1341

Prevalence and Diagnosis of Legionella Pneumonia: A 3-Year Prospective Study with Emphasis on Application of Urinary Antigen Detection Bernhard Ruf, Dirk Schiirmann, Ingeburg Horbach, Franz J. Fehrenbach, and Hans D. Pohle

From the II Department of Internal Medicine, Rudolf Virchow University Hospital (Wedding), Freie Universitas Berlin, and Department of Microbiology, Robert Kod: Institute, Federal Health OJfice, Berlin, Federal Republic of Germany

Since its first description in 1977 [1], legionellosis has emerged as an important cause of sporadic and epidemic pneumonia [2, 3]. However, incidences reported from previous prospective studies vary greatly [2-8]. In some studies legionellae were found to be the first or second most frequent cause of pneumonia; in others, only in a low percentage of pneumonia cases. Long-term studies, of which there are few, appear necessary because short-term changes in the prevalence of legionellosis seem to occur [2, 8]. Diagnosis in the early phase of the disease remains a problem because frequently neither clinical symptoms nor standard laboratory tests contribute to a causative diagnosis [2, 3, 9]. Diagnosis is also hampered by lack of or inadequate characterization of special procedures or lack of availability of suitable specimens [3, 9]. Recently, Legionella antigen detection in urine has been introduced as a potentially important diagnostic method [10-14]. Our goals were to determine the prevalence of legionellosis among community-acquired and nosocomial pneumonias in an unselected population ofhospitalized patients in a longterm study and to evaluate the sensitivities of different diagnostic tests. We focused on the diagnostic value of urinary antigen detection.

Received 27 April 1989; revised 18 June 1990. Reprints or correspondence: Dr. Bernhard Ruf, II. Department of Internal Medicine, Rudolf Virchow University Hospital (Wedding), Augustenburger Platz 1, 0-1000 Berlin 65, FRG. The Journal of Infectious Diseases 1990;162:1341-1348 © 1990 by The University of Chicago. All rights reserved. 0022-1899/90/6206-0018$01.00

Patients and Methods This prospective study was done at the Rudolf Virchow University Hospital from 1 January 1984 to 31 December 1986. The study population comprised adult in-patients with community-acquired or nosocomial pneumonia. Tests to diagnose legionella pneumonia were part of a pneumonia etiology diagnostic program and included clinical tests and in fatal cases also the examination oflung tissue obtained at autopsy. Tests included serum antibody detection, examination of respiratory secretions, and legionella antigen detection in urine. Patientenrollmentand sampling. Patients with pneumonia were registered centrally in the infectious diseases department, to which all patients with community-acquired pneumonia were admitted. Patients with nosocomial pneumonia were recruited through reports of pneumonia cases from different departments combined with weekly inquiries on wards to ensure reporting and by weekly review of chest radiograph reports in the radiology department for patients with pneumonic infiltrations as a cross-check. Chest radiography was done routinely on all patients on admission and additionally on hospitalized patients with lower respiratory tract symptoms or symptoms compatible with an acute infectious disease (e.g., fever, leukocytosis) of unexplained origin. Attending physicians were asked to obtain serum and urine specimens from all patients with pneumonia at least once after diagnosis and once before discharge and respiratory secretions before the start of antibiotic therapy. In a parallel autopsy study, lung tissue from all fatal pneumonia cases was obtained for examination for legionellae. Cases were also included when autopsy revealed clinically undiagnosed pneumonia. Diagnosis ofpneumonia. Pneumonia was diagnosed by radiographic demonstration of pulmonary infiltrates and symptoms compatible with pneumonia (fever >38°C and/or symptoms of acute respiratory tract infection). For nosocomial cases, the diagnosis was also accepted if it was made at autopsy. Patients with pulmonary tuberculosis, human immunodeficiency virus infection, or pulmonary infiltrates due to noninfectious causes (e.g., pulmonary infarction, congestive heart failure, complications secondary to tumor stenosis) were excluded.

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

During a 3-year period the frequency of legionellosis in hospitalized patients with communityacquired and nosocomial pneumonias was 3.4% (23/684 cases) and 5.9% (33/559), respectively. Of the diagnostic tests evaluated, detection of Legionella pneumophila serogroup 1 antigen in urine had the highest sensitivity, with 86% of culture-proven cases being positive. Sensitivities of serologic tests and examination of respiratory secretions (culture and direct immunofluorescence) were 36% and 26%, respectively. The diagnostic value of serology and of examination of respiratory secretions can be low when specimens are obtained and processed under the typical conditions of hospitalization. Urinary antigen detection represents an important diagnostic addition, and examination of postmortem lung tissue from fatal cases with pneumonia is an important adjunct for estimating the prevalence of legionellosis and for assessing the effectiveness of premortem diagnostic tests.

1342

Ruf et al.

Table 1. Annual frequency of legionellosis in pneumonia cases. Pneumonia, legionellosis/total (%) Year 1984 1985 1986 Total

Total

Community-acquired

Nosocomial

28/476 (5.9) 12/361 (3.3) 16/406 (3.9) 56/1243 (4.5)

121240 (5.0) 51206 (2.4)

16/236 (6.8) 7/155 (4.5) 10/168 (6.0) 33/559 (5.9)

6/238 (2.5) 23/684 (3.4)

of legionellae [24, 25]. These examinations were performed to exclude cross-reactivityofDFA conjugateswith non-legionella bacteria. Standard criteria were used to interpret results from serum antibody detection, culture, and DFA [9, 15]. A serum anti-Iegionella antibody titer of at least 1:256 was considered to be positive. A fourfold or greater titer rise was considered diagnostic, and a single or standing positive titer was considered presumptive evidence for legionellosis [9]. Details in interpretation of results from assays for detection of soluble legionella antigen have been outlined previously [11, 14]. Data analysis. Data were entered into the Axis data base system (Cortex GmbH, Berlin) and analyzed with Axis-SQL.

