Complex clinical and microbiological effects on Legionnaires’ disease outcome; A retrospective cohort study

Legionella pneumophila, especially serogroup 1 is an important cause of severe pneumonia which is usually acquired in the community and less commonly during hospitalization [1, 2]. Legionella pneumonia, also known as Legionnaires’ disease (LD), is characteristically a disease of middle age and elderly patients, and may be associated with a high mortality rate [3, 4]. Possible factors contributing to adverse outcomes include host susceptibility, appropriateness of therapy and possibly the virulence of infecting strain [5‐7].

Legionella spp. are resistant to beta-lactam antibiotics, which constitute the most commonly prescribed agent for patients with pneumonia. The recommended treatment for infections due to Legionella spp. includes macrolides, quinolones and tetracyclines [1]. Hence, appropriate antibiotic therapy may be often delayed, especially if Legionella is not a priori considered the etiologic agent. Another obstacle for the timely administration of appropriate therapy is delayed diagnosis. Respiratory culture-based diagnosis, the gold standard for diagnosis, requires specialized media and expertise, has a long turnaround time (7-14 days), and is not routinely performed in many clinical laboratories [8]. Serologic studies mostly require paired samples obtained two weeks apart [8] and largely have been abondoned in recent years.

The introduction of the urinary antigen tests (UAT) for detection of L. pneumophila serogroup 1, the most common etiologic agent of LD, has revolutionized LD diagnosis. In the US, between the years 2000–2009 UAT was used to confirm LD diagnosis in 97 % of cases, culture in 5 % and serology or direct fluorescent antigen testing in <1 % [9]. Similarly in Europe the average detection rate by UAT was 79 % and 10 % by culture [10]. UAT allows rapid and timely diagnosis and has been linked to decreasing mortality trends in the US [2]. The sensitivity of the current tests for L. pneumophila serogroup 1 is in the range of 70-90 % with high specificity (95–100 %) [8]. Some tests are able to detect other serotypes with lower sensitivity [8]. While being rapid and readily available in many laboratories, the increasing dependency on UAT for diagnosis of LD may result in substantial under-detection rate reaching up to 40 % of cases [8, 11]. Lower rates of culture-based diagnoses may limit our ability to study the full scope of LD epidemiology, perform outbreak investigation and study the possible impact of L. pneumophila phenotype (particularly non-1 serogroups) or genotype and various clinical variables on LD. [2, 8, 12]. The rapidly evolving molecular diagnostic methods are increasingly used for LD diagnosis but still are less informative than culture-based methods [13].

LD is a challenge to public health authorities worldwide [14]. The incidence in the US is on the rise and has reached 1.15 per 100,000 population in 2009 [9]. In EU countries the incidence has been stable in recent years. The overall notification rate in 2012 was 1.1 per 100 000 inhabitants with an associated 10 % case fatality rate [15]. In Israel, Legionella infection is an important, although uncommon cause of community-acquired and nosocomial pneumonia [16‐18]. In a recent national surveillance study the reported incidence was 0.5 to 0.9 cases per 100,000 population, corresponding to 38–63 annual case. The case fatality rate was 12.3 % [18]. The vast majority of cases (88 %) were documented by UAT, and only 7 % by culture [18].

In a large epidemiologic study from the US, Benin et al. [3] described the impressive trend of decreasing mortality from LD concomitant with the increasing use of UAT for diagnosis. However, when looking at each subgroup separately, the mortality trend was significant only among persons diagnosed by culture, but not among those diagnosed by UAT. It was argued that UAT may have selected persons with less severe illness or hastened the initiation of appropriate antibiotics.

We found that administration of appropriate antibiotics within the first two hospitalization days was negatively associated with fatal outcome, even after controlling for ICU admission. This finding is in agreement with a recent large study of LD from Spain, showing that inappropriate empirical antibiotic therapy was an independent risk factor for severe outcomes (e.g. ICU admission or death) [25]. Early coverage of Legionella has proven beneficial in a recent large prospective study of combination therapy for moderate to severe community acquired pneumonia [26]. This can be achieved by adherence to the Infectious Diseases Society of America/American Thoracic Society consensus guidelines for the empirical therapy of community acquired pneumonia recommending the administration of combination therapy including a beta-lactam antibiotic plus an agent covering atypical organisms [27]. Despite the controversy regarding the added value of an agent covering ‘atypical’ pathogens in the combination regimens [28‐33], its role in patients with LD has not been disputed [34]. Our institutional guidelines recommended combination therapy for empirical community acquired pneumonia therapy and specific anti-Legionella coverage for severe pneumonia. Unfortunately, the compliance with these guidelines was less then optimal. Only 56.3 % of the patients in the LD study group received appropriate antibiotic therapy on day 1.

