Comparison of wet-mount, Wright-Giemsa and Gram-stained urine sediment for predicting bacteriuria in dogs and cats

Abstract

This study assessed the standard urinalysis technique and sediment stain techniques as predictors of bacterial culture results for canine and feline urine. Canine (n = 111) and feline (n = 79) urine samples were evaluated using unstained wet-mount and air-dried Gram and Wright-Giemsa stained sediment; results were compared to aerobic bacterial culture. Eleven canine and 7 feline urine samples were culture positive. Unstained wet-mount and stained sediment had sensitivities of 89% and 83% and specificities of 91% and 99%, respectively. The specificity of using either stain was higher (P < 0.01) than wet-mount examination for detecting bacteriuria. There were significant differences among 3 technologists in detecting true positives (P < 0.01). Association of sediment and culture results used 112 canine and 81 feline samples. There was a negative association (P < 0.01) between lipid detection and wet-mount identification of bacteria.

Introduction

The identification of urinary tract infections (UTIs) in dogs and cats is typically made from microscopic examination of wet-mounts of urine sediment in combination with clinical signs. The examination of unstained or stained wet-mount sediment preparations in the traditional urinalysis method is not optimal for detection of bacteria. False positives (other structures misidentified as bacteria) and false negatives (failure to detect bacteria) are considered common. At the Atlantic Veterinary College (AVC) Diagnostic Services Laboratory, trace bacteria are often reported in the absence of cellular abnormalities or positive cultures, especially in feline urine samples. It is not known if this is due to the presence of dead or fastidious bacteria (1), if other structures are misidentified as bacteria, or if there are significant infections occurring. Alternative methods to identify bacteriuria may be superior. In 1 laboratory, examination of air-dried modified Wright (Diff Quik) stained sediment drop preparations had superior sensitivity and specificity over traditional urinalysis in the detection of bacteriuria in dogs and cats (2,3). Similar results were obtained from another laboratory utilizing air-dried Gram-stained sediment from dogs’ urine (4).

Other aspects of urine analysis may predict a positive culture, including estimates of increased white and red blood cells. One study (5) did not find an association between positive cultures and the presence of red blood cells (RBCs) in dogs, but other studies did in both dogs (2) and cats (6,7). It is unknown if other findings on routine urinalysis such as the presence of epithelial cells, sperm, crystals, or lipid may be associated with the presence of a UTI.

The overall goal of this study was to improve the quality of urinalysis for dogs and cats by comparing the current standard urinalysis method with other techniques to see which would be more predictive of bacterial culture results. The first specific aim was to compare the visual detection of bacteria in routine microscopic examination of unstained sediment on wet-mounts with that of air-dried sediment stained with Wright-Giemsa and Gram stains. The comparison included evaluation of sensitivity, specificity, negative and positive predictive value, test efficiency, and odds ratio. In addition, inter-individual differences among 3 technologists in visual detection of bacteria on wet-mounts were investigated. The second specific aim was to assess if red and white blood cell estimates could be predictive of urine culture results. The third specific aim was to assess other wet-mount urinalysis findings (epithelial cells, sperm, crystals, or lipid) to determine which, if any, may enhance the likelihood that a positive bacterial culture would be found. For all comparisons, aerobic culture of urine served as the reference method for determination of the presence or absence of bacteriuria.

Materials and methods

Sample selection and handling

Urine specimens were submitted as part of routine evaluation of canine and feline patients to the AVC Diagnostic Services Laboratory from April 2009 to February 2010. Urine was collected aseptically by cystocentesis. Samples were assigned one number for routine urinalysis and a second number for additional processing; this maintained blinding until data collection was complete.

Bacterial culture

All procedures were performed in a Class II Biosafety cabinet (Canadian Cabinets, Ottawa, Ontario). A 0.5-mL aliquot of urine was sterilely removed for bacteriologic culture and the remaining urine underwent routine urinalysis and additional processing. If multiple samples were collected from a single animal over the study period, only the results of the first submission were included. However, if a negative culture was initially obtained but a positive culture was subsequently found, the results of this first positive culture were instead selected for inclusion.

