Detection of Significant Bacteriuria by Use of the iQ200 Automated Urine Microscope

Abstract

In the microbiology laboratory, there is an augmented need for rapid screening methods for the detection of bacteria in urine samples, since about two-thirds of these samples will not yield any bacteria or will yield insignificant growth when cultured. Thus, a reliable screening method can free up laboratory resources and can speed up the reporting of a negative urine result. In this study, we have evaluated the detection of leukocytes, bacteria, and a new sediment indicator, the “all small particles” (ASP), by an automated instrument, the iQ200 urine analyzer, to detect negative urine samples that can be excluded from culture. A coupled automated strip reader (iChem Velocity), enabling the detection of nitrite and leukocyte esterase, was tested in parallel. In total, 963 urine samples were processed through both conventional urine culture and the iQ200/iChem Velocity workstation. Using the data, a multivariate regression model was established, and the predicted specificity and the possible reduction in urine cultures were calculated for the indicators and their respective combinations (leukocytes plus bacteria plus ASP and leukocyte esterase plus nitrite). Among all options, diagnostic performance was best using the whole microscopic content of the sample (leukocytes plus bacteria plus ASP). By using a cutoff value of ≥10⁴ CFU/ml for defining a positive culture, a given sensitivity of 95% resulted in a specificity of 61% and a reduction in urine cultures of 35%. By considering the indicators alone, specificity and the culture savings were both much less satisfactory. The regression model was also used to determine possible cutoff values for running the instrument as part of daily routine. By using a graphical representation of all combinations possible, we derived cutoff values for leukocyte, bacterial, and ASP count, which should enable the iQ200 microscope to screen out approximately one-third of the urine samples, significantly reducing the workload in the microbiology laboratory.

INTRODUCTION

Although traditional culture remains the “gold standard” for the diagnostic assessment of patients suspected to have a urinary tract infection (UTI), this methodology is costly, laborious, and time-consuming, and frequently, large proportions of the urine samples sent to the laboratory turn out to be negative. Thus, stepwise strategies in urine analysis have been developed to detect the presence of infection as quickly and reliably as possible, avoiding unnecessary culture testing, saving patient and laboratory expenses and maintaining efficiency within the microbiology department (1–3). In recent years, the speed of making the diagnosis has gained special economic attention too, since most European hospitals are subjected to reimbursement systems in which the duration of the patient's hospital stay is strictly limited. Traditional screening methods, such as dipstick testing for nitrite and leukocyte esterase as well as microscopic sediment analysis for bacteria and white blood cells, are fast but lack sensitivity (4). Moreover, manually performed methods are laborious and vulnerable to observer variation and imprecision. Thus, in order to obtain a more accurate analysis, there have been intensive attempts at exploring automated techniques for more-efficient UTI screening. The automated devices offer a high capacity of particle enumeration and can realize a great degree of labor and time savings compared to manually performed urine sediment methods (5). Several instruments were put on the market with the aim freeing up resources by rejecting negative samples quickly and reliably (2, 3). Nevertheless, results from earlier studies, which were done mostly with the flow cytometers of the Sysmex UF series, thus far have been mixed. While some authors have reported a fairly sufficient performance compared to urine culture (5–10), others completely denied the feasibility of the automated devices as a screening tool, mostly due to an unacceptably large number of false-negative results (11, 12). In addition, there is still an ongoing debate of what are the best cutoff values to discriminate samples in the positive and negative groups (10, 11, 13, 14).

Recently, a new automated instrument, the iQ200′ workstation (Iris Diagnostics), has been introduced. The main difference between this instrument and the flow cytometers is that the urine content is analyzed by assessment of digital images of the particles passing in the front of a microscope objective. The microscopic approach results in better performance in identifying the urine content, as there are more indicators on which the analysis can be based on (in the cytometers, there are only two measurement channels, one for bacteria and one for leukocyte detection). In this study, using 963 clinical samples which were sent to the LADR laboratory in Geesthacht, Germany, we compared the detection of bacteria, leukocytes, and other urine constituents (“all small particles” [ASP]) of the iQ200 system (and a coupled strip reader, iChem Velocity) to the gold standard, urine culture. The results were analyzed using different combinations of the indicators, which made it possible to find the best combination as well as to derive a suitable set of cutoff values.

