Usefulness of PCR and Antigen Latex Agglutination Test with Samples Obtained by Transthoracic Needle Aspiration for Diagnosis of Pneumococcal Pneumonia

Abstract

In a large number of cases, the etiology of community-acquired pneumonia (CAP) is not established. Some cases are probably caused by Streptococcus pneumoniae. Transthoracic needle aspiration (TNA) culture has a limited sensitivity which might be improved by antigen detection or gene amplification techniques. We evaluated the capacity of a PCR assay and a latex agglutination test to detect S. pneumoniae in samples obtained by TNA from 95 patients with moderate-to-severe CAP. Latex agglutination and PCR had sensitivities of 52.2 and 91.3%, specificities of 88.7 and 83.3%, positive predictive values of 62.3 and 65.6%, and negative predictive values of 83.3 and 96.5%, respectively, when culture techniques were used as the “gold standard.” When we considered expanded criteria for the diagnosis of pneumococcal pneumonia as a standard for our calculations, latex agglutination and PCR had sensitivities of 53.6 and 89.7%, specificities of 93.0 and 90.0%, positive predictive values of 78.9 and 81.3%, and negative predictive values of 80.3 and 94.7%, respectively. The additional diagnosis provided by the PCR assay compared to latex agglutination was 12.2% (95% confidence interval of the difference from 0.4 to 20.1%). PCR was more sensitive than TNA culture, particularly in patients who had received prior antibiotic therapy (83.3 versus 33.3%). Although PCR is a very sensitive and specific technique, it has not proved to be cost-effective in clinical practice. Conversely, latex agglutination is a fast and simple method whose results might have significant implications for initial antibiotic therapy.

Community-acquired pneumonia (CAP) continues to be a significant cause of morbidity and mortality worldwide. Streptococcus pneumoniae is the most commonly defined pathogen in nearly all studies of hospitalized adults (1, 8, 12). Current criteria for a definitive diagnosis of pneumococcal pneumonia require the isolation of S. pneumoniae from blood, pleural fluid, a metastatic-site specimen, or an uncontaminated respiratory sample obtained by invasive techniques. In a substantial number of cases, despite recent improvements in diagnostic methods, the etiology of CAP cannot be established; some of these cases are probably caused by S. pneumoniae. Lately, transthoracic needle aspiration (TNA) has been used to improve the diagnostic yield of pneumonia. In several studies, TNA has proved to be safe and has provided an etiologic diagnosis in 30 to 60% of cases by standard microbiological methods (7, 17, 18). However, its diagnostic effectiveness can be reduced by a number of factors, one of the most important of which is prior antibiotic therapy. New techniques, such as antigen detection methods and PCR, which do not depend on viable organisms, are probably less affected by the administration of antimicrobial agents and might improve the overall sensitivity of TNA (3, 6).

The aim of our study was to evaluate the ability of a PCR assay and a latex agglutination test to detect S. pneumoniae in samples obtained by TNA from patients with moderate-to-severe CAP.

MATERIALS AND METHODS

Setting and population studied.

The study was conducted at Bellvitge Hospital, a 1,000-bed university hospital in Barcelona, Spain. From February 1995 through May 1997 all patients with moderate-to-severe CAP requiring hospitalization were prospectively monitored at our institution. They were seen by a member of the study team who filled out a previously defined computer-assisted protocol and who provided medical advice when required. TNA was regularly performed at our institution during the study period because of the good results in terms of safety and the experience of our pneumologists over the last decade. During the study period, a total of 95 TNAs were performed among the 533 patients admitted in our hospital. Therefore, use of the TNA was not the usual standard of care, and the final decision relied on the emergency team attending each patient. For the purposes of this study, we identified all patients with moderate-to-severe CAP from whom TNA samples were obtained. TNA was performed if patients gave their consent and were able to collaborate in the TNA procedure; it was not performed in the presence of any of the following contraindications: low platelet count (≤60,000 cells/ml), a quick ratio of >1.8 or a quick time of <60%, severe pulmonary hypertension, mechanical ventilation, AIDS, and uncontrollable cough.

TNA procedure.