Quantitative culture was performed within 4 h of sample collection. Using aseptic technique, 1 μL of urine was inoculated onto blood agar (BA) and MacConkey agar plates and 10 μL was inoculated onto BA using calibrated inoculation loops (Astral Inoculation Loop; Bio Plas Inc, San Rafael, California, USA). The plates were incubated aerobically at 35°C to 37°C for incubator. If there was no growth within at least 24 h in a CO₂ 48 h, culture was considered negative. Any growth on either BA plate was interpreted as positive for bacteriuria.

Urinalysis (including wet-mount examination)

Routine urinalysis was performed by 1 of 3 laboratory technologists within 2 h of sample collection. The color, clarity, and volume of the sample were recorded. Urine was placed in a conical tube (Urinalysis Tube 15 mL; Fisher Scientific, Pittsburgh, Pennsylvania, USA) and centrifuged (International Equipment Company Clinical Centrifuge; Damon/IEC Division, Needham Heights, Massachusetts, USA) at approximately 1500 × g for 5 min. A drop of the re-suspended sediment was placed on a glass slide and a coverslip placed over it. The slide was examined and the routine sediment results recorded, including the identification and estimation of the number of white blood cells (WBCs), RBCs, and bacteria. Numerical values for WBCs and RBCs were divided into those with increased numbers [> 5/high power field (HPF) at 400×] and those with numbers considered normal (≤ 5/HPF). Numbers of bacteria were subjectively designated as trace, 1+, 2+, 3+, and 4+ as per standard laboratory protocol.

Evaluation of air-dried stained sediment

A drop of sediment was also placed in the center hole of a Cytopro cytocentrifuge filter (Wescor, Logan, Utah, USA) located on a glass slide; this paper served to wick away moisture to enhance rapid sediment drying. Two such air-dried slides were prepared for each sample. One was Gram-stained according to the packet insert (Becton, Dickinson and Company, Sparks, Maryland, USA) and the other was stained with Wright-Giemsa stain using a Hematek^® slide stainer (Siemens Healthcare Diagnostics, Frimley, Camberley, UK). An investigator (E. O’Neil), who was blinded to the routine urinalysis data and the culture results, examined the slides. Bacteria were counted using oil immersion (1000×) light microscopy. The total number of bacteria in 20 fields was classified as none, occasional (1 to 4 bacteria), few (5 to 9 bacteria), moderate (10 to 20 bacteria), or many (> 20 bacteria) (2).

Statistical analysis

All findings associated with positive cultures were compared with those with negative cultures. Categorical data were evaluated using Chi-square and Fisher’s exact tests. Sensitivity, specificity, positive and negative predictive values, test efficiency, and positive and negative likelihood ratios of unstained sediment wet-mounts and Gram and Wright-Giemsa stained air-dried sediment methods of detecting bacteriuria were calculated, with culture results used as the reference method. The McNemar test was used to examine the hypotheses that the Gram and/or Wright-Giemsa stain method had different diagnostic sensitivity and specificity for the detection of bacteriuria than those associated with the unstained wet-mount method. Odds ratios and 95% confidence intervals (95% CIs) for positive culture and false positive wet-mount results were calculated using logistic regression. Multivariable models were created for each outcome while controlling for confounding effects of group characteristics, including age, species, and gender; these were included as they have been reported to have an association with the development of bacteriuria (6,8). In the logistic regression, the significance of predictor variables was assessed by likelihood-ratio tests, and CIs were computed by the profile likelihood method. A computer software program (STATA 11, Data Analysis and Statistical Software; StataCorp LP, College Station, Texas, USA) was used to perform the statistical analysis. For all tests, a P-value < 0.05 was considered significant.

