Comparing the Performance of the New Fully Automated Urine Particle Analyzer UF-5000 with UF1000i and Gram Stain in Predicting Bacteria Growth Patterns in Women with Uncomplicated Urinary Tract Infections

Backgroud To compare the performance of the new ow cytometer UF-5000 with UF-1000i (Sysmex, Kobe, Japan) and Gram stain in predicting the bacterial patterns in urine samples Methods Women with symptoms suggestive of urinary tract infection were enrolled. Mid-stream urine sample was collected for gram staining, urine analysis and urine culture. Bacterial patterns were classied though UF1000i (none, cocci bacteria or rods/mixed growth), UF-5000 (none, cocci, rods or mixed growth) and Gram stain.

costs, waiting time and also to help clinicians improve patient care. Previous studies have examined the e cacy of the UF1000i (Sysmex UF-1000i; Sysmex Corporation, Kobe, Japan) in differentiating bacilli from cocci/mixed growth [7]. However, the UF1000i cannot differentiate cocci from mixed growth.
Recently, a new model of urine particle analyzer, the UF5000 (Sysmex Corporation, Kobe, Japan), was introduced where cocci are labeled separately from mixed growth bacteria [8,9]. Therefore, we did a prospective study to compare the accuracy of the gram staining, the UF1000i and the UF5000 in differentiating bacterial growth patterns (gram positive, negative or mixed growth) in midstream voided urine specimens in women visiting urological clinics for uUTI.

Methods
The study was approved by the Institutional Review Board in our hospital. From July, 2016 to June, 2019, we prospectively enrolled adult women (20 to 80 years old) visiting our clinic presenting with symptoms suggestive of urinary tract infection without fever >38°C, urolithiasis, pregnancy, congenital urinary tract anomaly, end stage renal disease under dialysis, neurogenic bladder under urethral catheterization, bladder cancer history, patients who were immunocompromised and recent antibiotics use (within seven days). After written informed consents were signed, patients were asked to complete a questionnaire including baseline characteristics (age, medical history including diabetes or hypertension, childbirth, previous abdominal surgery). They were also asked to complete a urinary tract infection symptoms assessment (UTISA [10]) which included 7 symptom categories and 7 quality of life categories, with scores for each assessment ranging from 0 to 3. Patients with total symptom scores of 4 or more were regarded as positive symptoms of urinary tract infection. At the clinics, the patients were asked to collect midstream voided urine in a sterile container for manual gram staining, routine urinalysis and urine culture. A study nurse instructed the patients on proper collection technique to attempt to reduce the contamination rate. Only specimen with a bacterial growth of ≧10 3 cfu/mL were included for comparison of bacterial growth pattern differentiation. From the sterile collection cup, 10 ml of sample was poured into urine sediment centrifuge tube (SY, Shih-Yung medical instruments Co., Ltd, Taipei, Taiwan) for automated urine particle analysis (Sysmex UF-1000i; UF-5000, Sysmex Corporation, Kobe, Japan) within half an hour after receiving the specimen. The gram staining of urine specimens was classi ed as gram positive, negative or mixed by two experienced clinical laboratory scientists with more than 10 years in the eld. For each analyzed sample, the bacteria scatter diagram was classi ed as rods, cocci/mixed growth or none on UF1000i. The bacteria scatter gram was classi ed as gram positive, negative, mixed growth or none on UF5000 [9,11].

Gram staining
The centrifuged urine from the urine sediment preparation was used to make slides for Gram staining. These slides were air dried, xed with heat, and then stained using the Gram stain procedure. Slides were assessed for the presence of bacteria and the staining characteristics were further described. Slides with bacteria were subsequently evaluated for bacterial morphology and whether these bacteria were Gram positive or Gram negative. Slides were classi ed as positive for bacteriuria if ≥1 bacteria/HPF was noted.
Then, the specimens were classi ed as gram positive, negative, mixed growth or none. [6] Microbiological analysis A 1 µl inoculation loop was applied to the commercial chromogenic agar medium (CPS® ID3, Biomerieux, I′Etoile, France) for urine culture. The culture plates were incubated at 35℃ for 18-24h aerobically. The quanti cation process of bacteria were by multiplying the dilution factor and the colonies on the agar plate. The growth of more than 2 species of bacteria within the urine culture without a dominant one were regarded as contaminated or mixed growth.

