Factors associated with hospital mortality in community-acquired legionellosis in France

RESULTS

During the study period, out of 1,595 LD cases attributed to Lp1 and reported to the InVS, 540 patients hospitalised for CA LD (33.9%) were enrolled in this study; 395 were male and 145 female, with a mean age of 60 yrs (range 17–100 yrs).

All cases were confirmed in hospitals where they were included, by positive urinary antigen detection of Lp1 by immunochromatographic test, mainly BinaxNOW® (Inverness Medical, Galway, Ireland). In addition, positive cultures and seroconversion were reported in 93 (17.2%) and 66 (12.2%) cases, respectively. The 93 clinical isolates were distributed among 36 genotypes (STs). About 50% of isolates were accounted for by only four STs, with ST23 being the most frequent (23.7%). Other common STs included ST1 (7.5%), ST146 (6.5%) and ST47 (11.8%). mAb 3/1-positive isolates were predominant (87 isolates); only six (6.5%) isolates did not react to mAb 3/1 (four Olda isolates and two Bellingham isolates).

The main risk factors for acquiring LD were smoking (53.5%), alcohol intake (18.7%), diabetes (15.7%), cancer/malignancy (6.3%), corticosteroid therapy (5.7%) and immunosuppressive therapy (3.3%). 74% of patients had at least one known risk factor of LD, regardless of age. No known risk factor was identified in the 139 (27.7%) remaining patients.

44 cases died during hospitalisation, corresponding to an overall CFR of 8.1% (44 out of 540). Four patients died within 24 h of diagnosis. Compared to LD cases reported to the InVS but not entered in this study, those included (n=540) did not differ statistically by sex, age or geographic distribution.

Interestingly, the CFR of culture-confirmed cases (15.1%, 14 out of 93) was greater than that among patients in whom the strain was not identified (6.7%, 30 out of 447; p=0.012).

Death was attributed solely to LD in 17 (38.63%) out of 44 patients and to LD complications in 27 (61.36%) out of 44 patients. Reported complications included: cardiovascular complications (n=10), respiratory complications (n=8), malignancy (n=3), shock (n=3), multiorgan failure (n=2), neurological complications (n=2), renal complications (n=1) and bacteraemia (n=1); some patients may have suffered more than one complication.

Demographic, clinical, biological and radiographic characteristics were compared between surviving and nonsurviving patients; nonsurvivors were older (p<0.001), more frequently female (p=0.02) and more often in the PSI high-risk group at admission (p<0.001) (table 1).

Corticosteroid therapy was significantly more frequent in nonsurvivors (p=0.03), whereas, surprisingly, smoking was significantly more frequent in survivors (p<0.001). Other potential risk factors, such as cancer/malignancy, diabetes and immunosuppression therapy, were not significantly associated with death.

Table 2 summarises the significant differences in clinical, biological and radiological characteristics in hospital survivors versus nonsurvivors.

Nonsurvivors were more often confused (p=0.02), and had dyspnoea (p=0.02), higher respiratory (p=0.03) and cardiac (p=0.02) frequencies, and lower median systolic (p<0.001) and diastolic (p=0.01) blood pressure. More frequently, they also had serum creatinine levels >160 μmol·L⁻¹ (p<0.001), serum CRP >500 mg·L⁻¹ (p=0.003) and higher neutrophil counts (p=0.009), but their lymphocyte levels were lower (p=0.001). They more frequently had multilobar or bilateral infiltrates (p=0.001) (table 2).

Survivors were clinically more symptomatic at admission, with a higher proportion having body temperature >38.5°C (p=0.001), chills (p=0.004), nausea (p=0.04) and headache (p<0.001).

Although only 14.4% (25 out of 174) of patients received appropriate empirical antibiotic treatment before hospital admission (most received β-lactams; data not shown), more than 70.8% (292 out of 412) and virtually all patients (99.8%, 537 out of 538) were given appropriate antibiotics upon admission (empirically) and after confirmation of LD diagnosis, respectively (table 3). The median time to appropriate antibiotic treatment was shorter in nonsurvivors (3.0 versus 4.0 days; p=0.1), but the difference was not significant.