Premedication with 0.5 mg of atropine was administered intramuscularly 30 min before the puncture. Puncture was performed without fluoroscopic or computed-tomography control, at the patient’s bedside, and before starting therapy. Intradermal and subcutaneous anesthesia with mepivacaine was administered. The procedure was carried out by using an ultrathin 25-gauge needle with its stylet. When it was believed to be on the target, a 20-ml syringe containing 5 ml of sterile saline was attached, and 4 ml was then injected. Suction was applied vigorously for at least 30 s. A second 20-ml syringe was then attached, and another 4 ml of saline was injected. One of the syringes was randomly used for conventional microbiological procedures and the latex agglutination test. The remaining sample was stored at −72°C for later PCR determination. For clinical reasons, if the amount of the TNA sample was insufficient, priority was given to standard microbiological studies.

Microbiological studies.

Prior to the initiation of therapy, two sets of blood cultures were drawn at the initial evaluation. Sputum samples were processed for Gram stain and culture, when available. Paired serum samples from the acute and convalescent phases (separated by 3 to 8 weeks) were also obtained for serological studies.

Cultures for conventional bacterial and fungal respiratory pathogens were carried out by standard methods in all TNA samples. Investigation of the pathogens in other specimens (blood, normally sterile fluids, sputum, etc.) was also done by conventional procedures. Isolation of Legionella pneumophila was attempted in TNA and sputum samples by using selective medium (BCYE-α; Oxoid, Basingstoke, England). Detection of L. pneumophila serogroup 1 antigen in urine was performed by an immunoenzymatic commercial kit (Legionella Urinary Antigen; Binax, Portland, Maine). Standard serological methods in our laboratory were used for determining antibodies to the following pathogens: Mycoplasma pneumoniae (indirect agglutination), Chlamydia psittaci (immunofluorescence [IF]), Chlamydia pneumoniae (micro-IF), Coxiella burnetii (IF), L. pneumophila serogroups 1 to 6 (enzyme immunoassay [EIA]), respiratory syncytial virus (EIA), parainfluenza 3 virus (EIA), and influenza A virus (EIA).

Latex agglutination for pneumococcal antigen in TNA samples.

After heating at 95°C for 5 min, TNA samples were centrifuged at 700 × g for 10 min. Agglutination was then performed on a 50-μl aliquot of the supernatant with the Slidex Pneumokit (bioMérieux, Marcy l’Etoile, France) according to the manufacturer’s instructions. Reading of the tests was almost immediate.

PCR assay for S. pneumoniae.

We used the method described by Zhang et al. (19), with minor modifications adapted to the specific conditions of our specimen. An 86 bp fragment of the PBP2b pneumococcal gene was amplified by this protocol. DNA extraction from TNA samples was accomplished with a commercial kit (QIAamp Blood Kit; Qiagen, Hilden, Germany) in order to avoid PCR inhibition by the hemoglobin frequently present in the specimen. The extraction procedure was completed after ca. 2 h of processing time.

Amplification was carried out in a PTC-100 thermal cycler (MJ Research, Inc., Watertown, Mass.) with primers JM01 (5′-ATG CAG TTG GCT CAG TAT GTA-3′) and JM02 (5′-CAC CCA GTC CTC CCT TAT CA-3′), which were supplied by Cruachem, Glasgow, United Kingdom. The primers were diluted in water to reach a 100 μM stock solution. The final concentration in the reaction was 1 pmol/μl. PCR was performed in a 50-μl volume (25 μl of reagent mixture and 25 μl of template) containing 0.2 μl of Taq DNA polymerase (5 U/μl; Life Technologies, Gaithersburg, Md.); 5 μl of 10× buffer and 5 μl of 25 mM MgCl₂ (both supplied with Taq enzyme); 5 μl of a mixture of digoxigenin-labeled deoxynucleoside triphosphates (PCR DIG Labeling Mix; Boehringer GmbH, Mannheim, Germany) containing 2 mM concentrations (each) of dATP, dCTP, and dGTP, 1.9 mM dTTP, and 0.1 mM dUTP; 0.25 μl of each stock solution primer; and 9.3 μl of bidistilled water. Amplification started with a denaturation step of 10 min at 94°C, followed by 30 cycles of 1 min at 94°C, 1 min at 55°C, and 1.5 min at 72°C. Finally, an additional extension cycle of 7 min at 72°C was carried out before proceeding to detection. The amplification step took about 4 to 5 h.