Results

Descriptive information

Samples from 193 animals contributed to most analyses; these consisted of 112 canine samples and 81 feline samples. A summary of traditional wet-mount urinalysis findings categorized by culture results is found in Table 1. Three samples did not have sediment preparations made, so the sediment examination comparisons were based on 190 samples (111 canine and 79 feline) detailed in Tables 2 and 3.

Table 1.

Urinalysis wet-mount sediment findings categorized by culture results for 112 dogs and 81 cats

	Canine (n = 112)		Feline (n = 81)
	Positive culture (n = 11)	Negative culture (n = 101)	Positive culture (n = 7)	Negative culture (n = 74)
Gender
Male	2	47	2	40
Female	9	54	5	34
Age (years) (mean ± SD)	9.6 ± 3.9	7.6 ± 4.3	10.7 ± 4.7	9.4 ± 5.4
Red blood cells (> 5/HPF)b	2 (18%)	16 (16%)	2 (29%)	21 (28%)
White blood cells (> 5/HPF)b	5 (46%)a	1 (1%)	4 (57%)a	1 (1%)
Bacteriuriab	10 (90%)a	8 (8%)	6 (86%)a	8 (11%)
Lipidb	1 (9%)	27 (27%)	1 (14%)	33 (45%)

^aP < 0.05 for the comparison between positive and negative culture results, SD — standard deviation, HPF — high power field.

^bn (%).

Table 2.

Comparison of urine culture with unstained and Gram and Wright-Giemsa stained sediment examination for the detection of bacteria in dogs and cats (n = 190)

		Unstained sediment examination		Gram and Wright- Giemsa stained air-dried sediment examination
	Culture	Positive	Negative	Positive	Negative
Combined canine and feline (n = 190)	Positive	16	2	15	3
	Negative	16	156	2	170
Canine (n = 111)	Positive	10	1	9	2
	Negative	8	92	1	99
Feline (n = 79)	Positive	6	1	6	1
	Negative	8	64	1	71

Table 3.

Test performance of wet-mount unstained sediment and Gram and Wright-Giemsa air-dried stained urine sediment examination against culture with 95% confidence intervals in the combined canine and feline group (n = 190)

Variable	Definition	Routine unstained wet-mount examination n (95% CI)	Gram and Wright-Giemsa air-dried stained urine sediment examination n (95% CI)
Sensitivity (%)	TP/(TP + FN) × 100	89 (65–99)	83 (59–96)
Specificity (%)	TN/(FP + TN) × 100	91 (85–95)	99 (96–100)a
PPV (%)	TP/(TP + FP) × 100	50 (32–68)	88 (64–99)
NPV (%)	TN/(TN + FN) × 100	99 (96–100)	98 (95–100)
Test efficiency (%)	(TP + TN)/(TP + FP + TN + FN) × 100	91 (85–94)	97 (94–99)
Likelihood ratio positive (LR+)	Sensitivity/(1 — specificity)	9.6 (5.8–15.7)	71.7 (17.8–289)
Likelihood ratio negative (LR−)	Specificity/(1 — sensitivity)	0.12 (0.03–0.45)	0.17 (0.06–0.47)
Odds ratio	LR+/LR−	78 (20–253)	425 (81–3752)

^aMcNemar’s test P < 0.01.

CI — confidence interval; TP — true positive; FN — false negative; TN — true negative; FP — false positive; PPV — positive predictive value; NPV — negative predictive value.

Urine culture

On bacterial culture, 18 (9%) of the 193 urine cultures yielded growth (Table 1). A single bacterial species was cultured in each of 11/112 canine and 7/81 feline urine samples; no sample yielded multiple species.

