Etiology of Community-acquired Pneumonia in Hospitalized Patients*

Etiology of Community-acquired Pneumonia in Hospitalized Patients* A 3-Year Prospective Study in Japan Tadashi Ishida, MD; Toru Hashimoto, MD; Machiko Arita, MD; Isao Ito, MD; and Makoto Osawa, MD Study objective: To compare the etiology of community-acquired pneumonia in Japan and Western countries, the causative pathogens were prospectively investigated in patients requiring hospitalization. Design: Prospective study over a 3-year pe ;od. Setting: A community general hospital in Japan. Patients: Three hundred twenty-six episodes of community-acquired pneumonia in 318 patients admitted to the hospital between July 1994 and June 1997. Methods: The microbiological diagnosis was based on the results of quantitative sputum culture, blood culture, and other invasive procedures, including transthoracic needle aspiration or bronchoscopic examination. Serologic tests for Mycoplasma pneumoniae, Chlamydia spp, Legionella spp, and viruses were also routinely performed. Results: Causative pathogens were identified in 199 episodes (61 %). Streptococcus pneumoniae was the most common pathogen (23%), followed by Haemophilus injluenzae (7.4%), M pneumoniae (4.9%), and Klebsiella pneumoniae (4.3%). The Streptococcus milleri group and Chlamydia pneumoniae were detected in 3.7 and 3.4% of the episodes, respectively. Pneumonia due to Legionella spp was recognized in only two patients. Conclusions: The etiology of community-acquired pneumonia in Japan did not differ markedly when compared with that of Western countries except for the low incidence of Legionella pneumonia. C pneumoniae and the S milleri group, which are emerging or newly recognized pathogens, were also significant causative microorganisms. (CHEST 1998; 114:1588-1593) Key w01 ds: community-acquired pneumonia; etiology; Legionella pneumonia; prospective study Abbreviations: CAP = community-acquired pneumonia; CFU = colony-forming units; COPD = chronic obstructive pulmona1y disease; ELISA = enzyme-linked immunosorbent assay; TNA = transthoracic needle aspiration community-acquired pneumonia (CAP) remains a common and serious illness with a significant morbidity and mortality despite the availability of potent new antibiotics and other effective therapies. For editorial comment see page 1509 In Japan, pneumonia is thefourth leading cause ofdeath, and 64 persons outof a 100,000 population died of this disease in 1995. 1 Management of pneumonia remains empiric and challenging at the time of initiation, because the precise etiology is usually unknown. Despite progress in diagnostic tools and *From the Department of Internal Medicine, Kurashiki Central Hospital, Kurashiki, Okayama, Japan. Manusc1ipt received March 5, 1998; r e\~ s iac"l>epted o n ]w1e 10, 1998. C01-respondence to: Tadashi Ishida, M.D, Department of Internal Medicin e, Ku-rashiki Central Hospital, 1-1-1 Mitca, Kurashiki, Okayrmw. 710, Japan 1588 laboratory tests, it takes at least a few days to identify causative microorganisms in blood or sputum samples, and the etiology of about half of all patients with CAP remains uncertain. Recently, several prospective studies on patients with CAP who required admission to hospital have been published in Western countries, 2-1 1 and guidelines for the management of CAP have been promoted. 12 13 These studies showed appreciable differences in the frequencies of causative agents, which seemed to be related to seasonal, geographical, and racial factors. On the other hand, there have been few prospective studies on the etiology of CAP in Asia. In this study, we prospectively investigated the etiology of CAP among patients admitted to a general hospital over 3 years in Japan and compared the results with data from Western countries. Clinical Investigations