Detection of PCR products was done by hybridization with a 5′ biotinylated internal probe and by using an immunoenzymatic commercial method (PCR-ELISA DIG Detection; Boehringer) according to the manufacturer’s instructions. The sequence for the probe was 5′-biotin-CAA ATA ATG GTG TTC GTG TGG CTC CTC GTA-3′ (Cruachem, Glasgow, United Kingdom). A green color appeared after the enzymatic assay, which was read at 405 nm. This enzymatic detection took more than 4 hours.

The complete PCR procedure was performed in duplicate for each sample. A positive control (DNA from a pneumococcal clinical isolate) and several negative controls (25 μl of bidistilled water) spotted between problem extracts were included in each run. A specimen was considered positive if its corresponding optical density was higher than twice the mean optical density value obtained for negative controls.

Limit of detection of the PCR procedure.

An overnight culture of a clinical strain of S. pneumoniae was suspended in a Mueller-Hinton broth to achieve a turbidity similar to that of the 0.5 McFarland standard. Then, serial 10-fold dilutions were made in the same broth supplemented with 5% human blood in order to mimic blood contamination in TNA samples. A 100-μl aliquot from each tube was plate subcultured before the remaining suspension was extracted by the same procedure as for TNA specimens and then amplified by PCR. Plates were incubated at 37°C for 24 h; the colonies were counted and referred to the exact number of CFU present in each tube suspension.

Definitions.

CAP was defined as a febrile, acute respiratory illness with the presence of a new infiltrate evident on a chest radiograph. Patients with a prior hospitalization within 2 weeks of a current diagnosis of pneumonia were excluded. Hospitalization was considered if one or more of the following conditions was present: age, ≥70 years; respiratory insufficiency (a PaO₂ level of ≤60 mm Hg or a PaO₂/FiO₂ ratio of ≤300); multilobar radiological involvement; shock; underlying diseases; and/or unresponsiveness to previous antibiotic therapy. To calculate the severity of the pneumonia we used the Simplified Acute Physiology Score described elsewhere (9).

Etiologic diagnosis was considered definitive when there was (i) isolation of a respiratory pathogen in normally sterile samples (blood, pleural fluid, and TNA samples); (ii) isolation of L. pneumophila or Mycobacterium tuberculosis in sputum; (iii) detection of L. pneumophila serogroup 1 antigen in urine; (iv) a fourfold increase in the antibody titer for M. pneumoniae, C. psittaci, C. pneumoniae, C. burnetii, or L. pneumophila (serogroups 1 to 6), (v) the presence of immunoglobulin M (IgM) antibodies (≥1/20) to C. pneumoniae, (vi) the presence of IgA antibodies (≥1/32) to C. pneumoniae, and (vii) seroconversion for the following respiratory viruses: respiratory syncytial virus, parainfluenza 3 virus, and influenza A virus. Etiological diagnosis was considered presumptive when a predominant microorganism that correlated with a predominant morphotype in the Gram stain from an acceptable specimen (presence of >25 polymorphonuclear leukocytes and <10 squamous cells per low-magnification field [×10]) was isolated in sputum. Presumptive aspiration pneumonia was diagnosed in the presence of a suggestive clinical and radiological picture. Cases that did not fulfill the etiologic diagnostic criteria described above were considered “pneumonia of unknown etiology.”

Besides these strict criteria, etiologic diagnosis was also obtained by using expanded criteria. With the expanded criteria, cases either with positive results for both latex agglutination test and PCR assay or with a single positive PCR or latex agglutination test plus the presence of gram-positive diplococci as the predominant morphotype in a Gram stain from an acceptable-quality sputum sample could also be considered to be pneumococcal pneumonia. Patients with positive results based solely on PCR or on the latex agglutination test were considered nonpneumococcal pneumonia cases (false positive for these assays).