Results We investigated 1243 cases of pneumonia; legionellosis was diagnosed in 56 (4.5%). Among nosocomial pneumonias, legionellosis was found in 33 (5.9%) of 559 cases and in 23 (3.4) of684 cases of community-acquired pneumonias (table 1). Nosocomial and community-acquired cases occurred sporadically; no yearly, seasonal, or monthly pattern was seen. Causative species were L. pneumophila in 55 (98 %) of 56 cases and L. dumoffii in 1 (2 %) of 56 cases. The distribution of different serogroups of L. pneumophila was serogroup 1 in 45 (80 %) of 56 cases, serogroup 4 in 4 cases (7 %), serogroup 5 in 2 cases (4 %), and serogroup 6 in 4 cases (7 %). Diagnostic testsfor legionellae. The different specialized laboratory tests and test combinations for legionellae performed on the 1243 patients are summarized in table 2. Of those, 1214 patients were tested while alive. Serum samples were examined from 1161 patients (1021 cases with more than one sample), urine from 989 patients (834 cases with more than one sample), and respiratory secretions from 633 patients (sputum in 572, tracheal aspirates in 37, and bronchoalveolar lavage fluid in 24 cases). Respiratory secretions were alwaysexamined by both culture and DFA. The complete collection of samples (serum, urine, and respiratory secretions) was examined for 560 patients, while both serum and urine samples but not respiratory secretions were available in 394 others. For 29 patients, all of whom had nosocomial infections, only lung tissue specimens obtained at autopsy were available. Of the 1243 patients, 325 died with pneumonia (26.1%); 311 of these were autopsied and had lung tissue specimens examined (table 2). Of these, 196 had lung tissue specimens

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

Pneumonias were designated nosocomial if onset occurred on or after the third day of hospitalization and community-acquired if onset occurred before the third day. Onset ofdisease was defined either by the first presentation of symptoms related to the infection (e.g., fever, respiratory symptoms, malaise) or by radiographic demonstration of pneumonic infiltrations, whichever began earlier. Handling ofspecimens obtainedforlegionella diagnosis. Serum, respiratory secretions (sputum, tracheal aspirates, bronchoalveolar lavage fluid), urine, and lung tissue (obtained at autopsies) were examined. Serum and urine samples were processed twice a week. Respiratory secretions were processed Monday through Friday once daily. All samples were stored at 4°C. From each autopsied patient, several lung tissue specimens from infected areas were examined for legionellae. For histologic processing, samples were fixedroutinely in 4 % formalin solution; unfixed specimens were taken for culture from all autopsies in 1984 and sporadically thereafter. Specimens were usually stored at -40°C and processed ~14 days after collection. Diagnostic procedures. Demonstration of legionella antibodies in serum was performed by an indirect fluorescence antibody assay (IFA) [9], using three different pools of heat-inactivated Legionella variants (L. pneumophila serogroups 1-7, L. dumoffii, L. micdadei, L. bozemanii, L. gormanii, L. jordanis, and L. longbeachae serogroups 1 and 2). To identify individual species, monovalent antigens were used. Components of the IFA were produced in the microbiology department of the Robert Koch Institute (Berlin). Strains were obtained from the Centers for Disease Control (CDC, Atlanta) except for the L. pneumophila serogroup 1 isolate, which was isolated from a patient in our hospital. For microscopic examination of respiratory secretions and tissues, the direct fluorescent antibody test (DFA) described by Cherry et al. [15] was used. Pooled fluorescent polyvalent antibodies against 13 Legionella species and serogroups (L. pneumophila serogroups 1-6, L. longbeachae serogroups 1 and 2, L. dumoffii, L. gormanii, L. micdadei, L. bozemanii, and L. jordanis) and the respective monovalent conjugates were used. For 1984 to mid-1985, antibodies to 12 variants (L. jordanis not included) were provided by the CDC (three pools with 4 Legionella variants each). For the remaining study period, commercially available pooled antibodies (Bios GmbH, Munich) were applied (two pools with 6 and 7 Legionella variants, respectively). For culture, respiratory secretions were pretreated with acid buffer to reduce contaminating microbes [16]. Tissue homogenates were prepared as outlined by Lattimer et al. [17]. Buffered charcoal yeast extract agar enriched with a-ketoglutarate [18, 19] was used for isolation. In addition, two semiselective media described by Edelstein [19] and Wadowsky and Vee [20], the latter modified according to Dennis et al. [21], were inoculated simultaneously. Legionella species and serogroups were identified by DFA [15]. Media were produced in the microbiology department of the Robert Koch Institute. Detection of L. pneumophila serogroup 1 antigen in urine was performed by a radioimmunoassay (RIA) and an ELISA, as outlined by Kohler et al. [11] and Fehrenbach et al. [14], respectively. To detect legionella antigen in lung tissue homogenates, specimens were processed according to Fehrenbach et al. [14]. Deparaffinized lung tissue sections were evaluated using hematoxylin-eosin, Gram's stain according to Brown and Hopps [22], and a modified Dieterle stain [23, 24]. Visualization of bacilli with the modified Dieterle stain but not with other stains is highly suggestive

JID 1990;162 (December)

110 1990;162 (December)

Table 2. Diagnostic tests and test combinations for legionella in No. of cases Clinical tests

Total

Autopsy *

Premortem studies S + U + C/O S+U S + C/O S U C/D U + C/O No premortem studies Total

1214 560 394 41 166 21 18 14 29 1243

282

29 311

* Deparaffinizedlung tissue sections studies by direct fluorescentantibody test (DFA) from all 311 cases; culture done on 196 of these. S. serum antibody detection; U. urinary antigen detection; CID, culture and DFA of respiratory secretions.

