological Diagnosis

Etiological diagnosis was achieved in all 84 patients. Serological testing was the method that accomplished more diagnoses (n = 29; 57% and n = 21; 64% from the community-acquired and nosocomial groups, respectively). Protected specimen brush cultures yielded L. pneumophila in four cases with CAP as compared with none from the other group. Diagnosis through bronchoalveolar lavage was confirmed in one case from each group. Sputum samples revealed the diagnosis of Legionella pneumonia in four of 51 (8%) community-acquired pneumonias and five of 33 (15%) nosocomial pneumonia. Table 1 shows the diagnostic methods used to diagnose the 84 episodes of Legionella pneumonia. No coinfection was observed in any case.

Prognostic Factors

The crude mortality rate in this study was 30% (24 of 84). Mortality rates of community-acquired pneumonia was 31% and 27% in the nosocomial type (p = 0.68). Among the nine patients who acquired the pneumonia in the ICU, three died (33%). Among all the variables recorded the following resulted as significant prognostic factors related to death in the univariate analysis: Chronic cardiac disease (RR: 6.8), diabetes mellitus (RR: 4.27), creatinine 1.8 mg/dl (RR: 3.95), septic shock (RR: 3.16) chest X-ray extension (RR: 3.07), mechanical ventilation (RR: 2.92), serum Na levels ( 136 mEq/L (RR: 2.63), PA_O2/FI_O2 < 130 mm Hg (RR: 2.34), and blood urea levels  30 mg/dl (RR: 2.23). On the other hand, the following two variables were related to better outcome: adequate treatment for Legionella (RR: 0.43), and pneumonia improvement (RR: 0.09). Table 4 shows the results of the bivariate analysis of prognosis. The logistic regression analysis demonstrated that serum Na levels  136 (RR: 21.3; 95% CI 1.11 to 408; p = 0.0237), and, APACHE II score > 15 at admission (RR: 11.5; 95% CI 1.75 to 76.1: p = 0.0251), were the only independent factors related to death. On the other hand, improving pneumonia is associated with better outcome in Legionnaires' disease than for patients not having improving pneumonia (RR: 0.02; 95% CI: 0.028 to 0.151; p < 0.0001). None of the remaining predictor variables improved the prognostic capacity of the model (Table 5).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The mortality of severe Legionella pneumonia in our study was 30%. The independent factors related to poor outcome were: hyponatremia  136 mEq/L, APACHE II score > 15, and absence of pneumonia improvement. The administration of adequate antibiotic treatment for Legionella with intravenous erythromycin was a factor of good prognosis in the univariate analysis. The clinical and radiographic presentations of both community-acquired and nosocomial pneumonia did not differ in our study. However, in the nosocomial type, patients had more COPD, chronic renal failure, cardiac disease, and alcoholism and presented more frequently unilateral chest X-ray involvement.

Overall, mortality of Legionella pneumonia is still high with a wide rate of variation. For instance, Marston and coworkers reported (13) a mortality rate of 24%. Nevertheless, this mortality rate accounted for different species and serogroups of Legionella spp. In a large study on community-acquired pneumonia, mortality due to Legionella pneumonia was 17% (7). Mortality of nosocomial Legionella pneumonia seems to be higher compared with community acquired, although there is variation among the reported series. For example, 18% in Pedro-Botet's series (14) versus 47% in Carratala's study (5). This variation, however, is difficult to explain, although it could be due to inclusion of immunosuppressed population in the latter. On the other hand, mortality of Legionella pneumonia admitted to ICU has been reported in few studies. For instance, Moine and colleagues (15) found a mortality rate of 25% as compared with 15% in another series (2). It is important to note that these rates were derived from only four cases in the first series, and five cases in the second. Our case fatality rate in both groups was 27% for nosocomial and 31% for community-acquired cases. These mortality rates are in keeping with the overall average rates of severe community-acquired pneumonia (2) and nosocomial pneumonia. However, both community-acquired and nosocomial pneumonia caused by high risk microorganismse.g., Pseudomonas aeruginosa present higher mortalities (2, 16).

Very few studies have focused on prognostic factors for Legionella pneumonia. Marston and coworkers (13) studying hospitalized and non-hospitalized patients with Legionella pneumonia identified the following factors as independently related to poor outcome: advanced age, male gender, hospital acquired infection, renal disease, malignancy, immunosuppression, and isolation of L. pneumophila serogroup 6.

