Diagnostic Utility of Urinalysis in Detecting Urinary Tract Infection in Hemodialysis PatientsVij R.a, b · Nataraj S.a, c · Peixoto A.J.a, d
aMedical Service, VA Connecticut Healthcare System, West Haven, Conn., bInternal Medicine Residency Program, Norwalk Hospital, Norwalk, Conn., cInternal Medicine Residency Program, University of Connecticut School of Medicine, Farmington, Conn., and dSection of Nephrology, Yale University School of Medicine, New Haven, Conn., USA Corresponding Author
Background/Aims: Urinary tract infections (UTIs) are common in dialysis patients. The diagnostic accuracy of pyuria in this population has been incompletely evaluated and is the object of this study. Methods: We assembled a historical cohort of male hemodialysis patients with urinalysis and urine microbiology data. Each matched pair of urinalysis and urine culture was considered as the unit of measurement. We evaluated the diagnostic performance and plotted receiver operating characteristic curves for different cutoff values of pyuria [>5, >10, >50, >100 white blood cells per high-power field (WBC/HPF)]. Results: Of 134 patients, 97 had at least one matched urinalysis and urine culture, providing 224 pairs for analysis. The prevalence of pyuria was 65 and 51% using cutoff values of >5 or >10 WBC/HPF, respectively. The presence of pyuria (>5 WBC/HPF) had a sensitivity between 82 and 89% and specificity of between only 53 and 55%. The presence of nitrites on dipstick had high specificity (94%) but very poor sensitivity (14–20%). All other urinary indices had poor diagnostic performance in the identification of a positive urine culture. Conclusion: In the absence of adequate specificity and positive predictive value in dialysis patients, urine culture should be obtained to guide further treatment.
Copyright © 2009 S. Karger AG, Basel
Urinary tract infections (UTIs) are common in dialysis patients, are associated with an increased rate of complications, and may be difficult to diagnose due to often subclinical presentations . Of special relevance is the limited diagnostic value of the presence of pyuria in the urinalysis, as pyuria is a common occurrence in dialysis patients (11–70%) and may be associated with the underlying etiology of the kidney disease rather than with the presence of infection . Accordingly, the diagnostic accuracy of pyuria for the presence of infection in this population is quite variable. In a review of the published literature, Fasolo et al.  estimated that the presence of pyuria had sensitivity of 50–82%, specificity of 69–99%, positive predictive value of 11–99% and negative predictive value of 87–96%. However, the analyzed studies [2,3,4,5,6] were limited by small sample size and variability of definitions. In addition, the literature has not addressed the diagnostic characteristics of other features of the urinalysis, such as dipstick leukocyte esterase and nitrites, in the identification of UTI in these patients.
In this report, we present data on the diagnostic properties of the urinalysis in the identification of UTIs in hemodialysis (HD) patients. The goal of this study was to address some of the shortcomings of the literature and provide a more reliable framework for the clinical interpretation of the urinalysis in this population.
We assembled a historical cohort of chronic HD patients from the VA Connecticut Healthcare System who received at least 3 months of chronic HD between 1995 and 2006. Recruitment was based on the availability of clinical and laboratory data in the electronic medical record. The local Human Investigation Committee approved the protocol and granted waivers of informed consent and research authorization in view of the retrospective nature of the study.
We abstracted data on age, gender, relevant medical history, time on dialysis, urinalysis results, complete blood counts and detailed urine microbiology results associated with each urinalysis. We also reviewed progress notes surrounding each urinalysis in search of symptoms and clinical presentation at the time. We excluded patients receiving peritoneal dialysis, and those who had indwelling bladder catheters or chronic urinary instrumentation. In patients who had received a renal transplant, we only collected data related to the period when they were on HD.
Pyuria was defined as >5 white blood cells per high-power field (WBC/HPF) in the centrifuged sediment. In addition, we analyzed data according to different cutoff values of pyuria (i.e., >5, >10, >50 and >100 WBC/HPF). We also collected data from dipstick leukocyte esterase and nitrites. For these items, results were analyzed dichotomously as ‘negative’ if reported as negative or trace, or as ‘positive’ if 1+ or greater. Significant bacteriuria qualifying as UTI was defined as >10,000colony forming units/milliliter (CFU/ml). We also performed separate analyses using 100,000CFU/ml as the cutoff for significant bacteriuria.
For the sake of this analysis, a matched pair was defined as a pair of urinalysis and urine culture obtained on the same day for an individual patient. When analyzing the diagnostic properties of the test, each of these pairs was considered as the unit of measurement.