Table 3. Positive results from tests to diagnose legionellosis in 56 patients with legionella pneumonia.

7 2 7 1 13 1 2 4 4 1

9 5

+ + + + -/ND -/ND -/ND -/ND -/ND -/ND -/ND -/ND

Respiratory secretions

Urinary antigen detection

Culture

DFA

Culture

DFA

-/ND -/ND + + + + + + + -/ND -/ND -/ND

-/ND -/ND -/ND + -/ND + -/ND -/ND -/ND + -/ND -/ND

-/ND + -/ND + -/ND -/ND + -/ND -/ND + -/ND -/ND

-/ND -/ND -/ND -/ND -/ND -/ND -/ND + -/ND -/ND + -/ND

-/ND -/ND -/ND -/ND -/ND -/ND -/ND + + -/ND + +

Lung tissue

NOTE. DFA, direct fluorescent antibody test; +, positive; -/ND, negative or not done. Diagnosis of legionellosis required at least one of the following: fourfold or greater rise of the serum anti-legionella antibody titer to at least 1:256, culture of legionellae from respiratory secretions or from lung tissue, antigen detection in urine, or demonstration of legionellae by DFA in deparaffinized lung tissue sections.

Table 4. Diagnostic tests and test combinations in 53 patients with legionellosis and number of cases with positive results. No. of positive cases

Tests S + U + C/O S+U S + C/D U + C/D S U C/D Total

No. of cases with tests performed

17 19 6 1 5 3 2 53

Total

16 17 1 1 3 3 1 42*

Serum antibody detection

7 6 1

Urinary antigen detection

12 16

Respiratory secretions Total

Culture

DFA

4

2

3

0 0

3 3

17

32

1 7

1 3

1 6

NOTE. For diagnostic criteria, see footnote to table 3. DFA, direct fluorescent antibody test; S, serum antibody detection; U, urinary antigen detection; C/D, culture and DFA of respiratory secretions. * Patients with positive premortem tests; remaining 11 patients diagnosed only by lung tissue examination at autopsy.

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

examined with both culture and DFA; for the remaining 115 only DFA was done. Laboratory diagnosis. Tests and test combinations that established the diagnosis in the 56 cases with legionella pneumonia are shown in table 3. A positive IFA was the sole positive test in 7 patients, urinary antigen detection in 13, and DFA at autopsy in 5. In the remaining 31 patients, two or more tests were positive. Premortem tests for legionellae were performed on 53 of 56 patients with legionella pneumonia (table 4). In 42 (79%) of 53 cases, premortem tests proved the diagnosis; in the remaining 11 cases (21%), diagnosis was established only by examination oflung tissue obtained at autopsy. Diagnosis was also established by lung tissue examination in the remaining 3 of 56 patients in whom no premortem tests had been performed. Among the 56 patients with legionella pneumonia, there were 22 fatal cases (39 %). In these, legionellae were present in lung tissue. These 22 cases represented 7.1% of the 311 fatal pneumonia cases examined at autopsy, Sensitivities of each clinical test and for various test combinations are illustrated in table 5. Using any positive test to define a case of legionellosis, the rates of positive results were 36% for serology (17/47 cases), 26% for examination of respiratory secretions (7/27), 11% for culture (3/27), and 22 % for DFA (6/27). Legionella antigen was detected in urine with the L. pneumophila serogroup 1 antigen assay in 80 % of cases (32/40) and in 88 % of those with L. pneumophila serogroup 1 pneumonia (29/33). Serum antibody detection. Of 17 cases with significant titers, 16 had a fourfold or greater rise of serum antibodies and one had a single titer of 1:512. Maximum titers were 1:256 (3 cases), 1:512 (5), 1:1024 (3), 1:2048 (4), and 1:4096 (2). Species involved were L. pneumophila in 16 cases (serogroup 1, 10; serogroup 4, 3; serogroup 6, 3), and L. dumoffii in 1 case (maximal titer 1:512). Sensitivity was only 6% (3/47 cases) using initial samples, all of which were drawn within 2 weeks after onset of symptoms; it was 3 % (1/30) for the

1243 pneumonia cases.

No. Serum of antibody cases detection

1343

Prevalence and Diagnosis of Legionellosis

Ruf et aI.

1344

110 1990;162 (December)

Table 5. Diagnostic results in 53 patients with legionella pneumonia evaluated by single tests and test combinations. No. of positive cases (% of total tested)

Tests

47 40 33 27 17 15 36 30 18 15 23

Total

Serum antibody detection

Urinary antigen detection

Respiratory secretions Total

Culture

OFA

7 (26) 4 (24) 4 (27)

3 (11) 2 (12) 2 (13)

6 (22) 3 (18) 3 (20)

5 (28) 5 (33) 5 (22)

2(11) 2 (13) 2 (9)

4 (22) 4 (27) 4 (17)

17 (36) 32 (80) 29 (88) 16 (94) 14 (93) 33 (92) 27 (90) 14 (78) 14 (93) 10 (43)

7 5 13 7

(41) (33) (36) (23)

8 (35)

12 (71) 12 (80) 28 (78) 26 (87) 13 (72) 13 (87)