Statistical methods.

The final diagnosis of cases according to both the strict and the expanded criteria, as described above, was used as the standard for determining the diagnostic usefulness of the PCR and latex agglutination methods in terms of sensitivity, specificity, and positive and negative predictive values. For the purposes of the calculations, patients were classified into two groups: (i) those with pneumococcal pneumonia and (ii) those with nonpneumococcal pneumonia, including cases with other known causes and pneumonia of unknown etiology.

The hypothesis of equivalence between PCR and latex agglutination was tested by McNemar statistics with continuity correction or by the binomial test, as appropriate. The magnitude of the difference was calculated in exact limits, when appropriate. A P value of <0.05 was considered statistically significant.

RESULTS

A total of 95 patients with moderate-to-severe CAP in whom TNA was performed were analyzed. There were 69 men and 26 women, with a mean age of 65 years (range, 19 to 87 years). The mean Simplified Acute Physiology Score was 8.7 (range, 1 to 27). In 34 patients, pneumonia involved more than one lobe on the initial chest radiograph evaluation. In 52 patients (54.7%) there were underlying diseases, mostly diabetes mellitus (n = 18), chronic heart diseases (n = 15), and chronic obstructive pulmonary disease (n = 13). In 18 patients there was more than one underlying disease. A total of 39 patients had received influenza vaccine, and only 2 patients had been immunized with pneumococcal vaccine in the previous 5 years. A total of 34 patients (37.8%) had received antibiotic treatment prior to hospital admission. Sputum samples were attempted from all study patients, but they were obtained only in 58 cases. In 28 of these the sample was considered an acceptable specimen for cultivation, as previously defined. A predominant microorganism was isolated in 14 cases, yielding S. pneumoniae in 9 cases and Haemophilus influenzae in 5 cases. In 3 additional cases in which the sputum sample was considered not acceptable for culture of conventional pathogens, a culture in selective media yielded L. pneumophila strains.

In 82 TNA samples, both the latex agglutination test and PCR assay were performed. In 3 of the total 95 specimens, there was not enough sample to perform both the latex agglutination test and the PCR (1 in the pneumococcal-pneumonia culture-positive group, 1 in the group with an etiology other than S. pneumoniae, and 1 in the unknown-etiology group). In three cases the sample amount was insufficient to perform PCR (one in the pneumococcal-pneumonia culture-positive group, two in the unknown-etiology group). Finally, in seven cases latex agglutination for pneumococcal antigen was not performed (one in the pneumococcal-pneumonia culture-positive group, one in the unknown-etiology group, and five in the other-etiologies group).

The TNA culture was positive in 29 cases (30.5%). S. pneumoniae was the most frequently isolated pathogen (16 cases), followed by L. pneumophila (9 cases), Haemophilus influenzae (2 cases), Escherichia coli (1 case), and mixed aerobic-anaerobic respiratory flora (1 case). In two cases, in addition to S. pneumoniae, a second pathogen (H. influenzae and Moraxella catarrhalis) was isolated in the TNA samples.

Overall, considering all standard techniques, 57 patients (60.0%) had an etiological diagnosis. TNA alone provided an etiological diagnosis in 10 cases (9 S. pneumoniae, 1 E. coli) and confirmed a presumptive diagnosis in 5 others (4 S. pneumoniae, 1 H. influenzae). Seven patients fulfilled strict diagnostic criteria for two different infectious agents.

Distributions of the final diagnosis and the PCR and latex agglutination test results are shown in Tables 1 and 2. In one case of M. pneumoniae pneumonia, the PCR determination was inconclusive (borderline readings after repeated testing), and it was considered true negative for the calculations.

TABLE 1.