All-cause 30-day in-hospital mortality rates in our LD study group reached 15.5 %. Comparison to mortality rates in other studies should be carried out with caution, taking into account differences in the patient mix, delay in appropriate antibiotic treatment and the methodology of outcome reporting. For example, in a recent large study including 215 hospitalized patients from a single center in Barcelona the in-hospital fatality rate was 6.1 % [25]. However, there were several important differences between the two studies: firstly, in the Spanish study the mean patients’ age was 58.2 years compared to 67.2 years in our study group. Secondly, patients with immunosuppressive conditions were excluded from the Spanish cohort. Thirdly, positive cultures were reported in only 22.7 % of the Spanish cohort compared to 40.8 % in our study group. Lastly, in-hospital fatality rate is not the same as all-cause 30-day mortality rate. The 2014 ECDC surveillance report [10] found a lower average case fatality rate of 9 % for cases acquired in the community and higher rates of 28 % for nosocomial cases. In addition, the majority of cases were diagnosed by UAT (79 %) and only 10 % by culture. On the other hand, in a recent national report from France [35] including 1,192 patients with culture-confirmed LD and a mean age of 59.4 years, the death rate of 18.6 % was closer to the rates reported in our study.

The proportion of serogroup 1 (80.0 %) among isolates obtained by culture in our study is similar to the global rates [10, 36]. The distribution of genotypes, with ST1 (50.1 %) and the OLDA/Oxford,ST1 phenon (23.3 %) the most frequently encountered isolates in our center, differed from many European localities. In a study from England and Wales [37] the leading genotypes from clinical isolates were ST47 (25.7 %) and ST37 (11.4 %), while ST1 accounted for only 4.8 % of isolates. However, ST1 was the leading genotype among environmental isolates (19.6 %). In another study from Belgium [38] the leading genotype among patient with community-acquired legionellosis was ST47, while ST1 was the most frequent isolate among nosocomial cases. In a recent French study the leading genotype was ST23 (19.8 %), while ST1 ranked third (9.1 %) [36]. Notably, in a study from Japan ST1 was the leading genotype (8.1 %) and OLDA,ST1 the most commonly encountered phenon [39].

Studies examining possible correlation between Legionella genotypes and clinical course or outcome of LD are scarce. In our study mortality rates among patients infected with ST1 was somewhat lower compared to the whole group of patients with positive cultures. Mortality rates were higher in patients with some other STs (100 % for STs 87 and nontypeable isolates and 50 % for ST 40), but the numbers in each group are too small for any generalization. A large epidemiologic study from France involving 1,090 patients and using pulsed field gel electrophoresis as the genotyping tool showed that mortality rates differed between genotypes, with Paris strain being most lethal [5]. Notably, the ‘Paris’ strain is in fact ST1 which was the most commonly encountered ST in our series. In a more recent study from the same group in France the following death rates were found: 30.9 % for ST1, 16.3 % for ST23, 13.2 % for ST47 and 18.6 % for all the other STs [35]. Although these studies suggest that ST1 is more virulent, however in our study ST1 was not associated with the highest mortality rates. One possible explanation is that less virulent strains may be associated with illness in immunocompromised patients and outcome in such patients may be grave due to delayed diagnosis or diminished immunity. Another possible explanation is that ST1 based on the SBT scheme may not be the same in all geographic localities. Using more discriminative next generation-based typing [40] in large epidemiologic studies may help to shed more light on this issue.

There are several limitations to our study. The study was of a retrospective nature, and may represent only a selected group of patients hospitalized with LD. For example, since UAT was not performed systematically in all admitted patients, it is possible that the treating physicians were more likely to look for LD in patients with more severe presentations. Such selection bias may result in higher ICU admissions and mortality rates. In addition, the study included a single center with a relatively small sample size of patients with positive cultures and genotype characterization.