NOTE. For diagnostic criteria, see footnote to table 3. Patients with L. pneumophila serogroup 1 pneumonia who had urine samples tested for legionella antigen were evaluated separately regarding the fact that an assay for detection of L. pneumophila SGI antigen was applied (indicated by indexSG1 assigned to the tests). DFA, direct fluorescent antibody test; S, serum antibody detection; U, urinary antigen detection; elD, culture and DFA of respiratory secretions.

first week and 12% (2/17) for the second week. Overall, serologic results were positive in 17 (36%) of 47 cases. Among patients who had more than one sample examined (including cases with a positive titer initially), 43 % (17/40) had a diagnostic titer. In 32 of 40 cases the interval between first and last sample was >14 days. For the 17 positive patients, diagnostic serum titers were first demonstrated during the first week after onset of symptoms in 1 patient, the second week in 2, the third in 4, the fourth in 6, the fifth in 3, and the eighth week in 1. Among the 30 patients with negative serologic results, the last sample was examined during the first 2 weeks after onset of symptoms in 10, the third week in 5, the fourth in 6, the fifth in 3, the sixth in 3, and >8 weeks in 3. Thus, in half of the patients with negative serologic results, the last serum sample examined was taken >3 weeks after onset of symptoms. Culture and DFA of respiratory secretions. Legionellae were detected by culture and/or DFA in 7 (26%) of 27 cases; 6 were positive by DFA and 3 by culture (2 cases by both culture and DFA, table 5). Culture was positive in 1 (5%) of 22 cases in which sputum was examined and in 2 (40 %) of 5 cases in which tracheal aspirates were examined. DFA was positive in 4 (18%) of 22 cases in which sputum was examined and in 2 (40 %) of 5 cases in which tracheal aspirates were examined. In all 7 positive cases, L. pneumophila serogroup 1 was demonstrated. Five additional patients' respiratory secretions contained bacteria that fluoresced with polyvalent anti-legionella conjugates in one sample each, but no positive results were found when the same samples were repeatedly examined with the appropriate monovalent conjugates. All other legionellaspecific tests in these cases were negative. Three patients died with pneumonia, and examinations of lung tissue specimens

were negative for legionellae (2/3 cases by culture, 3/3 cases by DFA). In these cases legionellosis was considered to be absent for purposes of this study. Legionella antigen detection in urine. In 1984, RIA was used, and RIA and ELISA were used concurrently until September 1985; subsequently, only ELISA was used. Samples that had been initially tested only with RIA were stored at -70°C and retested later with ELISA. ELISA has emerged as more sensitive than RIA using the same batch of antibodies. With ELISA, antigen was detected in 32 positive cases in unconcentrated urine samples, whereas RIA detected antigen in 2 of 32 cases only after urine samples had been concentrated 10-foldby lyophilization. Urinary antigen detection results reported here are from unconcentrated samples with ELISA. To ensure specificity of positive results, particularly in cases in which urinary antigen detection was the sole positive test, one positive sample per patient was retested after immersion of samples in a boiling water bath for 15 min. Rarely observed cross-reactivity with non-legionella pathogens has been reported to be eliminated by heat [10, 11], but positive results in our tests persisted after heat treatment. Of40 cases tested, 32 were urinary antigen-positive (table 5); antigen detection was the sole positive test in 13. Of the 19 cases in which additional tests were positive, 16 were caused by L. pneumophila serogroup 1 and 3 by other serogroups (serogroup 4 in 2 and serogroup 6 in 1), indicating cross-reactivity of urine antigen test antibodies with antigens of serogroups of L. pneumophila other than serogroup 1. Thus, 29 cases with antigenuria were considered to have L. pneumophila serogroup 1 infection. Of 40 patients whose urine samples were examined, 7 had infection due to legionellae other than serogroup 1 L. pneumophila as determined by other tests (serum antibody detec-

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

S U USG! C/O S + U + C/O S + U + C/OSGI S+U S + USGl U + C/O U + C/OSG! S + C/O

Total no. of cases

JID 1990;162 (December)

Prevalence and Diagnosis of Legionellosis

1 pneumonia were evaluated separately, because an assay for detecting primarily L. pneumophila serogroup 1 antigen was used. Serum antibody detection, urine antigen detection, and examination of respiratory secretions with culture and DFA were performed in combination in 15 cases with L. pneumophila serogroup 1 pneumonia. Using this combination, 14 (93%) of 15 cases were diagnosed. Serum was positive in 5 (33 %), respiratory secretions in 4 (27 %), and urine in 12 (80 %). Results of other test combinations are shown in detail in table 5.

Fatal cases with demonstration of legionellae in lung tissue. Legionellae were demonstrated in lung tissue obtained at autopsy from 22 patients; in 19, premortem tests had been performed. In none of 14 cases tested for serum antibody was a diagnostic serum antibody titer found «1:64 in 7/14 cases, 1:64 in 6, and 1:128 in 1). The latest serum samples after onset of illness were taken in the first week in 5, the second in 5, the third in 2 and the fourth in 2. Examination of respiratory secretions for legionellae was negative in all 7 patients in whom it was performed, sputum being negative in 5 and tracheal aspirates in 2. The sole positive premortem test was urinary antigen detection in 73 % of those for whom it had been done (8/11 cases). Urinary antigen detection was positive in 7 of 10 cases with L. pneumophila serogroup 1 pneumonia and in 1 case with L. pneumophila serogroup 6 pneumonia. Both serum and urine samples obtained from 8 patients with L. pneumophila serogroup 1 pneumonia were examined initially in the first week in 5 and in the second week in 3 patients after onset of symptoms, respectively; 6 (75 %) of 8 patients were positive by urinary antigen detection, whereas serologic results were negative before death. Discussion

