Article Text

Download PDFPDF

Legionella bacteria in shower aerosols increase the risk of Pontiac fever among older people in retirement homes
  1. M Bauer1,2,
  2. L Mathieu3,
  3. M Deloge-Abarkan1,2,
  4. T Remen1,2,
  5. P Tossa1,2,
  6. P Hartemann1,2,
  7. D Zmirou-Navier1,2
  1. 1
    Nancy-University School of Medicine, Département Environnement-Santé, Vandoeuvre-lès-Nancy, France
  2. 2
    INSERM, Vandoeuvre-lès-Nancy, France
  3. 3
    Ecole Pratique des Hautes Etudes (EPHE), Laboratoire d’Hydro-climatologie Médicale Santé Environnement, Vandoeuvre-lès-Nancy, France
  1. Dr L Mathieu, Laboratoire d’Hydro-climatologie Médicale Santé Environnement, School of Medicine, 9 avenue de la Forêt de Haye, 54500 Vandoeuvre les Nancy, France; laurence.mathieu{at}


Background: 828 elderly subjects residing in nursing homes were followed up during 4 months to ascertain incidence of symptoms associated with Pontiac fever (PF) in a non-epidemic setting.

Methods: The exposure situation was inhalation of Legionella bacteria while showering. An audit of the hot water system in all institutions allowed ascribing each subject to a water quality area wherefrom one shower was sampled for Legionella assays at the end of the follow-up period. Legionella were detected in water and aerosols using the culture (CFU, colony forming units) and in situ hybridization (FISH) techniques.

Results: Among 32 Pontiac-like episodes, 29 cases complied with the operational definition of PF elaborated for this study. Incidence density was 0.11 case/person–year (95% CI 0.07 to 0.15). Water concentrations greater than 105 Legionella FISH/l and 104 Legionella CFU/l were associated with an increased risk of PF (respectively RR 2.23, p = 0.05 and RR 2.39, p = 0.11, with significant dose–response patterns: p for trend <0.04). The condition also seems associated with aerosol concentrations above 103 Legionella FISH/l of air. A significantly higher risk of Pontiac-like episodes (RR 6.24, 95% CI 2.12 to 18.38) was seen for elderly subjects receiving corticosteroid therapy.

Conclusion: The water and threshold values identified in this research could be used to inform guidance measures aimed at protecting institutionalised older people from Legionnaires’ disease. Immunosuppressive therapy in the same population group can significantly enhance susceptibility to Legionella bacteria.

View Full Text

Statistics from

Legionella may elicit infection with two clinical presentations: Legionnaires’ disease, an acute pulmonary condition that may take the form of a serious pneumonia, and Pontiac fever (PF), a milder form that mimics influenza.13 The epidemiology of PF is not as well characterized as Legionnaires’ disease because of its benignity and lack of specificity.4 The reasons why contaminated water may result in Legionnaires’ disease or in PF are still not well understood;3 5 however, episodes of PF may serve as an indicator for the risk of Legionnaires’ disease.

Similar to many countries, regulations in France deal with Legionella pneumophila in water only.6 However, Legionella infection results from inhalation of a contaminated aerosol, not from ingestion.7 8 No known relationship has been established between water and bacteria concentrations in aerosols. In the past, few investigations have been carried out on Legionella concentrations in aerosols generated by contaminated water.913 A recent study conducted in the framework of a Legionnaires’ disease outbreak in the north of France, reported airborne L pneumophila concentrations spanning from 3.3×102 to 5.4×103 culturable Legionella (colony forming units (CFU)/m3).14 To date, no dose–response relationship has been established among humans. Infectious or lethal doses for guinea pigs exposed through inhalation of L pneumophila were highly variable across studies.1524

The risk of PF associated with shower water and aerosols contaminated by Legionella among older people living in nursing homes has been assessed in the framework of an epidemiological study conducted in the Lorraine region, in the north-east of France.


Study population

This retrospective follow-up study took place in 34 nursing homes from 2003 to 2006. The study population included 828 elderly subjects who were volunteers and signed informed consents. They were followed-up during 4 months in each home. A set of predefined symptoms that have been associated in the literature with PF were registered daily by the nursing staff. Information was also collected on demographic characteristics, current and/or past smoking habits, relevant medical history (respiratory and immunity-related conditions, such as diabetes and cancer) and current prescription of immunosuppressive therapy and antibiotics.

Because taking a shower represented the key exposure determinant in this study, only volunteers who declared taking at least one shower a week were included in the study. Neurodegenerative disorders were not a cause of exclusion insofar as subjects could take showers alone or with the assistance of the staff. Data about showering were recorded by the nursing home staff (the day and the room where the shower was taken, including if volunteers used different shower rooms during the follow-up period). A trained nurse dedicated to the study surveillance periodically checked the quality of registers.