Statistical analysis
The data within the manuscript was expressed as mean ± standard deviation. We used MedCalc Statistical Software version 19.1.3 (MedCalc Software bv, Ostend, Belgium;https://www.medcalc.org; 2019) for statistical analysis. Nominal or categorical data were compared with a X 2 test. Ordinal data were compared with Mann-Whitney test. Continuous data were compared with independent t-test, respectively. Agreement between two methods were evaluated with kappa statistics. The grading of agreement complied with Altman's recommendations [12] (<0.2: Poor agreement, 0.21 -0.40: Fair agreement, 0.41-0.60: Moderate agreement, 0.61-0.80: Good agreement, 0.81-1.00: Very good agreement). A p-value of less than 0.05 was considered as statistically signi cant.

Results
There were 85 patients with no UF5000 of UF1000i interpretation and 1 specimen with no growth on culture were excluded for nal comparisons. A total of 102 urine specimens from 102 women (mean age = 58.5 ± 18.5 years) with UTISA ≧ 4 and bacteria growth ≧ 10 3 cfu/mL were included for analysis. Table 1 summarizes the baseline characteristics of the 102 patients. The analyzed specimens included 10 grampositive cocci, 2 gram-positive bacilli, 66 gram-negative rods, and 24 specimens with two bacteria species or more that were regarded as mixed growth (Table 2). Gram-positive bacilli (Lactobacillus species) were excluded for analysis of agreement. Among specimens with single bacteria growth, there were Gram positive cocci ( 2 Streptococci spp., 3 Staphylococci spp., 1 Enterococci spp., 2 Group B Streptococci, 2 unclassi ed G(+)cocci) and Gram negative rods (53 Escherichia coli, 5 Proteus spp., 4 Klebsiella spp., and 4 Citrobacter spp.).  Table 2 The sensitivity of gram stain, UF1000i and UF5000 for speci c bacterial species There were only 97 specimens with gram stain results available. Of them, 29 urine specimens were classi ed as negative. Agreement levels between the results of gram stain and urine cultures are listed in the Table 3 with a kappa value of 0.48 (moderate, 95%CI: 0.36 to 0.60). The sensitivity and speci city of the gram stain for gram negative bacteria was 80.6% and 96.7%, respectively, and the sensitivity and speci city of the gram stain for cocci was 60% and 100%. The sensitivity and speci city of the gram stain for mixed growth was 18.2% and 97.4%, respectively.

UF5000
Agreement levels between the results of the UF5000 laser ow cytometry and urine cultures are listed in the Table 5 with a kappa value of 0.46 (moderate, 95CI: 0.34 to 0.58). The sensitivity and speci city of the UF5000 for GNB was 80.0% and 88.2%, respectively, and the sensitivity and speci city of the gram stain for cocci was 70% and 86.5%.. The sensitivity and speci city of the UF5000 for mixed growth was 4.5% and 94.9%, respectively.