Legionella pneumonia has been diagnosed worldwide [2, 3]. The responsible bacteria are found ubiquitously in aquatic environments and represent a common risk for human beings [3, 28]. Contaminated potable water supplies have been increasingly linked to this disease [28-31] and seem to represent the most frequent source of sporadic legionellosis, which probably accounts for most cases [2]. Generally valid data on incidence cannot be calculated. The risk of infection is low for healthy individuals but is increased in populations with certain predispositions, such as the immunocompromised [2, 3]. In the present study, which evaluated hospitalized patients with pneumonia acquired either in the community or in the hospital, prevalence oflegionellosis of 3.4 % and 5.9% , respectively, were found; these figures represent a minimal estimate, because it is possible that additional cases, in particular due to variants other than L. pneumophila serogroup 1, were missed for several reasons: (1) not all tests were applied in each patient, (2) the sensitivities of the different tests varied substantially and were low for some, and (3) only a limited

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

tion, 6; DFA of lung tissue, 1). The rate of positive results using the antigen detection test for L. pneumophila serogroup 1 cases was therefore 88% (29/33 cases, table 5). The sensitivity of urinary antigen detection in patients who had positive results for L. pneumophila serogroup 1 infection by any other test was 79% (15/19). The sensitivity in cultureproven cases and in cases with serum antibody detection was 86% each (6/7 cases in each of the two groups). Among fatal cases in which L. pneumophila serogroup 1 was demonstrated in lung tissue, sensitivity was 70% (7/10). In all 32 urinary antigen-positive cases but one, the first urine sample taken was positive. Antigen was detected in the first week of illness in 21 cases and in the second week in 11. Of 24 positive patients with serial urine specimens examined, 23 had more than one positive sample. In 16 of24, antigen was still detected in the last sample examined, and in 8 antigen shedding stopped during hospitalization. The longest interval after onset of symptoms that samples were positive in the 24 patients was the 1st week in 1, the 2nd in 1, the 3rd in 7, the 4th in 8, the 5th in 5, the 6th in 1, and the 10th in 1. In the 8 patients in whom antigen excretion stopped, the times that samples were last positive, then first negative (week/week of illness) were weeks 1/4 (1), 3/4 (1), 3/5 (1), 3/6 (3), 4/11 (1), and 519 (1). Eight of 40 patients from whom urine samples were examined had no antigen detected; more than one urine sample was tested for each. Four had L. pneumophila serogroup 1 infections; the remaining 4 were infected with other serogroups. Culture andDFA oflungtissue. DFA of deparaffinized lung tissue sections was positive in 22 fatal cases. In all cases moderate or large numbers of fluorescent typicallegionellalike rods were reproducibly demonstrated. The bacteria reacted specifically with one monovalent anti-Iegionella conjugate only. Typically, legionellae were demonstrated intraand extracellularly [25, 26, 27]. Bacteria were not found using routine histologic staining procedures, whereas the modified Dieterle stain revealed legionella-like rods. In 15 of 22 patients, native lung tissue was available and culture was positive in 13 (87 %) of these. Legionella antigen was detected in supernatants of tissue homogenates in all cases by ELISA. In three additional patients who died up to 12 weeks after recovering from clinically diagnosed legionellosis, lung tissue was negative by culture and DFA, but legionella antigen was detected in supernatants ofhomogenized specimens. During life, the diagnosis was made in one patient by both serum antibody detection (titer rise to 1:1024) and urinary antigen detection and in the other two patients by urinary antigen detection. Patientgroups with different test combinations. Because no single test detected all cases of legionellosis, diagnostic outcomes were evaluated using various test combinations (table 5). For evaluating urinary antigen detection in comparison with other tests, patients with L. pneumophila serogroup

1345

1346

Ruf et al.

largely dependent on the availability of appropriate specimens; for example, serum samples should be available from the late convalescent phase [2, 3, 9, 37]. Sputum samples may be not obtainable at all because cough is often nonproductive at the time of presentation in patients with legionella pneumonia [2, 3]. Examination of more representative specimens, such as tracheal aspirates, have a higher diagnostic yield [38, 39], but such specimens are infrequently obtained. Our results show that examination of specimens obtained under "typical" circumstances from hospitalized patients may lead to sensitivities considerably lower than those obtained under optimal conditions. Only 43 % of patients (l0/23 cases), in whom both serologic tests and examination of respiratory secretions were done had positive results. For serologic tests a sensitivity of rv75 % has been reported, provided that serum samples are obtained ~6 weeks after the onset of symptoms [9, 38]. Our considerably lower figure of 36 % probably reflects the influence of early discharge or death. In contrast to studies that found serology of value for early diagnosis with positive titers in 20 %-40 % of cases within the first week of illness [38, 39], we found that only 6% of cases were positive in initial samples. The sensitivity of respiratory secretion culture has been improved markedly by the use of semiselective media [37, 39]. Sensitivities of up to 44 % with sputum culture and 83 % with transtracheal aspirate culture were found in studies in which specimens were processed immediately [39]. In our study the considerably lower culture yield of 11% and 40 %, respectively, may be because our samples were not always processed immediately; however, others have reported positive cultures from samples that were stored for days under ordinary room conditions [38]. Using culture as the diagnostic standard, DFA sensitivities of 25%-70% have been reported [3, 37, 38]. Positive DFA results with concurrent negative cultures have also been seen consistently in individual cases within patient series [37, 39], but the overall higher sensitivity of DFA compared with culture in our study is in contrast to other reports [38, 39]. Positive DFA results using polyclonal antibodies must be judged critically because cross-reactivity with non-legionella bacteria has been reported [3, 40]. We observed five patients with bacteria in respiratory secretions that fluoresced with polyvalent polyclonal conjugates but not with the appropriate monovalent conjugates, and we considered the positive results possibly to be nonspecific. However, since DFA was not performed on the isolates obtained by routine cultures, the reason for the positive DFA results remains unclear. Results from additional specific tests for legionellae were negative, among them postmortem examination oflung tissue. Positive DFA results considered to be specific in our study were reproducible and were confirmed by additional tests. Urinary antigen detection appears to be a substantial improvement in tests for legionella diagnosis [10-14]. On the basis of data from ELISA examination of unconcentrated urine

