The Laboratory Evaluation of Opportunistic Pulmonary Infections

Dr. James H. Shelhamer (Critical Care Medicine Department, Clinical Center, National Institutes of Health [NIH], Bethesda, Maryland): Because the spectrum of processes that cause pulmonary disease in immunosuppressed patients is so broad, eminently logical reasons exist for obtaining specific information about the causative process—for example, to make certain that the patient receives appropriate therapy and that inappropriate, potentially toxic, and expensive therapies are avoided. However, these arguments for specific diagnosis are countered by valid concerns that the most useful diagnostic tests will be invasive and may ultimately be nondefinitive, thus incurring expense and morbidity without altering the clinician's initial management plan.

During the past several years, the ability to diagnose infectious causes of pulmonary disease has increased considerably. Sputum, tracheal secretion, bronchoalveolar lavage, blood, and even urine samples can be examined directly. The organism can be visualized by a tinctorial or fluorescent stain, or evidence of an organism may be detected by tests for specific antigens or nucleic acid. In some instances, with the use of newer methods, extremely sensitive assays can detect the presence of a pathogen within 24 or 48 hours of sample submission. Rapid-culture techniques can provide additional information within days about the presence of viruses, mycobacteria, or fungi as well as bacteria.

Major issues for clinicians who manage immunosuppressed patients have emerged. These issues include 1) the availability of assays for specific pathogens; 2) the sensitivity and specificity of each assay; 3) whether determining the presence of organism by culture, antigen detection, or detection of its nucleic acid proves that the organism is the cause of pulmonary dysfunction; 4) whether a particular test result is accurate and reliable enough to use to determine therapy and to make decisions about epidemiologic issues, including isolation precautions; and 5) whether the morbidity that some procedures entail, the staff time that is required for the collection and processing of the samples, and the expense that the procedures incur are truly warranted. In this Combined Staff Conference, we review some of the major advances that have been made in rapid identification of microbial pathogens, and we assess the role of these assays in clinical practice.

Dr. Vee J. Gill (Clinical Pathology Department, Clinical Center, NIH): As the range of pathogens that a microbiology laboratory can detect increases, clinicians need to recognize the optimal type of specimen to submit (Table 1) and the types of tests that can be done (Table 2). Although Gram stain and culture remain the mainstay of traditional microbiologic tests, the implementation of new methods in the clinical laboratory has led to more sensitive and more rapid detection techniques. These, in turn, can benefit patient care.

For routine bacteria, Gram stains of lower respiratory tract secretion or tissue samples are still important for providing rapid initial guidance on the morphologic characteristics of the bacteria. Although Gram stain of sputum has fallen into disfavor with some clinicians, an abundance of organisms (10 per high-power field) in an adequate specimen with neutrophils (25 per high-power field) is still considered highly suggestive of significance. However, both Gram stains and cultures are admittedly difficult to interpret in neutropenic patients or in those who have already received extensive treatment with antimicrobial agents. Some clinicians are enthusiastic about specimens obtained by protected brushes or bronchoalveolar lavage, both of which reduce upper airway contamination of specimens. Such specimens can then be cultured quantitatively [1, 2]. Quantitative culturing can add sensitivity and specificity to the diagnosis of bacterial pneumonia, but it is time consuming. Many clinicians and laboratory directors question whether the additional information that is gained justifies the added expense.

In subsequent sections, we present the current state of the art as well as new methods for the diagnosis of viruses, Legionella species, Mycoplasma pneumoniae, Chlamydia pneumoniae, Pneumocystis carinii, mycobacteria species, and fungi. For each of these major groups of pathogens, specific technical advances are ready to be incorporated into clinical laboratory procedures, and newer methods are being developed for use in the near future.

Rapid isolation and identification of respiratory viruses within 1 to 2 days can now be done using shell vial cultures that are stained with virus-specific fluorescent monoclonal antibodies. With the shell vial culture, the specimen is centrifuged onto a tissue culture monolayer contained within a screw-capped vial. The monolayer, which is grown on a coverslip within the vial, can be stained using various respiratory virus pools or virus-specific fluorescent monoclonal antibodies. By setting up multiple vials, one can stain a coverslip at 1-day or 2-day (or other) intervals as desired. This method is shown schematically in Figure 1. The technical simplicity of shell vial cultures has made it possible for them to be done in-house at many laboratories, providing useful information in a timely fashion.

For the laboratory diagnosis of Legionella pneumonia, bronchoalveolar lavage fluid samples and lung biopsy specimens remain the most reliable specimens and are preferred to expectorated sputum samples. Currently available methods include direct fluorescent monoclonal antibody staining and cultures using supplemented media for Legionella species. A urinary antigen radioimmunoassay for L. pneumophila serotype 1 was introduced several years ago but has not been readily available because the radioimmunoassay format has not been practical for clinical microbiology laboratories to set up. A commercially available urinary antigen assay that uses an enzyme-linked immunoassay now enables more laboratories to use this test. However, because this test is highly specific for L. pneumophila serotype 1, infection with a different serotype or species would go undetected. Successful use of the polymerase chain reaction on respiratory specimens for the diagnosis of Legionella infection has recently been reported.

Laboratory diagnosis of infection caused by M. pneumoniae and C. pneumoniae relies heavily on the detection of serologic conversion. Both agents can be cultivated, but with difficulty, so that this service is not routinely offered by clinical laboratories. In the future, polymerase chain reaction may be used to establish an early diagnosis for these pathogens.