MATERIALS AND METHODS

During our study, 963 samples of urine specimens were obtained. The urine specimens were routinely submitted to the microbiology division at the LADR laboratory, Geesthacht, Germany, for bacteriological culture (from March 2011 to May 2011). A total of 788 samples were collected from patients with midstream technique (clean-catch urine samples), and 175 samples were derived from indwelling urinary catheters. Urine was poured into sterile tubes not containing a preservative (Sarstedt Urine-Monovette; 10 ml) and sent to the microbiology laboratory within 3 to 6 h after collection, at maximum. Transport was performed in a cold shipping box maintaining a constant 2 to 8°C environment. After arrival in the laboratory, each sample was divided into two aliquots. Automated urine analysis was performed on the first aliquot using the iChem Velocity instrument (for analyzing the strip chemistries) and the iQ200 digital imaging system (for particle analysis). The indicators measured by iChem Velocity were nitrite and leukocyte esterase, and the indicators measured by the iQ200 module assessed were white leukocytes, bacteria, and the “all small particles” (ASP). The ASP application counts all the background particles of less than 3 μm and mainly reflects the presence of “small bacteria,” e.g., Gram-positive cocci (15). Urine culture was performed on the second aliquot by plating 10 μl of the sample on Columbia blood agar containing colistin-nalidixic acid and on chromogenic agar, the latter for growth assessment of gram-negative bacteria by color changes (Urin 3G-Agar II biplates, heipha Dr. Müller GmbH, Eppelheim, Germany). The plates were incubated under aerobic conditions at 36°C for 16 to 18 h. For the purpose of this study, we considered a culture positive if it contained one or two potential uropathogens at a concentration of ≥10⁴ CFU/ml as suggested by several authors (7, 15, 16). Potential uropathogens were defined as members of the family Enterobacteriaceae, Enterococcus species, Streptococcus agalactiae, Pseudomonas species, Candida species, Staphylococcus aureus, and non-S. aureus species. Specimens that yielded growth of ≥3 isolates (with no predominating organism) or samples that grew nonpathogens (e.g., Lactobacillus spp., Corynebacterium spp., or Neisseria spp.) were considered contaminated with commensal flora.

Statistical analysis.

An Excel spreadsheet was populated with data, including patient details, presence and quantitation of the chemistries (nitrite and leukocyte esterase) on the strip reader, presence and quantitation of the microscopic indicators (leukocytes, bacteria, and “all small particles”), urine culture results, and the types of microorganisms isolated, including yeasts. Bivariate association between the number of leukocytes, nitrite, leukocyte esterase, and ASP, bacteria, and colony counts were evaluated by estimating Pearson's correlation coefficients with 95% confidence limits and P values. Statistical analysis of the data was done using the R package (17). Using the gold standard definition, sensitivity, specificity, and the number of urine samples that can be excluded from culture were calculated for the microscopic indicators alone and their respective combinations. Assessment of the combinations was done using a multiple logistic regression model, followed by an analysis of the receiver operating characteristic (ROC) curves of the predicted probabilities from the combinations (17–19).

RESULTS

Sample characteristics and culture results.

Table 1 shows the culture results and the sample characteristics. In total, 515 (53.5%) samples were negative by culture (no growth or a bacterial count of <10⁴ CFU/ml), and 448 (46.6%) were positive, with bacterial counts of ≥10⁴ CFU/ml. A total of 465 (48.3%) specimens were from hospitalized patients (with a positivity rate of 41.8%), and 498 (51.7%) samples were from outpatients (positivity rate of 51.4%). As judged from the species pattern, the most common microorganisms identified were Escherichia coli (n = 304), Enterococcus spp. (n = 177), Gram-positive bacilli not specified due to small amount of growth (n = 63), Klebsiella spp. (n = 40), Gram-negative rods not specified due to small amount of growth (n = 33), Proteus spp. (n = 31), coagulase-negative staphylococci (n = 30), Pseudomonas spp. (n = 28), Candida spp. (n = 27), group B streptococci (n = 19), Staphylococcus aureus (n = 14), viridans streptococci (n = 5), Citrobacter spp. (n = 4), and Serratia marcescens (n = 3).

