Stark and Maki (19) found that .102 CFU/ml was a more valid index when patients were catheterized .... Wu, T. C., E. C. Williams, S. Y. Koo, and J. D. MacLowry.
Vol. 27, No. 4
JOURNAL OF CLINICAL MICROBIOLOGY, Apr. 1989, p. 716-720 0095-1137/89/040716-05$02.00/0 Copyright C 1989, American Society for Microbiology
Rapid Bioluminescence Method for Bacteriuria Screening MARIE T. PEZZLO,* VALERIE IGE, AMELIA P. WOOLARD, ELLENA M. PETERSON, AND LUIS M. DE LA MAZA Department of Pathology, Division of Medical Microbiology, University of California Irvine Medical Center, 101 City Drive, Orange, California 92668 Received 18 August 1988/Accepted 3 January 1989
A study was performed to evaluate the UTIscreen (Los Alamos Diagnostics, Los Alamos, N. Mex.), a rapid bioluminescence bacteriuria screen. The UTIscreen was compared with three other rapid bacteriuria screens: the Bac-T-Screen (Vitek Systems, Hazelwood, Mo.), an automated filtration device; the Chemstrip LN (Boehringer Mannheim Diagnostics, BioDynamics, Indianapolis, Ind.), an enzyme dipstick; and the Gram stain. A semiquantitative plate culture was used as the reference method. Of the 1,000 specimens tested, 276 had colony counts of >105 CFU/ml by the culture method. Of these, the UTIscreen detected 96% (265 of 276) using .5% of the integrated light output of the standard reading as a positive interpretive breakpoint, the Bac-T-Screen detected 96% (266 of 276), the Chemstrip LN detected 90% (249 of 276), and the Gram stain detected 96% (264 of 276). Of the 214 probable pathogens isolated at >105 CFU/mI, the UTIscreen detected 95% (204 of 214), the Bac-T-Screen detected 98% (210 of 214), the Chemstrip LN detected 92% (198 of 214), and the Gram stain detected 98% (209 of 214). The predictive values of negative test results at >105 CFU/ml for the UTIscreen, the Bac-T-Screen, the Chemstrip LN, and the Gram stain were 98, 97, 93, and 98%, respectively. The overall specificities at > 105 CFU/ml for the UTIscreen, the Bac-T-Screen, the Chemstrip LN, and the Gram stain were 70, 48, 51, and 69%, respectively. There were 532 specimens with colony counts of >103 CFU/ml, and of these, the UTIscreen, the Bac-T-Screen, the Chemstrip LN, and the Gram stain detected 72, 81, 76, and 73%, respectively. Of the 249 probable pathogens isolated at >103 CFU/ml, the UTIscreen, the Bac-T-Screen, the Chemstrip LN, and the Gram stain detected 91, 95, 89, and 93%, respectively. The overall specificities at >103 CFU/ml for these methods were 79, 55, 57, and 78%, respectively. The cost per test for detection was approximately $1.00 to $1.20 for the UTIscreen, the Bac-T-Screen, and the Gram stain and approximately $0.50 for the Chemstrip LN. Overall, the UTIscreen is rapid and easy to perform; its sensitivity compared favorably with those of other screening methods; it had a higher specificity than the Bac-T-Screen and Chemstrip LN; and it allowed for batching of specimens. semiautomated urine screens include bioluminescence and filtration methods. These systems are more rapid (1 to 15 min) than the growth detection methods and results of both compare favorably (2, 5, 6, 20, 21, 24, 25). The purpose of this investigation was to evaluate a bioluminescence method, the UTIscreen, and to compare it with other rapid bacteriuria screens: the Bac-T-Screen, an automated filtration method; the Chemstrip LN, an enzyme dipstick method; and the Gram stain. In this study, these methods were evaluated at various colony counts in an attempt to aid others in the selection of a laboratory approach to urine screening.