In addition to shell vial culturing, fluorescent antibody staining of smears made directly from specimens such as bronchoalveolar lavage fluid samples can yield quick results. Reagents that can be used for direct staining of specimens are available for respiratory syncytial virus, cytomegalovirus, herpes simplex virus, and varicella-zoster virus. Commercially available enzyme immunoassays have become more diverse and offer technically easy yet rapid testing; such tests are already available for respiratory syncytial virus and influenza A virus. Methods that provide results on the same day that the specimen is submitted, either by direct staining or by enzyme immunoassay, are attractive to clinical laboratories, but each assay must be assessed to ensure that its sensitivity and specificity are adequate.

Advances in the detection of P. carinii are attributable to improvement in rapid direct staining of induced sputum, bronchoalveolar lavage fluid, and lung biopsy specimens. The ability of microbiology laboratories to provide same-day staining with stains such as monoclonal fluorescent antibody stains has resulted in more sensitive and timely detection of pneumocystis pneumonia. Polymerase chain reaction has also been shown to increase detection of P. carinii, particularly when done on induced sputum specimens.

Mycobacterial isolation and identification have been significantly improved by the use of more rapid-culture techniques such as the radiometric BACTEC (Becton-Dickinson, Sparks, Maryland) system. Used in conjunction with new commercially available DNA probes (Accuprobe, Gen-Probe, San Diego, California), the time required for the isolation and identification of the significant mycobacterial pathogens has been greatly reduced. Figure 2 describes this innovative probe technology, which detects specific ribosomal RNA targets, allowing identification of an isolate within 1 day by a simple chemiluminescent assay.

Because direct detection of Mycobacterium tuberculosis in sputum specimens using molecular techniques such as polymerase chain reaction may yield even more sensitive and rapid diagnosis, continued progress in the use of these methods is especially warranted.

Advances in fungal diagnostics have been limited. Direct detection by smears, culture, and specific antigen assays (for Cryptococcus and Histoplasma species) are the methods on which clinicians currently rely. One improvement that has increased the utility of direct smears for fungi is the use of the calcofluor white stain. This is a rapid, easy-to-read stain in which fungal elements brightly fluoresce, but it requires the use of a fluorescent microscope. Other recent developments include commercial DNA probes (Gen-Probe) [2-8] that are available for the identification of H. capsulatum, Coccidioides immitis, and Blastomyces dermatitidis. Use of these probes greatly reduces the time and increases the accuracy of identification of these pathogens.

As we have mentioned, polymerase chain reaction or similar amplification techniques are currently under development for the detection of many infectious agents. In most instances in which polymerase chain reaction has been applied, the results suggest that amplification-based detection will greatly increase our ability to make specific diagnoses. The basic polymerase chain reaction is a powerful tool that may eventually permit more definitive and rapid identification of difficult-to-detect or slow-growing organisms as well as of organisms that cannot be cultured. Polymerase chain reaction results in a doubling of copies of the specified target DNA with each round of amplification, eventually resulting in million-fold levels of amplification. Figure 3 shows how this amplification is achieved. The product of amplification must then be detected; this can be done in various ways, including agarose gel electrophoresis followed by staining with ethidium bromide. Using a sequence-specific probe to confirm the specificity of the amplified DNA product also increases sensitivity, but traditionally this probing has required the use of radioisotopes. Alternative nonradioisotopic methods [such as chemiluminescent or enzyme-coupled probes] to detect specific amplification products are preferred for use in the clinical laboratory setting and will help speed the incorporation of polymerase chain reaction as a diagnostic tool. The immediate availability of polymerase chain reaction testing in the clinical laboratory has been hindered by 1) the complexity and time-consuming nature of the assays, 2) the need for personnel trained in molecular methods to do the testing, and 3) the need to define an optimal protocol for each organism and then to document the level of sensitivity and specificity of that protocol. Other important problems, such as contamination leading to false-positive reactions, the presence of inhibitors in clinical specimens, and assessing the clinical relevance of positive polymerase chain reaction assays, remain to be resolved. Polymerase chain reaction assays to improve the accuracy of the diagnosis of cytomegalovirus, L. pneumophila, Mycoplasma pneumoniae, C. pneumoniae, P. carinii, and Mycobacterium tuberculosis infections have been worked on by many investigators; these assays are discussed in the following sections.

Dr. Stephen W. Crawford [Fred Hutchinson Cancer Research Center, Seattle, Washington]: The diagnosis of viral pneumonia requires 1) clinical evidence of a pneumonia, 2) the presence of virus in the lung or blood, and 3) a causal link between the isolated virus and the pulmonary disease.

Viruses associated with pneumonia in immunosuppressed patients include double-stranded DNA viruses such as herpesviruses (cytomegalovirus, herpes simplex virus, varicella-zoster virus, Epstein-Barr virus, and human herpesvirus 6) and adenovirus as well as single-stranded RNA viruses such as influenza A and B, paramyxoviruses (respiratory syncytial virus and parainfluenza viruses 1, 2, and 3), measles, and picornaviruses. These are obligate intracellular parasites. Replication requires host cellular machinery, and release of completed versions may be associated with cell lysis. Thus, unlike bacteria and fungi, viral presence indicates infection, not colonization. The degree to which the infection is accompanied by an inflammatory response and by organ dysfunction determines whether the viral infection causes “disease.” Excretion of herpes viruses, such as cytomegalovirus, can occur without pneumonia [3-5]. Thus, culture of virus from respiratory specimens indicates infection, but not necessarily disease, and isolation of virus from sites remote to the lung does not necessarily confirm the cause of the pneumonia.