Patients MATERIALS AND METHODS All patients over 15 years of age with CAP admitted to Kurashiki Central Hospital (a 1,100-bed community general hospital) between July 1994 and June 1997 were evaluated for inclusion in the study. CAP was defined as new infiltrates of the chest revealed by radiographic examination on or within 24 h of admission, including at least two of the following: fever (temperature, 2: 37.8 C), production of purulent sputum, cough, and leukocytosis (WBC count, 2: 1 0, 0 0 0). 1 Patients ~ - L L were excluded from the study when abnormalities using chest radiographic examination were attributed to other causes such as congestive heart failure, pulmonaty infarction, or obstructive pneumonia due to lung cancer. Patients with leukemia or AIDS or receiving immunosuppressive therapy were also excluded. Prior to the study, the protocol was approved by the institutional ethics committee. Microbiological Studies On admission, two sets of blood cultures were obtained, and, if sputum was available, a Gram's stain test and a quantitative culture of sputum were obtained. Sputum data were only evaluated when the Gram's stain test revealed numerous leukocytes (> 25 in a 100 X microscopic Held) and few epithelial cells ( < 10 in a 100 X microscopic field). Certain invasive methods such as bronchoscopic examination, transthoracic needle aspiration (TNA), or thoracentesis were employed to obtain specimens in some patients after full explanation of the procedures. These specimens, including sputum, were also used for culture of Legionella spp and M.ycoplasnw pneumoniae on buffered charcoal-yeast extract alpha agar and pleuropneumonia-like organism agar, respectively. Pharyngeal swabs were obtained for culture of M pneumoniae when a diagnosis of Mycoplasma pneumonia was suspected. We tested the susceptibility of Streptococcus pneumoniae to oxacillin sodium using the disk-diffusion technique ( 1 - ~ -disk; L g BBL Microbiology Systems; Cockeysville, MD). Isolates producing a zone of inhibition of 2: 20 mm around the disk were defin ed as susceptible. Serologic Studies Serum samples were collected in the acute and! or convalescent periods. The samples were tested for complement fixation antibody to influenza A and B viruses, parainfluenza virus, adenovirus, cytomegalovirus, Chlamydia psittaci, and M pneumoniae and for indirect fluorescent antibody to Legionella spp. Antibodies to Chlamydia pnewnoniae were determined using the enzymelinked immunosorbent assay (ELISA). Criteria for Determination of Microbial Etiology The etiology of pneumonia was classified as "definitive," "presumptive," or "unknown" according to definitions established previously by Fang and c:olleagues.- 5 Bacteria were considered to be definitive causative agents when isolated from blood culture, TNA, or thoracentesis. \Ve considered the results of sputum culture in combination with the Gram's stain findings. An organism showing heavy growth (2: 10 7 colony-forming units [CFU]/mL) of a predominant bacterium on sputum culture was considered to be a presumptive pathogen. Moderate or light growth (10 6 or 10 5 CFU/mL) on the sputum culture was also considered as evidence of a presumptive pathogen if the Gram's stain test revealed a bacterium compatible with the culture results. Any microorganism isolated from bronchoalveolar lavage fluid was considered to be a presumptive pathogen when its concentration reached > 10 5 CFU/mL in quantitative culture. Bacteria isolated by culture of lung tissue obtained at autopsy or operation were designated as definitive etiologic agents. If Legionella spp., Mycobacterium tuberculosis, or M. pneumoniae was isolated fi om a specimen, that specimen was considered to be a definitive pathogen even if the culture showed little growth. For serologic tests, a fourfold increase in the antibody titer level between paired sera or an elevated single titer of 2: 256 X was considered definitive. Patient Characteristics RESULTS Three hundred thirty-nine patients who had a total of 347 episodes of suspected pneumonia were enrolled in the study. Twenty-one patients were subsequently excluded-9 with organizing pneumonia, 6 with obstructive pneumonia secondary to lung cancer, 3 with leukemia, 2 with eosinophilic pneumonia, and 1 with drug-induced pneumonitis. Thus, 318 patients (229 men and 89 women) with 326 episodes of pneumonia were finally analyzed. The mean age (± SD) of the patients was 65.2 ± 16.3 years (range; 18 to 93 years). Two hundred twenty-four patients (70.4%) had at least one underlying disease (Table 1). The major underlying diseases were diabetes mellitus (17.6%), COPD (15.7%), chronic heart disease (9.1 %), and cerebrovascular disease (9.1 %). In 124 episodes (38%), antibiotics had been administered before admission. Etiology of Pneumonia A microbiological diagnosis was established in 199 ( 61.0%) of the episodes of pneumonia. The classifi- Table!-Clinical Characteristics of 318 Patients With CAP Characteristic n % Smoking habits 165 51.9 Chronic alcoholism 20 6.3 Diabetes mellitus 56 17.6 COPD 50 15.7 Chronic heart disease 29 9.1 Cerebrovascular disease 29 9.1 Chronic liver disease 19 6.0 Postgastrectomy 15 4.7 Old pulmonary tuberculosis 14 4.4 Malignancy 14 4.4 Lung fibrosis 9 2.8 Psychological disorder 9 2.8 Bronchial asthma 8 2.5 Rheumatoid arthritis 7 2.2 Bronchiectasis 6 1.9 Miscellaneous 13 4.1 Previously healthy 94 29.6 CHEST I 114 I 6 I DECEMBER, 1998 1589

