Trends in Legionnaires Disease, : Declining Mortality and New Patterns of Diagnosis

Transcription

1 MAJOR ARTICLE Trends in Legionnaires Disease, : Declining Mortality and New Patterns of Diagnosis Andrea L. Benin, a Robert F. Benson, and Richard E. Besser Respiratory Diseases Branch, Division of Bacterial and Mycotic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia New diagnostic tests and empirical therapy for pneumonia may have important ramifications for the identification, treatment, and control of legionnaires disease (LD). To determine trends in the epidemiology of LD, we analyzed data for from the passive surveillance system of the Centers for Disease Control and Prevention. During this time period, there were 6757 confirmed cases of LD (median annual number, 360 cases/year). Diagnosis by culture and by direct fluorescent antibody and serologic testing decreased significantly; diagnosis by urine antigen testing increased from 0% to 69%. The frequency of isolates other than Legionella pneumophila serogroup 1 (LP1) decreased from 38% to 4% ( P p.003). The case-fatality rate decreased significantly, from 34% to 1% ( P!.001) for all cases, from 46% to 14% ( P!.0001) for nosocomial cases, and from 6% to 10% ( P p.05) for community-acquired cases. LD-related mortality has decreased dramatically. The decrease in culture-based diagnosis limits the recognition of non-lp1 disease and impairs outbreak investigation, because fewer Legionella isolates are provided for further examination. Legionella was first recognized as a pathogen in 1977, when Legionella pneumophila was identified as the agent responsible for an outbreak of severe pneumonia among attendees at the American Legion convention in Philadelphia in 1976 [1]. To understand the risk factors for infection, trends in disease incidence, and opportunities for prevention, the Centers for Disease Control and Prevention (CDC) began conducting passive surveillance for legionellosis in Previous analyses of CDC surveillance data have documented that elderly individuals, smokers, and people with underlying respiratory or immunocompromising Received 8 February 00; revised 4 June 00; electronically published 14 October 00. a Present affiliation: Robert Wood Johnson Clinical Scholars Program, Yale University School of Medicine, New Haven, Connecticut. Reprints or correspondence: Dr. Richard E. Besser, MSC-3, 1600 Clifton Rd., Atlanta, GA (rbesser@cdc.gov). Clinical Infectious Diseases 00;35: This article is in the public domain, and no copyright is claimed /00/ conditions are at increased risk for legionnaires disease (LD) []. More than one-half of the 146 species of Legionella and 68 distinct serogroups identified have been linked to disease [3 6]. The vast majority of reported cases are caused by L. pneumophila. The last analysis of data from the CDC surveillance system included cases from 1980 to 1989 []. Since then, several important changes in the diagnosis of infections with Legionella and treatment of pneumonia have taken place. These changes have had a direct impact on our ability to identify cases of LD and may have an impact on mortality from LD. A urine antigen test for the diagnosis of LD caused by L. pneumophila serogroup 1 (LP1) was made available in the early 1980s [7 9], and its use has increased considerably in recent years. This test is highly sensitive and specific, and it allows rapid diagnosis of LD. Rapidity of diagnosis is an important advantage of the urine antigen test, because it means that cases can be detected early in the course of infection, when treatment decisions can be affected, unlike diagnosis of LD by other modalities. Disadvantages of the urine antigen Trends in Legionnaires Disease CID 00:35 (1 November) 1039 Downloaded from on 08 April 018

2 test are that non-lp1 strains are not identified and that isolates of Legionella are not obtained, which impairs public health investigations of outbreaks of LD. During the 1990s, macrolide and fluoroquinolone antibiotics were increasingly recommended for empirical treatment of community-acquired pneumonia [10, 11]. Unlike b-lactam agents, these drugs provide good coverage for infections with Legionella. Thus, it is likely that these agents have been used as initial therapy with increasing frequency in treating patients with community-acquired pneumonia resulting from LD. To determine whether use of the urine antigen test has replaced other diagnostic methods and whether this change has led to a shift in reported cases caused by non-lp1 strains, we reviewed surveillance data for In addition, we analyzed trends in the case-fatality rates for LD