Bile composition of healthy cats and cats with suspected hepatobiliary disease using point-of-care analyzers: A prospective preliminary study

Abstract

Although hepatobiliary disease is common in cats, little is known about the bile composition in either these diseased states or in healthy cats. The objectives of this study were to evaluate several analytes from the bile of healthy cats and to investigate the usefulness of measuring these variables to predict bacterial cholangitis. Cats were prospectively enrolled and divided into 3 groups: 21 healthy cats (group 1) and 14 cats with suspected hepatobiliary disease: 9 without bacterial biliary infection (group 2) and 5 with bacterial biliary infection (group 3). Percutaneous ultrasound-guided cholecystocentesis was conducted on each cat. Bile cytology and culture were carried out and bile was analyzed for pH, lactate, and glucose levels using several point-of-care (POC) devices. Reference values for several bile analytes in healthy cats were calculated and are presented in this study. Neither the pH (P = 0.88) nor the lactate concentration (P = 0.85) was significantly different among the 3 groups. Sodium concentration was significantly higher in group 3 than in group 2 (P < 0.05). Bile pH, lactate, and glucose levels were unable to predict the presence of a bacterial infection in the bile.

Introduction

Cholangitis is a common disease in cats and neutrophilic and lymphocytic cholangitis are often found on histopathology (1). In fact, 3 clinical syndromes have been recognized in this species: neutrophilic, lymphocytic, and chronic cholangitis (2). Neutrophilic cholangitis is most often associated with ascending bacterial infections of gastrointestinal origin, lymphocytic cholangitis with an immune-mediated process, and chronic cholangitis with liver fluke infestation (1–3). Liver biopsy with histopathological evaluation is required for a definitive diagnosis, making an ante-mortem diagnosis a challenge in certain clinical situations (1,4). Unfortunately, clinical findings, hematology and serum biochemistry, particularly liver enzyme activity, do not predict bacterial infection within the bile (3,5).

Percutaneous ultrasound-guided cholecystocentesis (PUC) has been described for bile sampling and is a safe procedure in cats (6–8). Bile cytology may reveal white blood cell infiltrates and/or bacteria and bile culture is a more sensitive diagnostic tool to confirm a septic process compared to liver tissue culture (5,9). There is still some controversy, however, since bile specimens with a positive culture result are not always associated with cytological visualization of bacteria and vice versa (8,9). The evaluation of other diagnostic tests for rapidly predicting bacterial cholangitis in cats would therefore be beneficial in clinically managing these patients. Measuring various biochemical variables, particularly lactate, glucose, and pH, in certain body fluids of dogs and cats has been useful in detecting bacterial infection, but these analytes have never been tested in bile (10,11).

The main hypothesis of this study was that the presence of bacteria in bile would alter biliary composition and result in an increased lactate concentration, a decreased pH and a decreased glucose concentration compared with the bile of healthy cats. The aims of this study were to: establish reference values for several analytes within the bile of healthy cats, including pH, glucose and lactate, using point-of-care (POC) blood gas analyzers; evaluate the repeatability and level of agreement of these POC analyzers; and investigate whether any of these analytes could predict bacterial cholangitis in cats.

Materials and methods

This study was designed as a prospective clinical observational study and received approval from the institutional animal ethics committee.

The healthy group (group 1) consisted of cats older than 1 y of age enrolled from our institutional teaching colony. Cats were deemed healthy based on history, physical examination by a Board-certified internist or criticalist, complete blood (cell) count (CBC), serum biochemistry, measurement of total thyroxine concentration if they were ≥ 4 y of age or if a thyroid nodule was palpated, FeLV/FIV testing (Combo Plus FIV/FeLV; IDEXX Laboratories, Markham, Ontario), urinalysis and abdominal ultrasound. All healthy cats were fed the same diet (Veterinary Diet, Feline adult dry food, Royal Canin, Puslinch, Ontario) for at least 1 wk before enrollment.

Client-owned cats with suspected hepatobiliary disease (affected cats) were enrolled after informed consent had been obtained from the owners. To be considered for enrollment in this group, the cats had to have clinical signs that could be attributed to a hepatobiliary disease, i.e., anorexia, lethargy, vomiting, diarrhea, icterus, weight loss, or some combination of these signs. Moreover, they had to show increased activity of alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma glutamyl transferase (GGT), or total bilirubin on serum biochemistry, and/or demonstrate sonographic hepatobiliary abnormalities. Cats were excluded if they had received antibiotics within 7 d or oral corticosteroid therapy within 3 mo before enrollment in the study or if total thyroxine concentration was above the reference range.

