- Email: [email protected]
Evaluation of the Sysmex UF1000i flow cytometer for ruling out bacterial urinary tract infection
Clinica Chimica Acta 411 (2010) 1137–1142
Contents lists available at ScienceDirect
Clinica Chimica Acta j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c l i n c h i m
Short communication
Evaluation of the Sysmex UF1000i flow cytometer for ruling out bacterial urinary tract infection Rita De Rosa ⁎, Shamanta Grosso, Graziano Bruschetta, Manuela Avolio, Paola Stano, Maria Luisa Modolo, Alessandro Camporese Microbiology and Virology Department, Azienda Ospedaliera S. Maria degli Angeli, Via Montereale 24, 33170 Pordenone, Italy
a r t i c l e
i n f o
Article history: Received 22 June 2009 Received in revised form 22 March 2010 Accepted 22 March 2010 Available online 30 March 2010 Keywords: Urine culture Urinary tract infections (UTI) Bacteriuria screening Flow cytometer Urinalysis
a b s t r a c t Background: Urine culture is one of the most frequently requested tests in microbiology, and it represents the gold standard for the diagnosis of UTIs. Considering the high prevalence of negative results and the long TAT of the culture test, the use of a rapid and reliable screening method is becoming more and more important, as it reduces the workload, the TAT of negative results, and above all, unnecessary antibiotic prescription. Methods: The Sysmex UF1000i is a new urine flow cytometry analyzer capable of quantifying urinary particles, including BACT, WBCs, and YLCs. To evaluate the analytical performance of the UF1000i as a method for ruling out UTIs, we examined 1349 urine samples and compared the UF1000i results with standard urine culture results. Results: With instrument cut-off values of 170 BACT × 106/L and 150 WBCs × 106/L, we obtained a sensitivity of 98.8%, a specificity of 76.5%, a NPV of 99.5%, and four false negative results (1.2%), avoiding the culture of 57.1% of samples. Conclusion: The Sysmex UF1000i was capable of improving the efficiency of a routine microbiology laboratory by processing 100 samples/h and providing negative results in a few minutes, thus reducing unnecessary testing with an acceptable number of false negative results. In addition, the preliminary evaluation of B_FSC and B_FLH parameters from bacteria histograms seems to be useful for the distinction of bacterial strains detected (Gram-negatives versus Gram-positives). In fact when B_FSC was less than 30 ch, it allowed the distinction of Gram-negative bacteria in 97% of the samples. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Urinary tract infections (UTIs) represent the most frequently occurring infectious diseases in hospitals and community populations [1,2]. The urine culture test is the gold standard for identifying the aetiological agent of a UTI, estimating its concentration, offering susceptibility testing for antimicrobials, and following the effects of antimicrobial treatment [3]. However, these tests are labour-intensive, time-consuming, and do not provide the same-day results. Further, the frequency of UTIs generates a significant workload for the laboratory, and a large proportion of tested samples have negative results. Abbreviations: AUC, Area under the curve; BACT, Bacteria; B_FSC, Bacteria forward scatter; B_FLH, Bacteria fluorescent light intensity; Ch, Analytical channel; CFB, Colonyforming bacteria; CFU, Colony-forming unit; CI, Confidence interval; CLED, Cystinelactose electrolyte deficient; CNA, Colistin-nalidixic acid; LIS, Laboratory information system; NPV, Negative predictive value; PPV, Positive predictive value; RBC, Red blood cell; ROC, Receiver Operating Characteristic; TAT, Turn around time; UTI, Urinary tract infection; WBC, White blood cell; YLC, Yeast-like cell. ⁎ Corresponding author. Tel.: + 39 434399650; fax: + 39 434399170. E-mail address: [email protected] (R. De Rosa). 0009-8981/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cca.2010.03.027
Today, new automated methods based on advanced technologies are available for screening out negative samples (i.e., the rule out strategy) before processing them for culture [2,4–6]. Avoiding unnecessary cultures with a rapid screening method would provide a fast report of negative results, shortening the turn around time (TAT) as recommended by the leading quality process control authorities [7,8]. However, the priority of an optimal screening method in ruling out the diagnosis of a UTI is to guarantee a high sensitivity and a negative predictive value (NPV). The European Urinalysis Guidelines recommend an analytical sensitivity N90–95% to detect asymptomatic bacteriuria at 108 colony-forming bacteria/litre (CFB/L), equivalent to 105 colony-forming unit/millilitre (CFU/mL), by a rapid non-culture method, with a confirmatory culture of positive cases. The Sysmex UF1000i (Sysmex Co. Japan), supplied by Dasit SpA (Cornaredo, Italy), is a recently introduced fluorescence flow cytometer intended for urinalysis purposes which provides new analytical features that seem particularly suitable for microbiological diagnostics. Bacterial detection and counting are performed by the analyzer in a dedicated analytical channel with a specific reagent system. Additional technical parameters provide information on the
1138
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
size and staining properties of the particles that are classified and counted on this channel. In view of these facts, we evaluated the analytical performance of the Sysmex UF1000i as an automated and rapid method to rule out negative samples in the diagnosis of UTIs in comparison with the gold standard method represented by quantitative urine culture and the usefulness of the new technical parameters for bacterial morphology assessment. 2. Materials and methods 2.1. Patients and samples We studied 1349 consecutive urine samples collected from 870 females and 479 males, aged between 0 and 98 years (mean 54 years). These samples were examined by our microbiology laboratory with a specific request for quantitative urine cultures. Twenty-nine percent of the samples were from patients staying in the hospital or in long term care facilities, and 71.0% were from outpatients. This distribution was representative of the population tested over the past two years (27.0 and 73.0% in 2008 and 26.0 and 74.0% in 2007 from inpatients and outpatients, respectively). The samples from children younger than one year and from infants aged between 0 and 3 years accounted for 1.6 and 3.8% of the total number of samples, respectively. The samples were obtained from midstream urine (detailed instructions with illustrations for collection were provided) using a disposable, sterile, neutral container, screw lid with an integrated transfer straw (BD Vacutainer® Collection container, BD Diagnostics — Preanalytical systems, Milan, Italy). Immediately after the urine collection, a specific, sterile, preservative-free PET tube (BD Vacutainer® Urinalysis plus tube, BD Diagnostics — Preanalytical systems, Milan, Italy) was filled, through the straw of the container, with 11 mL of the urine sample. In infants older than one year, clean catch urine samples were obtained; in younger babies, a sterile collection bag was applied for a maximum of 30 min after carefully washing the genital region, and urine flow was frequently checked. All samples were collected on the morning of the examination, processed within 4 h after collection and within 30 min after their submission to the laboratory. For each specimen, a single vacuum test tube was used for microbiological examination and for analysis with the Sysmex UF1000i. 2.2. Sysmex UF1000i The Sysmex UF1000i is a fully automated fluorescence flow cytometer able to classify and count cells and formed particles [bacteria (BACT), white blood cells (WBCs), red blood cells (RBCs), yeast-like cells (YLCs), epithelial cells, crystals, casts, spermatozoa, small round cells, and mucus] in native uncentrifuged urine specimens. The previous generation of urine flow cytometers (Sysmex UF100), originally developed for urinalysis, used two fluorescent dyes (phenanthridine for staining cellular nucleic acids and carbocyanine for membrane staining), a single analytical channel, and a 488 nm argon laser. The particles are classified on the basis of their size, shape, volume, and staining features. The Sysmex UF1000i differs from the UF100 in its use of a 635 nm semi-conductor diode laser and two separated analytical channels with dedicated polymethine-based fluorescent dyes, each one at a specific temperature and incubation time. In the BACT channel, the urine specimen is mixed at a controlled temperature of 42 °C with a special diluent that increases the permeability of the bacterial cell membrane and facilitates the specific staining of the bacterial nucleic acids with the dedicated polymethine