during this period of changing prescribing patterns, to see whether there has been any change in mortality due to LD. SUBJECTS, MATERIALS, AND METHODS Surveillance for legionellosis. Voluntary passive surveillance for legionellosis has been carried out in the United States since State health departments submit a -page case-report form for each suspected case of LD. No isolates or other specimens are submitted. Although the case-report form has been modified several times, information about patient age, sex, race, travel history, specific laboratory test results for Legionella infection, and the species and serogroup of the infecting pathogen has been consistently collected. Information on outcome and hospital exposure has been collected, although the wording and responses have changed. As of 1999, reporting of LD was mandatory in all states except Oregon and West Virginia [1]. In 00, Oregon also began to require reporting of legionellosis. Definitions. We defined a case patient as a person with pneumonia that was diagnosed by chest radiography and who had at least 1 of the following: isolation of Legionella by culture; detection of Legionella by direct fluorescent antibody (DFA) testing of lung tissue or respiratory secretions; a 4-fold increase, to 1:18, in antibody titer to Legionella in serum between the acute phase and the convalescent phase; and detection of antigen in urine [13]. The currently used definition of nosocomial illness is LD that is confirmed by laboratory testing and occurs in a patient who has been hospitalized continuously for 10 days before the onset of illness [14]. Because of the nature of the questions on the reporting form and changes in the information collected over the surveillance period, we had to use a different definition for nosocomial disease. For this analysis, we defined a case of LD that occurred before 1983 as nosocomial if the affected individual had visited a hospital or been a patient in the hospital during the weeks before the onset of disease; cases that occurred between 1983 and 1987 were considered to be nosocomial if the individual had been a visitor, patient, or employee at a hospital during the weeks before the onset of disease; and cases that occurred between 1987 and 1998 were considered to be nosocomial if the individual had stayed or worked in the hospital as a patient or employee during the weeks before onset, had been hospitalized continuously for 3 days before the onset of infection, or had been discharged from the hospital 10 days before the onset of LD. Because no distinction about the cause of death was made until 1987, we defined case fatality in slightly different ways. From 1980 to 1985, a case was considered fatal if the answer to the question Case fatal? was recorded as yes. After 1987, a case was considered fatal if the more specific question Was death primarily due to LD? was answered yes. In the years between 1985 and 1987, both versions of the data-collection form were in circulation; responses were coded as above, depending on which form was used. We classified cases as outbreak related if it was so indicated on the case-report form. Cases were considered to be part of a nosocomial outbreak if they met the definition for nosocomial infection and if the field for nosocomial outbreak related on the case-report form was checked. Persons whose illness met the definition for nosocomial infection but for whom a non nosocomial outbreak source was listed were not considered to be part of a nosocomial outbreak. Persons who fit the definition for nosocomial infection were excluded from the communityoutbreak category, even if that field was marked on the casereport form. The response rates for the community-outbreak category were too low, however, to permit analysis of that variable. Persons whose reporting forms contained conflicting information about outbreak status were placed into the unknown/missing category. For the purposes of our analysis, persons who reported having traveled during the weeks before the onset of illness were classified as having travel-associated LD. We grouped the months of the year by season: December, January, and February as winter; March, April, and May as spring; June, July, and August as summer; and September, October, and November as fall. Statistical analysis. We compared the race, ethnicity, and sex distribution among persons with LD with that in the general US population, using 1990 census information [15]. We calculated yearly and age-adjusted rates of LD using actual and estimated census population data for Trends in cases per year, deaths per year, case-fatality rates per year, species and serogroup prevalence, and use of diagnostic testing methods were calculated by linear regression, using the PROC REG procedure in SAS version 8 software (SAS Institute) to determine whether the slope of the trend line differed from zero. To determine whether the distribution of 1040 CID 00:35 (1 November) Benin et al. Downloaded from on 08 April 018