Urine specimens represent the majority of samples received in the clinical microbiology laboratory for culture. The laboratory faces the challenge of rapidly identifying both positive and negative specimens. Although the semiquantitative plate culture method allows for the isolation and enumeration of most infectious agents, it does not provide for same-day reporting of negative specimens. For this reason, rapid urine screening tests have been developed. These tests not only provide for rapid reporting of negative specimens but also have the potential of reducing the cost of patient management. A number of rapid urine screens have been described (12). These include microscopic, enzymatic, filtration, and photometric methods. The most commonly used microscopic method is the Gram stain. As a urine screen, it is rapid, reliable, and correlates with colony counts of >105 CFU/ml (13, 22, 23). However, because the acceptable sensitivity of this method is at 105 CFU/ml, low-level bacteriuria may not be detected. Furthermore, the accuracy is greatly dependent on the expertise of the reader. Enzyme dipsticks have also been used as rapid bacteriuria screens. Although these tests are easy to perform, the overall sensitivity of these rapid enzyme dipsticks is too low (.90%) to be used alone as urine screens (15, 25). The first generation of semiautomated bacteriuria screens include photometric methods which require growth of the organism for detection; therefore test results are delayed (1 to 13 h). The second generation of *
MATERIALS AND METHODS Specimens. A total of 1,000 specimens which included 688 clean-voided and 312 catheterized urine specimens from both inpatients and outpatients submitted to the Medical Microbiology Laboratory at the University of California Irvine Medical Center were included in this study. Patients receiving antimicrobial therapy were not excluded. Collected urine was placed in a sterile tube, refrigerated (4°C), and processed within 8 h of collection. Semiquantitative culture. A semiquantitative plate count as described by Clarridge et al. (3) was used as the reference method. By using a calibrated platinum loop, a 0.001-ml sample of a well-mixed urine specimen was inoculated onto a 5% sheep blood agar plate (BBL Microbiology Systems, Cockeysville, Md.) and a biplate consisting of MacConkey agar and polymyxin B-nalidixic acid blood agar (Calscott, Inc., Carson, Calif.). An additional 0.1 ml of urine was
Corresponding author. 716
RAPID BACTERIURIA SCREENING BY BIOLUMINESCENCE
VOL. 27, 1989
TABLE 1. Number and percentage of positive test results Culture (CFU/ ml) (no. of specimens)
>105 (276) 1031-i1 (256) 10'-103 (175) 10' CFU/ml by the standard semiquantitative plate culture method. Of these, the UTIscreen detected 265 (96%) compared with 266 (96%) by the Bac-T-Screen, 249 (90%) by the Chemstrip LN, and 264 (96%) by the Gram stain (Table 1). Overall, 226 (82%) were detected by all methods, and at least one screening method was positive for each specimen. There were 256 (25.6%) specimens with colony counts of 103 to 105 CFU/ml. Of these, the UTIscreen detected 116 (45%), while the Bac-T-Screen, Chemstrip LN, and Gram stain detected 164 (64%), 156 (61%), and 124 (48%), respectively (Table 1). The remaining 468 specimens had 10
>103
>101
>105
>103
>lo0
>105
>103
>101
>105
>103
>101
96 70
72 79
62 85
96 48
81 55
73 58
90 51
76 57
68 57
96 69
73 78
61 80
55 98
79 71
91 48
42 97
67 72
81 47
40 93
67 68
79 43
54 98
79 72
88 46
718
J. CLIN. MICROBIOL.
PEZZLO ET AL.
TABLE 3. Probable pathogens at >105 CFU/ml detected by the test methods No. of isolates detected by':
Organism (no. of isolates)
Escherichia coli (108) Candida spp. (21) Klebsiella spp. (19) Enterococcus spp. (20) Pseudomonas spp. (10) Proteus mirabilis (11) Enterobacter spp. (5) Coagulase-negative staphylococci (7) Staphylococcus aureus (5) Streptococcus agalactiae (3) Serratia marcescens (2) Providencia stuartii (2) Citrobacter spp. (1)
UTIscreen
Bac-T-Screen
Chemstrip LN
Gram stain
107 18 19 17 8 10 5 7
108 20 19 19 8 il 5 7
105 18 15 19 7 10 5 7
105 21 19 19 9 il 5 7
5 3 2 2 1
5 3 2 2 1
4 3 2 2 1
5 3 2
1
" The percentages of all isolates detected were as follows: UTlscreen, 95: Bac-T-Screen, 98: Chemstrip LN, 92; and Gram stain, 98.