ICU admission, the need for mechanical ventilation and in-hospital complications (respiratory superinfection, renal and hepatic failure, and decompensation of pre-existing illness) were more likely to occur in nonsurvivors (table 3). As expected, patients in the high-risk group based on PSI (classes IV and V at admission) were at significantly higher risk of death (p<0.001) (table 1).

The log-rank test indicated that the probability of patient survival during the first 30 days after LD diagnosis was associated with male sex (p=0.01), age <60 yrs (p=0.02), and the presence of general (p=0.006) and digestive (p=0.05) symptoms, whereas ICU admission (p<0.001) and PSI class IV–V (p=0.004) were significantly associated with a lower probability of survival (table 4 and fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1–

Kaplan–Meier survival curves in patients with Legionnaires’ disease (LD) hospitalised in France (from April 1, 2006 to June 30, 2007). Survival was calculated from the time at which diagnosis of LD was confirmed until 30 days of hospitalisation. Survival according to a) sex (log-rank p=0.011) and b) intensive care unit (ICU) stay (log-rank p<0.001) is shown.

Independent factors linked with mortality in multivariate analysis (table 5) were age (per 10-yr increment; p<0.001), female sex (p=0.03), ICU admission (p=0.001), renal failure (p=0.003), corticosteroid therapy (p=0.04) and elevated CRP level (p=0.04).

DISCUSSION

To the best of our knowledge, this study of 540 patients is the largest prospective, hospital-based investigation to describe the clinical and biological features and outcome of confirmed CA LD in our modern era. For example, the biological parameters or the proportion of in-hospital complications among LD patients have rarely been documented in such a case series, and we consider our analysis to be a helpful description of sporadic LD cases. All diagnoses were confirmed by urinary antigen testing, a more reliable method than serological screening. However, reported values for pooled test sensitivity and specificity were 0.74 and 0.99, respectively, in a systematic review and meta-analysis [15], and LD caused by mAb 3/1-negative Lp1 were significantly less frequently diagnosed by commercially available assays [14].

Our sizeable patient population, enrolled over a limited time period, is representative of a larger group of 1,595 patients reported nationwide during the same period, providing an accurate description of CA LD in patients hospitalised in France.

The CFR of 8.1% is consistent with the mortality reported in routine national surveillance of LD in France (11%) and elsewhere. The LD fatality rate ranged from 5–25% in immunocompetent hosts [16] to 33% in severe LD [17], 8.5% in a study by Dominguez et al. [7] and 12.9% in the more recent investigation by Jespersen et al. [18].

In our study, the CFR for culture-confirmed cases was significantly higher than the global CFR, which could be related to the variability of diagnostic methods (culture of lung samples could be prescribed more often in severe cases of LD), significant dissemination of Legionella in more debilitated patients, or more virulent strains.

Various significant demographic risk factors for poor outcome were identified, including corticosteroid therapy and cancer/malignancy. However, diabetes, immunosuppression medication and smoking were not associated with death.

The poorer prognosis of older patients, and of those presenting with dyspnoea, high respiratory frequency, confusion, low blood pressure or multilobar infiltrates is consistent with the classical prognostic factors encountered in hospitalised patients with CA pneumonia (CAP) [19]. Serum creatinine >160 μmol·L⁻¹ and CRP >500 mg·L⁻¹ were significantly linked with mortality, whereas hyponatraemia was not retained by multivariate analysis. Elevated CRP values have been associated with mortality in patients with CAP [20] or pneumococcal pneumonia [21].

PSI class IV–V, ICU admission and the need for mechanical ventilation were coupled with mortality, consistent with findings in a prospective study of 84 LD patients requiring ICU admission, where variables linked with poor outcome by univariate analysis included cardiac disease, diabetes mellitus, serum creatinine ≥1.8 mg·dL⁻¹, septic shock, radiographic pulmonary infiltrate extension, mechanical ventilation, hyponatraemia (serum sodium <136 mEq·L⁻¹), P_a,O2/inspiratory oxygen fraction ratio <130 and blood urea levels ≥30 mg·dL⁻¹. Acute Physiology and Chronic Health Evaluation II score >15 at admission and serum sodium levels ≤136 mEq·L⁻¹ were the only independent factors related to death [22].