Discussion
This is the rst prospective study that compares the e cacy of automated urine ow cytometry system (UF1000i, UF5000), gram stain and urine cultures in urine specimens of women with uncomplicated urinary tract infections. The results showed that the UF5000 kept a good sensitivity (80.0%) in identifying gram negative bacteria with an acceptable speci city (88.2%). With regard to gram positive bacteria, the UF5000 outperformed the UF1000i in detecting gram positive cocci (UF5000 sensitivity: 70% and speci city: 86.5%) with good speci city which was comparable to gram staining (sensitivity: 60% and speci city: 100%). However, the sensitivity of the UF5000 for identifying mixed growth bacteria was poor.
The UF-5000 is an automated urine analyzer produced by Sysmex Corporation, which performs ow cytometry analysis with a higher level of accuracy and more precise data [13]. Several previous studies investigated the legacy models of automated urine particle analyzers (including the UF-500i and the UF-1000i) for screening urine cultures, while few studies have evaluated the ability of the newer UF5000 in differentiation of bacteria growth patterns. Compared with the legacy systems, the current study showed that the UF5000 kept the comparable sensitivity and speci city for GNB (80.0% and 88.2%, respectively). For speci c bacteria (Table 2), the current results revealed that the UF5000 will identify Klebsiella spp. Proteus spp., and Citrobacter spp. while only identifying 37 out of 49 E coli. However, the retrospective study by Kim et al [11] showed good performance of the UF5000 in identifying E coli. Further studies are required to check the performance of the UF5000 in identifying E coli. As for Gram-positive bacteria, the UF 5000 showed high sensitivity and speci city for Enterococcus spp. However, the sensitivity for Streptococci spp. was much lower. In our study, UF 5000 identi ed all staphylococci spp.(n = 3), enterococci spp. (n = 1), streptococci spp.(n = 3), except one Group B Streptococci (2*104 cfu/mL) and two gram (+) cocci (103 and 2 × 103 cfu/mL, respectively). With regard to gram positive bacteria, the UF5000 outperformed the UF1000i in detecting gram positive cocci which was comparable to gram staining.
Gram staining is associated with a sensitivity rate of 88%, a speci city rate of 95%, a negative predictive value of 96%, and a positive predictive value of 84% for identifying bacteriuria. [6,14] For differentiating bacterial growth patterns, gram stain yielded good sensitivity and speci city for gram negative (80.6% and 96.7%, respectively) and gram positive bacteria (60% and 100%, respectively). Although a real-time reporting of gram stain could reduce the blind initiation of antibiotics, and thus prevent unnecessary expenditure and drug treatment, gram staining is time consuming and labor-intensive. The UF-5000 offers comparable e cacy and a faster and far easier way to provide the same information as compared to the classic method of Gram staining.
Detecting Gram positive bacteria has signi cant clinical implications. The most commonly isolated Gram-positive uropathogens are Staphylococcus saprophyticus, Enterococcus faecalis, and Streptococcus agalactiae. One review published in the literature suggests that urologic diseases involving Gram-positive bacteria may be easily overlooked due to limited culture-based assays typically utilized for urine in hospital microbiology laboratories [5]. However, Hooton et al [12] found that only Staphylococcus saprophyticus correlated well with the catheterized urine. However, enterococci and streptococci correlated with catheterized urine culture poorly. Therefore, patients with gram positive bacteria shown on the UF5000 may not have uUTI by classical GNB and may not need empirical antibiotics. Second, in patients with Gram positive cocci determined by UF5000, urine culture is recommended and antibiotics targeting Gram positive bacteria, instead of empirical ones for gram negative bacteria. In this way, we may avoid unnecessary waiting times, overuse of antibiotics and increased medical costs. Third, immediately identifying Gram positive bacteria in ascites [13], cerebral spinal uids [14] and pleural uids may help clinicians make appropriate and timely antibiotic choice in these life threatening infections.
More clinical studies to explore the use of the UF-5000 in the situations is encouraged.
About 21.6% of urine cultures revealed mixed growth (n = 22, 21.6%) and lactobacillus (n = 2, 1.9%) which were regarded as contamination due to improper collection, transportation, preservation, and storage. As the study included female participants only, and this may explain the relatively higher contamination rate. Because the short urethral length and the urethra meatus is proximal to the vagina and anus, urine specimen from women were more easily contaminated than men. Our study shows that the sensitivity and speci city of the UF5000 for mixed growth was 4.5% and 94.9%, respectively. Further improvement of laser owcytometry in identi cation of mixed growth could help health care and avoid the waste of time, labor and money.
There existed some limitations in our study. Although the study is a prospective study that compared the e cacy of the three methods (gram stain, the UF1000i and the UF 5000i), the major limitation is that the number of samples was limited. Second, a signi cant proportion of patient specimens yielded gram negative bacteria and mixed growth culture, so the evidence supporting the promising e cacy of the UF5000 in detecting gram positive bacteria is limited due to the low number of specimens with gram positive bacteria. The strength of the study is its prospective nature of the study that compares two automated urine particle analyzers. Further studies are still warranted to evaluate the generalizability of the UF5000 in a larger subset of patients, institutions and populations. Also, the role of the UF5000 in detecting bacterial patterns in the other type of specimens, i.e. ascites, cerebral spinal uids and pleural effusion, should be investigated further.

Conclusions
The UF-5000 demonstrated potential utility for the rapid screening of bacterial morphology, which correlated well with the legacy analyzer UF1000i for GNB bacteria while showing improved sensitivity for detecting Gram-positive cocci.
Abbreviations GNB Gram negative bacilli UTISA urinary tract infection symptoms assessment score uUTI uncomplicated urinary tract infection