TABLE 1.

Culture results and sample characteristics

Culture result or sample characteristic	No. of samples (n = 963) [%]
Growth behavior of specimens
No growth, negative culture	315 [32.7]
≥102 CFU/ml–nonsignificant growth, negative culture	133 [13.8]
≥103 CFU/ml–nonsignificant growth, negative culture	67 [6.9]
≥104 CFU/ml–significant growth, positive culture	25 [2.6]
≥105 CFU/ml–significant growth, positive culture	71 [7.4]
≥106 CFU/ml–significant growth, positive culture	352 [36.6]
Patients
Male	307 [31.9]
Female	656 [68.1]
Hospitalized patients {% positive cultures}	465 [48.3] {41.8}
Outpatients {% positive cultures}	498 [51.7] {51.4}
Species pattern for all urine samples
Escherichia coli	304 [31.6]
Enterococcus spp.	177 [18.4]
Gram-positive bacilli, not specified	63 [6.5]
Klebsiella spp.	40 [4.1]
Gram-negative rods, not specified	33 [3.4]
Proteus spp.	31 [3.2]
Coagulase-negative staphylococci	30 [3.1]
Pseudomonas spp.	28 [2.9]
Candida spp.	27 [2.8]
Group B streptococci	19 [2.0]
Staphylococcus aureus	14 [1.5]
Viridans streptococci	5 [0.5]
Citrobacter spp.	4 [0.4]
Serratia marcescens	3 [0.3]

Performance of single indicators.

To determine the potential of the single indicators, we first assessed the predictive abilities of the indicators if used alone. As presented in the box plots and stacked bar charts (Fig. 1A to E), there was a significant direct relationship between the single indicators measured on the iQ200 system (leukocytes, bacteria, and “all small particles”) as well as on the iChem Velocity reader (leukocyte esterase and nitrite) and the results of quantitative urine culture. As expected, for each parameter, when taken individually, the values measured by the iQ200 instrument and the iChem Velocity were significantly higher in the groups of positive urine samples (containing ≥10⁴ CFU/ml) than in the groups of negative urine samples (containing <10⁴ CFU/ml) (Fig. 1A to E). When transformed and expressed in terms of specificity, however, the single indicators did not all exhibit a useful result except in the case of “all small particles”. As presented in Table 2, the “ASP” count is the only individual indicator that developed acceptable specificity, yielding a high number of samples which do not need further culture. For instance, at 95% sensitivity, the “ASP” channel exhibited a specificity of 44.2% and 25.8% of samples not needing further culture (Table 2).

FIG 1.

(A to C) Stacked bar charts of leukocyte esterase versus colony growth (A), nitrite versus colony growth (B) and bacteria versus colony growth (C). WBC, white blood cells; neg, negative; CI, 95% confidence interval. (D and E) Box plots of leukocytes versus colony growth (D) and “all small particles” (ASP) versus colony growth (E). The top and the bottom of each box represent the 75% and 25% quartiles, respectively, and the band inside the box represents the median. The lower and upper ends of the whiskers represent the lowest datum still within 1.5 interquartile ranges of the lower quartile and the highest datum still within 1.5 interquartile ranges of the upper quartile, respectively. cor. correlation.

TABLE 2.