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

number of variants were within the scope of our diagnostic tests (e.g., the urinary antigen test, which had the highest sensitivity of all tests performed, was established to detect primarily L. pneumophila serogroup 1 antigen). Most data on the incidence of legionella pneumonia have been obtained from hospital-based studies [2, 4-7]. Among community-acquired pneumonia cases, incidences between o and 32 % [2, 4, 5, 8] and among nosocomial cases of up to 47% [6] have been reported. This wide variation is probably due to differing study conditions, such as case enrollment. criteria, diagnostic tests applied, and epidemiologic circumstances (e.g., epidemic situations) [2,3]. Geographic differences in occurrence have also been proposed [2]. The relatively low incidence in our study may be explained by an unselected study population and the absence of an epidemic upsurge. Among patients with community-acquired pneumonia, considerably higher incidences oflegionella pneumonia have been found among those requiring hospitalization than among those treated outside the hospital [7, 32]. Thus, for assessment of the incidence of this disease among all patients with community-acquired pneumonia, results from hospital-based studies are of limited value. The difference in prevalence has been explained by a relatively frequent severe course in legionella pneumonia requiring hospitalization [32]. A factor contributing to severe courses may be the fact that pneumonias are often treated outside of the hospital empirically with ineffective antibiotics. The problem of nosocomial legionellosis, with prevalences frequently higher than for community-acquired cases [4, 5], arises from the accumulation of susceptible individuals in hospitals. The risk for this population probably increases with the degree of contamination of water supplies with legionellae [33]. The predominance of L. pneumophila (98 % of our cases), particularly serogroup 1 (80 % of our cases), agrees with results from other studies [2, 34], but the extent of the predominance is noteworthy. There is strong evidence that other Legionella variants were actually of minor importance and that this predominance was not only the effect of gaps in the clinical diagnostic approach for the following reasons. Lung tissue from patients with fatal pneumonia tested by culture and DFA did not reveal Legionella species other than L. pneumophila; likewise, in autopsy studies from earlier years, using IFA capable of detecting Legionella variants recognized to cause legionellosis, only L. pneumophila was identified [35, 36]. The exclusiveness of L. pneumophila in our nosocomial cases is convincingly explained by the fact that only this species was found in the hospital water supply, which was the source of our nosocomial infections [31]. We emphasized evaluating the diagnostic significance of different tests, in particular urinary antigen detection. For clinical diagnosis, serum antibody detection and examination of respiratory secretions are still the most widely used tests; however, these have considerable weaknesses. Sensitivities are

JID 1990;162 (December)

110 1990;162 (December)

Prevalence and Diagnosis of Legionellosis

histomorphologic characteristics, such as bacteria within macrophages [25, 26], and characteristic staining properties. Characteristically, legionellae are not seen in tissues by using conventional stains such as hematoxylin-eosin and Gram's, but they can be seen using the modified Dieterle stain [24, 25]. Thus, failure to see microbes using hematoxylin-eosin or Gram's stains in pneumonia cases alone justifies further investigation for legionellae. Our results show that Legionella is an important cause of pneumonia and that a broad spectrum of tests must be performed for optimal diagnosis. Urine antigen detection has proven to be an important test with a high sensitivity, which permits early diagnosis. The detection of only L. pneumophila serogroup 1 antigen is a handicap of most ,currently used urinary antigen tests; however, this is partly compensated for because L. pneumophila serogroup 1 is responsible for most legionella pneumonia cases [2, 34] and because some crossreactivity with other legionellae occurs [43, our study]. Examination of lung tissue specimens obtained after death is an important adjunct to gain data on the true prevalence of legionellosis and on the diagnostic significance of clinical tests.

Acknowledgments We thank the medical staff and the pathologists of the Rudolf Virchow University Hospital for their assistance and cooperation.

References

1. Fraser OW, Tsai TR, Orenstein W, Parkin WE, Beecham HJ, Sharrar RG, Harris J, Mallison GF, Martin SM, McDade JE, Shepard CC, Brachman PS, Field Investigation Team. Legionnaires' disease. Description of an epidemic of pneumonia. N Engl J Med 1977; 297:1189-1197 2. Bartlett CLR, Macrae AD, Macfarlane JT. Clinical aspects and diagnosis oflegionella infection. In: Bartlett CLR, Macrae AD, Macfarlane JT, eds. Legionellainfections. London: Edward Arnold, 1986:37-55 3. Winn WC Jr. Legionella and Legionnaires' disease: a review with emphasis on environmental studies and laboratory diagnosis. Crit Rev Clin Lab Sci 1985;21:323-381 4. Rudin JE, Wing EJ. Prospective study of pneumonia: unexpected incidence of legionellosis. South Med J 1986;79:417-419 5. Yu VL, Kroboth FJ, Shonnard J, Brown A, McDearman S, Magnussen M. Legionnaires'disease: new clinical perspective from a prospective pneumonia study. Am J Med 1982;73:357-361 6. Muder RR, Yu VL, McClure JK, Kroboth FJ, Kominos SO, Lumish RM. Nosocomial legionnaires'disease uncovered in a prospectivepneumonia study. Implications for underdiagnosis. JAMA 1983;249: 3184-3188 7. Woodhead MA, Macfarlane JT, McCracken JS, Rose DH, Finch RG. Prospective study of the aetiology and outcome of pneumonia in the community. Lancet 1987;1:671-674 8. Woodhead MA, Macfarlane JT, Macrae AD, Pugh SF. The rise and fall of legionnaires' disease in Nottingham. J Infect 1986;13:293-296 9. Wilkinson H. Hospital-laboratory diagnosis of Legionella infections. Atlanta: Centers for Disease Control, 1987 10. SathapatayavongsB, Kohler RB, Wheat U, White A, Winn WC Jr, Girod JC, Edelstein PH. Rapid diagnosis of legionnaires' disease by uri-