Isolation of respiratory viruses from either respiratory tissues or secretions, or, in some instances, from the blood, are of potential clinical importance. Detection of virus in lung tissue is the most convincing evidence of a viral cause of a pneumonia. The significance of the detection of viruses in the upper respiratory tree varies with the type of virus. Respiratory viruses that contain RNA genomes, such as myxoviruses and paramyxoviruses (influenza, respiratory syncytial virus, parainfluenza), do not establish persistent infection (latency), and, with these viruses, isolation almost always indicates active viral disease. These viruses are rarely found in the absence of symptoms of upper respiratory infection [6], and their presence is of limited duration. Although nasopharyngeal viral excretion strongly suggests that a respiratory virus is the cause of pneumonia, the true specificity of such isolation in determining the cause of pneumonia is unclear; this isolation may represent an upper respiratory pathogen unrelated to the process causing pneumonia.

Herpesvirus and adenovirus that contain DNA may result in lifelong persistence of infection and are capable of reactivation. Excretion from extrapulmonary sites, including the throat and blood, suggests disease but is not diagnostic for a pneumonia and may occur in asymptomatic patients [4-6].

Respiratory viruses can be detected by many techniques, including culture (conventional tube or shell vial cultures), detection of viral antigens by monoclonal antibody staining or by enzyme immunoassays, detection of viral nucleic acids by in situ hybridization or polymerase chain reaction, and serologic studies. Standard “tube” culture using cell types that are sensitive to viral infection and that show typical cytopathic changes when infected is the “gold standard” for respiratory virus detection. Tube culture may require weeks of incubation in some cases (for example, cytomegalovirus). Rapid culture of most respiratory viruses has been achieved with centrifugation (shell vial) cultures, in which the presence of viral replication in the tissue culture is confirmed with fluorescent monoclonal antibodies within 1 to 2 days of inoculation [7, 8]. The sensitivity and specificity of centrifugation culture of bronchoalveolar lavage fluid for the detection of cytomegalovirus (and presumably of other respiratory viruses) approximates that of culture of lung tissue [9]. Culture is sensitive for cytomegalovirus and herpes simplex virus, whereas respiratory syncytial virus and parainfluenza are successfully cultured less often. False-negative results due to technical factors (such as factors in blood and respiratory secretions that are cytotoxic for the tissue culture monolayer) are seen, and specialized laboratories are required to maintain appropriate cell cultures for viral incubation.

Serologic conversion provides evidence of exposure. However, it lacks sensitivity and specificity for a given episode of pneumonia, and documentation takes several weeks. Serologic evidence is often detected after the clinical need to establish a diagnosis. Serologic studies are of value in identifying immunosuppressed patients at risk for reactivation of latent viruses, such as cytomegalovirus and herpes simplex virus [10].

Characteristic histologic and cytologic abnormalities can be seen in cytologic and histopathologic specimens in some cases of viral pneumonia. “Owl's eye” Cowdry type A intranuclear inclusions are relatively specific for cytomegalovirus pneumonia, “smudge” cells for adenovirus, and the respiratory syncytial cell for respiratory syncytial virus infection, but these are insensitive markers. Such cytologic changes are specific when seen in cytospin preparations for bronchoalveolar lavage, but they are of low sensitivity [9, 11].

Specific and sensitive detection of viral antigens within infected cells can be achieved from respiratory secretion (including bronchoalveolar lavage fluid) and blood samples using labeled monoclonal antibodies by most virology and pathology laboratories within hours of sample collection [9, 11-13]. Direct fluorescent antibody staining of bronchoalveolar lavage or of nasopharyngeal secretions correlates closely with culture results, and the sensitivity probably is increased with the inclusion of pooled monoclonal antibodies to multiple antigenic sites [11].

Similarly, peroxidase-labeled antibody staining of the peripheral blood buffy coat is highly specific in some immunosuppressed hosts (such as patients receiving transplants) for tissue-invasive cytomegalovirus diseases, with the presence of antigenemia correlating with disease. The sensitivity of peroxidase-labeled monoclonal antibodies to the cytomegalovirus-encoded immediate-early antigen, pp65, is similar to that of culture in most patients and tends to be detected earlier than culture [14-21]. Commercially available enzyme immunoassay kits permit the detection of viral antigens in respiratory secretions. These assays appear to be most useful in the detection of respiratory syncytial virus, parainfluenza virus, and possibly influenza viruses in nasopharyngeal specimens [22, 23].

Detection of viral DNA is possible from tissue with in situ hybridization and from respiratory specimens and blood samples with polymerase chain reaction [24, 25]. Polymerase chain reaction of blood leukocytes for cytomegalovirus is more sensitive than either culture or assay for antigenemia [26-30]. However, because of its high sensitivity, it may detect a small copy number of virus genome with a resulting low positive but high negative predictive value for disease [31, 32]. Additionally, prompt processing (within hours) of clinical specimens is required. Thus, the clinical utility of peripheral blood polymerase chain reaction is probably limited to excluding cytomegalovirus as a potential cause of an episode of pneumonia. As research tools, polymerase chain reaction and in situ hybridization will help to clarify the biology of viral pneumonia.

It is sometimes difficult to determine with certainty that a virus isolated from the lung is the cause of a pneumonic process (Table 3). The strongest evidence is both detection of typical histopathologic changes (such as intranuclear inclusion bodies typical of cytomegalovirus infection) and isolation of the virus in culture in the absence of other identifiable causes of pneumonia. However, such manifestations of infection are not uniformly present [33], and lung tissue is not always available for pathologic evaluation. Often the diagnosis is presumed, on the basis of viral detection from appropriate fluids, in a patient at risk who has characteristic clinical signs.