point was decreased for all organisms to >103 CFU/ml, the sensitivities for the test methods ranged from 81 to 72% and from 73 to 61% at >101 CFU/ml. The specificities and predictive values of positive test results were lower for the Bac-T-Screen and the Chemstrip LN because of the high false-positive results by these methods. Overall, there were 272 probable pathogens isolated from 157 clean-voided specimens and 103 catheterized specimens. Of these, 248 specimens had pure cultures of probable pathogens and the remaining 12 were catheterized specimens with two probable pathogens. The most frequently isolated probable pathogens included Escherichia coli (n = 130), Candida spp. (n = 30), Enterococcus spp. (n = 21), Klebsiella pneumoniae (n = 21), Pseudomonas spp. (n = 19), and Proteus mirabilis (n = 16), accounting for the majority (79%) of isolates in this study. Of these, Pseudomonas spp., Enterococcus spp., and Candida spp. had the highest percentage of false-negative results for all the test methods. There were 214 probable pathogens with colony counts of >105 CFU/ml, 35 at 103 to 105 CFU/ml, and 23 at 101 to 103 CFU/ml. Of the 214 probable pathogens at >105 CFU/ml, the Bac-T-Screen and the Gram stain detected 98%, the UTI screen detected 95%, and the Chemstrip LN detected 92%, respectively (Table 3). There was no statistically significant difference for detection of any of the probable pathogens by the test methods (P > 0.1). When the interpretive breakpoint was decreased for probable pathogens to >103 CFU/ml, the sensitivities for the UTIscreen, the Bac-T-Screen, and the Gram stain were at least 90% compared with 85% for the
Chemstrip LN. The sensitivities for all methods at >101 CFU/ml ranged from 84 to 91%. The UTIscreen and the Gram stain had the highest specificities for all organisms, which ranged from 69% at >105 CFU/ml to 85% at >101 CFU/ml (Table 2). The specificities for the Bac-T-Screen and the Chemstrip LN ranged from 48 to 58% for colony counts of >105 to >10' CFU/ml. The predictive value of a positive test increased by approximately 40% for each test method as the interpretive breakpoint of the reference method decreased from >105 to >101 CFU/ml, whereas the predictive value of a negative test decreased (Table 2). Cost analysis. The total cost of screening by each method was determined (Table 4). Although salaries differ from laboratory to laboratory, the cost per test was based on the average salary of the technical personnel at our institution. For detection, the cost of the test methods ranged from $0.44 for the Chemstrip LN to $1.17 for the Gram stain compared with $1.63 for the reference culture method. However, when the average cost for both positive and negative specimens was calculated, the test methods were more expensive than the reference method, ranging from $0.12 higher for the UTIscreen to $0.49 higher for the Bac-T-Screen, with one exception, the Chemstrip LN, which was $0.20 lower. DISCUSSION The purpose of urine screening is to improve patient management by rapidly reporting negative results and identifying those specimens which do not warrant culture. Cri-
TABLE 4. Cost analysis of rapid urine screens at >105 CFU/ml Cost ($)/specimen by: Determination Culture method
Cost of supplies Cost of technical time" Cost of detection Avg cost/specimen, including positive and negative screen result Difference from culture
0.75a 0.88 1.63 1.63
UTIscreen
Bac-T-Screen
Chemstrip LN
Gram stain
0.75 0.22 0.97 1.75
0.75 0.33 1.08 2.12
0.20 0.22 0.44 1.43
0.05
+0.12
+0.49
-0.20
+0.34
1.12
1.17 1.97
a Includes 5% sheep blood agar and biplate of MacConkey agar and polymyxin-B-nalidixic acid blood agar. b Estimated at $13.00/h assuming 4.0, 1.0, 1.5, 1.0, and 5.1 min per sample for culture, UTIscreen, Bac-T-Screen, Chemstrip LN, and Gram stain, respectively.
VOL. 27, 1989
RAPID BACTERIURIA SCREENING BY BIOLUMINESCENCE
teria for selection of a rapid urine screen should include accuracy, reproducibility, detection time, ease of test performance, and low cost. Previous evaluations of microscopic and filtration methods have been favorable regarding sensitivities, predictive values of negative test results, and detection times (10, 13, 14, 22, 23). During recent years, the criterion of 2105 CFU/ml established by Kass (7) as the definition of a urinary tract infection has been challenged. Studies by Latham et al. (8) and Stamm et al. (18) found .102 CFU/ml to be a better predictor of infection in symptomatic women. Lipsky et al. (9) and Musher et al. (11) recommended that 2103 CFU/ml be used when evaluating voided urine specimens from men. Stark and Maki (19) found that .102 CFU/ml was a more valid index when patients were catheterized and had urinary symptoms or were immunosuppressed. Pfaller et al. (16) studied a random patient population and also found that lower count cultures most accurately identified urine specimens from infected patients. Because of the findings of these investigators, the rapid screens described here were evaluated at various colony counts. Previous evaluations employing bioluminescence assays for urine screening have demonstrated sensitivities of .85% at >i04 CFU/ml, with predictive values of negative test results of -95% (2, 5, 6, 20, 21, 24, 25). The overall sensitivity at >105 CFU/ml for the UTIscreen in this study was 96% when a 25% integrated light output was used as the positive breakpoint. At this interpretive breakpoint, the UTIscreen was as sensitive as the Bac-T-Screen and the Gram stain. All three methods had sensitivities of at least 90% at 103 CFU/ml for probable pathogens and 84% at 10'
CFU/ml.