cation was definitive in 72 episodes ( 22.1%) and presumptive in 127 episodes (39.0%). A pathogen was detected in only 55 of 124 episodes (44.4%), where antibiotic therapy had been given before admission; however, the patients \vithout prior therapy revealed a high diagnostic yield of 69.3% (140 of 202 episodes). The diagnostic methods employed and the results obtained are listed in Table 2. A blood culture was obtained in 279 episodes and was not taken in 47 episodes, and 14 cultures yielded a pathogen (S pneumoniae in 7, Klebsiella pneunwniae in 3, Flavobacterium spp in 2, Escherichia coli in 1, and the Streptococcus 1nilleri group \vith anaerobes in 1). Satisfactory sputum samples were obtained in 247 episodes, and 136 samples yielded causative microorganisms. Cultures of pharyngeal swabs were positive for M pneumoniae in 7 episodes. Specimens obtained by thoracentesis or bronchoscopic examination revealed a high positive culture rate (33.3 and 41.2%, respectively). Twenty-eight patients underwent TNA, and pathogens were detected in 14 of them (S pneumoniae in 6, the S milleri group \vith anaerobes in 3, Haemophilus injluenzae in 2, the S milleri group in 1, K pneumoniae in 1, and K pneumoniae \vith E coli in 1). The same organisms were cultured from sputum (6 patients) or blood (1 patient). Serologic tests were performed in 296 episodes (tests were not available in 30 episodes), and the results met the diagnostic criteria in 42 episodes. A significant rise of viral antibody titers in paired sera was observed in 14 (4.3%) of the patients. However, 7 patients clearly had bacterial pneumonia and were therefore excluded, leaving 7 \vith viral pneumonia. The most common pathogen was S pneumoniae, found in 75 episodes (23.0%), followed by H influenzae in 24 episodes (7.4%), M pneumoniae in 16 episodes ( 4. 9%), K pneumoniae in 14 episodes (4.3%), the S milleri group in 12 episodes (3.7%), and C pnewnoniae in 11 episodes (3.4%). Anaerobes were isolated in 8 episodes and C psittaci was found Table 2-Diagnostic Methods and Results Diagnostic Method Blood culture Sputum culture Pha1)'ngeal swab Thoracentesis Transthoracic needle aspiration Bronchoscopic examination Operation Autopsy Serologic examination No. of Episodes Studied 279 247 71 30 28 17 1 1 296 No. of Diagnostic Studies(%) 14 (5.0) 136 (55 0) 7 (9.6) 10 (333) 14 (50.0) 7 (41.2) 1 (1000) 1 (100.0) 42 (14.2) in 7. Pneumonia caused by M tuberculosis accounted for 5 episodes. Legionella pneumonia was only diagnosed in 2 episodes (Legionella pneumonphila and Legionella micdadei). Dual pathogens were detected in 14 episodes (4.3%) (Table 3). One 75-year-old patient, with Staphylococcus epidennidis pneumonia, also had cerebrovascular disease. This diagnosis was made because the organism was detected in the sputum at a concentration of 10 7 CFU/mL. Forty-one strains (54.7%) of S pneumoniae were susceptible to oxacillin sodium. Severity and Mortality The patients were categorized by severity of illness using methods established previously by Fine and colleagues. 14 Twenty-two episodes (6.7%) were class I (the lowest risk class); 5.5 (16.9%) were class II; 97 (29.8%) were class III; 101 (31.0%) were class IV; and 51 (15.6%) were class V (the highest risk class, which had a predicted probability of death of > 10%). Twenty-four patients (7.5%) underwent mechanical ventilation. Twenty patients (6.1 %) died during hospitalization. Pneumonia was believed to be the major cause Table 3-Microbiological Diagnosis of 326 Episodes of CAP Organism n* o/o Streptococcus pneumoniae 75 23.0 Streptococcus milleri group 12 3.7 Streptococcus pyogenes 1 0.3 Streptococws agalactiae 3 0.9 Staphylococcus aureus 7 2.1 Staphylococcus epider111idis 0.3 M oraxella cat a rrhalis 6 1.8 Haemophilus i n j l u e n ~ a e 24 7.4 Klebsiella pnewnoniae 14 4.3 Pseudomonas aeruginosa 8 2.5 Escherichia coli 3 0.9 Flavobacteriu1n 1neningosepticu1n 2 0.6 Acinetobacter spp 1 0.3 Legionella spp 2 0.6 Anaerobes 8 2.5 Mycobacterium tuberculosis 5 1.5 Mycoplasma pneumoniae 16 4.9 Chlamydia psittaci 7 2.1 Chlamydia pneumoniae 11 3.4 Virus 7t 21 Unknown 127 39.0 *Numbers include 14 episodes with dual organisms (the S milleri group with anaerobes, 4; S pneumoniae with H injluenzae, 3; S pneumoniae with M catarrhalis, 1; P pneumoniae with S aureus, l; S pneumoniae with the S milleri group, 1; S pneumoniae with C pneunumiae, 1; K pneumoniae with P aeruginosa, l; K pneumoniae with E coli, 1; K pneumoniae with C pneumoniae, 1). tinlluenza A, 3; influenza B, 1; parainfluenza, 2; cytomegalovirus, l. 1590 Clinical Investigations

of death for all these patients. Six of the patients who died were class IV and 14 were class V. Pathogens were detected in 13 of these patients. S pneumoniae was cultured in 3, K pneumoniae in 3, the S milleri group in 2, M tuberculosis in 2, Staphylococcus aureus in l, S pnewnoniae with S aureus in 1, and K pneumoniae with Pseudomonas aeruginosa in l. DISCUSSION The present study showed that pathogens could be detected in 61 % of patients requiiing admission to hospital for CAP, a percentage similar to that found in previous studies. 2-11 Despite extensive laboratory tests and various invasive procedures, the etiology of CAP was previously established in no more than 50 to 70% of patients at most except in a study from Israel, 8 in which causative agents were identified in over 80% of the patients. It is important to note that the number of definitive cases was small, and this indicated that it was difficult to make a precise diagnosis of the etiology of pneumonia. There are a number of possible reasons for this difficulty. First, antimicrobial agents may have been administered before hospitalization. Prior antibiotic therapy had been given to 38% of our patients, and this obviously lowered the diagnostic yield. Second, especially in the case of C pneumoniae infection, paired serum samples taken at a 2- to 3-week interval will not allow the detection of an increase in antibody levels, because this is often significantly delayed. Finally, there may have been additional causative pathogens not investigated. In our study, diagnoses were most often made from sputum cultures. Because of oropharyngeal contamination, many previous studies did not accept the results of sputum cultures for microbiological diagnosis. However, the reliability of sputum data can be improved by combining Gram staining and quantitative cultming of the sputum sample. It was reported that the bacterial flora, of sputum samples with fewer than 10 squamous epithelial cells per low-power field (X 100) in a Gram-stained smear, closely resembled the organism found in transtracheal aspirates. Is The culture of specimens obtained by TNA showed a high diagnostic yield, and complications of the procedure only occurred in two patients (mild pneumothorax in both cases). TNA is a safe and convenient technique that is vel)' useful for the diagnosis of CAP, especially in relation to the detection of anaerobes or microaerophilic streptococci such as the S milleri group. The most common pathogen in the present study S pneumoniae, as was the case in other coun- w a ~ tries. 2 - Ll However, only 7 of 75 cases revealed the positive blood cultures for S pneumoniae. This finding is lovver than that reported previously in other studies. 16 17 It is probably the main reason that 16 cases with S pneumoniae pneumonia received prior chemotherapy before hospitalization. The second most common organism was H influenzae. The incidence of H influenzae pneumonia varies considerably between countries, but tends to be higher in studies where many of the patients have underlying COPD. The incidence of M pneumoniae pneumonia (4.9%) was not very high. This low rate may also be attributed to the fact that patients with M pneumoniae pneumonia were hospitalized less frequently than those with other forms of CAP, because new macrolides and quinolones effective against mycoplasmas have made it easier to treat this type of pneumonia on an outpatient basis. Most episodes of K pneumoniae pneumonia occurred in patients with underlying diseases, and it was notable that