3 species and serogroups causing disease had changed over time, we analyzed species and serogroup trends for patients for whom positive culture results had been reported. None of the analyses of case-fatality rates were adjusted for underlying conditions. All reported P values are -tailed. P.05 was considered to be significant. In situations in which patients had data missing for the variable being analyzed, we only included in the calculations those patients for whom data were available. RESULTS From 1980 to 1998, 10,753 case reports were submitted to the CDC; 6757 (63%) of these reports met the case definition for LD. The annual number of cases ranged from 59 in 1986 to 485 in 1988; the median number of cases per year was 360 (figure 1). There was no temporal trend in the number of cases per 100,000 US residents reported per year ( P p.3; R p 0.06). Demographic characteristics. The median age of case patients was 60 years (interquartile range, years). The disease incidence peaked in the year-old age group (figure ). Sixty-five percent of patients were male, compared with 49% of the 1990 US population. Of persons for whom data on race and ethnicity were available, 87% were white/non- Hispanic, 11% were black/non-hispanic, 1.5% were Hispanic, and!1% were American Indian, Native Alaskan, Asian, or Pacific Islander. These proportions are comparable to the racial distribution of the US population in Cases of LD occurred throughout the year, although more were reported during the summer months: 33% of cases occurred in the summer, 6% in the fall, % in the spring, and 19% in the winter. Nosocomial LD. Between 1980 and 1998, 35% of cases (066/5934) met the definition for nosocomial infection. The proportion of cases that were nosocomial varied from a high of 45% in 1986 to a low of 5% in 1997; the proportion decreased over time ( P p.0004; R p 0.5). Of the nosocomial cases for which outbreak information was available, 8% (66/ 951) were associated with an outbreak. Case-fatality rate. Between 1980 and 1998, 0% of case patients (10/610) died. The case-fatality rate decreased dramatically, from a high of 34% in 1985 to 11.5% in 1998 (P!.001; R p 0.6) (figure 3). For nosocomial LD, the overall casefatality rate was 8% (59/1861). This rate also decreased markedly, from a high of 46% in 198 to 14% in 1998 ( P!.0001; R p 0.6). For community-acquired LD, the overall casefatality rate was 14% (496/354); the rate decreased from a high of 6% in 1985 to 10% in 1998 ( P p.05; R p 0.). Use of diagnostic testing. Diagnosis of infection with Legionella was made by urine antigen testing for 16% of patients (1087/6757), by a 4-fold rise in antibody titer for 9% of patients (1979/6757), by DFA testing of respiratory specimens for 31% of patients (083/6757), and by isolation of Legionella species by culture for 45% of patients (3045/6757). For 0% of patients, the diagnosis was made by 11 technique. The proportion of patients who had a positive urine antigen test increased significantly over time ( P!.0001; R p 0.8); the proportion with a 4-fold rise in serological titer decreased stead- Figure 1. Annual no. of cases of legionnaires disease (LD) reported and rate per 100,000 US residents, Trends in Legionnaires Disease CID 00:35 (1 November) 1041 Downloaded from on 08 April 018

4 Figure. No. of cases of legionnaires disease (LD) and rate per 100,000 US residents, by age group, ily over the time period ( P!.0001; R p 0.85). The use of DFA testing or culture to diagnose LD increased from 1980 to 1990 but then decreased after 1990 (DFA, P p.0008 and R p 0.8; culture, P p.0 and R p 0.6) (figure 4). The case-fatality rate over the course of the 19-year study period for persons who had positive results of culture for Legionella but did not have positive results of a urine antigen test reported was 7% (664/458). For persons who did have positive results of a urine antigen test reported but did not have a culture done, it was 10% (70/68). When the analysis was controlled for year of diagnosis, persons who had a culture but no urine antigen testing were significantly more likely to die than Figure 3. Case-fatality rate for community-acquired and nosocomial legionnaires disease in the US population, CID 00:35 (1 November) Benin et al. Downloaded from on 08 April 018