The UTIscreen missed Candida spp. and Enterococcus spp. more frequently than the Bac-T-Screen and the Gram stain did. A possible explanation may be that some strains of bacteria and yeast had lower ATP levels in spite of high colony counts (21). Another explanation may be that the bacterial releasing agent did not sufficiently lyse the cell walls of these gram-positive organisms (24). These factors may have contributed to a lower UTIscreen sensitivity for these species, which were detected by the other screening methods. The major difference between the semiautomated methods, the UTIscreen and the Bac-T-Screen, is in the specificities of these methods. The specificity of the UTIscreen at >105 CFU/ml was higher than that of the Bac-T-Screen (70 versus 48%); however, the sensitivities were the same (96%). Although the specificities were increased in all instances by lowering the interpretive breakpoint to either 103 or 10' CFU/ml, the specificity of the Bac-T-Screen remained lower than that of the UTIscreen. A possible explanation for decreased specificities by the test methods is that low-level bacteriuria may not be detected by the culture method. Also, it has been reported that the culture method may have an error rate of as high as 50% (1). Another contributing factor may have been the presence of antimicrobial agents which inhibit microbial growth but do not inhibit detection by bioluminescence or filtration (6). The Bac-T-Screen detects leukocytes as well as bacteria (14). The trapping of leukocytes along with other cells (i.e., squamous epithelial cells) by the filter card probably accounts for the lower BacT-Screen specificity. This may also account for lower specificities by the UTIscreen at the higher colony counts. If the somatic releasing agent does not destroy all of the somatic cell ATP, it will be detected by the luminometer. Macroscopically bloody urine specimens have been reported to be
719
positive by bioluminescence (2). The specificities of the UTIscreen at various colony counts are acceptable, considering that the main purpose of a urine screen is to rapidly identify negative specimens while reliably detecting positive specimens. In this study, the UTIscreen results were similar to the Gram stain results regarding sensitivity, specificity, and predictive values at various colony counts. Although the Gram stain compares favorably with other urine screens and test interpretation is subjective, it requires more labor time to perform than the UTIscreen, therefore resulting in a higher cost per test. In addition, it can be a tedious procedure because most urine specimens are negative. However, it can give a preliminary tentative identification of the organism group and the presence of leukocytes. Both the UTIscreen and the Bac-T-Screen employ instrumentation; therefore, interpretation of test results is objective, whereas the Gram stain interpretation is subjective, requinng technical expertise. Another difference between the instrument methods is the volume of urine required to perform the tests; 0.025 ml for the UTIscreen compared with 1.0 ml for the Bac-T-Screen. The ease of performance is an additional advantage of the
UTIscreen. Once the 0.025 ml of urine is added to the tube containing the dehydrated somatic cell releasing agent and the tube is allowed to incubate at room temperature for 10 min, technical time required to complete the test is minimal (15 s). Reagents are added and results are printed automatically. Specimens can be easily batched, whereas the other methods require individual handling. Cost is an important consideration and the UTIscreen compared closely with the culture method in this study when both positive and negative test results were considered. In conclusion, the UTIscreen compared favorably with other screening methods with respect to sensitivity, predictive value of a negative test, ease of performance, and cost. The advantages of this rapid urine screen include the ability to batch test, objective interpretation, higher specificity than the enzyme dipstick or the colorimetric filtration methods, and same-day results with cost similar to culture. LITERATURE CITED 1. Albers, A. C., and R. D. Fletcher. 1983. Accuracy of calibratedloop transfer. J. Clin. Microbiol. 18:40-42. 2. Bixier-Forell, E., M. A. Bertram, and D. A. Bruckner. 1985. Clinical evaluation of three rapid methods for the detection of
significant bacteriuria. J. Clin. Microbiol. 22:62-67. 3. Clarridge, J. E., M. T. Pezzlo, and K. L. Vosti. 1987. Cumitech 2A, Laboratory diagnosis of urinary tract infections. Coordinating ed., A. L. Weissfeld. American Society for Microbiology, Washington, D.C. 4. College of American Pathologists Workload Recording Committee. 1987. Manual for laboratory workload recording method, 1988 ed., p. 111-125. College of American Pathologists, Traverse City, Mich. 5. Drow, D. L., C. H. Baum, and G. Hirschfield. 1984. Comparison of the Lumac and Monolight systems for detection of bacteriuria by bioluminescence. J. Clin. Microbiol. 20:797-801. 6. Johnston, H. H., C. J. Mitchell, and G. D. W. Curtis. 1976. An automated test for the detection of significant bacteriuria. Lancet ii:400-402. 7. Kass, E. H. 1956. Asymptomatic infections of the urinary tract. Trans. Assoc. Am. Phys. 69:56-63. 8. Latham, R. H., E. S. Wong, A. Larson, M. Coyle, and W. E. Stamm. 1985. Laboratory diagnosis of urinary tract infection in ambulatory women. J. Am. Med. Assoc. 254:3333-3335. 9. Lipsky, B. A., R. C. Ireton, S. D. Fihn, R. Hackett, and R. E. Berger. 1987. Diagnosis of bacteriuria in men: specimen collection and culture interpretation. J. Infect. Dis. 155:847-854.