Recently, several authors have compared different CAP severity of illness (SOI) scores, such as PSI, CURB-65 (confusion, urea >7 mmol·L⁻¹, respiratory frequency ≥30 breaths·min⁻¹, systolic blood pressure <90 mmHg or diastolic blood pressure ≤60 mmHg, and age ≥65 yrs), CRB-65 (confusion, respiratory frequency ≥30 breaths·min⁻¹, systolic blood pressure <90 mmHg or diastolic blood pressure ≤60 mmHg, and age ≥65 yrs), and observed that higher risk patients were better identified with CURB-65 and CRB-65 scores, whereas PSI better distinguished lower-risk patients [23,24]. Unfortunately, our study was not designed to compare SOI scores, and because some variables included in these scores were not collected, any retrospective calculation or comparison is impossible.

We found that symptomatic patients with body temperature >38.5°C, chills, myalgia, nausea and headache had a better survival rate, which indicated that earlier diagnosis is facilitated by a typical clinical picture of infection and leads to better patient management. It also emphasises the need to suspect LD in cases with an incomplete or atypical clinical picture.

Surprisingly, outcome was poorer in females than in males. The impact of sex on mortality from infections is controversial. In an observational cohort study conducted in surgical units, mortality was significantly greater in females with lung infection [25]. Female sex independently predicted mortality in multivariate analysis of a large cohort of mechanically ventilated patients [26]. In another large cohort of patients with severe healthcare-associated infection who were hospitalised in ICUs, mortality was significantly higher among females [27]. Inversely, male sex was predictive of mortality within 90 days of hospitalisation for pneumonia [28]. Finally, sex-related outcome in infectious disease studies produced inconsistent results, depending on the type of infection and the target population. However, some data converged towards increased fatality in females with pulmonary infections. Further specific investigations are needed to clarify this observation.

A short time period to appropriate antibiotic treatment is usually indicated to improve the outcome in LD patients. This has been described previously in LD patients admitted to ICUs. Administration of quinolone or erythromycin within 8 h of arrival in the ICU has been associated with better survival [16]. Mortality may increase from 10% to 27% in patients who do not receive adequate antibiotic treatment as part of empirical therapy on admission [29]. Surprisingly, in our study, median time to appropriate treatment after the onset of clinical signs of LD was shorter (but not statistically significant) for nonsurvivors.

Our investigation has some limitations. Patients were selected on the basis of positive Lp1 testing, which could have introduced a bias as we have no information on other serogroups, and because of limited sensitivity of the test, as discussed previously.

The number of deaths in the cohort was rather low, so we could only identify the most relevant risk factors. Selection bias may have prevailed because: 1) less severe cases of LD not requiring hospital care were not included; and 2) even if the study was not designed to monitor the quality of clinical practices, physicians may have been less likely to report fatal cases in such a study. Moreover, the survival rate did not take into account patients who may have died after the 30-day follow-up period. Patient survival was not calculated from the date of onset, but rather from the date of diagnosis. The date of onset is often difficult to obtain because of memory bias towards more severe initial symptoms. In addition, this information may be difficult to collect in patients admitted to the ICU. However, we performed multivariate analysis, including the time interval between onset and diagnosis, and found that it was not linked with survival (p=0.975), indicating that such a bias, if present, was very limited. Finally, some confounding variables may have been missed. Our analysis aimed to test a large number of variables and, consequently, increased the chances of a wrong conclusion. A lack of power could also be acknowledged for some factors, such as genotypes, obtained in a subsample of the cohort. Additional analyses are still required to better document the impact of genotype on case severity, and the time to initiation of appropriate therapy must be further explored.

In conclusion, age, female sex, ICU stay, renal failure, corticosteroid treatment and increased CRP level were significantly associated with mortality in hospitalised CA LD patients. Less symptomatic patients had a poorer outcome. LD remains a significant cause of pneumonia and the associated mortality can be high in weakened persons. The identification of high-risk groups in terms of mortality could increase practitioner awareness and contribute to its reduction. Specific, prospective LD studies, comparing different SOI scores, should be performed in this field.