Wet-mount unstained and Wright-Giemsa and Gram-stained sediment examination for bacteriuria with comparison to culture

The detection of bacteria on wet-mount sediment and on Gram- and Wright-Giemsa stained air-dried sediment examinations compared with culture is summarized in Table 2. Bacteria were observed in unstained wet-mounts in 32/190 (17%) urine samples. Sixteen percent of canine and 18% of feline urine samples were positive on wet-mounts for bacteriuria and together had 8% true positive, 82% true negative, 8% false positive, and 1% false negative results. When the species were separated, true positives in canine and feline samples were 9% and 8%, respectively, and true negatives were found in 83% of canine samples and 81% of feline samples. False positive results were found in 7% of canine samples and 10% of feline samples. In both species, 1% of the samples were classified as false negatives.

Bacteria were well-preserved and easily identified using either Wright-Giemsa or Gram stains (Table 2). Detection of bacteria in either stained preparation was identical. Therefore, these are reported together and the sample type termed as stained air-dried sediment. Bacteria were observed in 17/190 (9%) from the combined canine and feline group and in 9% of both canine and feline samples. Bacteria were ≥ 20 per oil immersion field in all samples in which they were identified. In the combined canine and feline group, these preparations had classifications of 8% true positive, 90% true negative, 1% false positive, and 2% false negative. When species results were examined separately, 8% of the air-dried sediment examinations were true positives in both canine and feline groups. True negative results were found in 89% of canine samples and 90% of feline samples. There were false positives for 1% of the samples for both species and there were false positives and false negatives for 2% of canine samples and 1% of feline samples.

The test performance of the examination of unstained wet-mount and air-dried stained sediment is found in Table 3. The positive predictive value of air-dried stained sediment in detecting bacteriuria was higher (88%) than that for unstained wet-mount examination (50%). The specificity of air-dried sediment examination (99%) was significantly different (P < 0.01) from the 91% specificity of the wet-mount examination. The sensitivities of both air-dried sediment and wet-mount examinations were high at 83% and 89%, respectively; these were not significantly different. The odds ratio of air-dried stained sediment examination in which bacteria were seen was 425 in predicting a positive culture compared with 80 for unstained wet-mount examination.

Comparison of bacterial detection in wet mount examinations by 3 technologists

Summary data on technologist performance are found in Table 4. There was a significant difference (P = 0.01) between the 3 technologists in the accuracy of identifying bacteria in wet mount examinations using culture as the reference method. There was also a significant difference (P = 0.01) between them in the rate of false positives in the wet-mount detection of bacteria in those samples that were negative using culture as the reference method. There were too few positive cultures to make comparison between technologists. Technologists 1, 2, and 3 had 0%, 8%, and 22% false positives, 14%, 7%, and 4% true positives, 83%, 85%, and 70% true negatives and 2%, 0%, and 4% false negatives, respectively.

Table 4.

Comparison of detection of bacteria on wet-mount examination compared to culture results by 3 technologists

Technologista Culture	Bacteria noted on wet-mount examination
	1		2		3
	Yes	No	Yes	No	Yes	No
Positive n (%)	6 (14)	1 (2)	8 (7)	0 (0)	1 (4)	1 (4)
Negative n (%)	0 (0)	35 (83)	10 (8)	106 (85)	5 (22)	16 (70)
Total	42		124		23

^aTechnologist could not be identified for 4/193 samples.

Evaluation of red and white blood cells as predictors of positive cultures

For the combined canine and feline group as well as for the separate species groups, no association was found between increased numbers of RBCs on sediment examination (> 5 RBCs/HPF) and culture results. There was a significant association with the presence of > 5 WBCs/HPF and positive culture for both canine and feline urine samples (P < 0.01).

Evaluation of other urinalysis findings as predictors of positive cultures

There was no association between culture results and the detection of sperm, epithelial cells, or crystals. There was a significant negative association (P < 0.01) between lipid detection and bacteriuria on wet-mount examination; bacteria were therefore more commonly identified in samples that did not have lipid present.