720
J. CLIN. MICROBIOL.
PEZZLO ET AL.
10. Murray, P. R., R. B. Smith, and T. C. McKinney. 1987. Clinical evaluation of three urine screening tests. J. Clin. Microbiol. 25:467-470. 11. Musher, D. M., S. B. Thorsteinsson, and V. M. Andriole. 1976. Quantitative urinalysis: diagnosing urinary tract infection in men. J. Am. Med. Assoc. 236:2069-2072. 12. Pezzlo, M. T. 1988. Detection of urinary tract infections by rapid methods. Clin. Microbiol. Rev. 1:268-280. 13. Pezzlo, M. T., G. L. Tan, E. M. Peterson, and L. M. de la Maza. 1982. Screening of urine cultures by three automated systems. J. Clin. Microbiol. 15:468-474. 14. Pezzlo, M. T., M. A. Wetkowski, E. M. Peterson, and L. M. de la Maza. 1985. Detection of bacteriuria and pyuria within two minutes. J. Clin. Microbiol. 21:578-581. 15. Pfaller, M. A., and F. P. Koontz. 1985. Laboratory evaluation of leukocyte esterase and nitrite tests for the detection of bacteriuria. J. Clin. Microbiol. 21:840-842. 16. Pfaller, M. A., B. Ringenberg, L. Rames, J. Hegeman, and F. Koontz. 1987. The usefulness of screening tests for pyuria in combination with culture in the diagnosis of urinary tract infection. Diagn. Microbiol. Infect. Dis. 6:207-215. 17. Ransohoff, D. F., and A. R. Feinstein. 1979. Problems of spectrum bias in evaluating the efficiency of diagnostic tests. N. Engl. J. Med. 299:926-930. 18. Stamm, W. E., G. W. Counts, K. R. Running, S. Fihn, M.
19. 20. 21.
22. 23. 24.
25.
Turck, and K. K. Holmes. 1982. Diagnosis of coliform infection in acute dysuric women. N. Engl. J. Med. 307:463-468. Stark, R. P., and D. G. Maki. 1984. Bacteriuria in the catheterized patient: what quantitative level of bacteriuria is relevant? N. Engl. J. Med. 311:560-564. Thore, A., A. Lundin, and S. Ansehn. 1983. Firefly luciferase ATP assay as a screening method for bacteriuria. J. Clin. Microbiol. 17:218-224. Thore, A., S. A. A. Lundin, and S. Bergman. 1975. Detection of bacteriuria by luciferase assay of adenosine triphosphate. J. Clin. Microbiol. 1:1-8. Tilton, R. E., and R. C. Tilton. 1980. Automated direct antimicrobial susceptibility testing of microscopically screened urine cultures. J. Clin. Microbiol. 11:157-161. Washington, J. A., II, C. M. White, M. Laganiere, and L. H. Smith. 1981. Detection of significant bacteriuria by microscopic examination of urine. Lab. Med. 12:294-296. Welch, W. D., L. Thompson, M. Layman, and P. M. Southern, Jr. 1984. Evaluation of two bioluminescence-measuring instruments, the Turner design and Lumac systems, for the rapid screening of urine specimens. J. Clin. Microbiol. 20:1165-1170. Wu, T. C., E. C. Williams, S. Y. Koo, and J. D. MacLowry. 1985. Evaluation of three bacteriuria screening methods in a clinical research hospital. J. Clin. Microbiol. 21:796-799.