5 Figure 4. Diagnostic testing for Legionella, percentage of all reported cases meeting case definition, by type of test, DFA, direct fluorescent antibody; 4-Fold rise, 4-fold increase in antibody titer. those who had urine antigen testing but no culture or any other diagnostic test (relative risk,.14; 95% CI, 1.6.9; P!.0001). Although mortality decreased over time among persons who had culture and either DFA or serologic testing ( P!.0001; R p 0.78), culture alone ( P!.0001; R p 0.7), or DFA testing alone ( P!.0001; R p 0.75), mortality showed no temporal trend among persons with urine antigen testing alone ( P!.3; R p ) (figure 5). Species and serogroups. Most reported cases of LD were caused by LP1. The next most common isolates were Legionella micdadei, LP6, LP3, and Legionella longbeachae (table 1). When all cases for which serogroup data were available were included in the analysis, the percentage of cases caused by LP1 increased steadily ( P p.01; R p 0.3), especially after 1990 ( P!.0001; R p 0.95). Conversely, the percentage of cases that were not caused by LP1 but for which the serogroup was known decreased, from a high of 8% (60/18) in 1983 to 4% (1/ 33) in 1998 ( P p.001; R p 0.4) (figure 6). Figure 5. Annual mortality rate among US residents with legionnaires disease, by type of diagnostic test, *Denominator is!10 tests per category per year. Trends in Legionnaires Disease CID 00:35 (1 November) 1043 Downloaded from on 08 April 018

6 Table 1. Proportion of legionnaires disease caused by each serogroup and species of Legionella, United States, Species, serogroup All isolates, % (n p 340) Isolates from community-acquired infections, % (n p 159) Isolates from nosocomial infections, % (n p 890) Legionella pneumophila Unknown L. bozemanii L. dumofii L. gormanii L. micdadei L. feeleii L. longbeachae L. jordanis NOTE. Only isolates identified by culture are included. When the analysis was restricted to patients who had positive culture results, LD was caused by LP1 in 51% (118/340); the percentage ranged from 58% (56/97) in 1990 to 91% (98/108) in 1998 and had no temporal trend. An additional 3% (750/ 340) of these patients had LD caused by L. pneumophila with a serogroup that was not specified (table 1). Travel-associated LD. Twenty-one percent (959/4610) of case patients met the definition for travel-associated LD. The percentage of travel-associated disease ranged from 14% in 1983 to 5% in ; there was no temporal trend ( P p.; R p 0.1). DISCUSSION This analysis of surveillance data documents important findings: a dramatic shift in diagnostic testing for LD, from a predominant use of serologic testing to the use of a urine antigen assay, and a 66% decrease in mortality for nosocomial and community-acquired legionellosis combined. The increase in the use of the urine antigen test has had important implications for outbreak control because it effectively allows the identification of outbreaks of LD due to LP1 in hospitals and the community. Recently, urine antigen test results led to the detection and rapid control of an outbreak of LD in an automotive plant that otherwise would have been missed [16]. The introduction of urine antigen testing also has led to the detection of previously unrecognized, ongoing transmission in hospitals. In these hospitals, detection of nosocomial transmission has been followed by corrective action that halted the spread of disease [17]. What are the perils of relying on this test? Although the urine antigen test can be used to effectively diagnose infections with LP1, it is neither licensed nor sensitive for diagnosis of infections caused by other species of Legionella or other L. pneumophila serogroups. Although LP1 is the leading cause of LD in the United States, 10 species of Legionella can cause disease [4]. The use of urine antigen testing has resulted in the preferential diagnosis of LP1 infections. The percentage of non- LP1 cases decreased by 79% between 1983 and The shift in diagnostic testing impairs our ability to detect non-lp1 cases and, as a consequence, our ability to detect and stop outbreaks. After a recent cluster of cases due to L. longbeachae was reported [18], a review of CDC surveillance data for L. longbeachae infections revealed that reports of these infections had decreased during the 1990s as the use of culture for diagnosis of LD had waned (data not shown). Many outbreaks of LD may be unrecognized because of the decrease in the use of culture for diagnosis of pneumonia. The drawbacks to exclusive use of urine antigen testing are not limited to poor detection of LD. Although this test identifies LP1 infections and is more sensitive than culture, public health investigations of outbreaks of LD rely on having both clinical and environmental Legionella isolates. When an outbreak of LD is detected, public health departments conduct both epi CID 00:35 (1 November) Benin et al. Downloaded from on 08 April 018