Bone marrow transplant recipients with diffuse pneumonia have a high mortality rate if cytomegalovirus is cultured from the lung, regardless of the presence of typical cytomegalovirus infection histopathology [10, 29]. Also, isolation of cytomegalovirus from bronchoalveolar lavage fluid specimens or blood samples is associated with a high (> 50%) incidence of subsequent pneumonia among asymptomatic bone marrow transplant recipients [4, 6]. In addition, quantitation of cytomegalovirus in bronchoalveolar lavage fluid specimens does not correlate with disease [34]. Therefore, the operational definition of cytomegalovirus pneumonia after marrow transplantation is detection of cytomegalovirus from lungs in the presence of pneumonia. The diagnosis of viral pneumonia among recipients of solid organ transplantation may be similarly viewed. However, many centers restrict the diagnosis of definite viral pneumonia to those cases with typical cytologic changes in which no other pathogens have been identified [12]. This approach may under-represent the true incidence of viral pneumonia. Cytomegalovirus rarely causes pneumonia in patients with the acquired immunodeficiency syndrome (AIDS) despite the frequent isolation of the virus in these patients. The presence of cytomegalovirus does not appear to influence mortality, regardless of treatment [35, 36].

All immunosuppressed patients with diffuse pneumonia should be evaluated for viral pneumonia. Among patients seropositive for cytomegalovirus, an antigenemia test with negative results probably excludes cytomegalovirus pneumonia. Positive results from nasopharyngeal swab by enzyme immunoassay, direct fluorescent antibody, or culture may increase suspicion of respiratory viral infection (respiratory syncytial virus, influenza, parainfluenza) among patients with concomitant symptoms of upper respiratory infection, but they do not confirm the cause. Minimal evaluation should include bronchoscopy with bronchoalveolar lavage. The degree to which lung biopsy helps in the diagnosis of viral infection is unclear. All bronchoalveolar lavage fluid specimens should be examined for viruses by routine cytologic studies and, ideally, for cytomegalovirus and herpes simplex virus by centrifugation culture. During endemic seasons, cultures for respiratory syncytial virus, influenza, and parainfluenza should be done. If no viral culture facilities are available, monoclonal fluorescent antibody staining with probes specific for these viruses should detect most cases.

Dr. Thomas C. Quinn (National Institute of Allergy and Infectious Diseases, NIH): Legionella species, Mycoplasma pneumoniae, and C. pneumoniae are now well-recognized pathogens that warrant immediate therapy. None of the organisms grow on standard bacteriologic media; none are identified on Gram stain; and none respond to the standard empiric choice of penicillin or cephalosporin antibiotic therapy frequently used to treat pneumonia.

Fifteen serogroups of L. pneumophila based on surface antigen analysis and 33 other Legionella species have now been identified, many of which have been shown to be responsible for pneumonia [37-39]. Serogroup 1 is responsible for about 80% of disease caused by L. pneumophila, with serogroups 4 and 6 responsible for most other cases [38]. Infections caused by Legionella species other than L. pneumophila are uncommon, constituting less than 20% of infections. Of these pathogens, L. micdadei, L. longbeachae, L. dumoffii, and L. bozemanii appear to be the most common.

Five methods are currently used for the laboratory diagnosis of Legionella infections (Table 4). These include isolation of the organism on selective culture medium, determination of antibody level, detection of the bacterium in tissue or body fluid specimens using direct fluorescent antibody tests, detection of antigenuria, and detection of DNA by the polymerase chain reaction.

Legionella species can be isolated from clinical specimens on selective medium using supplemented charcoal yeast extract medium (BCYE α) [38, 40]. The organism has been successfully isolated from sputum, transtracheal aspirates, endotracheal suction specimens, blood, biopsied lung tissue, pleural fluid, bronchial lavage fluid, pericardial fluid, and peritoneal fluid and from the respiratory sinuses. The disadvantage of culture is that it takes 5 to 10 days for colonies to appear, and the specimens must be processed in a precise manner on the abovementioned selective media for optimal isolation.

Acute and convalescent serologic studies are of diagnostic value, but it takes 6 to 8 weeks for the antibody titers to increase, thereby limiting the diagnostic value of these studies early in the course of the infection [41]. In immunocompromised patients, the antibody response may be severely impaired. Most laboratories use the indirect immunofluorescent antibody technique to determine antibody concentrations. About 75% of immunologically normal patients with culture-proven legionnaire's disease caused by L. pneumophila serogroup 1 develop a fourfold increase in titer by 8 weeks after onset of illness. Because up to 30% of healthy populations sampled have L. pneumophila serogroup 1 antibody titers of 1:128 or greater, only a fourfold increase in titer can be considered significant [41, 42]. Cross-reactions have also been reported in patients with other causes of pneumonia, and specificity has been estimated to be 90% for a fourfold titer increase.

The direct fluorescence antibody test is a rapid-detection test that uses a monoclonal antibody to all serogroups of L. pneumophila[38, 43]. This technique has been used successfully with expectorated sputum, endotracheal suction aspirates, biopsied lung tissue, and transtracheal aspirates. Use of secretions or biopsied tissue obtained by bronchoscopy has not resulted in high yield. The true sensitivity of the direct fluorescence antibody test is unknown, although 25% to 70% of patients with culture-proven L. pneumophila infection have positive results from direct fluorescent antibody tests [44]. The test specificity is greater than 90%, but cross-reactivity to Bordetella pertussis occurs. Results of the direct fluorescence antibody tests of sputum remain positive for 2 to 4 days after initiation of specific antibiotic therapy for L. pneumophila.

L. pneumophila serogroup 1 antigenuria can be detected using a commercial radioimmunoassay [45]. Cross-reactions between serogroups are uncommon, thereby limiting the usefulness of this test for the diagnosis of other Legionella infections. The advantage of this urine test is its high sensitivity, estimated to be 95% in culture-proven cases and 80% in patients with serologically proven disease. Specificity is also very high (estimated to be 99%). An enzyme-linked immunoassay is now commercially available. Test results can be obtained the day of testing, although it should be noted that they may remain positive for weeks to months after recovery from pneumonia.