Case definition

In a previous paper based on preliminary data of this study, we proposed an operational definition for epidemiological investigation and surveillance of PF.4 According to it, the case definition of PF was based on two criteria: (1) occurrence within 3 days after a shower of at least one clinical symptom among headache, myalgia, fever and shivers, possibly associated with other “minor” symptoms; this event could last up to a maximum of 8 days (minimum 2 days); and (2) contamination of the water and/or aerosol in the shower enclosure by Legionella bacteria detected either by culture or FISH techniques. No serological or microbiological evidence of infection was required as part of the case definition for practical and ethical reasons because we could not take repeated sputum or blood or urine samples in this at-risk population. Some recent papers showed that during epidemics, blood and sputum samples for Legionella culture were negative,3 25 as was L pneumophila serogroup 1 urinary antigen.26 27

Exposure assessment

The methodology and detection techniques used to assess exposure are detailed elsewhere.28 Because all showers could not be sampled, for practical and financial reasons, the selected showers were chosen to be representative of exposure of a group of volunteers. To this end, hot water systems in each nursing home were evaluated in order to assess their criticity in terms of Legionella risk, and to choose the sampling points (showers) that best fitted the volunteers’ exposure. Samples were taken immediately after the 4 months follow-up period, meaning that exposure was assessed retrospectively, for evident ethical reasons. This study design was approved by the ethical committee of Nancy University Hospital.

One litre of hot water was collected in sterile bottles after about 7 minutes of water flushing.29 Airborne Legionella were collected using two bioaerosol samplers in parallel for each shower: a MAS-100 (impaction onto an agar medium, GVPC) and an SKC Impinger (impaction in liquid medium: sterile distilled water).28

Airborne and water Legionella spp were quantified by FISH using a mix of three probes validated for Legionella (LEG705, LEG226 and LEGPNE1).30 31 In addition, the culturable fraction of Legionella was evaluated according to the French standard method.32 Although only Legionella spp bacteria are identified using the FISH technique, culturable Legionella can be split according to species (L pneumophila or Legionella spp), as recommended by the French standard.32

Data analysis

A subject could take showers in locations with different microbial water qualities throughout the 4 months of follow-up, a feature that was accommodated through five exposure scenarios. Table 1 describes the key characteristics of the three scenarios that assessed exposure according to water Legionella data, and of the two scenarios for airborne Legionella. In order to study the dose–response relationship, the reference exposure level was chosen below the detection limit value, which was 6×103 Legionella cells/l of water for the FISH technique and 100 CFU/l for the culture technique, according to the analysed water volumes. In the aerosol, the FISH detection limit was 3.7×102 Legionella cells/l of air. Data were input with EPI INFO software and analysed with the SAS software.

Table 1 Exposure of elderly residents in nursing homes to Legionella bacteria in water and aerosols while showering: five tested exposure scenarios


Elderly population

The study included 828 subjects (29.9% men, 70.1% women) across 34 nursing homes. The total number of residents in the nursing homes was 2504, meaning that 33% complied with the study selection criteria and volunteered (participation rate ranging from 13% to 78% according to nursing homes). The average age was 81 years (standard deviation (SD) = 10; extreme values 46–102), women being slightly older than men (respectively 83 and 78 years).9 10 Non-smokers composed 81% of the population; 26% of subjects had neurological disorders and cognitive impairments (dementia, debility, Parkinson syndrome), 16% had diabetes and 35% had co-morbidities (renal, liver, non-cognitive neurological or psychiatric conditions). Antecedents did not differ between men and women, except previous (p<10−4) and current (p = 5×10−4) smoking habits and alcoholism (p = 2×10−4) that were more frequent among men.

Pontiac fever cases

Among these 828 subjects, 32 met the clinical definition of PF; they will hereafter be referred to as “PF-like events”. There were no cases of Legionnaires’ disease diagnosed during the study period. Contaminated shower water samples for Legionella spp were found among 29 of these 32 subjects, irrespective of the detection method (culture or FISH). Hence, these 29 subjects strictly comply with the study definition of PF.4 Women composed 2/3 of these cases. The average age was 82 years (SD = 9.3). The mean episode duration was 3.5 days and the average incubation period (ie the time elapsed after the most recent shower) was 2.1 days. Among the PF cases, the most frequent symptoms were fever (75.9%), cough (44.8%), diarrhoea (27.6%) and headache (17.2%) (table 2). The median number of symptoms per case was 2. The overall incidence density of PF in this population was 0.11 case/person–year (95% CI 0.07 to 0.15). Cumulative incidence over 4 months was 3.5%. Among these 29 cases, 11 were associated with culturable Legionella and 21 with FISH Legionella. In three instances, identifying bacteria under microscopic counting was unsuccessful due to a high background fluorescence noise. When only contaminated aerosols were considered, the number of cases reduced to 10 and only the FISH technique was able to detect the bacteria.