The logistic regression analysis showed that APACHE II score > 15 at admission, and serum Na levels  136, were the only independent factors related to poor prognosis. On the other hand, improving pneumonia is associated with better outcome in Legionnaires' disease than for patients not having improving pneumonia. Of these factors, the most interesting and amenable of medical intervention is the presence of hyponatremia. Some years ago Yu and colleagues (17) described hyponatremia as a characteristic biochemical feature of Legionella pneumonia, although others did not find the same features (18). The mechanism of hyponatremia is yet undefined, but inappropriate secretion of antidiuretic hormone has been suggested. The relationship between hyponatremia and death is difficult to explain. However, hyponatremia could be a marker of systemic involvement in Legionnaires' disease.

The univariate analysis disclosed several factors related to poor outcome. Among these, several are well described in the literature: underlying condition, renal failure, bilateral chest X-ray involvement, septic shock, and need for mechanical ventilation. Other studies dealing with severe community-acquired or nosocomial pneumonia have also described these findings (2, 19, 20). An important finding in the univariate analysis was that inappropriate specific treatment for L. pneumophila was related to poor outcome. This finding has been also described in ventilator-associated pneumonia (21) and specifically in Legionella pneumonia (22). In a study on severe Legionella community-acquired pneumonia mortality increased from 10% to 27% if patients were not treated initially with erythromycin (23). The inclusion of erythromycin in all severe community-acquired pneumonia cases and in nosocomial pneumonia cases with specific risk factors for acquiring Legionella has been standardized (6, 24). Following strictly these guidelines, with special emphasis in areas with high prevalence of this microorganisms could reduce mortality of Legionella pneumonia.

Some studies have focused on differences between community-acquired and nosocomial types of Legionella pneumonia (25). As mentioned above some authors have found differences as regards mortality, although we did not observe similar mortality rates. In a recent study, Pedro-Botet and colleagues found that smoking habit, cough, thoracic pain, and extrapulmonary manifestations were more prevalent in community-acquired Legionella pneumonia (14). On the other hand chronic lung diseases and cancer were more prevalent in nosocomial legionellosis. In addition, oxygen and corticosteroid therapy and renal failure were also more frequent in the nosocomial group. In our study, we found that patients with nosocomial Legionella infection were more likely to have had chronic lung disease (64%) as compared with the community-acquired group (41%). By the same token, those patients with underlying chronic renal failure, cardiac diseases, and alcoholism were more liable to acquire Legionella pneumonia in the hospital. Our study shows also a clear predominance of unilateral chest radiographic involvement in the nosocomial Legionella pneumonia at the initial presentation. On the other hand, bilateral chest X-ray involvement on admission was more common in the community-acquired type.

The definite and most reliable method for the diagnosis of Legionella pneumonia is the isolation of L. pneumophila from respiratory secretions (26). A major disadvantage of any culture method is the time required for growth, which in the cases of Legionella can range from 3 to 14 d. Multiple plates containing selective dye-containing media, are more sensitive than using only BCYE- which was the case in our study. Direct immunofluorescence is a widely used test for the detection of Legionella from clinical specimens (sensitivity 43%, specificity 88%). Nevertheless, the serologic detection of antibodies is still the most widely used test. The delay inherent in serologic diagnosis constitutes its major disadvantage. The period of time needed to obtain a 4-fold titer rise usually ranges from 4 to 8 wk. Around 20 to 30% of patients suffering from Legionnaires' disease, however, will not develop an antibody increase (26). Therefore, the use of IgM assay improves the sensitivity of the test. An important characteristic of our study is that microbiologic diagnoses were carefully assessed and we only selected the cases with definite diagnosis of L. pneumophila pneumonia. Finally, urine antigen detection appears to have high specificity, but sensitivity needs to be confirmed in a prospective manner (27). Unfortunately, we did not perform this methodology in our study.

In summary, severe Legionella pneumonia is a cause of life-threatening admission to ICU. Furthermore, although studies have been performed regarding prognostic factors of community-acquired pneumonia, yet very little information is available as concerns this issue in Legionella pneumonia, especially in severe cases admitted to ICUs. This is the first study using a multivariate approach identifying prognostic factors for severe Legionella pneumonia. A better understanding of the prognostic factors in cases of severe Legionella pneumonia will optimize our therapeutic approach in this disease and help to decrease both its mortality and morbidity rates.