Correspondence: Anthony S. Kupelian, Research Department of Clinical Physiology, Department of Medicine, University College London, Clerkenwell Building, Archway Campus, 2–10 Highgate Hill, London N19 5LW, UK.
What's known on the subject? and What does the study add?
Microscopic pyuria is widely used as a surrogate marker of infection, although there is little data supporting its use in patients who present with non-acute LUTS. The effects of urinary storage, preservation, and the use of laboratory methods to enhance leucocyte detection, are also unclear.
This large, prospective study highlights the poor performance of dipstick urine analysis, and direct microscopy, as surrogate markers of UTI in patients with LUTS. A series of laboratory analyses also examine the effects of urine handling and processing on test integrity, which have important implications for clinical practice.
To evaluate the diagnostic performance of pyuria as a surrogate marker of urinary tract infection (UTI) in patients with chronic lower urinary tract symptoms (LUTS), and determine the impact of sample storage, cytocentrifugation, and staining techniques, on test performance.
Patients and Methods
Between 2008 and 2011, we recruited 1223 patients (120 men; 1103 women; mean age 54 years) with one or more LUTS from a specialist urological outpatient service. We conducted a prospective observational study to determine the performance of microscopic pyuria ≥10 wbc/μL as a surrogate marker of UTI in patients with LUTS.
All patients provided clean-catch midstream urine (MSU) samples for analysis, and routine microbiological cultures were used as our reference standard. We also scrutinised the performance of dipstick leucocyte esterase ≥ ‘trace’ in the detection of microscopic pyuria.
The influence of sample handling and processing on test performance was examined in a series of laboratory studies.
The effects of storage on leucocyte decay were determined using repeated microscopic assessments of individual urine samples, to plot temporal changes in leucocyte numbers. This study used varied storage conditions (≈20 °C and 4 °C), and boric acid preservation.
Paired microscopic assessments were used to determine the effects of centrifugation on leucocyte salvage in spun/unspun samples (relative centrifugal force range 39–157 g). Similar methods were used to assess microscopic leucocyte quantification in stained/unstained urine (Sternheimer-Malbin protocol).
The positive predictive value (PPV) and negative predictive value (NPV) of pyuria as a surrogate marker of UTI were 0.40 (95% confidence interval [CI] 0.37–0.43) and 0.75 (95% CI 0.73–0.76), respectively.
The dipstick was unable to identify significant microscopic pyuria (≥10 wbc/μL) in 60% of the samples: PPV 0.51 (95% CI 0.48–0.55); NPV 0.75 (95% CI 0.73–0.76). Microscopic pyuria performed poorly as a surrogate of UTI defined by bacterial culture.
Whilst refrigeration and preservation did retard leucocyte loss (F = 11; DF = 2; P < 0.001), 40% of cells were still lost by 4 h.
Centrifugation had an unpredictable influence on cell salvage (coefficient of variation 5750%) and the use of staining to improve leucocyte detection proved ineffective (Z = –0.356; P = 0.72).
Pyuria performs badly as a surrogate of UTI in patients with LUTS. This is exacerbated by cell loss during storage, and neither centrifugation, nor staining, appears to confer any diagnostic advantage.
Clinicians should be alerted to the significant limitations of these tests.
LUTS are a global problem. Large population-based studies suggest that 2 billion individuals are affected by one or more LUTS, a figure expected to rise . The exclusion of UTI is cited universally as a critical step in the assessment of LUTS [2, 3], although the precise role of infection in generating these symptoms is unclear. Nonetheless, the exclusion of infection in all symptomatic patients is widely advocated.
Whilst bacteriological culture remains the accepted reference standard for the detection of UTI, the use of pyuria as a surrogate marker has replaced bacterial culture in many clinical services. Whether evaluated by microscopy, urinary dipstick, or automated methods, it is used as a stand-alone surrogate, or used to triage samples submitted for bacteriological culture. In the latter case, the absence of ‘significant’ pyuria is considered definitive evidence of the absence of UTI. Whilst the widespread use of pyuria has imparted a sense of confidence in its performance, little data exist relating to its use in LUTS assessment. Questions also remain about the influence of specimen handling, and strategies used to optimise diagnostic accuracy.