Performance of (single) indicators and indicator combinations^a

Method(s) and indicator(s)b	Parameterc	Value for parameter at the following sensitivity (%):
		90	91	92	93	94	95	96	97	98	99
Urine chemistry (iChem Velocity)
Leukocyte esterase	% SPE	51.2	47.4	42.1	36.9	31.6	26.3	21.1	15.8	10.5	5.3
	% NNC	31.9	29.4	26.1	22.9	19.6	16.3	13.1	9.8	6.5	3.3
Nitrite	% SPE	19.1	17.2	15.3	13.4	11.5	9.5	7.6	5.7	3.8	1.9
	% NNC	14.8	13.3	11.9	10.4	8.9	7.4	5.9	4.4	3.0	1.5
Nitrite + LE	% SPE	61.4	57.5	53.5	49.5	42.9	35.7	28.6	21.4	14.3	7.1
	% NNC	37.3	34.8	32.2	29.6	25.6	21.3	17.1	12.8	8.5	4.3
Urine microscopy (iQ200)
All small particles	% SPE	70.0	68.1	64.7	57.5	49.8	44.2	42.5	36.6	24.1	14.4
	% NNC	41.9	40.4	38.1	33.9	29.3	25.8	24.4	20.9	13.8	8.1
Bacteria	% SPE	37.0	33.3	29.6	25.9	22.2	18.5	14.8	11.1	7.4	3.7
	% NNC	24.3	21.9	19.5	17.0	14.6	12.2	9.7	7.3	4.9	2.4
Leukocytes	% SPE	37.8	34.0	29.1	27.0	24.8	18.1	15.3	10.8	6.0	3.0
	% NNC	24.8	22.3	19.2	17.6	16.0	12.0	10.0	7.1	4.1	2.0
ASP + bacteria + leukocytes	% SPE	73.5	71.8	69.0	66.6	64.0	61.0	50.9	40.9	35.8	21.1
	% NNC	43.7	42.4	40.4	38.7	36.8	34.8	28.9	23.2	20.0	11.7
Urine chemistry + urine microscopy, iQ200 + iChem Velocity
	% SPE	76.1	73.7	71.6	70.5	66.4	60.6	54.7	47.4	39.4	21.8
	% NNC	45.1	43.4	41.8	40.7	38.1	34.5	31.0	26.6	21.9	12.0

^aThe specificities for indicators and combinations of indicators and percentages of samples not needing culture, predicted to a given sensitivity.

^bLE, leukocyte esterase; ASP, all small particles.

^c% SPE, percent specificity; % NNC, percentage of sample that do not need further culture.

Performance of the indicator combinations.

As with the Iris system, multiple indicators are available, and it is very interesting to evaluate whether a combination of the indicators gets better results for making a diagnosis. Thus, we investigated whether an indicator combination (combination 1, leukocyte esterase plus nitrite; combination 2, ASP plus bacteria plus leukocytes; or combination 3, leukocyte esterase plus nitrite plus ASP plus bacteria plus leukocytes) might improve the capabilities of identifying samples that can be excluded from urine culture. In order to assess the indicator combinations, a multiple logistic regression model was established. Using the predicted probabilities which were received from the logistic function, the data were translated into ROC curves, and the key statistical figures were derived. As presented in Table 2, the estimated specificities and the percentages of cultures saved when the indicator combinations would have been used as screening tests were calculated as a function of sensitivity (in the range of 91 to 99%, which is typical for screening purposes). As one can see from the table, the results with the parameter combinations were convincing. For both techniques (automated microscopy and strip testing), there was a clear benefit in going from the single indicators to the combinations. However, when comparing the combinations, our calculations showed that the microscopy indicators (ASP plus bacteria plus leukocytes) allow the prediction of positivity in culture more specifically and with greater chance to reduce the number of unnecessary cultures than using strip testing (leukocyte esterase plus nitrite). At sensitivity levels of 95% (99%), we obtained 61% (21.1%) specificity and 34.8% (11.7%) of the samples that do not require further culture in the case of iQ200 microscopy and reached only 35.7% (7.1%) specificity and 21.3% (4.1%) of the samples not requiring further culture with iChem Velocity strip chemistries. Thus, with regard to reducing the number of samples that are urine culture candidates, iQ200 microscopy was clearly superior to iChem Velocity strip chemistries. Interestingly, the results of the whole workstation, iQ200 plus iChem Velocity, (leukocyte esterase plus nitrite plus ASP plus bacteria plus leukocytes) were almost identical to that obtained for the better technique alone (microscopy), and thus, in this data set, it does not appear that using chemical measurement in addition to microscopy adds substantially to the systems capability to diagnose urinary tract infection.