Final diagnosis and latex agglutination and PCR test results for the detection of S. pneumoniae in TNA samples from 95 patients with moderate-to-severe CAP^a

Diagnosis	Total no. of cases	LA and PCR				LA only		PCR only
		No. tested	LA Pos	PCR Pos	Both Pos	No. tested	Pos	No. tested	Pos
Pneumococcal pneumonia	25	22	12	20	12	1	0	1	1
Definitive	19	17	9	15	9	0		1	1
Presumptive	6	5	3	5	3	1	0	0
Nonpneumococcal pneumonia	32	26	3	2	0	0		5	1
Definitive	28	23	3	2	0	0		4	1
Presumptive	4	3	0	0	0	0		1	0
Pneumonia of unknown etiology	38	34	4	7	2	2	0	1	1
Total	95	82	19	29	14	3	0	6	3

^aPos, positive; LA, latex agglutination. 

TABLE 2.

Results of the PCR assay and latex agglutination test for the detection of S. pneumoniae in patients with nonpneumococcal pneumonia or pneumonia of unknown etiology

Diagnosis	Total (n = 70)	Latex agglutination (n = 62)			PCR (n = 66)
		No. tested	Positive	Negative	No. tested	Positive	Negative
Definitive	28	23	3	20	27	3	24
L. pneumophila	14	12	0	12	14	0	14
C. pneumoniae	7	6	3	3	7	1	6
C. psittaci	1	1	0	1	1	0	1
M. pneumoniae	1	1	0	1	1	0	1
C. burnetii	1	0			0
H. influenzae	2	1	0	1	1	0	1
E. coli	1	1	0	1	1	1	0
Aspiration pneumonia	1	1	0	1	1	1	0
Presumptive	4	3	0	3	4	0	4
H. influenzae	3	3	0	3	3	0	3
Aspiration pneumonia	1	0			1	0	1
Pneumonia of unknown etiology	38	36	4	32	35	8	27
Gram-positive diplococci in sputum Gram stain		3	0	3	4	3	1

The calculated limit of detection for our PCR assay was between 2 and 27 CFU/ml. As Table 3 shows, PCR was more sensitive than latex agglutination. Focusing on the 82 samples in whom both diagnostic tests were performed, the additional diagnosis provided by PCR assay was 32.7% (95% confidence interval of the difference from 9.5 to 36.4%) in the subgroup of culture-proven pneumococcal-pneumonia group and 12.2% (95% confidence interval of the difference from 0.4 to 20.1%) when calculated for all pneumonias. Table 4 shows clinical and laboratory data of false-positive results for either the PCR assay or the latex agglutination test when the strict diagnostic criteria are used.

TABLE 3.

Comparison of latex agglutination test and PCR gene amplification for the detection of S. pneumoniae in TNA samples from patients with moderate-to-severe CAP

Parametera	Latex agglutination		PCR
	Strict criteria	Expanded criteria	Strict criteria	Expanded criteria
Sensitivity (%)	52.2	53.6	91.3	89.7
Specificity (%)	88.7	93.0	83.3	90.0
PPV (%)	63.2	78.9	65.6	81.3
NPV (%)	83.3	80.3	96.5	94.7

^aPPV, positive predictive value; NPV, negative predictive value. 

TABLE 4.

Summary of clinical and laboratory data of cases showing false-positive results in PCR and/or latex agglutination tests by using strict criteria

Patient no.	Final diagnosis	Sex, age (yr)	Prior antibiotic therapy	Diagnostic method	Sputum Gram staina	LAb	PCR
1	C. pneumoniae infection	Male, 79	Oral amoxicillin	IgM (+) and high IgG titers	MOF	ND	+
2	E. coli infection	Male, 84	No	TNA culture	ND	−	+
3	Aspiration pneumonia	Male, 83	No	TNA and blood culture	ND	−	+
4	C. pneumoniae infection	Male, 72	No	IgM (+) and high IgA titers	ND	+	−
5	C. pneumoniae infection	Male, 59	Intramuscular penicillin-procaine	Seroconversion	ND	+	−
6c	C. pneumoniae infection	Male, 63	Oral roxithromycin	IgM (+) and high IgG titers	GPDC	+	−
7c	Pneumonia of unknown etiology	Male, 66	No		SR	+	+
8c	Pneumonia of unknown etiology	Male, 39	No		ND	+	+
9c	Pneumonia of unknown etiology	Male, 77	Oral ciprofloxacin		GPDC	ND	+
10c	Pneumonia of unknown etiology	Male, 79	No		GPDC	−	+
11c	Pneumonia of unknown etiology	Male, 76	No		GPDC	−	+
12	Pneumonia of unknown etiology	Male, 71	Oral erythromycin		ND	−	+
13	Pneumonia of unknown etiology	Male, 66	No		ND	−	+
14	Pneumonia of unknown etiology	Female, 84	No		ND	−	+
15	Pneumonia of unknown etiology	Male, 25	Oral amoxicillin		SR	+	−
16	Pneumonia of unknown etiology	Male, 59	No		SR	+	−