Table 2 Frequency of the clinical symptoms among the 29 cases of Pontiac fever

Exposure characterization

The key exposure variable was “to take a shower”. The average number of weekly showers was 1.1 and the average follow-up 122 days. None of the risks are significantly in excess with the culture method (table 3), but an increased risk of PF is suggested (although not significantly) for scenario 1, starting with a water concentration of culturable Legionella spp ⩾104 CFU/l (rate ratio, RR 2.39; p = 0.11). On average, 78.3% of the culturable Legionella bacteria present in hot water showers were L pneumophila. All exposure scenarios for Legionella spp in hot water measured by the FISH technique show an increased risk over a concentration of 105 Legionella cells/l, whereby subjects had at least twice a greater risk to develop PF than if exposure was lower (table 3), with significant dose–response patterns (p for trend  = 0.04 and <0.001 for scenarios 2 and 3).

Table 3 Relative risk of Pontiac fever associated with Legionella concentrations in the hot water of showers according to the exposure scenarios and the detection methods

Dealing with bioaerosols, a borderline significant risk is associated with concentrations of hybridized (FISH) Legionella over 104 Legionella cells/l of air according to exposure scenario 4 (RR 2.05; p = 0.1), and over 103 Legionella cells/l air with scenario 5 (RR 1.77; p = 0.1) (table 4).

Table 4 Risk of Pontiac fever according to Legionella concentrations measured by fluorescent in situ hybridization (FISH) technique on the shower aerosol samples collected by liquid impaction

Predisposing factors

In table 5, immunosuppressive therapy is associated with a 6.2-fold increase in the risk of PF (p = 0.01). Among women, a history of respiratory conditions is mildly associated with a 2.5 greater risk (p = 0.1), a result that does not stand for men nor for the total study population. Cognitive impairments seem to exhibit a borderline “protective” effect (RR 0.47; p = 0.12). This unexpected result is not due to cases taking a smaller number of showers than non-cases (p = 0.45). Also, the incidence of PF did not differ according to whether subjects took more or less than 18 showers (ie the median value) over the whole study span (RR 0.95). These putative risk factors were included into a multiple logistic model, with age, gender, immunotherapy, cognitive conditions and water Legionella concentrations. Only two showed independent effects: water concentration threshold of 105 Legionella cells/l detected by FISH (OR 2.3; p = 0.03) and immunotherapy (OR 4.0; p = 0.04) (data not shown).

Table 5 Risk of Pontiac fever according to different factors, by gender


The study relates to 828 elderly volunteers (average age 82 years, 2/3 women) residing in 34 nursing homes and followed-up during 4 months. Showering was the main route of exposure to Legionella, bacteria being measured in hot water and in the shower aerosol. Among 32 subjects who exhibited Pontiac fever-like symptoms, 29 were exposed to Legionella spp in the shower hot water. The overall incidence density of PF was 0.11 case/person–year (95% CI 0.07 to 0.15). Despite lack of specificity, the combination of symptoms we used to qualify episodes as “Pontiac-like” showed an association with contamination of the shower water: Legionella spp concentrations above 104 CFU/l (culture) (of which 2/3 were L pneumophila) and 105 hybridized Legionella/l (FISH) were associated with an increased risk to develop PF. An increased risk was also suggested when airborne Legionella concentrations exceeded 103 Legionella-cells/l of air (FISH).

Pontiac fever: characteristics, incidence and risk factors

The incidence of PF is underestimated, with fewer than 20 epidemics described during the last 40 years, undoubtedly because of a lack of appropriate surveys and reporting definition. The clinical symptoms described, heterogeneous across studies, are those of a pseudo-influenza syndrome: nauseas, myalgias, fever, shivers, headaches, cough.13 26 27 33 34 26 27 33 34Although no single clinical definition is accepted to date, our work uses for the first time an operational definition of PF associating clinical symptoms and Legionella measurements in water and aerosols.