Finding optimal cutoff values for the sediment indicators when used in combination.

In order to derive cutoffs that give well-balanced results in terms of sensitivity/reject rate, we used the data from the logistic regression model of the iQ200 microscopy and transformed them in a graphical representation. In order to reduce the complexity, the impact of “bacteria” has been displayed as scenario a or bacteria present (see Fig. S1a in the supplemental material) and scenario b or bacteria absent (Fig. S1b). This procedure generated two grid patterns of value pairs that were defined by the ASP values (on the y axis) and the leukocyte counts (on the x axis). In the figures, each point on the grid represents a double value consisting of (i) the predicted sensitivity (numerator) and (ii) the predicted percentage of samples that do not require further culture (denominator). Finally, using the grid patterns, we identified a suitable set of cutoff value combinations which enable both a sensitivity of greater than 95% and a percentage of samples not needing culture of at least 30.4 to 35.9% (Table 3; Fig. S1a and Fig. S1b).

TABLE 3.

Optimized cutoff values for the iQ200 automated urine microscope

Scenario a (in the presence of bacteria)		Scenario b (in the absence of bacteria)
All small particles (no. of pcls/μl)a	Leukocytes (no. of cells/μl)	All small particles (no. of pcls/μl)	Leukocytes (no. of cells/μl)
>2,000	>15	>8,000	>10
>1,700	>20	>6,500	>30
>1,400	>50	>5,000	>75
>1,200	>100	>4,000	>150
>1,000	>150	>3,000	>600
		>2,500	>1,000
		>2,000	>2,500

^apcls, particles.

DISCUSSION

The main objective of using an automated device for the screening of urine samples is to reduce the number of specimens cultured. This in turn allows financial and labor savings and helps to speed up the reporting of a negative urine sample. The iQ200 instrument is a new automated urine analyzer that uses a combination of three indicators (leukocytes, bacteria, and “all small particles”) to diagnose urinary tract infection. As there is an additional indicator, “all small particles”, on which the diagnosis can be based on, the advanced technique holds promise of better results for reducing the workload within the laboratory than the previous generation of automated urine microscopes. However, since there is still uncertainty of what are the best instrument settings when the parameters are used together (15), we were particularly interested in which indicator combinations would give the best diagnostic performances and how much gain could be realized by adjustment. In the beginning, we looked at the manufacturer's recommended settings (i.e., ≥5 bacteria/μl, ≥25 leukocytes/μl, and/or ≥3,000 ASP/μl) (Alice Airaud, personal communication), which showed excellent sensitivity (98.9%) but did not yield an acceptable culture savings rate (only 21.5%). Thus, we opted to assess various combinations of leukocyte, bacterial, and ASP counts to achieve a better performance. The statistical treatment of the data comprised a two-step procedure, first establishing a multivariate regression model and then running a threshold optimization procedure. This approach enabled us to find those threshold combinations for the indicators that yielded both a sensitivity of at least 95% and a urine culture savings rate of at least 30.4 to 35.9%. Remarkably, this savings rate would not have been possible without the ASP count. This new indicator mainly reflects the presence of Gram-positive bacteria that cannot be tracked well in the “bacteria” channel. While other studies see no extra value for “all small particles” and thus consider this piece of information dispensable (15), we can clearly claim that this indicator contributes a significant amount to the detection performance of the automated urine microscope. As judged from the performance data of the individual parameters, the ASP count is even the best single indicator among the sediment constituents (Table 2).