Although not commercially available yet, amplification of Legionella nucleic acids by polymerase chain reaction has the potential to offer rapid results and to increase the sensitivity of current detection methods used on respiratory samples within a 24-hour period [46]. Polymerase chain reaction has been proven effective in the detection of L. pneumophila in bronchoalveolar lavage fluid specimens, nasopharyngeal swabs, and sputum samples. In a recent study of 40 immunocompromised patients with pneumonia at the NIH, we were able to use bronchoalveolar lavage fluid samples to detect L. pneumophila in 4 patients (10%) [47]. Polymerase chain reaction is also currently used for the detection of Legionella species in environmental samples [48]. Because primers are used for the detection of the 5s rRNA gene of Legionella species, polymerase chain reaction is capable of detecting Legionella species other than L. pneumophila.

The laboratory diagnosis of Mycoplasma pneumoniae infection can be made either by culture on selective media or by detection of an appropriate increase in specific antibody titer (Table 5). Mycoplasma pneumoniae can be isolated from throat washings, sputum, or throat swabs 7 to 10 days after inoculation in broth media [49]. A presumptive identification of Mycoplasma pneumoniae can be made if colonies show heme absorption of red blood cells.

Serologic diagnosis depends on a fourfold increase in complement fixation titer for acute and convalescent sera [50]. IgM antibody first appears 1 week after infection, peaks at 4 to 6 weeks, and does not start to decrease until 4 to 6 months later. One third to three fourths of patients with Mycoplasma pneumoniae develop a fourfold or greater increase in titers of cold hemagglutinin [51], but these IgM antibodies are not diagnostic of Mycoplasma pneumoniae infection because they can be associated with various other infections as well as with connective tissue and neoplastic disorders. Polymerase chain reaction for Mycoplasma pneumoniae infection has also recently become available [52, 53]. In a recent study of 34 patients with respiratory illness and evidence of pneumonia, evidence of Mycoplasma pneumoniae infection was obtained in 10 patients (29%); in 8 of these 10 patients, results of both polymerase chain reaction and serologic studies were positive. However, it should be noted that the detection of Mycoplasma pneumoniae in the respiratory tract does not necessarily correlate with respiratory disease, at least not in an immunocompetent patient. Consequently, serologic tests should be used in addition to polymerase chain reaction to distinguish between acute and persistent infections.

All three Chlamydia species—C. psittaci, C. trachomatis, and C. pneumoniae—have been associated with pulmonary infection [54]. Chlamydia psittaci is responsible for psittacosis, a disease of birds, of which man is an accidental host. Chlamydia trachomatis is a common cause of pneumonitis in infants born to women with genital infection. Of the three species, C. pneumoniae, formerly known as TWAR, is the most common cause of acute respiratory disease, responsible for 5% to 15% of cases of community-acquired pneumonia [55], and according to one survey, for 10% of cases of pneumonia in immunocompromised patients [56].

Laboratory diagnosis usually depends on isolation of the organism, an increase in serologic titer, or polymerase chain reaction (Table 6). Because C. pneumoniae is a fastidious organism, it is difficult to isolate and requires in vitro tissue culture. It has been isolated in either HL cells or HEp-2 cells [57, 58]. Nasopharyngeal swabs provide the best specimen, but they must be transported in specialized transport media and inoculated immediately in the cell culture. Culture results are not usually positive for 72 hours or longer.

Most investigators have relied on serologic diagnosis using the microimmunofluorescence test. Grayston and colleagues [55] proposed criteria for serologic diagnosis of C. pneumoniae infection that have been used by many laboratories and clinicians. Patients with acute infection have a fourfold increase in the IgG titer, a single IgM titer of 1:16 or greater, or a single IgG titer of 1:512 or greater. Patients with past or preexisting infection have an IgG titer of at least 1:16 and less than 1:512. Because the microimmunofluorescence assay is not widely available, the complement fixation test has been used as an alternative. This test is genus-specific and should be used primarily for the diagnosis of lymphogranuloma venereum and psittacosis. Grayston and colleagues [55] found that about one third of patients with C. pneumoniae had detectable complement fixation antibody.

Because of the limitations of both culture and serologic studies for the establishment of early diagnosis, polymerase chain reaction has recently been developed, providing a sensitive and specific assay for early detection [56, 59]. Using primers for either the major outer membrane protein or the 16s rRNA gene of C. pneumoniae, polymerase chain reaction has been shown to be highly sensitive and specific. In one study at the Clinical Center, National Institutes of health, we screened 132 culture-negative bronchoalveolar lavage fluid specimens from 108 immunocompromised patients (34% of whom tested positive for human immunodeficiency virus [HIV] infection) and 7 healthy volunteers [56]. Thirteen specimens (9.8%) from 12 immunocompromised patients (11.1%) yielded positive results. No healthy volunteer had a positive polymerase chain reaction in a bronchoalveolar lavage specimen. Only 2 of these 12 infected patients had other microbiologic agents, including respiratory syncytial virus and P. carinii, implicated in the pneumonia. Only 4 of the 13 patients had diagnostic titers of antibody to C. pneumoniae. The inability of patients to respond to specific antigenic stimuli promptly or at all may have been caused by their immunocompromised state. Thus, more rapid tests such as polymerase chain reaction may be a useful addition to the more conventional methods.

Dr. Joseph A. Kovacs (Critical Care Medicine Department, Clinical Center, NIH): Despite the extensive use of prophylaxis, Pneumocystis carinii remains one of the most common causes of opportunistic pneumonia in immunosuppressed patients, especially patients with HIV infection. Human P. carinii cannot be cultured in vitro at present; therefore, unambiguous diagnosis of P. carinii pneumonia requires direct detection of the organism in an appropriate clinical specimen. Furthermore, because available data suggest that P. carinii is not a colonizing pathogen, detection of P. carinii in pulmonary samples from patients who have not received treatment, at least by standard tinctorial or immunofluorescent assays, is currently equivalent to diagnosis of P. carinii pneumonia.