Regression models for detection of bacteria by wet-mount sediment and air-dried stained sediment

Regression models were analyzed to evaluate predictors for testing true positives (on culture) and false positives on wet-mount examinations. There were too few positive culture results to evaluate possible predictors for testing false negatives. The only predictor affecting the probability of testing true positives by culture was the presence of > 5 WBCs/HPF (P < 0.01) which increased the odds of detecting bacteriuria by a factor of 85 (95% CI: 16 to 453). There was high (97%) test efficiency (Table 3) between Gram- and Wright-Giemsa-stained air-dried sediment and culture results, and too few discordant results to analyze the impact of predictors on false positive or false negative results. There was also good (91%) test efficiency (Table 3) between wet-mount examinations and culture results, but lipid and RBCs had significant associations with false positive wet-mount examination results. The absence of lipid (P = 0.01) and the presence of RBCs (P = 0.02) (Table 5) significantly increased the probability of detecting bacteria in culture negative samples.

Table 5.

Logistic model of wet-mount bacterial detection of culture negative canine and feline urine (n = 172)

	Dependent variable
Wet-mount bacteria	Odds ratio	95% CI	P
Presence of lipid	0.13	0.007–0.71	0.01
> 5 RBCs/HPF	3.5	1.2–10.6	0.02

CI — confidence interval; P — probability; RBCs — red blood cells. P < 0.05 for the comparison between positive and negative culture results. HPF — high power field (400×).

Discussion

This study met the first specific aim of comparing the detection of bacteria in unstained sediment on wet-mounts with that of air-dried sediment stained with Wright-Giemsa and Gram stains using bacterial culture as the reference. The air-dried stained sediment examination using either stain proved to be superior, with a markedly higher positive predictive value of detecting bacteria on either of the air-dried sediment preparations (88%) compared with that of routine unstained wet-mount examination (50%). The negative predictive value was high for both the wet-mount and either air-dried stained methods of urine examination. This suggests that urine samples negative by either method of microscopic examination, at least in the laboratory tested, are unlikely to require culturing. It is important to note that these positive and negative predictive values are dependent on the prevalence of bacteriuria in the study population and may not translate to all populations. The 9% prevalence of bacteriuria (defined by positive culture) in the canine population in the present study was similar to a previous study that reported UTIs in 10% of dogs admitted to veterinary hospitals (9). The 9% culture positive prevalence for the feline population was intermediate with that of previous studies which reported prevalences of 3% (10), 6% (3), 14% (6), and 33% (11). Variation in case selection and population, and limited information on the collection and culture methods of previous studies make it difficult to compare these prevalences to that in the current study.

This is the first study to utilize both Wright–Giemsa and Gram stains in the evaluation of urine sediment in both dogs and cats. The specificity of the air-dried stained sediment examination in detecting bacteriuria using either stain was higher (99%) compared with wet-mount examination (91%). Sensitivity was actually higher for wet-mounts (89%) compared to air-dried stained sediment examination (83%) but the difference was not significant. This contrasts with previous studies of canine urine (2,4), in which evaluation of air-dried sediment stains had higher sensitivity and specificity in detecting bacteriuria than wet-mount examination. This may reflect regional population differences, case selection, inclusion of feline samples, or different techniques in preparing the sediment. Veterinary laboratories typically have special cytospin instruments in which fluids are gently centrifuged to leave sediment on slides. Cytospin technology is not available to the average veterinary practice and the present study endeavored to assess a simple drop method that could be used in that setting. In another study evaluating occult UTIs in cats (7), examination of both Gram-stained sediments and wet-mounts, or either of these resulted in visualization of bacteria in 97% of culture positive samples and no visualization in culture negative samples. It was not specified whether the slide preparation was done using sediment drops or a cytospin technique and the study did not include all culture results, making it difficult to compare with the present study.