fluorescent dye. The particles are then classified and counted on the basis of their size and staining characteristics, using the forward scatter and fluorescence light intensity emitted by each cell, and presented in specific histogram and scattergram patterns generated from the BACT channel. Two additional parameters available from this channel, the B_FSC (bacteria forward scatter) and the B_FLH (bacteria fluorescent light intensity), expressed in arbitrary units (analytical channel, ch), provide information on the size and the nucleic acid content, respectively, of the bacteria particles. The Sysmex UF1000i can theoretically analyse up to 100 samples/ h, requiring a volume of 4.0 mL of uncentrifuged native urine sample in an automated mode (where the sample is automatically mixed) or 1.0 mL in a manual mode. All the samples were analyzed with the Sysmex UF1000i in accordance with the manufacturer's recommendations, immediately after inoculation of the cultures. A conductivity value b6 mS/cm as an indicator of sample dilution was set for excluding samples from evaluation. After the analysis was completed, all the results were available for transmission to the laboratory information system (LIS) for real-time reporting immediately after data validation. 2.3. Urine culture A standard quantitative urine culture was performed on all the samples. Well-mixed urine specimens were inoculated with a 1-μL calibrated loop onto non-selective cystine-lactose electrolyte deficient (CLED) agar plates (Kima, Padua, Italy) and selective colistinnalidixic acid + 5% sheep blood (CNA) agar (Kima, Padua, Italy). CLED was used as a quantitative reference, and CNA was used for improved isolation and preliminary identification of Gram-positive bacteria and for making the distinction of contaminants from uropathogenic species easier. The plates were incubated aerobically at 37 °C for 18–24 h and examined for significant bacteriuria, and one or two potentially pathogenic microorganisms. The limit of significant bacteriuria was considered to be 107 CFB/L, corresponding to the conventional 104 CFU/mL. For children younger than one year, significant bacteriuria was set at 106 CFB/L. Samples showing the growth of more than two species of bacteria, without a predominant organism, were considered as positive cultures but classified as contaminated and not subjected to the identification procedure. Bacterial identification was performed by using the Vitek 2 automated system (bioMerieux, Florence, Italy). In cases of insufficient discrimination of the strains, the identification was obtained by conventional biochemical reactions or by the API identification system (bioMerieux, Florence, Italy). 2.4. Data analysis The results for the Sysmex UF1000i BACT and WBC counts were compared with the results of urine cultures using Receiver Operating Characteristic (ROC) curve analysis. Different cut-off values for bacteria and WBCs were evaluated to determine the sensitivity, specificity, and predictive values for predicting significant bacteriuria, with respect to the reference urine culture test at a limit of ≥107 CFB/L. Considering the morphological differences between the bacteria most frequently found in UTIs (Gram-negative rods versus Grampositive cocci), we evaluated the usefulness of the B_FSC and B_FLH research parameters in discriminating between Gram-negative and Gram-positive microorganisms. Samples with a pure culture of Gram-negative (220 samples) or Gram-positive (45 samples) bacteria or with two isolated species that were both Gram-negative (nine samples) were considered, for a total of 274 samples. The data from the Gram-negative bacteria (229 samples) and Gram-positive bacteria (45 samples) were evaluated, and the distribution of B_FSC and B_FLH values in the two groups was
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
1139
compared by the Mann–Whitney test. A p-value b 0.05 was considered statistically significant. All data were recorded on Microsoft Excel spreadsheets. Statistical analysis was performed with the Analyse-It version 2.13 Clinical Laboratory module (Analyse-It Software, Leeds, England) program and Microsoft Excel for Windows in the Microsoft Office Professional 2000 package. 3. Results 3.1. Screening of significant bacteriuria Out of 1349 urine specimens evaluated, 346 bacterial culture samples were positive (25.6%) and 1003 samples were negative (74.4%). Of the 346 positives cultures, 210 were positive at 109 CFB/L, 65 were positive at 108 CFB/L, and 71 were positive at 107 CFB/L. Among the 346 positive samples, a single microorganism was identified in 268 samples, two pathogens were identified in 18 samples, and 60 samples had a mixed growth (10 at 109 CFB/L, 17 at 108 CFB/L, and 33 at 107 CFB/L) with more than two species of bacteria and no predominant organism. The pathogens of this “mixed flora” were not subjected to the subsequent species identification procedure. Although we considered 106 CFB/L as the cut-off for positive culture results in infants younger than one year, no positive sample at this level was found in the 21 children in this age group. The microorganisms identified (304 in total), in a decreasing order of prevalence, were: Escherichia coli (n = 184), Enterococcus faecalis (n = 35), Proteus mirabilis (n = 19), Klebsiella pneumoniae (n = 19), Streptococcus agalactiae (n = 11), Pseudomonas aeruginosa (n = 7), Staphylococcus coagulase negative (n = 5), Citrobacter koseri (n = 4), Staphylococcus saprophyticus (n = 3), Enterobacter aerogenes (n = 3), Candida albicans (n = 3), Klebsiella oxytoca (n = 2), Morganella morganii (n = 2), Staphylococcus aureus (n = 1), and other Gramnegative bacteria (n = 6). We isolated 10 strains of E. coli (three at bacterial growth of 109 CFB/L, two at 108 CFB/L, and five at 107 CFB/L), three strains of P. mirabilis (two at 109 CFB/L and one at 108 CFB/L), and one strain of M. morganii (109 CFB/L) from infants under three years of age. Three specimens presented a growth of mixed flora (one at 108 CFB/L and two at 107 CFB/L), and no Gram-positive strains were isolated from this population. In nine samples with a growth of one Gram-positive and one Gramnegative bacteria, the isolation of the Gram-positive bacteria from the CNA plate was immediately possible without the need for re-spreading. In one case in which P. aeruginosa and E. faecalis were isolated, the growth of Enterococcus was evident only on the CNA plate. These results reflect data on the etiology of uropathogen isolates over the past two years from hospital inpatients and from community outpatients in our laboratory. Our results are also similar to those reported in another Italian study [9]. Our case report is represented, for the most part, by communityacquired UTIs, particularly in the elderly. For this reason, as well as the lack of information on symptoms and antibiotic treatment, the 107 CFB/L threshold for significant bacteriuria, the 1-μL disposable loop inoculum, and the 24-h incubation time were considered acceptable for routine use. The ROC curves for the Sysmex UF1000i BACT and WBC counts are shown in Fig. 1, in which culture results ≥107 CFB/L were taken as the reference for significant bacteriuria. The area under the curve (AUC) for the BACT count is 0.96 (95% CI = 0.95–0.97), which is higher than that for the WBC count (0.83; 95% CI = 0.80–0.86). The performance of the Sysmex UF1000i was examined for a range of cut-off values for the BACT count, along with the best cut-off point obtained from the ROC curve analysis. The true positive, false positive and negative values were determined for delineating the best cut-off for significant bacteriuria. The results are shown in Table 1.
Fig. 1. ROC curve for UF1000i bacterial count (BACT UF) and UF1000i leucocyte count (WBC UF) versus urine quantitative culture in 1349 specimens with 346 cases positive in culture. Urine cultures were considered positive if the bacterial growth was ≥107 CFB/L.
The best cut-off point of 440 × 106/L (with 91.3% sensitivity and 90.5% specificity from ROC analysis) and most sensitive cut-off point of 20 × 106/L BACT were excluded due to the number of false negatives and false positives, respectively, that would have affected the clinical outcome and efficiency of the screening process (see Table 1). Among the other cut-off values for bacterial counts, the best compromise between sensitivity and specificity was selected at 170×106/L. At this value, the number of false negatives was 47% lower than the 17 false negative samples at 265×106 BACT/L. Because the presence of leukocytes is an important parameter in UTI detection (and to evaluate the success of treatment) [10], the sensitivity, specificity, PPV, and NPV of the Sysmex UF1000i were also evaluated by considering different cut-off values of the WBCs count to achieve a low number of false negatives. Table 2 shows the sensitivity, specificity, PPV, NPV and true and false positive and negative values calculated when we used an algoritm in which (alternatively or in combination) the positivity for 170 × 106 BACT/L and for different cut-offs for the WBC count were obtained. Five of the nine false negative samples with a BACT count lower than 170 × 106/L had a WBC count greater than 150 × 106/L. Because lower WBC values increased only the number of false positives without a further reduction of false negatives, 150 × 106 WBC/L was chosen along with 170 × 106 BACT/L as the optimum cut-off values for ruling out significant bacteriuria in our population.
Table 1 Diagnostic performance of Sysmex UF1000i at different cut-off values for bacterial count versus urine quantitative culture in 1349 specimens with 346 cases positive in culture. Bacterial count (×106/L)
Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) No. of true positives No. of false negatives No. of true negatives No. of false positives
440
265
170
20
91.3 90.5 76.9 96.8 316 30 908 95
95.1 85.0 68.7 98 329 17 853 150
97.4 79.8 62.4 98.9 337 9 800 203
99.1 43.6 37.7 99.3 343 3 437 566
1140
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
Table 2 Analytical performance of Sysmex UF1000i versus urine quantitative culture results (1349 specimens with 346 cases positive in culture) at three different cut-off values for WBC count and equal cut-off for BACT count: 170 × 106/L. Test is considered positive when at least one of the parameters, BACT count or WBC count, exceeded the cut-off value.