this pneumonia revealed a high mortality (4 of 14 episodes were fatal) The rate of P aeruginosa pneumonia was quite high. The 8 patients in the present study were all aged 65 years or older. The underlying disease was COPD in 4 cases, cerebrovascular disease in 1, chronic liver disease in 1, old tuberculosis in 1, and diabetes in l. In all cases, the diagnosis was based on quantitative culture of sputum. One of the patients died, but the others improved. The prevalence of tuberculosis is still high in Japan, and we experienced five patients with caseous pneumonia during the study period. They initially received treatment on the basis of a diagnosis of common bacterial pneumonia. They were later referred to our hospital and were found to have tuberculosis. Because of the delay in reaching the correct diagnosis, tuberculosis exacerbated and caused the death of two patients. The S milleri group (Streptococcus anginosus, Streptococcus constellatus, and Streptococcus intermedius) colonizes the oral cavity and the upper respiratmy tract, and little attention in the literature has been paid to these organisms as pathogens in CAP. However, interest has focused in recent years on their role in pulmonary infection-these organisms are said to account for 2 to 12% of CAP.6,18,I9 It seems that the detection rate for the S milleri group is low unless TNA is done comprehensively. Use of TNA in the present study certainly increased detection of these organisms to a level of 3.7%. Further investigations will be needed to determine whether there is a difference between Japan and Western countries in the incidence of S milleri pneumonia. Many of our patients with S milleri pneumonia had CHEST I 114 I 6 I DECEMBER, 1998 1591

cornorbidities such as diabetes lllellitus or alcoholism. Five of 12 patients developed empyema. Mixed infection with anaerobk bacilli was recognized in tour patients. C pneumoniae is an emerging pathogen that has occupied a higher rank in the etiology of CAP in recent studies. A study pe1fonned in Finland during epidemic years showed that over 40% of CAP was caused by C pneumuniae. 20 However, few reports have ntentioned the incidence of C pneuttwniae pneumonia among patients with CAP in Jap<UI. The present study showed that the incidence of C pneunwniae pneumonia was 3.4% of CAP, indicating that it was one of the significant causes also in Japan. The ELISA laboratmy test for antibodies is widely available nowadays 21 22 and has made more screenings possible. The rnost conspicuous difference between our study and those done overseas is the low frequency of both viral and Legionclla pneumonia. The low incidence of viral pneumonia was attributed to the Ltct that we defined an episode as viral pneumonia only when it showed an elevation of antibodies for virus and yielded no positive bacterial cultures. The incidence of Legionella pneumonia in Japan remains far lower than in \Vestern countries despite thorough testing at many institutions in this country. 2 a Only two patients had Legionella pneumonia in the present series. The reason why Legionella pneumonia remains rare in Japan is unce1iain. lt might be because the design of buildings in Japan, particularly with regard to the air-conditioning systems, is different fi om that of Western countries. In Japan, most private homes have their own individual air-conditioning units, so that air does not pass through a cooling tower. Transmission of Legionella through air-conditioning systems is therefore unlikely to occur in p1ivate households. In most cases of conullunity- acquired Legionella pneumonia reported in Japan where the route of infection was identified, huntidifiers and hot spring water were responsible. The detection rate of Legionella spp. by culture is low, and the diagnosis is basej most often on antibody titers. The number of cases detected may increase if sensitive urinary antigen tests become more available for screening. Tbe present study may be the first prospective one to in vestigate the etiology of CAP in Japan. The results indicate that the etiology of CAP does not differ rnarkedly between Japan and Western countries except with regard to Legionclla pneumonia. S pneumouiae was the most common pathogen, and emerging or newly recognized pathogens, such as C pnew noniae and the S tnilleri group, were also significant. Racial