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

specimens, a sensitivity of 70%-86%, depending on case definition (86% for culture-proven cases) was found, which is in the range of data from earlier studies [10, 11, 41]. The fact that initial samples were positive for urinary antigen in all but one antigen-positive case supports the usefulness of this test for early diagnosis. The specificity of urinary antigen detection has previously been shown to be very high [10, 11,41], and we considered a positive result to be diagnostic for legionella pneumonia even when it was the sole positive test. Non-legionella-related cross-reactivity, reported in rare cases, can be extinguished by boiling urine samples [10, 11], whereas the positivity in our samples resisted boiling. No positive results were found with our test when urine samples from patients with urogenital tract infections of various causes were tested [42]. Crossreactivity in our test was seen, however, with urinary antigens from patients infected with L. pneumophila other than serogroup 1, as has been reported earlier [14, 43]. For our purposes this is a useful cross-reactivity. The high specificity of our test is also supported by results from examination of supernatants from homogenized lung tissue specimens obtained at autopsy (unpublished data). Samples from 475 cases (196of this study) were examined; positive results were seen in 21 patients, 18 ofwhom also had positive culture and/or DFA. The other 3 patients had a recent history of clinically proven legionella pneumonia. These patients died up to 12 weeks after having recovered from legionellosis, and autopsy did not reveal the continued presence of pneumonia. These results demonstrate that antigen detection may permit postmortem diagnosis of legionellosis in patients who died late in the convalescent phase or with a remote history of infection, when legionellae cannot be demonstrated in lung tissue by culture or DFA [26, 44]. The examination of lung tissue obtained at autopsy is an important adjunct to clinically performed tests in diagnosing legionella pneumonia, since we found a number of cases diagnosed only at autopsy. Examination of respiratory secretions and serologic tests were negative in all cases with a postmortem diagnosis in which these tests had been done. The sole test giving positive results in 73 % of these cases was urinary antigen detection, which underlines its importance in clinical diagnosis. The prevalence of legionellosis of 7.1% among our autopsied cases is in the range found in other studies [45, 46]. Microbiologic examination of lung tissue is typically done only on special request. Histopathologic evaluation can be important in diagnosing legionella pneumonia as well. In our study deparaffinized lung tissue sections were examined using DFA, which has been found to be highly specific and sensitive. All patients from whom legionellae were cultured were also positive by DFA. Our assessment that positive DFA results were specific without culture having been done is supported by the following characteristics found in all our cases: demonstration of fluorescent legionella-like rods with typical

1347

1348

Ruf et al.

29. Meenhorst PL, Reingold AL, Groothuis DG, Gorman GW, WIlkinson HW, McKinney RM, Feeley IC, Brenner DI, van Furth R. Waterrelated nosocomial pneumonia caused by Legionellapneumophila serogroups 1 and 10. I Infect Dis 1985;152:356-364 30. Neill MA, Gorman OW,Gilbert C, RousselA, Hightower AW, McKinney RM, Broome CV. Nosocomiallegionellosis, Paris, France. Evidence for transmission by potable water.Am I Med 1985;78:581-588 31. Ruf B, Schiirmann D, Horbach I, Seidel K, Pohle HD. Nosocomial legionella pneumonia: demonstration of potable water as the source of infection. Epidemiol Infect 1988;101:647-654 32. Woodhead MA, Macfarlane IT, Rodgers FG, Laverick A, Pilkington R, Macrae AD. Aetiologyand outcomeof severecommunity-acquired pneumonia. I Infect 1985;10:204-210 33. Edelstein PH. Control of Legionella in hospitals. I Hosp Infect 1986;8:109-115 34. Reingold AL, Thomason BM, Brake BI, Thacker L, Wilkinson HW, Kuritsky IN. Legionella pneumonia in the United States: the distribution of serogroups and species causing human illness. I Infect Dis 1984;149:819 35. SchiirmannD, RufB, Fehrenbach FI, Jautzke G, Pohle HD. Fatal legionnaires' pneumonia: frequency of legionellosis in autopsied patients with pneumonia from 1969 to 1985. I Pathol 1988;155:35-39 36. RufB, Schiirmann D, Pohle HD. Fatallegionellapneumonia:retrospective examination of lungtissueusingdirectandindirectfluorescent-antibody methods. Zentralbl Bakteriol MikrobiolHyg [A] 1987;266:443-448 37. Edelstein PH. Laboratory diagnosisofinfections caused by legionellae. Eur I Clin Microbiol 1987;6:4-10 38. EdelsteinPH, Meyer RD, FinegoldSM. Laboratory diagnosisoflegionnaires'disease. Am Rev Respir Dis 1980;121:317-327 39. Zuravleff JJ, Yu VL, Shonnard JW, Davis BK, Ribs ID. Diagnosis of legionnaires'disease. An updateof laboratory methods with newemphasis on isolation by culture. lAMA 1983;250:1981-1985 40. TenoverFC, Edelstein PE, Goldstein LC, Sturge IC, Plorde JJ. Comparison of cross-staining reactions by Pseudomonas spp. and fluorescein-labeled polyclonaland monoclonal antibodiesdirected against Legionellapneumophila. I Clin Microbiol 1986;23:647-649 41. Aguero-Rosenfeld ME, Edelstein PH. Retrospective evaluation of the Du Pontradioimmunoassay kit fordetectionof Legionella pneumophila serogroup 1 antigenuria in humans. I Clin MicrobioI1988;26:17751778 42. Horbach I, RufB, Schiirmann D, Fehrenbach F, Pohle HD. Demonstration oflegionella antigen in the urine: significancefor differentialdiagnosis. Zentralbl Bakteriol Mikrobiol Hyg [A] 1987;264:102-104 43. Kohler RB, Wheat U, French ML, Meenhorst PL, Winn WC Jr, Edelstein PH. Cross-reactiveurinaryantigensamongpatients infected with Legionella pneumophila serogroups 1 and 4 and the Leiden 1 strain. I Infect Dis 1985;152:1007-1012 44. Hernandez FI, Kirby BD, StanleyTM, EdelsteinPH. Legionnaires'disease. Postmortem pathologic findings in 1 of 20 cases. Am I Clin Pathol 1980;73:488-495 45. FayD, Baird1M, Aguirre A, HaegrenV,WerlingK, Bell TW. Unrecognized legionnaires' disease as a cause of fatal illness. lAMA 1980;243:2311-2313 46. Gerber JE, Casey CA, Martin P, Winn WC Jr, Legionnaires'disease in Vermont, 1972-1976. Am I Clin Pathol 1981;76:816-818