Table 3 Results obtained at B_FSC cut-off value of 30 ch: number and percentage of culturepositive samples with the identification of only Gram-negative or Gram-positive bacteria. Total no. of samples
WBC count (×106/L)
Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) No. of true positives No. of false negatives No. of true negatives No. of false positives
150
100
50
98.8 76.5 59.17 99.5 342 4 767 236
98.8 75.6 58.26 99.5 342 4 758 245
98.8 73.3 56.07 99.5 342 4 735 268
Because we did not measure antibiotic concentrations in the samples, we cannot rule out that, in the 33 negative cultures selected from WBC count N150 × 106/L, there may be some cases of low bacterial load (i.e., currently false negative cultures caused, for example, by unsuccessful treatment). Four samples were found to be culture-positive and Sysmex UF1000i-negative (false negatives, 1.2%). The culture results for these four samples were: mixed commensal flora, 107 CFB/L (one sample); E. faecalis, 107 CFB/L (one sample); P. mirabilis, 108 CFB/L (one sample); and E. coli, 108 CFB/L (one sample). The Sysmex UF1000i, at the aforementioned combined cut-off values, demonstrated a sensitivity of 98.8%, a specificity of 76.5%, a PPV of 59.2%, a NPV of 99.5%, and an agreement of 81.8% with the culture method. We have evaluated the between specimens mean carryover rate for bacterial counting that emerged to be 0.015%. 3.2. B_FSC and B_FLH indices
Gram neg Gram pos Total
Samples with B_FSC b 30 ch
Samples with B_FSC ≥30 ch
No.
%
No.
%
No.
%
229 45 274
83.6 16.4 100
163 5 168
97.0 3.0 100
66 40 106
62.3 37.7 100
(44.6 ch, p b 0.0001) bacteria. The 95% central percentile range of B_FSC was 11.5–70.4 ch in Gram-negative bacteria and 17.6–127.5 ch in Gram-positive bacteria. Of the 274 samples considered in this study, 168 (61.3%) had a B_FSC value b30 ch, 163 (97.0%) were from samples with Gramnegative bacteria and five (3.0%) from samples with Gram-positive bacteria. The results are summarized in Table 3. The distribution of B_FLH values in Gram-negative and Grampositive bacteria is shown in Fig. 3. The median B_FLH was significantly lower in Gram-negative (85.5 ch) than in Gram-positive (94.6 ch, p = 0.027) bacteria, but the 95% central percentile range of B_FLH in the two groups was similar (72.3–179.4 ch in Gramnegatives and 75.2–168.8 in Gram-positives), as can be observed in Fig. 3. As shown in Fig. 3, due to the overlap between the two groups, the B_FLH parameter seems to be less effective than B_FSC in discriminating between Gram-negative and Gram-positive bacteria. The distribution of B_FSC and B_FLH was also evaluated in the 69 culture samples positive for two species (one Gram-negative and one Gram-positive) and for mixed flora. The median B_FSC in this group (37.2 ch) was significantly higher than that of the Gram-negative group (p b 0.0001) but lower than that of the Gram-positive group
Morphological assessments were made experimentally, for the first time in our study, on the basis of the B_FSC and B_FLH parameters. The distribution of B_FSC values in Gram-negative and Grampositive bacteria is shown in Fig. 2. The median B_FSC was significantly lower in Gram-negative (22.1 ch) than in Gram-positive
Fig. 2. Distribution of B_FSC (forward scatter) values in 220 Gram negative (Gram NEG) and 45 Gram positive (Gram POS) bacterial strains identified in culture-positive samples. The data were evaluated using non-parametric Mann–Whitney statistical analysis. Boxes represent quartile ranges of B_FSC, with a horizontal line for the median value.
Fig. 3. Distribution of B_FLH (fluorescent light intensity) values in 220 Gram negative (Gram NEG) and 45 Gram positive (Gram POS) bacterial strains identified in culturepositive samples. The data were evaluated using non-parametric Mann–Whitney statistical analysis. Boxes represent quartile ranges of B_FLH, with a horizontal line for the median value.
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
(p = 0.0165). The 95% central percentile range of B_FSC in these samples (14.4–108.8 ch) was similar to the range observed in Grampositive bacteria. The median B_FLH was significantly higher in the 69 samples with a mixed flora (median 104.2 ch) than in Gram-negative (pvalue b 0.0001) and in Gram-positive bacteria (p-value = 0.0178), even if the 95% central percentile range of B_FLH (74.2–169.2) was similar to that observed in Gram-negative and Gram-positive bacteria. In all the nine samples with a growth of one Gram-positive and one Gram-negative bacterium, an E. coli strain was identified. In five of these, the B_FSC value was lower than 30 ch and in the other four the maximum value obtained for B_FSC was 47.9. Of the 60 mixed-flora samples, 10 had a B_FSC value lower than 30 ch. However, as the Gram status of the microorganisms of these samples were not subjected to the identification procedure, they were not considered for subsequent experiments in the study. 4. Discussion In conventional clinical microbiology, bacteria culture detection remains the gold standard technique for the diagnosis of UTI, species identification and susceptibility testing. However, this method is timeconsuming and often unnecessarily applied to negative samples. A clinical useful screening method for UTI should be rapid, inexpensive, easy to perform and must have the highest values of sensitivity and NPV. This would mean a prompt reporting of normal samples, an improvement in the efficiency and quality of microbiological diagnoses reducing TAT, without losing valuable time in treating the patients. Moreover the use of automation will allow large numbers of specimens to be processed with reduced technical labour. In the present study we evaluated the performance of Sysmex UF1000i in comparison with the urine culture method for screening urine samples for UTI. We assumed a cut-off ≥107 CFB/L as significant bacteriuria for culture test. The selection of this criterion for the diagnosis of UTI was based on the heterogeneous patient population as well as the lack of information on clinical symptoms and antimicrobial treatment. Possible false negative cultures could be caused by the presence of dead bacteria in the urine due to treatment or a low bacterial load. However, in cases where the presence of UTI symptoms and/or patients taking antibiotic treatments is known, the threshold for diagnosis of UTI can fall to 106 CFB/L or less in particular cases. These cases should therefore be subjected to a culture test, bypassing the screening process. Since high sensitivity and NPV are the priority in ruling out the diagnosis of UTI (the false positive can be correctly diagnosed by culture test), the best cut-off for the BACT count along with WBC count values was examined to obtain the sensitivity required to screen out the negative samples (Table 1). At a cut-off value of 170 × 106 bacteria/L, we obtained nine false negative results (Table 1) that were further reduced to only four samples (1.2%) by the use of the WBC count at a cut-off value of 150 × 106/L when bacteria are counted under the established limit (Table 2). In this setting, the diagnostic performance of the Sysmex UF1000i for all clinically significant uropathogens was considered acceptable for routine use. Earlier studies demonstrated that UF1000i could accurately detect bacteria obtaining a sensitivity of 96.4%, a specificity of 89.1% and a NPV of 97.6% when compared with bacterial cultures at a cut-off ≥107 CFB/L. In addition, correlation of leucocyte count against microscopic chamber comparison method was very good (r = 0.9592) [11]. Our results are also similar to those from a previous evaluation reporting UF1000i sensitivity of 96.5%, specificity of 86.7% and NPV of 98.5% at a BACT limit of 100 × 106/L without considering WBC count [12]. The effect of the screening process by UF1000i on the work-flow in our microbiology laboratory can be estimated from the figures in
1141