In the present study, false positive rates for bacterial detection in urine wet-mounts were higher than false negative rates; similar findings were reported in 3 previous studies (2–4). Reasons for false positives could include the presence of non-viable, fastidious, or anaerobic bacteria or the misidentification of small particles as bacteria. Such particles could include small lipid droplets, cytoplasmic organelles, amorphous crystals, or debris (3). The differences between the 3 technologists in their ability to correctly identify bacteria may be multi-factorial. Individual bias, experience, and quality of training could be factors in both the laboratory and in private veterinary practices. Concern about potentially missing UTIs may result in the subconscious tendency to call equivocal structures bacteria. This may lead to unnecessary requests for bacterial culture or unnecessary antibiotic therapy. The Gram and Wright-Giemsa staining methods may avoid this bias as the stained bacteria are easier to differentiate from other structures.

Lipid droplets were more commonly observed in those samples that had no bacteria identified on wet-mount examination. It seems, therefore, that small droplets were identified as either bacteria or lipid, but not both. If lipid droplets were erroneously judged to be bacteria, this could contribute to false positive results. The differentiation of small lipid droplets from bacteria may be aided by the addition of Oil Red O stain to wet-mount slides.

The second specific aim of the study was to assess if RBC and WBC estimates could predict culture status. The study found that the increased number of RBCs on wet-mount evaluation was not associated with positive culture. This agrees with a previous study (5) but contradicts other studies in which the presence of blood was associated with positive urine cultures in dogs (2) and cats (6,7). While increased RBCs often are seen in UTIs, this may simply reflect vascular trauma during cystocentesis or hematuria from other causes, such as urolithiasis. A surprising finding in the current study was the association between the presence of RBCs on wet-mount examination and the visual identification of bacteria in samples that were culture negative. This suggests that technologists in the current study may have identified equivocal structures as bacteria instead of debris when RBCs were present.

The presence of increased numbers of WBCs on wet-mount examination had a strong association with positive urine cultures for both species as has been previously reported (2–7,12). However, the present study supported the observation that increased WBCs is an imperfect screening tool for the detection of bacteriuria. Approximately half of the culture positive samples had ≤ 5 WBC/HPF (Table 1). These samples would therefore not have been cultured if one relied on the finding of increased WBCs to trigger suspicion for a possible UTI. Similar observations regarding the poor predictive value of the presence or absence of WBCs with UTIs have been made in other studies (2–5,8). Increased WBCs in the absence of bacteriuria may be due to other causes of inflammation and WBCs may not be increased in early UTIs or in those in immunosuppressed patients.

The third specific aim was to assess if the presence of epithelial cells, sperm, crystals, or lipid on wet-mount examination could enhance the likelihood that a positive bacterial culture would be found. No positive relationship was found. As noted previously, the presence of lipid droplets may have been misidentified as bacteria in the present study and contributing to false positive results.

In summary, the results of this study showed that the application of either a Gram or Wright-Giemsa stain to air-dried sediment was superior to evaluating unstained wet-mounts for predicting urine culture results. One of these stained preparations could be a routine step in all urinalyses or could be used to further evaluate samples with structures suggestive of bacteria on initial wet-mount preparations. Gram stain was not shown to be superior to Wright-Giemsa stain for the detection of bacteria, so either could be chosen, although the information about the Gram reaction may be useful (4). In the laboratory herein, false positives were more frequent than false negatives on wet-mount examination, possibly reflecting a bias derived from concern about potentially missing UTIs. The possible misidentification of lipid droplets as bacteria may explain some of the false positive results. Differences in the ability of technologists to reliably detect bacteria on wet-mount examinations were found and may be a concern in other laboratories and clinical practices. No association between culture results and RBC estimates was seen. While WBC estimates were statistically correlated with culture results, this finding was inconsistent and the presence or absence of WBCs should not be used to predict bacteriuria. The results of this study, including the recommendation to utilize air-dried stained sediment examination, could be applied to large veterinary diagnostic laboratories as well as to smaller clinic settings.