Affected cats were further subdivided into those with noninfected bile (group 2) and those with bacterial-infected bile (group 3) according to biliary cytology and culture results. Affected cats were included in group 3 if a positive culture was obtained or if bacteria were observed on cytology by a Board-certified clinical pathologist blinded to the history, clinical findings and culture results of the animal. A complete abdominal ultrasound was carried out by a single Board-certified radiologist, with a specific focus on size and echogenicity of the liver, diameter and wall thickness of the common bile duct, appearance and thickness of the gallbladder wall and appearance of the bile in the gallbladder, i.e., presence of biliary sludge.

The bile specimen was obtained via PUC under anesthesia using a short-acting intravenous anesthetic protocol. Healthy cats received of buprenorphine (Vetergesic; Champion Alstoe Animal Health, Whitby, Ontario), 0.02 mg/kg body weight (BW), IV and alfaxolone (Alfaxan; Jurox Animal Health, Rutherford, New South Wales, Australia), up to 2 mg/kg BW, IV to effect. The anesthetic protocol for the affected cats was established individually at the discretion of the attending anesthesiologist. The gallbladder was emptied as completely as possible during the procedure and the cat was excluded from the study if < 1 mL of bile was obtained.

Several drops of bile were introduced into an anaerobic transport medium (Starswab Anaerobic Transport System; Starplex, Etobicoke, Ontario) and immediately transported to the laboratory, where 200 μL of bile were placed in an ethylenediaminetetraacetic acid (EDTA) tube and direct and centrifuged bile smears were prepared immediately after collection. The remainder of the bile was used for analyte measurements with a lactate analyzer (Lactate Pro; Arkray, Edina, Minnesota, USA), a glucometer (AlphaTrak 2; Zoetis, Kirkland, Quebec) and POC analyzers: i-STAT (EC8+ cartridge, i-STAT Handheld Analyzer; Abbott Point of Care, Mississauga, Ontario) and Enterprise Point-of-Care (EPOC) (Element POC; Heska Corporation, Barrie, Ontario). All analytes were measured within 20 min of specimen collection. Biochemical variables were measured in the following manner: the lactate and glucose concentrations were measured 3 times on each sample with the handheld analyzers, then measurements with POC analyzers were carried out one after the other, beginning with the EPOC device.

Although pH, lactate and glucose were the biochemical variables of primary interest in this study, the POC analyzers also provided additional results. The POC analyzers measured lactate and glucose concentration via amperometry, as did the handheld lactate and glucose analyzers. Furthermore, pH, sodium (Na⁺), potassium (K⁺), ionized calcium (iCa²⁺) (only on the EPOC), chloride ion (Cl⁻) and partial pressure of carbon dioxide (PCO₂) were measured via potentiometry and hemoglobin was measured via conductometry on both POC analyzers. Bicarbonate (HCO₃⁻), base excess (BE), total carbon dioxide (TCO₂), hematocrit (HCT) and anion gap (AG) are calculated values on both POC analyzers.

A power calculation was not carried out at the start of this study due to the lack of clinical information available to enable this calculation. The number of animals required to observe a statistically significant difference in the lactate variable was therefore calculated based on preliminary results of the study. At the time of the calculation, 8 healthy cats and 5 affected cats had been enrolled. A sample size calculation revealed that a minimum of 7 cats in the healthy group and 7 cats in the affected group were required in order to demonstrate a statistically significant difference in lactate concentration at an alpha threshold of 5% with a t-test for 2 independent samples at a power of 85%. Given this calculation, our goal was to enroll 20 affected cats so as to be able to conduct a method comparison experiment between the POC analyzers and the handheld lactate analyzer and to compare differences in analytes among the various clinical groups. Furthermore, we aimed to enroll a minimum of 20 healthy cats so that reference values (RV) for the biochemical variables could be calculated.