Table 2. A total of 771 samples (=767 + 4 cases; 57% from all specimens) were reported as negative with a high NPV of 99.5%, avoiding unnecessary cultures and clinical delays in treatment. Only 236 samples (17.5% from all specimens) were cultured unnecessarily, while 342 positive cases (25.4% from all specimens) were accurately diagnosed. The number of false positive cases could be reduced if the information on clinical symptoms and antimicrobial assumption is obtained, indicating true infection at lower growth, thus enhancing specificity. In fact, in some cases which were positive at the screening, the absence of growth could be explained by the presence of non-vital bacteria due to antibiotic treatment, which was counted using the UF1000i, or by the persistence of high WBC counts caused by an unsuccessful treatment [10]. Using an algorithm on the LIS to select samples to be seeded, the reduction in manual labour could be estimated at 0.5 full time equivalents (FTE) each day in our laboratory. This change would lead to an improvement in TAT and efficiency because, in addition to avoiding seeding almost 57% of the samples, reading an equal amount of plates and manually entering these results into the information system (LIS) could also be avoided. The cut-off value for BACT counts found in our study was significantly lower than the data reported on several studies showing the performance of the Sysmex UF100 for detection and quantification of bacteriuria and for UTI screening [5,6,13–19]. In these studies the best cut-off value for UF 100 was reported to range from 1000 to 8000 × 106 bacteria/L, most frequently around 3000 × 106 bacteria/L. The cut-off value observed for WBC counts ranged from 25 × 106/L to 111 × 106/L; the sensitivities obtained by one or both parameters varied from 73 to 94.4% and NPVs ranged from 96% to 97.9% [5,6,14,18,19]. In addition to the development of software in UF1000i analysers, the differences are mostly explained by the new analytical channel for specific detection of bacteria that allows the Sysmex UF1000i to exclude the identification of debris, mucus and cell fragments from particles thus providing more sensitive and linear results for bacterial counting along with increasing the specificity of the counts by both the routine and specific BACT channels, as reported in earlier publications [11,12,20]. These enhanced analytical features in bacterial detection, together with the first filtering on BACT, could also explain the higher cut-off value found in this study for WBC count (150 × 106/L) in comparison to the above mentioned studies with Sysmex UF100. However, the cut-off values of BACT and WBC counts depend on the patient population studied, the type of specimens, the selected threshold for significant counts in culture, and thus must be investigated and reported by each laboratory. The preliminary results of the B_FSC and B_FLH indeces provided by the Sysmex UF1000i (evaluated for the first time in this study) seem to be very interesting in the morphological assessment of the bacteria. In fact, B_FSC values showed significant differences between Gramnegative and Gram-positive bacteria (Fig. 2, Table 3). These differences in B_FSC values, from a morphological point of view, could be due to the characteristics of the Gram-positive bacteria aggregating in irregular chains and/or irregular grape-like clusters, which are detected and classified by the Sysmex UF1000i as particles with a larger size, with B_FSC values greater than 30 ch in a majority of the samples. B_FLH seems to be less effective in discriminating between Gramnegative and Gram-positive bacteria, showing a significant overlap in their fluorescent staining properties (Fig. 3). This behaviour could be explained by the similar affinity of the polimethine fluorescent dye for staining the nucleic acids of Gram-negative and Gram-positive bacteria. The results obtained in this study indicate that the Sysmex UF1000i is an accurate and cost-effective automated method for ruling out samples that do not need to be cultured. Based on our experience, the analyzer could significantly contribute to the improvement of TAT in the
1142
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
clinical microbiology laboratory, according to the indications of the latest international guidelines [3,7,8]. In fact, Sysmex UF1000i significantly shortens the TAT of negative samples from 24 h to less than 1 h from the time of arrival to the laboratory, as more than 50% of the results can be quickly reported to the clinicians. Considering our workload in one year, this means a real-time reporting of results for approximately 11,000 patients. This could mean an improvement in the quality of patient care, reducing the blind initiation of antibiotic therapies, thus reducing costs and preventing resistances. From a financial point of view, when a negative result is obtained, the test costs are substantially less than the urine culture tests, both for the laboratory and for the patient. Finally, the preliminary results of the B_FSC parameter in discriminating between Gram-negative and Gram-positive bacteria are interesting and require further investigation in extensive studies. Besides the prompt and safe reporting of negative results, positive samples that need to be cultured could be immediately evaluated by looking at the value of the B_FSC index: if lower than the chosen best cut-off value (i.e. 30 ch in our preliminary study) the presence of Gramnegative bacteria seems to be very likely. Given our experience, more than 63% of the organisms (isolated in the community and in hospitals) that cause urinary tract infections were E. coli and altogether more than 80% of the organisms were Gram-negative. In view of the above findings, this new analytical information could represent a useful diagnostic tool to establish a more specific antimicrobial treatment in real time while awaiting results for sensitivity testing. Acknowledgement The authors thank Aurelio Pacioni, Marketing Scientific Manager of Clinical Laboratory (Dasit SpA, Italy) for his technical assistance. References [1] Foxman B. Epidemiology of urinary tract infections: incidence, morbidity, and economic costs. Am J Med 2002;113:5S–13S. [2] Huicho L, Campos-Sanchez M, Alamo C. Metaanalysis of urine screening tests for determining the risk of urinary tract infection in children. Ped Infect Dis J 2002;21: 1–11.
[3] Kouri T, Fogazzi G, Gant G, et al. European Urina1ysis Guidelines. Scand J Clin Lab Invest 2000;60(Suppl 231):1–97. [4] Manoni F, Valverde S, Antico F, Salvadego M, Giacomini A, Gessoni G. Field evaluation of a second-generation cytometer UF-100 in diagnosis of acute urinary tract infections in adult patients. Clin Microbiol Infect 2002;8:662–8. [5] Kim SY, Kim YJ, Lee SM, et al. Evaluation of the Sysmex UF100 Urine Cell Analyzer as a screening test to reduce the need for urine cultures for community-acquired urinary tract infection. Am J Clin Pathol 2007;128:922–5. [6] Evans R, Davidson MM, Sim LRW, Hay AJ. Testing by Sysmex UF100 flow cytometer and with bacterial culture in a diagnostic laboratory: a comparison. J Clin Pathol 2006;59:661–2. [7] Joint Commission International Accreditation and Certification. International Standards for Hospitals, http://www.jointcommissioninternational.org (last consultation 12.06.2009). [8] International Standard ISO 15189: 2007 (E). Medical Laboratories — Particular requirements for quality and competence, http://www.iso.org (last consultation 12.06.2009). [9] De Francesco MA, Ravizzola G, Peroni L, Negrini R, Manca N. Urinary tract infections in Brescia, Italy: etiology of uropathogens and antimicrobial resistance of common uropathogens. Med Sci Monit 2007;13:BR136–44. [10] Ottiger C, Schaer G, Huber AR. Time-course of quantitative urinary leukocytes and bacteria counts during antibiotic therapy in women with symptoms of urinary tract infections. Clin Chim Acta 2007;379:36–41. [11] Okada H, Horie S, Inoue J, Kawashima Y. The basic performance of bacteria counting for diagnosis of urinary tract infection using the fully automated urine particle analyser UF 1000i. Sysmex J Int 2007;17:95–101. [12] von Wulffeln H. Evaluation of a new automated urine cell analyzer (Sysmex UF1000i) for bacteriological urinalysis. 17th European Congress of Clinical Microbiology and Infectious Disease ICC, Munich, Germany, 31 Mar–4 Apr 2007 (Abstract). [13] Ben-Ezra J, McPherson RA. Evaluation of the Sysmex UF100 automated urinalysis analyzer. Clin Chem 1998;44:92–5. [14] Fenili D, Pirovano B. The automation of sediment urinalysis using a new urine flow cytometer (Sysmex UF100). Clin Chem Lab Med 1998;36:909–17. [15] Kouri TT, Kähkönen UM, Malminiemi KH, Vuento R, Rowan RM. Evaluation of Sysmex UF-100 Urine Flow Cytometer vs Chamber Counting o supravitally stained specimens and conventional bacterials cultures. Am J Clin Pathol 1999;112:25–35. [16] Regeniter A, Haenni V, Risch L, et al. Urine analysis performed by flow cytometry: reference range determination and comparison to morphological findings, dipstick chemistry and bacterial culture results. A multicenter study. Clin Nephrol 2001;55: 384–92. [17] Hannemann-Pohl K, Kampf SH. Automation of urine sediment examination: a comparison of the Sysmex UF100 automated flow cytometer with routine manual diagnosis (microscopy, test strips and bacterial culture). Clin Chem Lab Med 1999;37: 753–64. [18] Koken T, Aktepe OC, Serteser M, Samli M, Kahraman A, Dogan N. Determination of cut-off values for leucocytes and bacteria for urine flow cytometer (Sysmex UF100) in urine tract infections. Int Urol Nephrol 2002;34:175–8. [19] Gessoni G, Valverde S, Penzo L. Diagnosis of acute urinary tract infections using Sysmex UF100. Sysmex J Int 2004;14:18–22. [20] Nanos NE, Delanghe JR. Evaluation of Sysmex UF1000i for use in cerebrospinal fluid analysis. Clin Chim Acta 2008;392:30–3.