The identification of urinary leucocytes using light microscopy was first described in 1893. Early pioneers studied the centrifuged deposits of large volumes of urine , although doubts were raised about the veracity of this approach by some [5, 6]. In 1928, Dukes  described a new method of assessment, using a cell counting chamber to evaluate fresh, uncentrifuged urine. His study of 300 midstream urine (MSU) samples from asymptomatic controls produced estimates for normal mean leucocyte counts of 1.6 wbc/μL and 5.4 wbc/μL for males and females, respectively. Despite significant dispersion around these mean estimates (range 0–50 wbc/μL), he proposed a threshold of <10 wbc/μL as the upper limit of normal pyuria excretion. His experiments were not replicated until the 1950s, when several groups reported similar results [6, 8, 9], thereby binding the <10 wbc/μL threshold into clinical practice.
Although not the first to explore the utility of quantitative bacterial counts in the diagnosis of UTI [10, 11], Kass (1957)  was credited with the development of a definitive diagnostic criterion. Even today, his culture threshold of ≥105 colony-forming units (cfu)/mL exerts a dominating influence on the diagnosis of UTI. It is also the key reference standard in work evaluating surrogate markers of UTI, including pyuria.
Kass'  original study compared 74 women with pyelonephritis, and 337 asymptomatic control subjects, in an attempt to define a quantitative bacterial threshold indicative of infection. Whilst Kass' contribution was undeniably significant, his work did have limitations. The groups included in the analysis were not representative of the wider population of symptomatic patients. Kass also assumed dominant pathogenicity from coliform organisms, considering mixed bacterial growth as likely contamination. However, the medical community is becoming aware of the pathogenic significance of polymicrobial infection in human disease of the bladder .
The prevalence of infection varies considerably amongst different symptomatic populations [12, 14], exerting a significant influence on diagnostic test performance. Stamm et al. (1982)  produced compelling evidence of the inherent problems associated with a single, dichotomous diagnostic threshold for UTI across all LUT syndromes, although his work is largely overlooked. Further challenges have since been fielded, but they have gone unheard [15, 16].
The urinary dipstick has been promoted to limit reliance on direct microscopy and microbiological culture. Whilst its reliability in excluding UTI has been convincingly disputed [16-19], such criticism has done little to dent enthusiasm for its use. Its poor performance in most clinical settings may be in part a consequence of inappropriate microscopic and microbiological reference standards used during development and calibration.
Whilst the acute onset of classic cystitis symptoms permits accurate diagnosis based on history alone , chronic LUTS present greater difficulty. Here, the diagnosis of UTI hinges on corroborative testing, but there is very little evidence to support the utility of pyuria in this setting. Much of the work evaluating pyuria in the diagnosis of UTI shows significant heterogeneity, with poor reporting of the patient population studied, and a lack of clarity relating to presenting symptoms. These factors are critical when interpreting test performance.
Despite the lengthy transit of specimens from clinic to laboratory, the literature is silent on the effects of storage and preservatives on leucocyte survival. Data are similarly sparse on the effects of centrifugation and staining as methods for enhancing cell detection.
We determined the diagnostic performance of microscopic pyuria, and dipstick leucocyte esterase, in a large, prospective study of adult patients presenting with LUTS. We wanted the study to be pragmatic, and we used MSU sampling, and analysis methods common to ordinary clinical practice, to achieve this. We approached the effects of urine handling and processing by devising a series of laboratory studies to examine each of these variables.
Patients and Methods
Recruitment and data collection were undertaken in a single, specialist incontinence service in the UK, between October 2008 and November 2011. Adult patients presenting with one or more LUTS were invited to participate. Patients provided consent before study inclusion and ethical approval was granted by the Whittington and Moorfields Research Ethics Committee. Patient demographics and symptoms were recorded in a bespoke clinical database. This assessment included storage, voiding, postmicturition, and pain symptoms (Table 1). Patients with acute frequency-dysuria, symptoms suggestive of pyelonephritis, symptoms of <3 months duration, and those who were pregnant, were excluded. Clinicians undertaking data entry were ‘blinded’ to the results of urine analysis. Control data were collected from asymptomatic volunteers who met the inclusion criteria described below (Table 2).