7 Figure 6. Percentage of legionnaires disease (LD) cases caused by serogroups and species other than Legionella pneumophila serogroup 1 for which the serogroup was known, demiological and environmental investigations to localize the source of transmission. Because attack rates for legionellosis are low and epidemiological studies alone can rarely completely pinpoint the source of transmission, environmental sampling is used to supplement epidemiological associations by providing Legionella isolates that can be compared with clinical isolates by use of molecular subtyping techniques [19]. Without clinical isolates, it is impossible to interpret the findings of environmental investigations, because legionellae are commonly found in the environment; it is not uncommon during outbreak investigations to identify multiple strains of Legionella, including several strains of LP1. What accounts for the documented decrease in mortality among patients with LD? Because the urine antigen test is more sensitive than culture, serologic testing, or DFA testing, it is possible that its use has led to the detection of disease in patients with milder forms of legionellosis, for whom case-fatality rates are lower. Indeed, mortality decreased over time among patients for whom diagnosis was made using culture but not among those for whom diagnosis was made by urine antigen testing; this finding suggests that patients for whom the diagnosis was made by urine antigen testing either were less severely ill or received treatment more quickly. Reporting during the early years of the time period may have been skewed by the application of diagnostic testing to only the sickest patients. Nevertheless, the decrease in mortality persists even when we separate the patients by type of diagnostic test. Another reason for the decrease in mortality may be the changes in empirical antibiotic treatment of patients hospitalized with pneumonia. The American Thoracic Society, in 1993 [11], and the Infectious Diseases Society of America, in 1998 [10], published recommendations for empirical therapy for pneumonia. These guidelines included the use of fluoroquinolones and macrolides, antimicrobial agents that are effective for the treatment of LD. Correspondingly, the use of these agents has steadily increased in the United States. Data from 3 recent, large outbreaks of LD from The Netherlands, Spain, and Australia suggest that countries using these agents as empirical therapy for patients hospitalized for community-acquired pneumonia have lower mortality rates during outbreaks of LD than do countries using b-lactam antibiotics as first-line agents [0]. These data must be interpreted with caution, because surveillance techniques for LD in the United States are insensitive. One population-based study in Ohio estimated the national incidence of LD to be ,500 cases/year [1]. This would mean that only.5% 4.5% of cases are reported to the CDC. Reasons for underreporting could include poor recognition by clinicians of LD as a cause of pneumonia, a lack of routine diagnostic testing, a lack of reporting to health departments, and limited reporting by health departments to the CDC. However, despite underreporting, surveillance data can reveal trends that have important public health implications. What diagnostic tests should be used for suspected cases of LD? Several groups have issued recommendations for the diagnosis of community-acquired and nosocomial pneumonia [14, 4]. These recommendations suggest that, for patients for whom LD is a possible diagnosis and for all patients with Trends in Legionnaires Disease CID 00:35 (1 November) 1045 Downloaded from on 08 April 018