Contents lists available at ScienceDirect
Clinica Chimica Acta j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c l i n c h i m
Short communication
Evaluation of the Sysmex UF1000i flow cytometer for ruling out bacterial urinary tract infection Rita De Rosa ⁎, Shamanta Grosso, Graziano Bruschetta, Manuela Avolio, Paola Stano, Maria Luisa Modolo, Alessandro Camporese Microbiology and Virology Department, Azienda Ospedaliera S. Maria degli Angeli, Via Montereale 24, 33170 Pordenone, Italy
a r t i c l e
i n f o
Article history: Received 22 June 2009 Received in revised form 22 March 2010 Accepted 22 March 2010 Available online 30 March 2010 Keywords: Urine culture Urinary tract infections (UTI) Bacteriuria screening Flow cytometer Urinalysis
a b s t r a c t Background: Urine culture is one of the most frequently requested tests in microbiology, and it represents the gold standard for the diagnosis of UTIs. Considering the high prevalence of negative results and the long TAT of the culture test, the use of a rapid and reliable screening method is becoming more and more important, as it reduces the workload, the TAT of negative results, and above all, unnecessary antibiotic prescription. Methods: The Sysmex UF1000i is a new urine flow cytometry analyzer capable of quantifying urinary particles, including BACT, WBCs, and YLCs. To evaluate the analytical performance of the UF1000i as a method for ruling out UTIs, we examined 1349 urine samples and compared the UF1000i results with standard urine culture results. Results: With instrument cut-off values of 170 BACT × 106/L and 150 WBCs × 106/L, we obtained a sensitivity of 98.8%, a specificity of 76.5%, a NPV of 99.5%, and four false negative results (1.2%), avoiding the culture of 57.1% of samples. Conclusion: The Sysmex UF1000i was capable of improving the efficiency of a routine microbiology laboratory by processing 100 samples/h and providing negative results in a few minutes, thus reducing unnecessary testing with an acceptable number of false negative results. In addition, the preliminary evaluation of B_FSC and B_FLH parameters from bacteria histograms seems to be useful for the distinction of bacterial strains detected (Gram-negatives versus Gram-positives). In fact when B_FSC was less than 30 ch, it allowed the distinction of Gram-negative bacteria in 97% of the samples. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Urinary tract infections (UTIs) represent the most frequently occurring infectious diseases in hospitals and community populations [1,2]. The urine culture test is the gold standard for identifying the aetiological agent of a UTI, estimating its concentration, offering susceptibility testing for antimicrobials, and following the effects of antimicrobial treatment [3]. However, these tests are labour-intensive, time-consuming, and do not provide the same-day results. Further, the frequency of UTIs generates a significant workload for the laboratory, and a large proportion of tested samples have negative results. Abbreviations: AUC, Area under the curve; BACT, Bacteria; B_FSC, Bacteria forward scatter; B_FLH, Bacteria fluorescent light intensity; Ch, Analytical channel; CFB, Colonyforming bacteria; CFU, Colony-forming unit; CI, Confidence interval; CLED, Cystinelactose electrolyte deficient; CNA, Colistin-nalidixic acid; LIS, Laboratory information system; NPV, Negative predictive value; PPV, Positive predictive value; RBC, Red blood cell; ROC, Receiver Operating Characteristic; TAT, Turn around time; UTI, Urinary tract infection; WBC, White blood cell; YLC, Yeast-like cell. ⁎ Corresponding author. Tel.: + 39 434399650; fax: + 39 434399170. E-mail address: [email protected] (R. De Rosa). 0009-8981/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cca.2010.03.027
Today, new automated methods based on advanced technologies are available for screening out negative samples (i.e., the rule out strategy) before processing them for culture [2,4–6]. Avoiding unnecessary cultures with a rapid screening method would provide a fast report of negative results, shortening the turn around time (TAT) as recommended by the leading quality process control authorities [7,8]. However, the priority of an optimal screening method in ruling out the diagnosis of a UTI is to guarantee a high sensitivity and a negative predictive value (NPV). The European Urinalysis Guidelines recommend an analytical sensitivity N90–95% to detect asymptomatic bacteriuria at 108 colony-forming bacteria/litre (CFB/L), equivalent to 105 colony-forming unit/millilitre (CFU/mL), by a rapid non-culture method, with a confirmatory culture of positive cases. The Sysmex UF1000i (Sysmex Co. Japan), supplied by Dasit SpA (Cornaredo, Italy), is a recently introduced fluorescence flow cytometer intended for urinalysis purposes which provides new analytical features that seem particularly suitable for microbiological diagnostics. Bacterial detection and counting are performed by the analyzer in a dedicated analytical channel with a specific reagent system. Additional technical parameters provide information on the
1138
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
size and staining properties of the particles that are classified and counted on this channel. In view of these facts, we evaluated the analytical performance of the Sysmex UF1000i as an automated and rapid method to rule out negative samples in the diagnosis of UTIs in comparison with the gold standard method represented by quantitative urine culture and the usefulness of the new technical parameters for bacterial morphology assessment. 2. Materials and methods 2.1. Patients and samples We studied 1349 consecutive urine samples collected from 870 females and 479 males, aged between 0 and 98 years (mean 54 years). These samples were examined by our microbiology laboratory with a specific request for quantitative urine cultures. Twenty-nine percent of the samples were from patients staying in the hospital or in long term care facilities, and 71.0% were from outpatients. This distribution was representative of the population tested over the past two years (27.0 and 73.0% in 2008 and 26.0 and 74.0% in 2007 from inpatients and outpatients, respectively). The samples from children younger than one year and from infants aged between 0 and 3 years accounted for 1.6 and 3.8% of the total number of samples, respectively. The samples were obtained from midstream urine (detailed instructions with illustrations for collection were provided) using a disposable, sterile, neutral container, screw lid with an integrated transfer straw (BD Vacutainer® Collection container, BD Diagnostics — Preanalytical systems, Milan, Italy). Immediately after the urine collection, a specific, sterile, preservative-free PET tube (BD Vacutainer® Urinalysis plus tube, BD Diagnostics — Preanalytical systems, Milan, Italy) was filled, through the straw of the container, with 11 mL of the urine sample. In infants older than one year, clean catch urine samples were obtained; in younger babies, a sterile collection bag was applied for a maximum of 30 min after carefully washing the genital region, and urine flow was frequently checked. All samples were collected on the morning of the examination, processed within 4 h after collection and within 30 min after their submission to the laboratory. For each specimen, a single vacuum test tube was used for microbiological examination and for analysis with the Sysmex UF1000i. 2.2. Sysmex UF1000i The Sysmex UF1000i is a fully automated fluorescence flow cytometer able to classify and count cells and formed particles [bacteria (BACT), white blood cells (WBCs), red blood cells (RBCs), yeast-like cells (YLCs), epithelial cells, crystals, casts, spermatozoa, small round cells, and mucus] in native uncentrifuged urine specimens. The previous generation of urine flow cytometers (Sysmex UF100), originally developed for urinalysis, used two fluorescent dyes (phenanthridine for staining cellular nucleic acids and carbocyanine for membrane staining), a single analytical channel, and a 488 nm argon laser. The particles are classified on the basis of their size, shape, volume, and staining features. The Sysmex UF1000i differs from the UF100 in its use of a 635 nm semi-conductor diode laser and two separated analytical channels with dedicated polymethine-based fluorescent dyes, each one at a specific temperature and incubation time. In the BACT channel, the urine specimen is mixed at a controlled temperature of 42 °C with a special diluent that increases the permeability of the bacterial cell membrane and facilitates the specific staining of the bacterial nucleic acids with the dedicated polymethine