Bronchoscopy using either bronchoscopic biopsy or bronchoalveolar lavage has a very high diagnostic yield. The major advantage of bronchoscopic biopsy is that multiple small specimens can be obtained and histopathologically examined for pathogens other than P. carinii. For P. carinii pneumonia, the sensitivity of transbronchial biopsy or bronchoalveolar lavage individually is more than 90% [60-62].

Sputum induction using hypertonic saline has replaced bronchoscopy as the initial diagnostic procedure at many centers. This noninvasive procedure has a diagnostic yield of 50% to 90%. Increased experience in both obtaining and examining the sample has resulted in yields of more than 80% at many institutions [63, 64]. It is a rapidly done technique with relatively few risks, although it may not detect other potential causes of pneumonia. Negative results may require follow-up bronchoscopy if clinical circumstances warrant definition of the causative process.

For many years, the standard method for detecting P. carinii in clinical samples has been tinctorial staining [65]. Tinctorial stains include the cyst wall stains, such as toluidine blue-O and methenamine silver stains, which will stain only the cysts of P. carinii and not the more numerous trophozoite form of the organism. Cyst wall stains also stain fungi. The second category of tinctorial stains, Giemsa and Diff-Quik, stain not only the intracystic sporozoite, but also the more numerous trophozoite form; unfortunately, they will also stain other organisms, such as yeast, as well as cells and subcellular debris, and this makes them more complex to interpret. Recently, monoclonal antibodies to human P. carinii, which react with both cysts and trophozoites, have been found to be very useful diagnostically [64]. Prospective blinded studies have shown that immunofluorescent assays have a higher sensitivity than tinctorial stains, primarily in the examination of induced sputum samples (69% to 92% compared with 28% to 80%, respectively) [64, 66-68]. These monoclonal antibodies are not known to cross-react with any other microbial forms likely to be found in human specimens and thus are highly specific.

Detection of organisms using the polymerase chain reaction has recently been applied to P. carinii[69, 70]. The most commonly used primers are based on the mitochondrial ribosomal RNA [69]. The major advantage of polymerase chain reaction is an increased sensitivity compared with tinctorial or immunofluorescent stains, but this comes at a cost of decreased specificity, which may be due to low numbers of organisms that are not causing the disease or to polymerase chain reaction-related technical problems. In a recent study comparing immunofluorescence to an enzyme-based polymerase chain reaction assay for examination of both sputum and bronchoalveolar lavage fluid specimens, the sensitivity in sputum specimens increased from 78% with immunofluorescence to 100% with the polymerase chain reaction assay with virtually no loss of specificity [71]. For bronchoalveolar lavage fluid samples, the two techniques were equally sensitive at 100% and virtually identical in terms of specificity. Based on these and other data, it appears that in bronchoalveolar lavage fluid examination, polymerase chain reaction will have no role in diagnosis of P. carinii infection, but it may have a diagnostic role in sputum examination. This test may also have a role in the typing of strains as well as in epidemiologic and environmental studies.

A final issue in the diagnosis of P. carinii pneumonia is the effect that aerosol pentamidine prophylaxis has on the detection of P. carinii. The use of aerosol pentamidine prophylaxis has been reported to be associated with an increase in upper lobe disease and a concomitant decrease in the sensitivity of both routine bronchoalveolar lavage and induced sputum examinations [72, 73]. Sensitivity in this situation can be improved to more than 90% by multilobe lavage including the upper and the middle lobe [74, 75]. The rare cases of extrapulmonary disease, which also appear to be more common in patients receiving aerosol pentamidine, can be diagnosed either clinically (for example, choroiditis) or histopathologically [76, 77].

Dr. Henry Masur [Critical Care Medicine Department, Clinical Center, NIH]: To diagnose mycobacterial disease in immunosuppressed patients, three basic issues must be addressed: 1) whether mycobacterial disease is present, 2) whether the mycobacterial isolate is Mycobacterium tuberculosis, and 3) whether the mycobacterium is clinically important. Mycobacterium tuberculosis receives special consideration not only because of its pathogenicity, since other mycobacteria can also cause life-threatening disease, but because of its potential for person-to-person transmission and the epidemiologic and economic need to determine if respiratory isolation for the patient is or is not necessary.

To determine if mycobacterial disease is present, most laboratories still rely on direct microscopy to detect organisms in smears from appropriate specimens [78, 79]. Indirect tests relying on detection of IgG or IgM antibodies in serum, on detection of enzymes such as adenosine deaminase or antigens in various types of clinical specimens, or on detection of cell wall lipids by chromatographic techniques have evoked considerable interest, but none of these tests shows a combination of sensitivity, specificity, and practicality adequate to warrant its routine use [78]. Direct microscopy is most often done on respiratory secretions, although it can obviously be applied to any clinical specimen. For most adults, expectorated or induced sputum specimens are screened initially, with bronchoalveolar lavage or lung biopsy necessary only in difficult diagnostic situations [78, 80]. With adequate encouragement, at least 80% of patients should be able to produce an expectorated or induced sputum sample. Gastric lavages are not commonly used, except in young children who cannot produce sputum; these specimens are useful for culture and for examination by direct techniques, despite earlier concerns about the frequent occurrence of nonpathogenic mycobacterial saprophytes in the stomach [81]. The sensitivity of direct microscopy depends on the stain used, the quality of the specimen, and the time and expertise involved in examining it; generally, at least 10⁵ organisms/mL of specimen must be present, although some laboratories appear to be capable of detecting 10² to 10³ organisms/mL [78]. This technique is very specific for identifying the presence of mycobacteria, but not for species identification. Laboratories are encouraged to use a fluorochrome stain such as auramine rhodamine for screening rather than a tinctorial stain such as the Ziehl-Neelson stain because the former is faster for screening specimens, requires less expertise to be read accurately, and is believed by many to be more sensitive.