Table 1. Lower urinary tract symptom prevalence matrix for patients.
*Urinary frequency defined as ≥8 episodes/24 h; †Nocturia defined as ≥2 episodes.
Table 2. LUTS profile for asymptomatic control subjects.
No urinary urgency
No perception of increased urinary frequency
No urinary incontinence
No voiding symptoms
No postmicturition symptoms
No pain attributed to the urinary tract
Urine Collection and Analysis
We obtained MSU samples using the midstream clean-catch method . Patients were carefully instructed in the method and provided with written guidance. Trained researchers determined microscopic leucocyte counts using a Neubauer haemocytometer counting chamber  and an Olympus CX41 light microscope (x 200; Olympus, Southend-on-Sea, UK); automated dipstick analysis was conducted using Multistix® 8 SG reagent strips paired with a Clinitek® Status analyser (Siemens, Munich, Germany). Routine laboratory culture was undertaken in a local accredited NHS laboratory by biomedical scientists using standard methods . A positive result was indicated by the growth of ≥105 cfu/mL of a single recognised urinary pathogen after 24 h culture. All samples were processed immediately after collection and the results recorded in a secure database.
Leucocyte Storage and Preservation
We examined leucocyte survival in urine under different environmental conditions, and when stored with a preservative agent. We selected boric acid after a series of pilot experiments, which tested ethanol, acetic acid, and hydrochloric acid. Fresh MSU samples containing ≥10 wbc/μL were divided into 2 mL aliquots, and stored as follows: (i) storage at room temperature (≈20 °C); (ii) refrigeration (4 °C); (iii) storage at room temperature with 2% boric acid; and (iv) refrigeration with 2% boric acid. A microscopic leucocyte count was performed at 2, 4, 6, 24, and 48 h after collection.
Centrifugation and Leucocyte Salvage
We determined the effects of centrifugation on leucocyte salvage by subjecting MSU samples to variable centrifugation forces. Each MSU was subject to a microscopic leucocyte count before centrifugation. We then transferred the urine into four sample tubes, each containing 3 mL urine. The tubes were centrifuged for 5 min in a Denley BR401 centrifuge (RMAX 140 mm) using four different protocols (39, 77, 100, and 157g). The samples were processed in randomly generated sequences (Research Randomizer, http://www.randomizer.org).
After centrifugation, we counted the leucocytes in the supernatant to determine the efficacy of centrifugation in concentrating leucocytes from whole urine into the sediment pellet. We then re-suspended the sediment pellet thoroughly within the supernatant, and repeated the leucocyte count to assess the effects of centrifugation on cell integrity, and consequent total leucocyte salvage.
Staining to Optimise Diagnostic Accuracy
We used a KOVA™ stain (a modified Sternheimer-Malbin, Gram-based stain) to examine its effects on the sensitivity of microscopic leucocyte enumeration . We divided MSU samples into two aliquots and determined the microscopic leucocyte counts in stained and unstained urine in a random sequence. The stain was applied by incubating 1 mL urine with 150 μL stain in a sample bottle for 8 min at room temperature.
Microscopic Pyuria as a Surrogate Marker
The design, conduct, and reporting of this analysis were conducted in accordance with the Standards for Reporting of Diagnostic Accuracy (STARD) initiative . The performance of microscopic pyuria as a surrogate marker of UTI was determined by a comparison of microscopic leucocyte enumeration, and routine laboratory urine culture. Positive microscopy was defined as the presence of ≥10 wbc/μL on microscopic assessment, whilst the culture reference standard was the growth of ≥105 cfu/mL of a single recognised uropathogen. We also collected data on samples that manifest low-level microscopic pyuria 1–9 wbc/μL, and cultures reported as ‘mixed growth of doubtful significance’ (attributed to contamination, poor sampling, or failure to refrigerate the sample).
We also determined the performance of dipstick urine analysis as a surrogate marker of microscopic pyuria. A positive dipstick was defined as ≥ ‘trace’ leucocyte esterase, and the microscopy reference standard ≥10 wbc/μL.