fluorescent dye. The particles are then classified and counted on the basis of their size and staining characteristics, using the forward scatter and fluorescence light intensity emitted by each cell, and presented in specific histogram and scattergram patterns generated from the BACT channel. Two additional parameters available from this channel, the B_FSC (bacteria forward scatter) and the B_FLH (bacteria fluorescent light intensity), expressed in arbitrary units (analytical channel, ch), provide information on the size and the nucleic acid content, respectively, of the bacteria particles. The Sysmex UF1000i can theoretically analyse up to 100 samples/ h, requiring a volume of 4.0 mL of uncentrifuged native urine sample in an automated mode (where the sample is automatically mixed) or 1.0 mL in a manual mode. All the samples were analyzed with the Sysmex UF1000i in accordance with the manufacturer's recommendations, immediately after inoculation of the cultures. A conductivity value b6 mS/cm as an indicator of sample dilution was set for excluding samples from evaluation. After the analysis was completed, all the results were available for transmission to the laboratory information system (LIS) for real-time reporting immediately after data validation. 2.3. Urine culture A standard quantitative urine culture was performed on all the samples. Well-mixed urine specimens were inoculated with a 1-μL calibrated loop onto non-selective cystine-lactose electrolyte deficient (CLED) agar plates (Kima, Padua, Italy) and selective colistinnalidixic acid + 5% sheep blood (CNA) agar (Kima, Padua, Italy). CLED was used as a quantitative reference, and CNA was used for improved isolation and preliminary identification of Gram-positive bacteria and for making the distinction of contaminants from uropathogenic species easier. The plates were incubated aerobically at 37 °C for 18–24 h and examined for significant bacteriuria, and one or two potentially pathogenic microorganisms. The limit of significant bacteriuria was considered to be 107 CFB/L, corresponding to the conventional 104 CFU/mL. For children younger than one year, significant bacteriuria was set at 106 CFB/L. Samples showing the growth of more than two species of bacteria, without a predominant organism, were considered as positive cultures but classified as contaminated and not subjected to the identification procedure. Bacterial identification was performed by using the Vitek 2 automated system (bioMerieux, Florence, Italy). In cases of insufficient discrimination of the strains, the identification was obtained by conventional biochemical reactions or by the API identification system (bioMerieux, Florence, Italy). 2.4. Data analysis The results for the Sysmex UF1000i BACT and WBC counts were compared with the results of urine cultures using Receiver Operating Characteristic (ROC) curve analysis. Different cut-off values for bacteria and WBCs were evaluated to determine the sensitivity, specificity, and predictive values for predicting significant bacteriuria, with respect to the reference urine culture test at a limit of ≥107 CFB/L. Considering the morphological differences between the bacteria most frequently found in UTIs (Gram-negative rods versus Grampositive cocci), we evaluated the usefulness of the B_FSC and B_FLH research parameters in discriminating between Gram-negative and Gram-positive microorganisms. Samples with a pure culture of Gram-negative (220 samples) or Gram-positive (45 samples) bacteria or with two isolated species that were both Gram-negative (nine samples) were considered, for a total of 274 samples. The data from the Gram-negative bacteria (229 samples) and Gram-positive bacteria (45 samples) were evaluated, and the distribution of B_FSC and B_FLH values in the two groups was
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
1139
compared by the Mann–Whitney test. A p-value b 0.05 was considered statistically significant. All data were recorded on Microsoft Excel spreadsheets. Statistical analysis was performed with the Analyse-It version 2.13 Clinical Laboratory module (Analyse-It Software, Leeds, England) program and Microsoft Excel for Windows in the Microsoft Office Professional 2000 package. 3. Results 3.1. Screening of significant bacteriuria Out of 1349 urine specimens evaluated, 346 bacterial culture samples were positive (25.6%) and 1003 samples were negative (74.4%). Of the 346 positives cultures, 210 were positive at 109 CFB/L, 65 were positive at 108 CFB/L, and 71 were positive at 107 CFB/L. Among the 346 positive samples, a single microorganism was identified in 268 samples, two pathogens were identified in 18 samples, and 60 samples had a mixed growth (10 at 109 CFB/L, 17 at 108 CFB/L, and 33 at 107 CFB/L) with more than two species of bacteria and no predominant organism. The pathogens of this “mixed flora” were not subjected to the subsequent species identification procedure. Although we considered 106 CFB/L as the cut-off for positive culture results in infants younger than one year, no positive sample at this level was found in the 21 children in this age group. The microorganisms identified (304 in total), in a decreasing order of prevalence, were: Escherichia coli (n = 184), Enterococcus faecalis (n = 35), Proteus mirabilis (n = 19), Klebsiella pneumoniae (n = 19), Streptococcus agalactiae (n = 11), Pseudomonas aeruginosa (n = 7), Staphylococcus coagulase negative (n = 5), Citrobacter koseri (n = 4), Staphylococcus saprophyticus (n = 3), Enterobacter aerogenes (n = 3), Candida albicans (n = 3), Klebsiella oxytoca (n = 2), Morganella morganii (n = 2), Staphylococcus aureus (n = 1), and other Gramnegative bacteria (n = 6). We isolated 10 strains of E. coli (three at bacterial growth of 109 CFB/L, two at 108 CFB/L, and five at 107 CFB/L), three strains of P. mirabilis (two at 109 CFB/L and one at 108 CFB/L), and one strain of M. morganii (109 CFB/L) from infants under three years of age. Three specimens presented a growth of mixed flora (one at 108 CFB/L and two at 107 CFB/L), and no Gram-positive strains were isolated from this population. In nine samples with a growth of one Gram-positive and one Gramnegative bacteria, the isolation of the Gram-positive bacteria from the CNA plate was immediately possible without the need for re-spreading. In one case in which P. aeruginosa and E. faecalis were isolated, the growth of Enterococcus was evident only on the CNA plate. These results reflect data on the etiology of uropathogen isolates over the past two years from hospital inpatients and from community outpatients in our laboratory. Our results are also similar to those reported in another Italian study [9]. Our case report is represented, for the most part, by communityacquired UTIs, particularly in the elderly. For this reason, as well as the lack of information on symptoms and antibiotic treatment, the 107 CFB/L threshold for significant bacteriuria, the 1-μL disposable loop inoculum, and the 24-h incubation time were considered acceptable for routine use. The ROC curves for the Sysmex UF1000i BACT and WBC counts are shown in Fig. 1, in which culture results ≥107 CFB/L were taken as the reference for significant bacteriuria. The area under the curve (AUC) for the BACT count is 0.96 (95% CI = 0.95–0.97), which is higher than that for the WBC count (0.83; 95% CI = 0.80–0.86). The performance of the Sysmex UF1000i was examined for a range of cut-off values for the BACT count, along with the best cut-off point obtained from the ROC curve analysis. The true positive, false positive and negative values were determined for delineating the best cut-off for significant bacteriuria. The results are shown in Table 1.
Fig. 1. ROC curve for UF1000i bacterial count (BACT UF) and UF1000i leucocyte count (WBC UF) versus urine quantitative culture in 1349 specimens with 346 cases positive in culture. Urine cultures were considered positive if the bacterial growth was ≥107 CFB/L.
The best cut-off point of 440 × 106/L (with 91.3% sensitivity and 90.5% specificity from ROC analysis) and most sensitive cut-off point of 20 × 106/L BACT were excluded due to the number of false negatives and false positives, respectively, that would have affected the clinical outcome and efficiency of the screening process (see Table 1). Among the other cut-off values for bacterial counts, the best compromise between sensitivity and specificity was selected at 170×106/L. At this value, the number of false negatives was 47% lower than the 17 false negative samples at 265×106 BACT/L. Because the presence of leukocytes is an important parameter in UTI detection (and to evaluate the success of treatment) [10], the sensitivity, specificity, PPV, and NPV of the Sysmex UF1000i were also evaluated by considering different cut-off values of the WBCs count to achieve a low number of false negatives. Table 2 shows the sensitivity, specificity, PPV, NPV and true and false positive and negative values calculated when we used an algoritm in which (alternatively or in combination) the positivity for 170 × 106 BACT/L and for different cut-offs for the WBC count were obtained. Five of the nine false negative samples with a BACT count lower than 170 × 106/L had a WBC count greater than 150 × 106/L. Because lower WBC values increased only the number of false positives without a further reduction of false negatives, 150 × 106 WBC/L was chosen along with 170 × 106 BACT/L as the optimum cut-off values for ruling out significant bacteriuria in our population.
Table 1 Diagnostic performance of Sysmex UF1000i at different cut-off values for bacterial count versus urine quantitative culture in 1349 specimens with 346 cases positive in culture. Bacterial count (×106/L)
Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) No. of true positives No. of false negatives No. of true negatives No. of false positives
440
265
170
20
91.3 90.5 76.9 96.8 316 30 908 95
95.1 85.0 68.7 98 329 17 853 150
97.4 79.8 62.4 98.9 337 9 800 203
99.1 43.6 37.7 99.3 343 3 437 566
1140
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
Table 2 Analytical performance of Sysmex UF1000i versus urine quantitative culture results (1349 specimens with 346 cases positive in culture) at three different cut-off values for WBC count and equal cut-off for BACT count: 170 × 106/L. Test is considered positive when at least one of the parameters, BACT count or WBC count, exceeded the cut-off value.