8 nosocomial pneumonia, urine antigen testing and culture of appropriate respiratory secretions be performed. Performing both tests allows rapid diagnosis of LP1 infections, later diagnosis of infection with other legionellae, and availability of isolates. Given that empirical use of fluoroquinolones and extendedspectrum macrolides is successful in treating patients with undiagnosed LD, cost considerations may limit the use of tests to diagnose LD. Formal cost studies are needed to assess the implications of various approaches to the diagnosis of pneumonia. Cost equations must include the important public health implications of not diagnosing LD, including the continued presence of a disease-transmission source in the hospital or community. There has been a remarkable reduction in LD mortality since To continue to reduce mortality, our efforts must be directed at disease prevention. Prevention relies on the identification of sources of transmission and thus requires a diagnostic approach to pneumonia that includes both urine antigen testing and culture for legionellae. References 1. Fraser DW, Tsai TR, Orenstein W, et al. Legionnaires disease: description of an epidemic of pneumonia. N Engl J Med 1977; 97: Marston BJ, Lipman HB, Breiman RF. Surveillance for legionnaires disease: risk factors for morbidity and mortality. Arch Intern Med 1994; 154: Lo Presti F, Riffard S, Meugnier H, et al. Legionella taurinensis sp nov, a new species antigenically similar to Legionella spiritensis. Int J Syst Bacteriol 1999; 49: Benson RF, Fields BS. Classification of the genus Legionella. Semin Respir Infect 1998; 13: Fields BS, Benson RF, Besser RE. Legionella and legionnaires disease: 5 years of investigation. Clin Microbiol Rev 00; 15: Adeleke AA, Fields BS, Benson RF, et al. Legionella drozanskii sp nov, Legionella rowbothamii sp nov and Legionella fallonii sp nov: three unusual new Legionella species. Int J Syst Evol Microbiol 001; 51: Kohler RB, Zimmerman SE, Wilson E, et al. Rapid radioimmunoassay diagnosis of legionnaires disease: detection and partial characterization of urinary antigen. Ann Intern Med 1981; 94: Berdal BP, Farshy CE, Feeley JC. Detection of Legionella pneumophila antigen in urine by enzyme-linked immunospecific assay. J Clin Microbiol 1979; 9: Tilton RC. Legionnaires disease antigen detected by enzyme-linked immunosorbent assay. Ann Intern Med 1979; 90: Bartlett JG, Breiman RF, Mandell LA, File TM Jr. Community-acquired pneumonia in adults: guidelines for management. Infectious Diseases Society of America. Clin Infect Dis 1998; 6: Niederman MS, Bass JB, Campbell GD, et al. Guidelines for the initial management of adults with community-acquired pneumonia: diagnosis, assessment of severity, and initial antimicrobial therapy. American Thoracic Society. Medical Section of the American Lung Association. Am Rev Respir Dis 1993; 148: Roush S, Birkhead G, Koo D, Cobb A, Fleming D. Mandatory reporting of diseases and conditions by health care professionals and laboratories. JAMA 1999; 8: Centers for Disease Control and Prevention. Case definitions for infectious conditions under public health surveillance. MMWR Morb Mortal Wkly Rep 1997; 46: Centers for Disease Control and Prevention. Guidelines for prevention of nosocomial pneumonia. MMWR Morb Mortal Wkly Rep 1997; 46: United States Department of Commerce Census of population United States. Washington, DC: US Government Printing Office, Centers for Disease Control and Prevention. Outbreak of legionnaires disease among automotive plant workers Ohio, 001. MMWR Morb Mortal Wkly Rep 001; 50: Lepine LA, Jernigan DB, Butler JC, et al. A recurrent outbreak of nosocomial legionnaires disease detected by urinary antigen testing: evidence for long-term colonization of a hospital plumbing system. Infect Control Hosp Epidemiol 1998; 19: Centers for Disease Control and Prevention. Legionnaires disease associated with potting soil California, Oregon, and Washington, May June 000. MMWR Morb Mortal Wkly Rep 000; 49: Whitney CG, Hofmann J, Pruckler JM, et al. The role of arbitrarily primed PCR in identifying the source of an outbreak of legionnaires disease. J Clin Microbiol 1997; 35: Navarro C, Garcia-Fulgueiras A, Kool J, et al. Update on the outbreak of legionnaires disease in Murcia, Spain. Eurosurveillance Wkly 001; 9: Available at: Accessed 30 September Marston BJ, Plouffe JF, File TM, et al. Incidence of communityacquired pneumonia requiring hospitalization results of a population-based active surveillance study in Ohio. Arch Intern Med 1997; 157: Niederman MS, Mandell LA, Anzueto A, et al. Guidelines for the management of adults with community-acquired pneumonia: diagnosis, assessment of severity, antimicrobial therapy, and prevention. Am J Respir Crit Care Med 001; 163: Bartlett JG, Dowell SF, Mandell LA, File TM Jr, Musher DM, Fine MJ. Practice guidelines for the management of community-acquired pneumonia in adults. Infectious Diseases Society of America. Clin Infect Dis 000; 31: Heffelfinger J, Dowell S, Jorgensen J, et al. Management of communityacquired pneumonia in the era of pneumococcal resistance: a report from the Drug-Resistant Streptococcus pneumoniae Therapeutic Working Group. Arch Intern Med 000; 160: CID 00:35 (1 November) Benin et al. Downloaded from on 08 April 018