^aMOF, mixed oral flora; ND, not done; SR, specimen rejected because of poor quality; GPDC, gram-positive diplococci. 

^bLA, latex agglutination. 

^cCases considered as true positive in the analysis with the expanded criteria. 

PCR was more sensitive than TNA culture, particularly in patients who had received prior antibiotic therapy. Of the 34 patients who had received antibiotic treatment prior to hospital admission, 6 fulfilled the criteria for pneumococcal pneumonia. As shown in Table 5, TNA culture yielded S. pneumoniae in two cases (33.3%), while PCR was positive in five cases (83.3%). For its part, latex agglutination was positive in one of the four cases in which it was performed (25.0%).

TABLE 5.

Patients with a final diagnosis of pneumococcal pneumonia who had received previous antibiotic therapy

Patient no.	Previous antibiotic therapy	TNA			Sputum Gram stainb	Other positive cultures for S. pneumoniae
		Culture	LAa	PCR
17	Ciprofloxacin	+	ND	+	ND	No
18	Amoxicillin-clavulanated	+	−	+	ND	No
19	Norfloxacin	−	−	+	SR	Blood
20	Ciprofloxacin	−	−	+	GPDC	Pericardial fluid
9c	Ciprofloxacin	−	ND	+	GPDC	No
6c	Roxithromycin	−	+	−	GPDC	No

^aLA, latex agglutination; ND, not done. 

^bSR, specimen rejected (because of poor quality); GPDC, gram-positive diplococci. 

^cCases considered to be pneumococcal pneumonia by the expanded criteria (they are also listed in Table 4 with the corresponding numbers). 

^dMIC of amoxicillin = 4 μg/ml. 

The latex agglutination test result was available for physicians in the emergency room in the majority of cases and provided information before the initial antibiotic therapy was selected. In fact, only two patients with a positive latex agglutination test were admitted on combined β-lactam and macrolide therapy. In none of the patients admitted on a single β-lactam therapy was a macrolide added to therapy after a negative latex agglutination result.

DISCUSSION

Although previous studies (7, 11, 17, 18) have shown TNA to be a highly specific technique for the diagnosis of pneumonia, current indications of TNA in the CAP setting have not been well defined. Our study did not address this issue. We studied a large number of hospitalized patients with moderate-to-severe CAP in whom TNA was performed and evaluated the effectiveness of the latex agglutination test and PCR in these specimens. As expected, TNA improved the diagnostic yield in our study; in fact, more than a quarter of the definitive etiologic diagnosis (15 of 57) were made by culture of TNA samples. As for our two false-negative PCR results, they were from non-bloody samples, and so a possible negative effect by hemoglobin was reasonably ruled out. However, the possibility of a negative test resulting from the use of separate syringes for doing the different microbiological studies should be born in mind.

It is assumed that culture-positive pneumococcal pneumonia represents only a portion of the cases of pneumonia caused by this pathogen, so there is considerable room for alternative, more-sensitive diagnostic tests. The new laboratory techniques used in the diagnosis of pneumococcal pneumonia mainly confront two challenges related to their sensitivity and specificity. The first is the lack of a satisfactory standard for comparison. This is especially true of methods based on gene amplification, which is assumed to have a very low limit of detection. The second is the increasing number of coinfections reported in several recent studies (2, 4, 10). This phenomenon may lead us to consider cases of undiagnosed coinfections as false-positive results of these techniques.