Table 3 Results obtained at B_FSC cut-off value of 30 ch: number and percentage of culturepositive samples with the identification of only Gram-negative or Gram-positive bacteria. Total no. of samples
WBC count (×106/L)
Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) No. of true positives No. of false negatives No. of true negatives No. of false positives
150
100
50
98.8 76.5 59.17 99.5 342 4 767 236
98.8 75.6 58.26 99.5 342 4 758 245
98.8 73.3 56.07 99.5 342 4 735 268
Because we did not measure antibiotic concentrations in the samples, we cannot rule out that, in the 33 negative cultures selected from WBC count N150 × 106/L, there may be some cases of low bacterial load (i.e., currently false negative cultures caused, for example, by unsuccessful treatment). Four samples were found to be culture-positive and Sysmex UF1000i-negative (false negatives, 1.2%). The culture results for these four samples were: mixed commensal flora, 107 CFB/L (one sample); E. faecalis, 107 CFB/L (one sample); P. mirabilis, 108 CFB/L (one sample); and E. coli, 108 CFB/L (one sample). The Sysmex UF1000i, at the aforementioned combined cut-off values, demonstrated a sensitivity of 98.8%, a specificity of 76.5%, a PPV of 59.2%, a NPV of 99.5%, and an agreement of 81.8% with the culture method. We have evaluated the between specimens mean carryover rate for bacterial counting that emerged to be 0.015%. 3.2. B_FSC and B_FLH indices
Gram neg Gram pos Total
Samples with B_FSC b 30 ch
Samples with B_FSC ≥30 ch
No.
%
No.
%
No.
%
229 45 274
83.6 16.4 100
163 5 168
97.0 3.0 100
66 40 106
62.3 37.7 100
(44.6 ch, p b 0.0001) bacteria. The 95% central percentile range of B_FSC was 11.5–70.4 ch in Gram-negative bacteria and 17.6–127.5 ch in Gram-positive bacteria. Of the 274 samples considered in this study, 168 (61.3%) had a B_FSC value b30 ch, 163 (97.0%) were from samples with Gramnegative bacteria and five (3.0%) from samples with Gram-positive bacteria. The results are summarized in Table 3. The distribution of B_FLH values in Gram-negative and Grampositive bacteria is shown in Fig. 3. The median B_FLH was significantly lower in Gram-negative (85.5 ch) than in Gram-positive (94.6 ch, p = 0.027) bacteria, but the 95% central percentile range of B_FLH in the two groups was similar (72.3–179.4 ch in Gramnegatives and 75.2–168.8 in Gram-positives), as can be observed in Fig. 3. As shown in Fig. 3, due to the overlap between the two groups, the B_FLH parameter seems to be less effective than B_FSC in discriminating between Gram-negative and Gram-positive bacteria. The distribution of B_FSC and B_FLH was also evaluated in the 69 culture samples positive for two species (one Gram-negative and one Gram-positive) and for mixed flora. The median B_FSC in this group (37.2 ch) was significantly higher than that of the Gram-negative group (p b 0.0001) but lower than that of the Gram-positive group
Morphological assessments were made experimentally, for the first time in our study, on the basis of the B_FSC and B_FLH parameters. The distribution of B_FSC values in Gram-negative and Grampositive bacteria is shown in Fig. 2. The median B_FSC was significantly lower in Gram-negative (22.1 ch) than in Gram-positive
Fig. 2. Distribution of B_FSC (forward scatter) values in 220 Gram negative (Gram NEG) and 45 Gram positive (Gram POS) bacterial strains identified in culture-positive samples. The data were evaluated using non-parametric Mann–Whitney statistical analysis. Boxes represent quartile ranges of B_FSC, with a horizontal line for the median value.
Fig. 3. Distribution of B_FLH (fluorescent light intensity) values in 220 Gram negative (Gram NEG) and 45 Gram positive (Gram POS) bacterial strains identified in culturepositive samples. The data were evaluated using non-parametric Mann–Whitney statistical analysis. Boxes represent quartile ranges of B_FLH, with a horizontal line for the median value.
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
(p = 0.0165). The 95% central percentile range of B_FSC in these samples (14.4–108.8 ch) was similar to the range observed in Grampositive bacteria. The median B_FLH was significantly higher in the 69 samples with a mixed flora (median 104.2 ch) than in Gram-negative (pvalue b 0.0001) and in Gram-positive bacteria (p-value = 0.0178), even if the 95% central percentile range of B_FLH (74.2–169.2) was similar to that observed in Gram-negative and Gram-positive bacteria. In all the nine samples with a growth of one Gram-positive and one Gram-negative bacterium, an E. coli strain was identified. In five of these, the B_FSC value was lower than 30 ch and in the other four the maximum value obtained for B_FSC was 47.9. Of the 60 mixed-flora samples, 10 had a B_FSC value lower than 30 ch. However, as the Gram status of the microorganisms of these samples were not subjected to the identification procedure, they were not considered for subsequent experiments in the study. 4. Discussion In conventional clinical microbiology, bacteria culture detection remains the gold standard technique for the diagnosis of UTI, species identification and susceptibility testing. However, this method is timeconsuming and often unnecessarily applied to negative samples. A clinical useful screening method for UTI should be rapid, inexpensive, easy to perform and must have the highest values of sensitivity and NPV. This would mean a prompt reporting of normal samples, an improvement in the efficiency and quality of microbiological diagnoses reducing TAT, without losing valuable time in treating the patients. Moreover the use of automation will allow large numbers of specimens to be processed with reduced technical labour. In the present study we evaluated the performance of Sysmex UF1000i in comparison with the urine culture method for screening urine samples for UTI. We assumed a cut-off ≥107 CFB/L as significant bacteriuria for culture test. The selection of this criterion for the diagnosis of UTI was based on the heterogeneous patient population as well as the lack of information on clinical symptoms and antimicrobial treatment. Possible false negative cultures could be caused by the presence of dead bacteria in the urine due to treatment or a low bacterial load. However, in cases where the presence of UTI symptoms and/or patients taking antibiotic treatments is known, the threshold for diagnosis of UTI can fall to 106 CFB/L or less in particular cases. These cases should therefore be subjected to a culture test, bypassing the screening process. Since high sensitivity and NPV are the priority in ruling out the diagnosis of UTI (the false positive can be correctly diagnosed by culture test), the best cut-off for the BACT count along with WBC count values was examined to obtain the sensitivity required to screen out the negative samples (Table 1). At a cut-off value of 170 × 106 bacteria/L, we obtained nine false negative results (Table 1) that were further reduced to only four samples (1.2%) by the use of the WBC count at a cut-off value of 150 × 106/L when bacteria are counted under the established limit (Table 2). In this setting, the diagnostic performance of the Sysmex UF1000i for all clinically significant uropathogens was considered acceptable for routine use. Earlier studies demonstrated that UF1000i could accurately detect bacteria obtaining a sensitivity of 96.4%, a specificity of 89.1% and a NPV of 97.6% when compared with bacterial cultures at a cut-off ≥107 CFB/L. In addition, correlation of leucocyte count against microscopic chamber comparison method was very good (r = 0.9592) [11]. Our results are also similar to those from a previous evaluation reporting UF1000i sensitivity of 96.5%, specificity of 86.7% and NPV of 98.5% at a BACT limit of 100 × 106/L without considering WBC count [12]. The effect of the screening process by UF1000i on the work-flow in our microbiology laboratory can be estimated from the figures in
1141