No other study has evaluated the incidence of PF in a non-epidemic framework: in a nursing home population. In this setting, cumulative incidence is low (3.6%) compared with other studies; the latter were all performed during epidemics, with incidences ranging from 11% to 94%.1 35 Hence, it is difficult to compare such different data, both in exposure settings and population characteristics. Our results reflect a background noise of PF within a sensitive population. Further, the relatively benign symptoms that are associated with PF episodes can remain unnoticed, leading to the underestimation of incidence.25 We accounted for possible intercurrent epidemics during the follow-up. No influenza case was recorded in the nursing homes, nor did the French national institute for health surveillance report influenza epidemics during the study period.36

Analysis of risk factors showed that only cognitive disorders and corticotherapy were associated with Pontiac fever. No other studies have looked at the relationship between cognitive disorders and PF. Consequently, we cannot compare this finding and this requires further research. Our study also highlights a strong relation between corticotherapy and PF (RR 4.0; 95% CI 1.1 to 14.9) that had never been defined as a risk factor of PF, although already described during Pontiac epidemics.33 35

No other health condition registered in this study proved a risk factor for PF; in accord with previous studies, gender was not found to be associated with the condition.1 35 37 38 Studies on smoking habits are contradictory, some suggesting that smoking does not influence the risk of PF,1 37 38 in contrast with what has been described for Legionnaires’ disease,39 whereas other authors report relative risks ranging from 1.2 to 5.1.27 33 In accordance with the former studies, our results do not show an association between smoking and PF.

Aerosols from showers: a source of exposure to airborne Legionella

In France, the aim is to prevent L pneumophila proliferation and to limit concentrations under a threshold value of 103 CFU/l in hot water networks.6 Few studies have tried to detect airborne Legionella,9 13 40 but no relation between Legionella concentrations in water and counts of bacteria dispersed into aerosols has been yet established. As a consequence, the management of the hazard posed by Legionella faces serious limitations since inhalation of airborne Legionella is the route of exposure.

In our study, showering was the key exposure determinant. Other sources of Legionella dissemination, such as air-conditioning or surrounding cooling towers (up to a radius of 10 km) were looked for. None of the study nursing homes had air-conditioning during the study period. Cooling towers were present near most nursing homes, leading to potential sources of Legionella discharge; they could not be investigated, however. We nevertheless think that aerosol generated while showering is the main exposure route in these health facilities because (1) patients very seldom went out of their institution, preventing then from direct outdoor aerosols exposure, (2) some of the nursing homes that presented several PF cases were not surrounded by cooling towers, and (3) the plausibility for cooling towers to be associated with increasing concentrations of Legionella in shower water and aerosol is low.

For ethical reasons, Legionella concentrations were only carried out at the end of the follow-up, encompassing, in each institution, one measurement for hot water, and one for bioaerosols. Using this point information to assess exposure relies on the assumption of the stability of Legionella levels in the hot water networks and in the generated aerosols. Two nursing homes underwent a follow-up of hot water quality, one with a high level of contamination (18 days, 21 water samples), the other one with less severe contamination (5 days, 42 water samples). In the latter, no cultivable or hybridized Legionella was detected during the follow-up, whatever the point of shower (no graph presented). Figure 1 exhibits the results for the former (two showers). Shower point A was located close to the production of hot water; its temperature was on average 46.2°C. Shower point B was located further away from hot water production, with an average temperature of 50.1°C. This figure shows a variability of Legionella concentrations (FISH) about 1 log on average, which is close to the uncertainty of the detection method and in accord with microbial fluctuations usually found in water distribution systems.4143 One reason might be that all our water samples were taken after 7 minutes of flushing, meaning they represented water flowing in the drain, not merely water accumulated in the final shower pipe. Thus, the assumption of stability of the bacterial concentrations over the study period seems reasonable.

Figure 1 Repeated measurements of bacterial concentrations (total number of cells measured by DAPI staining, number of hybridized Legionella and number of culturable Legionella in the hot water of two showers (A and B) in one nursing home (follow-up of 18 days).

Another limitation of this study resides in the semi-ecological character of exposure assessment, since groups of older people were ascribed a ward or building estimate of Legionella exposure, after careful assessment of the hot water system homogeneity. This study design tends to lessen statistical power but does not bias point estimates when average group exposure is correctly represented by the measured concentrations.44 45

Airborne Legionella exposure related to Pontiac fever

The originality of this study is to use a 4 month follow-up of an older population living in nursing homes rather than use an epidemic situation. To our knowledge, this is the first study that has assessed Legionella exposure in a non-epidemic framework. Several scenarios were elaborated to assess exposure while showering. Scenarios with culturable Legionella detected in hot water showed no association with PF, except a non-significant risk increase above a culturable Legionella spp exposure threshold of 104 CFU/l (RR 2.4). Among them, 78.3% were Legionella pneumophila, the remaining 21.7% were non-pneumophila Legionella species that we have not been able to identify. This is in the scope of those described during PF epidemics, ranging from 9×104 CFU/l to 3×108 CFU/l (Legionella pneumophila).3 33 38