In all of the laboratory-based analyses that required repeated microscopic leucocyte counts, enumeration was undertaken by the same researcher. Any experimental processing of urine was undertaken by another researcher. All urine samples were presented for analysis identified only by a randomly generated, four-digit study number. Those undertaking cell quantification were ‘blinded’ to the results of previous assessment. The work that examined the performance of pyuria as a surrogate marker of UTI required the involvement of multiple research staff due to the volume of samples assessed. Those undertaking microscopy were ‘blinded’ to the results of the dipstick analyses.
We used IBM® SPSS® SamplePower® (IBM, New York, USA) to determine sample size for the diagnostic efficacy analyses. The standard deviation (sd) of the log leucocyte count, derived from existing urine microscopy study data, was two (sd 2). A sample size of 400 in each group would yield 83% power to detect a clinically significant between-group difference of 0·5 at the 1% level (α = 0·01), allowing for multiplicity. Statistical analysis was conducted using IBM® SPSS® Statistics 19 (IBM, New York, USA). The data were reported as counts and proportions where appropriate, and summarised using standard descriptive statistics. Missing data points were excluded from the analysis. We checked the data for normality using Q-Q plots, and between group differences were assessed using anova, factorial repeated-measures anova, and the Wilcoxon signed-ranks test.
Urinary Leucocytes Decay Rapidly during Storage
In all, 200 MSU samples with pyuria (≥10 wbc/μL) on initial urine microscopy were included. Leucocyte survival decreased over time for all samples (Fig. 1). Within 2 h, the count in the bench storage (room temperature) group fell rapidly to ≈60% of its initial count, whilst refrigeration maintained the count at ≈80%. Boric acid retarded leucocyte decay most effectively at all time-points and its action was temperature-independent (Fig. 1); nonetheless, cell loss at 4 h still approached 40%. A repeated-measures anova showed significant effects associated with refrigeration and the use of boric acid (F = 11; DF = 2; P < 0.001), and duration of storage (F = 282; DF = 3; P < 0.001). None of the strategies used to prevent cell loss were effective enough to be clinically useful.
Centrifugation Produces Unpredictable Effects on Cell Salvage
In all, 20 MSU samples were included. Initial microscopy showed mean (sd, range) leucocyte counts of 207 (253, 0–845) wbc/μL. The results of centrifugation and reconstitution of the samples are tabulated below (Table 3). Cell sedimentation was less effective when the 39g protocol was used (F = 5; DF = 3; P = 0·003), although even greater forces were unable to remove all cells from the urinary supernatant. Whilst there was a trend towards better cell concentration at higher speeds, overall cell salvage was variable. Optimal cell recovery was associated with the 100g protocol, with higher speeds causing a net reduction in cell salvage.
Table 3. The effects of centrifugation on urinary leucocyte salvage using different centrifugal forces.
Mean proportion of leucocytes salvaged % (95% CI)a
Salvage from urinary supernatant
Salvage after sediment re-suspended
aQuantitative leucocyte recovery after centrifugation, expressed as a percentage of the initial count, before processing; RCF, relative centrifugal force.
The wide variance associated with the mean estimates of proportionate cell recovery indicates an unpredictable influence from centrifugation. When the analysis was confined to the 100g protocol, the coefficient of variation was marked at 5750%.
Staining Does Not Improve Diagnostic Accuracy
In all, 100 MSU samples that showed pyuria were included. We found no difference between the stained and unstained counts. The median difference in leucocyte counts between the two methods was zero with an interquartile range of 2.0 (Z = –0.356; P = 0.72).
Pyuria Performs Poorly as a Surrogate Marker of UTI
In all, 1223 patients (120 men and 103 women; mean age 54 years; 95% CI 53–55) and 36 asymptomatic control subjects (eight men and 28 women; mean age 41 years; 95% CI 36–46) provided 5081 MSU samples, which underwent routine laboratory culture. After the exclusion of 706 samples, which were not subject to microscopic evaluation (n = 257) or dipstick urinalysis (n = 449), 4375 samples were included in the final analysis. Patients had widespread urinary urgency (mean [sd] urgency score 3.38 [3.15]) and incontinence symptoms (mean [sd] daily incontinence 0.93 [1.79]) at presentation; mean (sd) 24-h urinary frequency was 9.3 (5.00). Symptoms were longstanding with a mean (sd) duration of 4.6 (3.91) years.