Table 2. A total of 771 samples (=767 + 4 cases; 57% from all specimens) were reported as negative with a high NPV of 99.5%, avoiding unnecessary cultures and clinical delays in treatment. Only 236 samples (17.5% from all specimens) were cultured unnecessarily, while 342 positive cases (25.4% from all specimens) were accurately diagnosed. The number of false positive cases could be reduced if the information on clinical symptoms and antimicrobial assumption is obtained, indicating true infection at lower growth, thus enhancing specificity. In fact, in some cases which were positive at the screening, the absence of growth could be explained by the presence of non-vital bacteria due to antibiotic treatment, which was counted using the UF1000i, or by the persistence of high WBC counts caused by an unsuccessful treatment [10]. Using an algorithm on the LIS to select samples to be seeded, the reduction in manual labour could be estimated at 0.5 full time equivalents (FTE) each day in our laboratory. This change would lead to an improvement in TAT and efficiency because, in addition to avoiding seeding almost 57% of the samples, reading an equal amount of plates and manually entering these results into the information system (LIS) could also be avoided. The cut-off value for BACT counts found in our study was significantly lower than the data reported on several studies showing the performance of the Sysmex UF100 for detection and quantification of bacteriuria and for UTI screening [5,6,13–19]. In these studies the best cut-off value for UF 100 was reported to range from 1000 to 8000 × 106 bacteria/L, most frequently around 3000 × 106 bacteria/L. The cut-off value observed for WBC counts ranged from 25 × 106/L to 111 × 106/L; the sensitivities obtained by one or both parameters varied from 73 to 94.4% and NPVs ranged from 96% to 97.9% [5,6,14,18,19]. In addition to the development of software in UF1000i analysers, the differences are mostly explained by the new analytical channel for specific detection of bacteria that allows the Sysmex UF1000i to exclude the identification of debris, mucus and cell fragments from particles thus providing more sensitive and linear results for bacterial counting along with increasing the specificity of the counts by both the routine and specific BACT channels, as reported in earlier publications [11,12,20]. These enhanced analytical features in bacterial detection, together with the first filtering on BACT, could also explain the higher cut-off value found in this study for WBC count (150 × 106/L) in comparison to the above mentioned studies with Sysmex UF100. However, the cut-off values of BACT and WBC counts depend on the patient population studied, the type of specimens, the selected threshold for significant counts in culture, and thus must be investigated and reported by each laboratory. The preliminary results of the B_FSC and B_FLH indeces provided by the Sysmex UF1000i (evaluated for the first time in this study) seem to be very interesting in the morphological assessment of the bacteria. In fact, B_FSC values showed significant differences between Gramnegative and Gram-positive bacteria (Fig. 2, Table 3). These differences in B_FSC values, from a morphological point of view, could be due to the characteristics of the Gram-positive bacteria aggregating in irregular chains and/or irregular grape-like clusters, which are detected and classified by the Sysmex UF1000i as particles with a larger size, with B_FSC values greater than 30 ch in a majority of the samples. B_FLH seems to be less effective in discriminating between Gramnegative and Gram-positive bacteria, showing a significant overlap in their fluorescent staining properties (Fig. 3). This behaviour could be explained by the similar affinity of the polimethine fluorescent dye for staining the nucleic acids of Gram-negative and Gram-positive bacteria. The results obtained in this study indicate that the Sysmex UF1000i is an accurate and cost-effective automated method for ruling out samples that do not need to be cultured. Based on our experience, the analyzer could significantly contribute to the improvement of TAT in the
1142
R. De Rosa et al. / Clinica Chimica Acta 411 (2010) 1137–1142
clinical microbiology laboratory, according to the indications of the latest international guidelines [3,7,8]. In fact, Sysmex UF1000i significantly shortens the TAT of negative samples from 24 h to less than 1 h from the time of arrival to the laboratory, as more than 50% of the results can be quickly reported to the clinicians. Considering our workload in one year, this means a real-time reporting of results for approximately 11,000 patients. This could mean an improvement in the quality of patient care, reducing the blind initiation of antibiotic therapies, thus reducing costs and preventing resistances. From a financial point of view, when a negative result is obtained, the test costs are substantially less than the urine culture tests, both for the laboratory and for the patient. Finally, the preliminary results of the B_FSC parameter in discriminating between Gram-negative and Gram-positive bacteria are interesting and require further investigation in extensive studies. Besides the prompt and safe reporting of negative results, positive samples that need to be cultured could be immediately evaluated by looking at the value of the B_FSC index: if lower than the chosen best cut-off value (i.e. 30 ch in our preliminary study) the presence of Gramnegative bacteria seems to be very likely. Given our experience, more than 63% of the organisms (isolated in the community and in hospitals) that cause urinary tract infections were E. coli and altogether more than 80% of the organisms were Gram-negative. In view of the above findings, this new analytical information could represent a useful diagnostic tool to establish a more specific antimicrobial treatment in real time while awaiting results for sensitivity testing. Acknowledgement The authors thank Aurelio Pacioni, Marketing Scientific Manager of Clinical Laboratory (Dasit SpA, Italy) for his technical assistance. References [1] Foxman B. Epidemiology of urinary tract infections: incidence, morbidity, and economic costs. Am J Med 2002;113:5S–13S. [2] Huicho L, Campos-Sanchez M, Alamo C. Metaanalysis of urine screening tests for determining the risk of urinary tract infection in children. Ped Infect Dis J 2002;21: 1–11.
[3] Kouri T, Fogazzi G, Gant G, et al. European Urina1ysis Guidelines. Scand J Clin Lab Invest 2000;60(Suppl 231):1–97. [4] Manoni F, Valverde S, Antico F, Salvadego M, Giacomini A, Gessoni G. Field evaluation of a second-generation cytometer UF-100 in diagnosis of acute urinary tract infections in adult patients. Clin Microbiol Infect 2002;8:662–8. [5] Kim SY, Kim YJ, Lee SM, et al. Evaluation of the Sysmex UF100 Urine Cell Analyzer as a screening test to reduce the need for urine cultures for community-acquired urinary tract infection. Am J Clin Pathol 2007;128:922–5. [6] Evans R, Davidson MM, Sim LRW, Hay AJ. Testing by Sysmex UF100 flow cytometer and with bacterial culture in a diagnostic laboratory: a comparison. J Clin Pathol 2006;59:661–2. [7] Joint Commission International Accreditation and Certification. International Standards for Hospitals, http://www.jointcommissioninternational.org (last consultation 12.06.2009). [8] International Standard ISO 15189: 2007 (E). Medical Laboratories — Particular requirements for quality and competence, http://www.iso.org (last consultation 12.06.2009). [9] De Francesco MA, Ravizzola G, Peroni L, Negrini R, Manca N. Urinary tract infections in Brescia, Italy: etiology of uropathogens and antimicrobial resistance of common uropathogens. Med Sci Monit 2007;13:BR136–44. [10] Ottiger C, Schaer G, Huber AR. Time-course of quantitative urinary leukocytes and bacteria counts during antibiotic therapy in women with symptoms of urinary tract infections. Clin Chim Acta 2007;379:36–41. [11] Okada H, Horie S, Inoue J, Kawashima Y. The basic performance of bacteria counting for diagnosis of urinary tract infection using the fully automated urine particle analyser UF 1000i. Sysmex J Int 2007;17:95–101. [12] von Wulffeln H. Evaluation of a new automated urine cell analyzer (Sysmex UF1000i) for bacteriological urinalysis. 17th European Congress of Clinical Microbiology and Infectious Disease ICC, Munich, Germany, 31 Mar–4 Apr 2007 (Abstract). [13] Ben-Ezra J, McPherson RA. Evaluation of the Sysmex UF100 automated urinalysis analyzer. Clin Chem 1998;44:92–5. [14] Fenili D, Pirovano B. The automation of sediment urinalysis using a new urine flow cytometer (Sysmex UF100). Clin Chem Lab Med 1998;36:909–17. [15] Kouri TT, Kähkönen UM, Malminiemi KH, Vuento R, Rowan RM. Evaluation of Sysmex UF-100 Urine Flow Cytometer vs Chamber Counting o supravitally stained specimens and conventional bacterials cultures. Am J Clin Pathol 1999;112:25–35. [16] Regeniter A, Haenni V, Risch L, et al. Urine analysis performed by flow cytometry: reference range determination and comparison to morphological findings, dipstick chemistry and bacterial culture results. A multicenter study. Clin Nephrol 2001;55: 384–92. [17] Hannemann-Pohl K, Kampf SH. Automation of urine sediment examination: a comparison of the Sysmex UF100 automated flow cytometer with routine manual diagnosis (microscopy, test strips and bacterial culture). Clin Chem Lab Med 1999;37: 753–64. [18] Koken T, Aktepe OC, Serteser M, Samli M, Kahraman A, Dogan N. Determination of cut-off values for leucocytes and bacteria for urine flow cytometer (Sysmex UF100) in urine tract infections. Int Urol Nephrol 2002;34:175–8. [19] Gessoni G, Valverde S, Penzo L. Diagnosis of acute urinary tract infections using Sysmex UF100. Sysmex J Int 2004;14:18–22. [20] Nanos NE, Delanghe JR. Evaluation of Sysmex UF1000i for use in cerebrospinal fluid analysis. Clin Chim Acta 2008;392:30–3.