Power calculation: Prevalence of pyuria in dialysis patients is on an average 30–40%, while its sensitivity and specificity for the detection of UTI in this population is 50–88 and 69–99%, respectively . We approached the sample size calculation through an estimation of the number of paired urinalysis and urine cultures required to guarantee a precise estimate (i.e. the confidence interval) of sensitivity and specificity of pyuria . Using a prevalence of 35%, and anticipated sensitivity of 80% and specificity of 90%, we approximated that ∼200 pairs of urinalysis and urine culture would allow us to confirm sensitivity and specificity within a tight 10% confidence interval, with a false-positive level (α) set at 5% (see Carley et al.  for details of the method).
We calculated descriptive statistics for the population. Sensitivity, specificity, positive and negative predictive values, positive and negative likelihood ratios were calculated for leukocyte esterase and nitrites, in relation to microbiologically proven UTI. Descriptive statistics were calculated and receiver operating characteristic (ROC) curves were plotted for different cutoff values of pyuria. All statistical analyses were performed using Microsoft Excel 2002 (Microsoft Corp., Redmond, Wash., USA).
Patients were all men, aged 70 ± 10 years (mean ± SD, range 45–91 years). All were on stable chronic HD (at least 3 months) performed three times a week. Table 1 describes the baseline characteristics of the cohort, including clinical features that are relevant to the occurrence of pyuria and/or UTIs. There were 134 patients with available urinalyses, of which 97 had at least one matched urinalysis and urine culture. We defined each available pair of urinalysis and urine culture as the unit of measurement. In total, there were 224 such pairs available for analysis, and this is the denominator used for most of the results presented.
|Table 1. Baseline characteristics of the cohort|
In most cases, we were unable to confirm the presence of symptoms. However, in 29 of 97 patients (30%), at least one symptom possibly related to UTI was uncovered. In addition, 13 of 97 patients (13%) were either hospitalized with a UTI or had one such infection as a complicating diagnosis during a hospital admission. In 12 of the 97 cases (12%), an admission for UTI was complicated by septic shock.
A total of 90 cultures had ≥10,000 CFU. Of these, the following organisms were isolated: 31 Enterococcus faecalis (6 vancomycin-resistant), 11 Candida albicans, 3 Candida glabarata, 8 Escherichia coli, 5 Klebsiella pneumonia, 3Klebsiella oxytoca, 4 Enterobacter aerogenes, 4 Pseudomonas aeruginosa, 2 Staphylococcus aureus (1 methicillin-resistant), 3 Proteus mirabilis, 1 Providencia retgerri, 2 Morganella morganii, 2 Serratia marcescens, 2 Citrobacter freundi and 1 Citrobactor diversus. Other positive results were not specific for the final organism and were reported as either non-lactose fermenting Gram-negative rods or yeast. There were 16 positive urine cultures without pyuria (WBC <5 WBC/HPF) on urinalysis.
As shown in table 1, positive nitrites were observed in 6.1% of all urinalyses, whereas a positive dipstick for leukocyte esterase was seen in 41.4% of samples, and pyuria was demonstrated in 58 or 37% of samples (using the 5 or 10 WBC/HPF threshold, respectively). Of the available matched urinalysis-urine culture pairs, positive nitrites were observed in 9.4%, positive leukocyte esterase was found in 58%, and pyuria was demonstrated in 61.6–47.8% of samples (using the 5 or 10 WBC/HPF threshold, respectively).
Tables 2 and 3 describe sensitivity, specificity, positive predictive and negative predictive values and likelihood ratios for the presence of UTI in the cohort. Results are based on the 224 pairs of urinalyses and urine cultures, and provide separate information for leukocyte esterase, nitrites, and for the presence of pyuria based on different cutoff values (5, 10, 50 and 100 WBC/HPF). We also present separate results according to the number of colonies observed in the positive urine cultures (either >104 or >105 CFU/ml).
|Table 2. Diagnostic performance of different urinalysis indices in the identification of a positive urine culture with >104 CFU/ml|
|Table 3. Diagnostic performance of different urinalysis indices in the identification of a positive urine culture with >105 CFU/ml|
Presence of pyuria (>5 WBC/HPF) had a sensitivity between 82 and 89% and specificity of between only 53 and 55%. Figure 1 depicts the ROC curves for different thresholds of pyuria in the identification of a positive urine culture. The diagnostic performance of pyuria was similar among patients with diabetic renal disease and those with other etiologies (data not shown). The presence of nitrites had high specificity (94%) but very poor sensitivity (14–20%). All other urinary indices had poor diagnostic performance in the identification of a positive urine culture.