Nucleic acid probes for detecting mycobacteria directly in clinical specimens initially elicited considerable interest but did not prove to be sufficiently sensitive to detect culture positive specimens reliably. Nucleic acid amplification techniques such as polymerase chain reaction have a greater potential for detecting fewer organisms and have been shown to be quite useful in detecting Mycobacterium tuberculosis; polymerase chain reaction based on the insertion element 156110 [81-83] and the rRNA amplification system developed by Gen-Probe, the Amplified Mycobacterium Tuberculosis Direct Test [82, 84], have shown very high sensitivity for smear-positive specimens, and have detected Mycobacterium tuberculosis in a substantial fraction of smear-negative samples. These techniques are specific for Mycobacterium tuberculosis and do not give a positive signal when other mycobacteria are present, but they may give a positive signal in some patients with previously treated Mycobacterium tuberculosis disease or who are purified protein derivative (PPD)-positive without active disease, which complicates the interpretation of a positive result [82]. A negative amplification result is useful for determining that acid-fast organisms, if present, are not Mycobacterium tuberculosis and that Mycobacterium tuberculosis is unlikely to be present, which can be epidemiologically important. A positive result may indicate an active, inactive, or latent process. The use of polymerase chain reaction and other molecular methods is an exciting prospect for the rapid, accurate diagnosis of mycobacterial disease. Experience thus far has been limited to a few laboratories that have the expertise to develop their own methods. As yet, unfortunately, no simple test is available commercially in the United States to enable laboratories and physicians to benefit from the potentially increased sensitivity and speedy detection that these methods permit.

Culture of a specimen is the most sensitive technique for establishing the presence of active mycobacterial disease. Specimens can be plated on solid agar such as Lowenstein-Jensen culture medium, or in liquid medium such as Middlebrook 7H/11. The time until growth is detected depends on the size of the inoculum, the specific mycobacterial species and strain, and the presence of inhibiting drugs. Liquid systems are generally faster than solid medium systems for detecting growth [78, 79]. Radiometric culture systems that operate on the principle of a radiolabeled substrate being metabolized into a detectable gas are considerably faster and are being used by an increasing number of large laboratories; results are generally available 7 to 10 days earlier than with conventional solid media [79].

Respiratory specimens are most often the material cultured, but any potentially infected body fluid or tissue should be considered. In some instances, certain mycobacteria other than Mycobacterium tuberculosis may grow, but this may represent colonization or even contamination of the specimen. Mycobacterium avium complex can often be the former, and Mycobacterium gordonae, the latter. Pleural fluid and blood should not be overlooked as specimens for culture. In patients with human immunodeficiency virus infection, for example, blood cultures for Mycobacterium avium complex can be positive in many patients, especially those with low CD4+ lymphocyte counts [85]. If the BACTEC radiometric culture system is used, medium 13A should be used for blood samples rather than 12B, which is used for all other specimens.

Although nucleic acid probes are not used for direct testing of specimens, they have become a major advance in the laboratory's ability to identify specific mycobacterial groups or species accurately, once there is adequate growth of mycobacteria in a solid or liquid medium. Commercially available probes (such as Accuprobe) can identify Mycobacterium tuberculosis complex, Mycobacterium avium complex, Mycobacterium kansasii, and Mycobacterium gordonae within hours [78, 79]. This novel molecular approach (Figure 2), using DNA probes directed against ribosomal sequences, has dramatically improved identification of mycobacteria, reducing identification time from several weeks to 1 day once sufficient growth is present.

Is the mycobacterium identified from a specimen always important? Although the presence of Mycobacterium tuberculosis always represents disease that must be treated, and Mycobacterium kansasii usually represents disease requiring treatment as well, there are no universal rules for determining when other mycobacteria are pathogens, colonizers, or contaminants. The quantity of mycobacteria, the species, the host, and the clinical situation must be assessed. Some generalizations are helpful. As noted above, Mycobacterium gordonae has been identified as a true pathogen only rarely: It is most often a waterborne contaminant. Mycobacterium avium complex rarely causes localized or disseminated disease in any population other than patients with HIV infection. (However, a few cases of Mycobacterium avium complex causing parenchymal lung disease in persons with no underlying processes do occur.) In patients with HIV infection, blood isolates almost always represent disseminated disease, yet sputum, urine, or stool isolates often represent colonization and have not been shown to predict efficiently the future development of dissemination in studies done to date.

Dr. Frederick P. Ognibene (Critical Care Medicine Department, Clinical Center, NIH): The most common fungal pathogens responsible for pulmonary infections in immunocompromised hosts and the tests that may be helpful in the diagnosis of pneumonia caused by these organisms are shown in Table 7. A major frustration in pulmonary diagnostics has been the inability to develop useful tests for the fungi of most concern, Candida species and Aspergillus species.

The isolation of Aspergillus species from respiratory secretions had been regarded as limited in usefulness in the antemortem diagnosis of invasive pulmonary aspergillosis. The detection of tissue invasion by fungal hyphae was considered the “gold standard” for diagnosis. However, Yu and colleagues [86] published data that indicated that in all of their patients with leukemia or neutropenia or both who ultimately had documented invasive aspergillosis, either A. fumigatus or A. flavus was isolated from cultures of the respiratory tract. They concluded that the isolation of Aspergillus organisms from sputum or lavage culture was highly predictive of invasive pulmonary infection, and many physicians follow their approach despite uncertainty about the specificity of this finding [86, 87]. Histopathologic evidence of fungi with the characteristic 45-degree angle branching and transverse septae confirm the diagnosis. No serologic studies for Aspergillus species are adequately sensitive or specific [88, 89].