All 43 control MSU samples were negative for microscopic pyuria and routine laboratory culture (not tabulated). The diagnostic performance of microscopic pyuria in detecting a positive routine laboratory culture of ≥105 cfu/mL is presented in Table 4. Summary performance statistics are presented in Table 5. Microscopic pyuria did not perform well as a surrogate marker of UTI.
Table 4. The performance of microscopic pyuria in the detection of a positive routine laboratory culture of ≥105 cfu/mL.
Almost two-thirds of MSU samples that showed microscopic pyuria ≥10 wbc/μL, showed no leucocytes on dipstick testing. These data and summary performance statistics are reported in Tables 6 and 7. The use of leucocyte esterase dipstick testing failed to detect significant pyuria in most of the samples included in this study.
Table 6. The performance of dipstick ≥ ‘trace’ leucocyte esterase in the detection of microscopic pyuria ≥10 wbc/μL.
Microscopic pyuria result, n (%)
Pyuria 1–9 wbc/μL
Pyuria ≥10 wbc/μL
Dipstick ≥ ‘trace’
Total, n (%)
Table 7. Summary diagnostic performance statistics for dipstick ≥ ‘trace’ leucocyte esterase in the detection of microscopic pyuria ≥10 wbc/μL.
Microscopic pyuria is our key diagnostic surrogate of infection. Whilst microscopy remains a component of urine analysis in most diagnostic laboratories, the dipstick has all but replaced the microscope in the clinic setting. Irrespective of which method is favoured, these data have important implications for those who use these tests to guide their management.
Urine needs to be evaluated for pyuria immediately after collection, as rapid leucocyte lysis occurs in the hours after sampling. This cell destruction appears to be retarded by boric acid, although significant cell loss appears inevitable. Urinary centrifugation affects cell salvage so variably that it is inappropriate for use in clinical practice. Staining appears to confer no significant influence on leucocyte detection. These data also cast considerable doubt on the veracity of dipstick urine analysis, and microscopic pyuria, as clinicallyuseful indicators of urinary infection in this patient population.
The present study, which examined the diagnostic performance of pyuria as a surrogate marker of UTI, used several researchers to accommodate the volume of samples processed. Whilst the microscopic evaluation of pyuria may have been subject to subtle inter-observer differences, the large sample size and narrow CIs for the test performance statistics suggest that any variation was minimal. The patients in the present study were recruited from a single specialist centre and could be subject to selection bias. Nonetheless, the distribution of LUTS in the present study sample are similar to those reported by large population-based, symptom prevalence studies , which ought to permit the application of these findings to the wider population of patients with LUTS.
The deterioration of urinary leucocytes is of particular concern. Boric acid has been advocated as a bacteriostatic urinary preservative in samples submitted for microbiological culture, at similar concentrations to those used in the present study . Sample tubes with a boric acid additive are commercially available, but not widely used. One small post-marketing study reported the utility of these additive tubes in the preservation of urinary leucocytes , although our data challenge this conclusion.
The comparative MSU culture and microscopy data challenge the diagnostic threshold of ≥10 wbc/μL convincingly. This threshold was conceived using data from the asymptomatic, applied universally, and validated using potentially inappropriate culture reference standards [7, 29]. Differences in disease prevalence amongst symptomatic groups mean that assumptions relating to test performance cannot be upheld without careful evaluation in representative patient populations.
Whilst microscopic pyuria has been accepted as an important surrogate of urinary infection for many years, it has serious limitations. Leucocytes are subject to marked deterioration when stored, even for short periods, and pyuria demonstrates poor sensitivity in patients with LUTS. Neither centrifugation, nor staining, appears to confer any diagnostic advantage.
This work was generously funded by project grants from the International Urogynecology Association, and Research into Ageing.