|Fig. 1. ROC curves for different cutoff values of pyuria to detect positive urine culture (>10,000 or >100,000 CFU/ml). a ROC curve for different cutoff values of pyuria and positive culture (>10,000 CFU/ml). Area under curve (AUC): 0.73. SE AUC: 0.015. b ROC curve for different cutoff values of pyuria and positive culture (>100,000 CFU/ml). AUC: 0.76. SE AUC: 0.0153. c ROC curve for nitrite, leukocyte esterase and pyuria >5 WBC/HPF. AUC: 0.70. SE AUC: 0.0162.|
In this report we have presented data on the diagnostic performance of selected components of the urinalysis in the detection of UTI in dialysis patients. Our results indicate that pyuria has adequate sensitivity but poor specificity (and therefore, high negative predictive but low positive predictive value). In addition, we demonstrated that the leukocyte esterase dipstick performed slightly less well than pyuria, and that the nitrite dipstick had excellent specificity but dismal sensitivity.
The prevalence of pyuria in our study was 65 and 51% using the cutoff values of >5 or >10 leukocytes per high-power field, respectively. These results are relatively similar to previous findings in the literature, where the observed prevalence of pyuria in dialysis patients ranged from 11 to 70% . Our data add to the literature by analyzing the prevalence of abnormalities related to dipstick testing for leukocyte esterase (41%, similar to pyuria) and nitrites (6%, distinctively uncommon). We are unaware of any previous studies analyzing dipstick abnormalities in patients with kidney disease or on dialysis.
Overall, the urinalysis is a mediocre test for the identification of UTI in dialysis patients. This is not much different from the general population. Several possible reasons lie behind the limited diagnostic value of pyuria in end-stage kidney disease. Cabaluna et al.  were the first to analyze pyuria in a small group of 25 dialysis patients. Their results indicated a profile of frequent white cell clumps that develop over time on dialysis. These clumps were composed primarily of lymphocytes and were not directly associated with infection. Possible explanations included progressive damage of the kidneys or mucosal abnormalities in the urinary tree, especially the bladder, though neither of these possibilities was formally evaluated . Another possibility is that lower urine volume predisposes to more pronounced pyuria, as suggested by the results of Saitoh at al.  in 90 HD patients, although this hypothesis was not corroborated by the methodologically stronger study of Chaudry et al.  in 32 HD patients.
Some limitations apply to our study. We used a retrospective design, therefore the presence of symptoms could not be fully ascertained. Because we know that the presence of symptoms alters the operating characteristics of pyuria , it is possible that this limitation may have hampered our ability to more precisely define the diagnostic role of pyuria. Likewise, we were unable to determine the residual urine volume for each patient and its relationship to abnormalities in the urinalysis. This is a question that remains unanswered in the literature. Lastly, all of our patients were men, thus limiting the generalizability of our data.
In the absence of adequate specificity and positive predictive value in dialysis patients, our conclusion is that a urine culture should be obtained to guide further treatment. However, given the relatively high negative predictive value of pyuria, our results suggest that a patient who has atypical symptoms (or no symptoms) has a low likelihood of infection if the urinalysis has no pyuria, leukocyte esterase or nitrites. We believe our results are important in that we addressed several of the previous limitations of the literature, especially those related to sample size, case mix (dialysis and non-dialysis), and the lack of data on the dipstick. Urinalysis is a commonly obtained test, and our data help clarify issues in interpretation of this test in dialysis patients. Prospective studies may further clarify the utility of the urinalysis in patients on HD.
This material is the result of work supported with resources and the use of facilities at the VA Connecticut Healthcare System, West Haven, Conn., USA.
Aldo J. Peixoto, MD
Medical Service – 111, 950 Campbell Avenue
West Haven, CT 06516 (USA)
Tel. +1 203 932 5711, ext. 5907, Fax +1 203 937 3425
Received: November 30, 2008
Accepted: January 27, 2009
Published online: September 3, 2009
Number of Print Pages : 5
Number of Figures : 1, Number of Tables : 3, Number of References : 9
Nephron Clinical Practice
Vol. 113, No. 4, Year 2009 (Cover Date: November 2009)
Journal Editor: El Nahas M. (Sheffield)
ISSN: 1660-2110 (Print), eISSN: 1660-2110 (Online)
For additional information: http://www.karger.com/NEC