In addition to better sensitivity and good specificity, other important advantages of the new laboratory tests when applied on a routine basis are rapidity, simplicity, and cost-effectiveness. This is particularly the case when latex agglutination and PCR are compared for the diagnosis of pneumococcal pneumonia. Data from the literature have shown variable results from PCR assays with regard to sensitivity. In cases of bacteremic pneumococcal pneumonia, sensitivity with blood specimens has ranged from 37.5 to 100.0%, depending on the technical procedure used (5, 14, 16, 19). Overall, sensitivity is lower when PCR was applied to whole blood than when applied to buffy coat preparations or sera, probably as a consequence of the inhibitory effect of hemoglobin on the amplification. This could also partially explain why other authors have reported PCR assays to be less sensitive than pneumococcal antigen detection in TNA samples (13, 15).

In an attempt to avoid this negative effect and to increase sensitivitiy, we used a column DNA extraction method and amplicon detection by hybridization with a biotinylated probe. In our study, detection of pneumococcal DNA by PCR proved to be more sensitive than latex agglutination, although, when considering all patients, the difference was limited. In contrast, latex agglutination compares favorably with the PCR assay in terms of cost, simplicity, and rapidity. For instance, latex agglutination can be performed in less than half an hour of processing time. PCR assay, on the other hand, is labor-intensive; the complete procedure takes about 10 h, which precludes its use as a realistic alternative method to conventional diagnostic techniques.

There were three false-positive results in the latex agglutination test among patients with other known etiologies when the strict criteria were applied. In all three cases the patients had a final diagnosis of C. pneumoniae pneumonia. In one case, the clinical and laboratory data and the presence of gram-positive diplococci in a sputum Gram stain with growth of normal flora supported the assumption of a coinfection. This patient was considered as true positive in the analysis with expanded criteria. False-positive results of latex agglutination in patients with C. pneumoniae pneumonia have been reported elsewhere (15). It should be noted that we used serology as the standard diagnosis of C. pneumoniae infection; this is considered a sensitive technique, although its specificity in diagnosing acute infection is currently under debate.

As for PCR, there were also three false-positive results in patients with other known etiologies. One patient previously treated with amoxicillin had a serologically diagnosed C. pneumoniae pneumonia, so a possible coinfection cannot be excluded. The second patient presented with aspiration pneumonia in which mixed aerobic and anaerobic bacteria of the upper respiratory tract were isolated by TNA. There are two possible explanations for this result. First, the presence of multiple bacteria might have inhibited the growth of a small inoculum of S. pneumoniae which could be detected by gene amplification. Second, this might be a real false-positive result, given the relationship in the PBP-2b genes of viridans streptococci and S. pneumoniae, although Zhang et al. (19) found no such cross-reactivity with the same primers and probe as the ones we used in our PCR assay. E. coli pneumonia was diagnosed in the third patient because of TNA-positive culture for this microorganism and negative blood cultures (a sputum sample was not obtained). PCR test results were repeatedly positive for S. pneumoniae. Because the E. coli strain was not available for further studies, we considered this PCR result a false positive.

Clinical presentation of all patients with pneumonia of unknown etiology and positive results in PCR and/or latex agglutination tests was compatible with bacterial pneumonia, and it is our understanding that most of these cases are, in fact, caused by S. pneumoniae. In many cases etiologic diagnosis was not performed because of the impossibility of obtaining a sputum sample of good quality or because of other circumstances, such as previous antibiotic treatment. In fact, PCR was more sensitive than TNA culture and latex agglutination in this subgroup of patients. For these reasons, we considered the expanded criteria to be more suitable for evaluating the real performance of both PCR and latex agglutination assays.

In summary, our study shows that latex agglutination and PCR improve the diagnostic yield of TNA. Latex agglutination is a simple, inexpensive technique whose results can be made readily available in the emergency room setting. Although less sensitive than PCR, it appears to offer practical advantages for the choice of initial antibiotic therapy. PCR is a very sensitive and specific technique which, due to its complexity, should perhaps be reserved for research purposes. A sensitive and specific PCR assay such as the one we used here could be a good standard for future evaluations of other diagnostic techniques.