Downloaded from http://jid.oxfordjournals.org/ at University of Birmingham on July 19, 2015

nary antigen detection. Comparison of ELISA and radioimmunoassay. Am I Med 1982;72:576-582 11. Kohler RB, Zimmerman SE, Wilson E, Allen SD, EdelsteinPH, Wheat U, White A. Rapidradioimmunoassaydiagnosisoflegionnaires' disease: detection and partial characterization of urinary antigen. Ann Intern Med 1981;94:601-605 12. Kohler RB, Winn WCJr, Wheat U. Onset and duration of urinary antigen excretion in legionnaires' disease. I Clin Microbiol 1984;20: 605-607 13. Kohler RB, WinnWC Ir, Girod IC, Wheat U. Rapiddiagnosis of pneumonia due to Legionella pneumonia serogroup 1. I Infect Dis 1982;146:444 14. Fehrenbach FI, Horbach I, RufB, Schiirmann D, Pohle HD. Rapid detection of legionella antigen in tissues and body fluids. Israel I Med Sci 1986;22:706-710 15. CherryWB, PittmanB, HarrisPP, HebertGA, ThomasonBM, Thacker L, Weaver RE. Detection of legionnaires' disease bacteria by direct immunofluorescent staining. I Clin MicrobioI1978;8:329-338 16. Bopp CA, Summer JW, Morris GK, Well IG. Isolation of Legionella spp. from environmentalwater samples by low-pHtreatment and use of a selective medium. I Clin Microbiol 1981;13:714-719 17. Lattimer GL, Rhodes LV,Salventi IF, Cepil BR. Isolation of Legionella pneumophila from clinical specimens: salutary effectsof lung tissue dilution. Am Rev Respir Dis 1980;122:101-105 18. Pasculle AW, Feeley IC, Gibson RI, Cordes LG, MyerowitzRL, Patton CM, Gorman GW, Carmack CL, Ezzell IW, Dowling IN. Pittsburgh pneumonia agent. Direct isolation from human lung tissue. I Infect Dis 1980;141:727-732 19. EdelsteinPH. Improved semiselective medium for isolationof Legionella pneumophila from contaminated clinical and environmental specimens. I Clin Microbiol 1981;14:298-303 20. Wadowsky RM, YeeRB.A glycine-containingselectivemediumfor isolation of Legionellaceaefrom environmental specimens. Appl Environ Microbiol 1981;42:768-772 21. Dennis PI, Bartlett CLR, WrightAE. Comparison of isolation methods for Legionella spp. In: Thornsberry C, Balows A, Feeley IC, JakubowskiW,eds. Legionella. Proceedings of the 2nd International Symposium. Washington,DC: American Society for Microbiology, 1984:294-296 22. Brown RC, Hopps HC. Staining of bacteria in tissue sections. A reliable Gram stain method. Am I Clin PathoI1973;60:234-240 23. VanOrden AE, Greer PW.Modificationof the Dieterle spirochetestain. I HistotechnoI1977;1:51-53 24. Chandler FW, Hicklin MD, Blackmon IA. Demonstration of the agent of Legionnaires'disease in tissue. N Engl I Med 1977;297:1218-1220 25. Schiirmann D, Grosse G, Horbach I, Fehrenbach FI, RufB, Pohle HD. Legionella Pneumonienin autoptischem Untersuchungsgut. Zur Diagnostik und Pathologieder Legionellose. Pathologe1987;8:156-162 26. Hicklin MD, Thomason BM, Chandler FW, Blackmon IA. Pathogenesis of acute legionnaires'disease pneumonia. Am I Clin Patho11980; 73:480-487 27. Winn WC Jr, MyerowitzRL. The pathology of the Legionella pneumonias. A review of 74 cases and the literature. Hum PathoI1981;12: 401-422 28. Bartlett CLR, Macrae AD, Macfarlane IT. Environmental sources of infection and modes of transmission. In: Bartlett CLR, Macrae AD, Macfarlane IT, eds. Legionella infections. London: Edward Arnold, 1986:90-119

JID 1990;162 (December)