We chose to contrast increasing Legionella concentrations to a common “no exposure” level in a view to assess whether risks were augmented above a given threshold, information that would be meaningful in terms of risk management. A limit value of 103 culturable Legionella (CFU)/l of water has been set by the European Commission and the French Ministry of Health to prevent the risk of Legionnaires’ disease.46 47 Our results do not show an increased risk of Pontiac fever at this level, but, putatively, beyond 104 Legionella spp (CFU)/l of water. Because our population is composed of elderly subjects, one might infer that this Legionella level of 103 CFU/l of water protects the general population against the risk of PF. Whether this 103 Legionella CFU/l threshold also offers a margin of safety for Legionnaires’ disease warrants further evaluation. To assess this hypothesis one would need, among others, information about the differential pathogenesis of PF and Legionnaires’ disease.

These Legionella thresholds are established on culture-based approaches. Even if culture is the standard method for environmental samples, a number of studies have proved that it underestimates the true number of viable microorganisms in environments, including aerosols.28 31 The FISH technique, as an alternative to the culture technique, was used to detect Legionella. An excess of PF fever incidence was obtained above 105 hybridized Legionella cells/l of hot water, as it was above 104 CFU/l for culturable Legionella spp. The 1 log difference between the values for FISH and culture is not surprising and highlights that (1) environmental bacteria have a low culturability, (2) a fraction of the Legionella population in hot water are physiologically active (ie have enough intact rRNA to be detected by FISH), but are not culturable within 10 days. FISH might yield useful information on the Legionella physiology even if it is not considered as a viability marker.28

Of interest is the finding that the incidence of PF might increase with airborne Legionella spp concentrations above 103 Legionella cells/l of air (p for trend  = 0.14, table 4). This analysis, strictly based on cases associated with aerosolized Legionella, backs the view that inhalation of aerosol is the true route of exposure.

To conclude, this study is the first one, to our knowledge, that quantifies the risk related to airborne Legionella produced by shower aerosols. According to the study definition of PF, it suggests excesses of PF incidence for Legionella spp concentrations above 104 CFU/l and 105 hybridized Legionella cells/l (FISH) of water and, more putatively, above 103 hybridized Legionella cells/l of air (FISH). However, extrapolation of these findings to Legionnaires’ disease or to the general population should be proposed with caution. They need confirmation in various settings and for different populations.

What this study adds

  • Epidemiology of Pontiac fever (PF) is not well characterised and the reasons why Legionella contaminated water may result in Legionnaires’ disease or in PF are still not well understood.

  • There is a risk of PF in a non-epidemic framework in a nursing homes population to airborne Legionella produced by shower aerosols.

  • The PF incidence increased for Legionella spp concentrations above 104 CFU/l (measured by culture) and 105 hybridized Legionella cells/l (measured by in situ hybridization, FISH) of water, and, more putatively, above 103 hybridized Legionella cells/l of air (FISH).

  • A significantly higher risk of Pontiac-like episodes was seen for elderly subjects receiving corticosteroid therapy.


This study was funded by AFSSET (French Agency for Environmental and Occupational Health Security), the French Ministry of Health and Veolia Environnement. M. Deloge-Abarkan was recipient of a doctoral scholarship from ADEME and EDF Medical studies department. Enric Robine, from CSTB, the French Building Research Center, contributed to the design of the aerosol exposure assessment part of the study. The water system critical point analysis procedure was elaborated by Veolia Environnement. Lory Mouchot and Lahoucine Benamghar, from INSERM ERI 11, participated to the statistical analysis. This study could not have been undergone without the help of the local public health and medical authorities and without the participation of the many elderly persons who volunteered. We also thank the staff in the nursing home; they cannot all be cited. Brigitte Risse and Maryse Fiorani, from the Clinical Investigations Center of Nancy University Hospital (coordinator: Anca Radauceanu) were responsible of the follow-up; their role was pivotal.


View Abstract


  • Funding: This study was funded by AFSSET (French Agency for Environmental and Occupational Health Security) (Grant number: RD-2002-015 and RD-2003-009), the French Ministry of Health and Veolia Environnement. Magali Deloge-Abarkan was recipient of a doctoral scholarship from ADEME and EDF. These funding sources were not involved in the study design, in the collection, analysis and interpretation of the data, or in the decision to submit the paper for publication.

  • Competing interests: None.

  • Ethics approval: Ethics approval was obtained.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • In this issue
    Mauricio L Barreto