and geographical differences may have little influence on the etiology of CAP. Recog- nitiou of the results will lead us to treat patients with prornpt antimicrobial therapy. REFERENCES Ht alth and Welfare Statistics Association of Japan. Annual Heport of National Health and Hygiene in Japan. Kousei No Shihyo 1996; 43:Sl -S.50J 2 Woodhead MA, Macfarlane JT, McCracken JS, et al. Prospective study of tlte aetiology and outwme of pneumonia in the community. Lancet 1987: 1:671-674 3 Research Committee of the British Thoracic Society and the Puhlic Health LaboratOJ)' Service. Community-acqu ired pueurnonia in adults in British hospitals in 1982-83: a survey of aetiolo!,')', rnmiality, prognostic factors and outcome. Q J Med 1987; 62:195-220 4 MarTie TJ. Durant H, Yates L. Community-acquired pneumonia requiring hospitalizatjon:.5-year prospective study. Rev Infect Dis 1989; 11 :586-599.5 Fang CD. Fine M, Orloff J, et al. New and emerging etiologies for commtmity acquired pneumonia with implications for therapy: a prospective multicenter study of 3.59 cases. Medicine 1990; 69:307-316 6 Bates JH, Campbell GO, Barron AL, eta!. Microbial etiology of acute pneumonia in hospitalized patients. Chest 1992; 101 :100.5-1012 7 Steinhoff D, Lode 1-l, Hnckdeschel C, et al. Chlamydia pnc11moniac as a cause of community-acquired pneumonia in hospitalized patients in Berlin. Clin infect Dis 1996; 22:9.58-964 f) Lieberman D, Schlaeffer F, Boldur 1, et al. Multiple pathogcr.s in adult patients admitted \lith community-acquired prre rnnonia: a one year prospective study of 346 consecutive patients. Thorax 1996;.51:179-184 9 Bohte R, va n Furth R, van d en Broek PJ. Aetiology of community-acquired pneumonia: a prospective study among adults requiring admission to hospital. Thorax 1995;.50:.543-.547 JO :--reill A\IL Martin IR, Weir H, et al. CornrTrunity-acquired pneumonia: aetiology and usefuln ess of severity criteria on admission_ Thorax HJ96;.51: 1010-1016 11 Ortqvist A, Valtonen M, Cars 0, et al. Oral empiric treatment of community-acquired pneumonia: a multicenter, doubleblind, randomized study comparing sparfl oxacin with ro>. itlrrornycin. Chest 1996; 110:1499-1506 L2 British Thoracic Society. Guidelines fih- the management of community acquircd pneumonia in adu lts admitted to hospita l. Br J Hosp Med 1993; 49:346 350 1:3 American 'fhoracic Society. Guidelines for the initial managf'ment of adults \lith community-acqui red pneumonia: diagnosis. assessment of, evf rity, and initial antimicrobial therapy. Am Hev Hc;,pir Db 1993; 14&: 1418 --1426 14 Fine MJ, Auble TE, Yealy OM, et al. A prediction rule to identify low-risk patients with community-acquired pneumoni fl. :\ Engl J Med 1997; 336:243-2.50 1.5 Mu rray PR, Washington JA. Microscopic and bacteriologic analys is o f expectorated sputu m. Mayo C lin Proc 19/.5; 50:339- :344 16 Chalasani l\"p, Valdccanas MA, Gopal AK, et al. Clinical utility of blood cu ltures in adult patients wi th communityacquired pneumonia without def1ned unde rlying risks. Chest l9\j5; 108:9:32-936 17 Lippmann ML. Goldberg SK, Walke11ste irr MD, et al. Bacte ri <U pneumococcal prreurnonia: a community hospital expcri t nce. Chest 199.5; 108:1608 --1613 18 Wo11g CA, Donald F,.VI acfi:u-lane JT. Streptococcus milleri 1592 Clinical Investigations

pulmonary disease: a r eview and clinical description of 25 patients. Thorax 1995; 50:1093-1096 19 Shinzato T, Saito A. The Streptococcus milleri group as a cause of pulmonmy infections. Clin Infect Dis 1995; 21: S238-S243 20 Kauppinen MT, Herva E, Kujula P, e t al. The etiology of community-acquired pneumonia among hospitalized patients during a Chlamydia pneumoniae epidemic in Finland. J Infect Dis 1995; 172:1330-1335 21 Brade L, Brunnemann H, Ernst M. Occurrence of antibodies against chlamydiallipopolysacchalide in human sera as measured by ELISA using an artificial glycoconjugate antigen. FEMS Immunol Med Microbiol 1994; 8:27-41 22 Kishimoto T, Kubota Y, Matsushima T, eta!. Assay of specific anti-chlamydia pneumoniae antibodies by ELISA method: evaluation of ELISA kit using outer membrane complex [in Japanese]. Kansenshogaku Zasshi 1996; 70:821-829 23 Arakawa M, Inamatsu T, Ezaki T, e t a!. A summary of Legionella pneumonia in Japan from 1979 to 1992 [in Japanese]. Kankyo Kansen 1993; 8:1-10 CHEST I 114 I 6 I DECEMBER, 1998 1593