Zygomycetes genera, including Mucor and Rhizopus, may also cause serious pulmonary disease in severely immunocompromised patients. A sputum or bronchoalveolar lavage specimen with a positive wet mount showing characteristic morphologic features or a positive culture strongly suggests that these agents are the cause of the pulmonary disease. Histopathologic specimens of the lung reveal broad, nonseptate hyphae with irregular right-angle branching. No reliable serologic tests exist for diagnosis.

Infection with Fusarium species may mimic the presentation of aspergillosis. Patients may have associated skin lesions or orbitofacial involvement. Blood cultures are frequently positive and provide a definitive diagnosis. No useful serologic studies exist for this pathogen [90].

For cryptococcosis, a sputum or bronchoalveolar lavage specimen that is either smear- or culture-positive for Cryptococcus neoformans is considered diagnostic because this encapsulated organism does not occur as a commensal. Although serologic tests may yield positive results for cryptococcal antigen in as many as 90% of patients with cryptococcal meningitis, results are positive in only 50% of patients with pneumonia alone [91]. A histologic specimen is often necessary to establish the diagnosis of cryptococcal pulmonary disease, because sputum and pleural fluid cultures may be negative in half of patients.

The diagnosis of disseminated histoplasmosis depends on a positive culture, typically from blood. For a pulmonary infection, a positive bronchoalveolar lavage smear or culture is diagnostic. However, it is important to note that a 2- to 4-week period may be required for Histoplasma capsulatum to grow in cultures. The histopathologic detection of intracellular organisms in viable pulmonary tissue is diagnostic evidence of disseminated disease. In nonimmunosuppressed patients, serologic studies are highly sensitive and provide confirmatory diagnostic evidence of Histoplasma infection [92]. Complement fixation has been used to detect antibody, and a titer of 1:32 or greater or a fourfold increase in titer is evidence of active or recent infection. Immunodiffusion studies are more specific but less sensitive than complement fixation. The presence of both immunoprecipitation bands (called M and H) is the most specific serodiagnostic finding, but it is also rare. A positive M band alone is a fairly specific finding. These serologic responses may occur in less than 50% of immunosuppressed patients. Detection of H. capsulatum antigen in specimens of blood and urine is done at only a few centers, but detection of antigen is believed to be useful for the diagnosis of disseminated histoplasmosis [93].

Coccidioides immitis can cause an acute, progressive pneumonia that may lead to dissemination to extrapulmonary foci. A positive culture of sputum or bronchoalveolar lavage is diagnostic. Wet mount staining is often helpful when the characteristic spherules with endospores are detected. Elevated complement fixing antibody studies are a hallmark of disseminated disease; titers of at least 1:32 indicate disseminated disease in most patients [92].

For the accurate diagnosis of Candida pneumonia, routine sputum and bronchoalveolar lavage cultures are not useful. Candida pneumonia is probably an unusual cause of pulmonary dysfunction, although the lung may be extensively involved with microemboli in patients with disseminated candidiasis. Histologic evidence of inflammation in the presence of Candida organisms is believed to be the most convincing evidence for Candida pneumonia. Blood cultures are not always positive in invasive disease caused by Candida species. In many patients with disseminated disease (not just pneumonia), blood cultures may be negative [94]. No reliable serologic studies exist for this pathogen [89].

Dr. Shelhamer: For an increasing number of community-acquired and opportunistic pathogens, new diagnostic tests that permit rapid detection of organisms in clinical specimens are becoming available. These test include direct detection systems that allow identification of organisms in clinical specimens within 24 hours by microscopy using tinctorial or immunologic staining techniques, by detection of antigen, or by detection of nucleic acid. Culture systems are also becoming more sensitive and more rapid, often providing results within a few days.

Should every reference diagnostic laboratory do all of these tests? In many cases, these tests are expensive and require considerable expertise. Some laboratories may not have the volume or the appropriate number of positive specimens during the course of a year to warrant doing each test. Some tests may not provide accurate, sensitive, or specific enough information to alter clinical management and thus are not practical for many cost-conscious facilities.

Even if these tests are done, clinicians need considerable expertise to interpret them appropriately. Each test has very specific meaning; unfortunately, these tests may signify different clinical information in different patient populations, and thus no easy, universally appropriate algorithm can be devised for their interpretation.

It is important for clinicians to be cognizant of the availability of these evolving tests. Some of the tests can be extremely valuable in certain well-defined situations, even if all of them cannot be routinely done. For instance, any center that has a large population of allogeneic bone marrow transplant recipients might consider doing rapid cytomegalovirus tests, such as some combination of shell vial cultures, buffy-coat antigen tests, or even polymerase chain reaction. A hospital with a high rate of Legionella infection, tuberculosis, or respiratory syncytial virus might consider instituting the appropriate rapid diagnostic test to ensure appropriate therapy in the case of legionellosis and to prevent nosocomial transmission as well as ensure appropriate therapy in the cases of tuberculosis and respiratory syncytial virus infection. Thus, institutions will have to pick and choose from this increasingly complex but useful menu, and develop tests that are appropriate for their patient population, their epidemiologic trends, and their budget.

Dr. Gill: Clinical Pathology Department, Clinical Center, Building 10, Room 2-C-385, National Institutes of Health, Bethesda, MD 20892.

Dr. Quinn: 1159 Ross Research Building, 720 Rutland Avenue, Baltimore, MD 21205.

Dr. Crawford: Pulmonary and Critical Care Medicine, Fred Hutchinson Cancer Research Center, 1124 Columbia Street, Seattle, WA 98104.