As a main finding of our calculations, it became clear that the suitability of an automated urine device is subject to circumstances that are beyond the analytical process itself. A key influence is exerted by the threshold value applied to define a positive urine culture. As has already been pointed out, approximately 30.4 to 35.9% of the urine samples can be excluded from being cultured with the iQ200 microscope. A reduction of one-third is similar to the saving rates that other working groups have recently published. However, this work was done almost exclusively with the flow cytometers of the Sysmex UF series (6, 7, 9, 10, 15, 16), and in many of these studies, which were mostly conducted in hospital surroundings (9, 12, 15, 16, 20), the prevalence of UTIs was in the range of 13 to 30% (8–11, 13, 15, 16). Our trial gives strong evidence that automated urine microscopy not only works in the setting of a hospital laboratory but is also a valuable tool in the nonhospital laboratory, delivering fast screening results to general practitioners. Furthermore, in our study, the prevalence of positive urine samples, by using a cutoff value of ≥10⁴ CFU/ml, was an above average 46.6%. The possibility for culture savings was thus lower. Given this obstacle, the exclusion rate predicted from our calculations (30.4 to 35.9%) seems fairly good, as this corresponds to approximately two-thirds (57 to 67%) of all the negative samples (53.5%) that can be excluded from culture.

Facing the ongoing discussion of what is suitable in diagnosing urinary tract infection, it is very interesting to see what happens when one changes the thresholds for defining a positive urine culture. In a recent publication with very low thresholds, the saving effect was greatly reduced to 20%, of which 14% were false-negative results, compared to 52% (28%) by using a cutoff value of ≥10⁵ (≥10⁴) CFU/ml (12). If we had changed the settings in a similar manner (like the authors of reference 12 did), we also would observe deterioration of the culture savings to 15% for a decrease in the threshold to ≥10³ CFU/ml (Fig. 2). Vice versa, we would observe an improvement to 40% culture savings rate for a threshold of ≥10⁵ CFU/ml (Fig. 2). However, 10⁵ CFU/ml is much too high for certain “high-risk” populations (e.g., pregnant women, children, patients with urological disorders, request for repeat culture), in which 10³ CFU/ml or even lower colony counts could represent significant bacterial infection. Thus, automated testing that is tested against higher microbial counts (i.e., ≥10⁴ CFU/ml or ≥10⁵ CFU/ml) must not be applied at all.

FIG 2.

Comparison of sensitivity (x axis), specificity (left-hand y axis) and reject rate of urine samples (right-hand y axis) at different gold standard definitions, using model 7 (iQ200 automated microscopy).

For the application in a nonhospital laboratory, however, we think that ≥10⁴ CFU/ml is adequate. In Germany at least, most doctors in ambulatory medicine sent 70 to 80% of their samples from patients who do not have a fully expressed UTI risk profile, but rather the bladder or the kidneys are only two of several other sources of infection which should be excluded. If specimens that are not amenable to automated testing are strictly excluded, automated testing can reliably be done on the ≥10⁴ CFU/ml level. By using the optimized cutoffs derived from the present study, utilization of the iQ200 microscope would decrease the number of bacterial cultures performed by 30.4 to 35.9%. Extrapolating these results to the approximately 28,000 requests for urine culture/year that our laboratory processes translates into 7,500 to 8,750 negative urine cultures that would be avoided. Necessarily, this success is shadowed by the fact that one has to expect a small number of false-negative results per year. We think that this is acceptable if the knowledge about the limitation is carefully communicated to the physicians so that they can react accordingly if needed (e.g., request a repeat culture). Rather, if most general practitioners act upon the results of automated urine testing, which are available almost immediately after the sample has reached the laboratory, the advantages of the iQ200 analyzer are highlighted, as not only can it reduce the number of negative cultures but it can also help to avoid the use of antibiotics.

Finally, we feel confident that the statistical approach in this study holds promise of better results than using the default settings. As many studies in the past worked with the cutoff recommendations of the manufacturer, there could have been some underachieving in terms of the culture savings rate that can probably be improved by a threshold optimization procedure.