This study has shown that detection of urinary pneumococcal antigen by using the Binax NOW S. pneumoniae
antigen test is an extremely useful technique for the rapid diagnosis of bacteremic pneumococcal infections in adults. Our evaluation of this test was limited to adults because several other recent studies have shown it to have very poor specificity for young children, particularly where there is a high rate of pneumococcal carriage. Dowell et al. found that the test result was no more likely to be positive among 88 children (age, ≤5 years) with radiographically confirmed pneumonia than among 198 controls (frequencies of positive results, 35 and 34%, respectively) (7
). In a study of 210 healthy children, in several age groups of ≤5 years, Hamer et al. found that the test gave positive results for 11 to 21%, with test positivity and pneumococcal carriage decreasing with increasing age (11
). Other investigators reported that 28 to 50% of children had positive antigen tests (1
). In all of these studies, a positive result was very much more likely if concurrent pneumococcal carriage was demonstrated by nasopharyngeal culture.
When the Binax NOW S. pneumoniae
urinary antigen test has been evaluated for adults, the results have been good, although the reported numbers of patients with pneumococcal bacteremia have been low. Murdoch et al. studied 420 adults with community-acquired pneumonia, including 20 patients with pneumococcal bacteremia, 16 of whom (80%) had detectable urinary antigen levels (14
). Dominguez et al. detected pneumococcal antigen in urine specimens of 82% of 28 patients (precise ages not given) with bacteremic pneumococcal pneumonia (6
). In another study, 75% of 16 adults with bacteremic pneumococcal pneumonia had positive urinary antigen test results (9
In our evaluation, pneumococcal urinary antigen was detected in 82% (88 of 107) of bacteremic pneumococcal cases overall, a finding in agreement with those of the other smaller studies. However, we found that antigen detection was significantly greater for those with pneumonia (87%) than for those without pneumonia (70%). This may well result from the greater total bacterial load associated with pneumonic infections.
In determining the test specificity, Murdoch et al. used 169 adult control patients with an admission diagnosis other than a respiratory or infectious disease and found that none had detectable pneumococcal antigen (specificity, 100%) (14
). It is highly unlikely that the test would ever be used for this type of patient, although it is reassuring that there were no false positives that might have been related to pneumococcal carriage. In our study we deliberately selected clinically relevant controls, i.e., adult patients with nonpneumococcal community-acquired bacteremic infections. Urinary pneumococcal antigen was detected in 3 of 106 patients, giving a specificity of 97%. The study by Dominguez et al. also used control patients with pneumonia or bacteremia due to other organisms and also reported a test specificity of 97% (2 of 71 tests positive) (6
Of the three false-positive results in our study, two occurred with patients who might be considered at high risk for pneumococcal infection. One alcoholic patient had an E. coli urinary tract infection and bacteremia. Although radiography was not performed, there was no clinical evidence of a chest infection. Another patient with alcoholic cirrhosis had Enterobacter sp. bacteremia. Although the admission chest radiograph was clear, she died 5 days later, and evidence of early pneumonia was found upon postmortem examination. The third patient was an elderly lady admitted with K. pneumoniae bacteremia and a normal chest radiograph. The first and third patients had not received pneumococcal vaccine recently; the vaccine status of the second was unknown. None of these patients had nasopharyngeal swabs taken to detect pneumococcal carriage. It is possible that these patients had coinfection with S. pneumoniae.
Selection bias may possibly have affected our sensitivity and specificity results. We did not include all adult pneumococcal cases and other-organism controls admitted to the participating units during the period of this study. Selection depended on the submission of a urine sample taken prior to, or within 24 h of, the start of antibiotic treatment. Other factors included the availability of a member of the study group and obtaining consent to participate.
In an attempt to improve the sensitivity of the NOW S. pneumoniae
urinary antigen test, we tried a slight modification to the manufacturer's protocol. This was to read the test after 1 h, which resulted in just one additional positive result in the pneumococcal bacteremia cases but no additional false positives among the controls. Murdoch et al. found that urine concentration resulted in a slight increase in the sensitivity of pneumococcal antigen detection without any deleterious effect on test specificity (14
). It should be noted that neither of these protocol modifications is recommended by the manufacturer and both result in delay to an otherwise very rapid test.
Studies of other pneumococcal antigen tests for bacteremic patients, using latex agglutination or counterimmunoelectrophoresis, have revealed highly variable detection rates from 0 to 88% (3
), and test specificities have often been poorly defined. A recent serotype-specific tube latex agglutination assay has been evaluated thoroughly in adults; it was found to have a sensitivity of 57% for the 10 serotypes included and a specificity of 98% (16
). Despite the excellent specificity, this assay does not perform as well as the Binax NOW test: it has much lower sensitivity and is more labor-intensive.
We also wanted to find out how long urinary pneumococcal antigen was detectable in order to determine how useful the NOW S. pneumoniae
test might be after the commencement of appropriate antibiotic treatment. This was studied with 45 patients selected at random from those who were antigen positive on initial testing. Detectable urinary antigen was still present in 83% of those retested on treatment day 3 and persisted for at least 7 days in many patients. Cerosaletti et al. described similar posttreatment results obtained by using latex agglutination (4
We were unable to collect samples from all patients in the case group at the designated times in our original study protocol (i.e., treatment days 2, 3, 5, and 7). Some patients with severe infections died, while some with mild infections were discharged, before day 7. However, nearly all patients had a urine sample tested on treatment day 2 and/or treatment day 3. Patients were less likely to have further samples taken after the test had become negative on one or two occasions. This resulted in lower numbers of samples being tested on treatment days 5 and 7 but a higher-than-expected proportion of antigen-positive samples. However, for the 45 cases that were initially antigen positive, if some of the missing data could be assumed (i.e., a missing sample between two positive tests is positive; missing samples between a positive and a negative test have no result; and once a sample tests negative, further samples remain negative), 86% would have detectable antigen on day 3 and 73% would have detectable antigen on day 7 (data not shown). In clinical practice it is quite possible that patients may receive antibiotics, perhaps at home, for a day or two before the urinary antigen test is performed, although it is unlikely that a patient would be tested after this time.
The Binax NOW S. pneumoniae antigen test is a valuable and rapid tool for the early diagnosis of bacteremic pneumococcal infections in adults, even after appropriate antibiotic treatment has commenced. Further studies are needed to compare the diagnostic utility of this immunochromatographic urinary antigen test with alternative nonculture techniques in bacteremic and nonbacteremic pneumococcal infections.