Reference values (RVs) were established and confidence intervals (CIs) were calculated around the upper and lower reference limits following guidelines from the American Society for Veterinary Clinical Pathology (ASVCP) (12) with Reference Value Advisor, Version 2.1 macro (Microsoft Corporation, Redmond, Washington, USA) (13). Outliers in the data were identified using Dixon and Tukey methods. Parametric or robust methods were used and 90% CIs around the reference values were determined using bootstrap methods. Confidence intervals (CIs) provide an estimate of the uncertainty of the RV’s limits and should not exceed 0.2× the width of the reference intervals (RIs); WCI/WRI ratios (width of the CI/width of the RVs) were calculated and considered acceptable if < 0.2 (12).

The reliability of each machine was assessed based on the calculated coefficient of variation (CV) for each analyte (14) and compared to published precision targets in the blood (15–17). A method comparison study was carried out to evaluate the interchangeability of the POC machines. Pearson’s coefficients of correlation (r) were calculated among the machines for each analyte of interest to determine if there was a significant correlation among them and which regression model was to be applied for data analysis. If there was no significant correlation among the machines, no regression analysis was done. Regression analysis was used to detect the presence of a proportional and/or constant error and was carried out using a web-based software (App URL: bahar.shinyapps.io/method_compare/).

To graphically evaluate the data, Bland-Altman plots were constructed for the following analytes: pH, PCO₂, Na⁺, K⁺ and lactate, by plotting %Difference (i-STAT and EPOC) against the mean value of both methods for each variable using GraphPad Prism 8 software (GraphPad Software, San Diego, California, USA) (18). The method of comparison between the POC analyzers was analyzed by the Bland-Altman method and by applying preset clinically acceptable levels of agreement (17,19,20). As there are no published preset analytical specifications for bile, the authors subjectively created clinically acceptable target values for agreement based on published whole blood results for the analytes of interest before the study began (15–17,21). For the 2 methods to be interchangeable, 95% of the data points had to fall within the clinical limits of agreement (19).

In order to evaluate the difference among analytes in the 3 groups, a Kruskal-Wallis test was used for the initial comparison of the various analytes and the Mann-Whitney test was used for follow-up pairwise comparisons if the P-value of the Kruskal-Wallis test was < 0.05. Each P-value was divided by 3 to get the Bonferroni-corrected P-value for the Mann-Whitney test. All analyte data were evaluated for normality using Anderson-Darling test. The software SPSS (SPSS Statistics 25; IBM, Markham, Ontario) was used for this analysis.

Results

For group 1, 40 healthy cats were recruited and 21 were included in the study. Cats were excluded for the following reasons: stomatitis (n = 1); heart murmur associated with an obstructive hypertrophic cardiomyopathy (n = 1); abnormalities on CBC or biochemistry (n = 2); abnormalities detected on the abdominal ultrasound (n = 11: 9 cats with significant changes in intestinal wall thickness, 1 cat with a duodenal mass and 1 cat with signs consistent with feline triaditis); volume of collected bile < 1 mL (n = 1); cytology not consistent with bile (n = 1); or positive bile culture with bacterial skin contaminants (n = 2).

The affected groups consisted of 14 client-owned cats: 9 in group 2 and 5 in group 3. White bile was present in 2 cats (1 in group 1 and 1 in group 2). These cats were excluded from the study because cytology slides did not contain any material, so it could not be confirmed that the fluid was bile. The affected cats were significantly older than the healthy cats (P < 0.0001). The median age for the healthy cats was 2.9 y (1 to 5.2 y) and 10.9 y (2.8 to 14 y) for the affected cats. In the healthy group, there were 4 castrated males, 13 spayed females and 4 females, and in the affected group, there were 3 castrated males and 11 spayed females. No statistical difference in sex distribution was observed between the healthy and affected cats (P = 0.45). The clinical signs and findings of the physical examinations most frequently observed in the affected cats were anorexia (n = 13), vomiting (n = 12), lethargy (n = 9), dehydration (n = 8), hyperthermia (n = 5), weight loss (n = 5) and icterus (n = 4).

The calculated coefficient of variation (CV) for pH, PCO₂, HCO₃⁻, Na⁺, K⁺, glucose and lactate for each of the POC or handheld analyzers are listed in Table I. As glucose concentration was undetectable in most of the bile specimens, further data analysis was not carried out. The CV for each of the analyzers was below that of the recommended precision targets on blood, thus demonstrating that the machines were repeatable and reliable for measuring the analytes of interest in bile.

Table I.

Recommended precision target values for blood gas and electrolyte analytes on blood samples compared to obtained coefficient of variation (CV) for point-of-care (POC) analyzers used to assess bile.

	Precision target values (CV%)	i-STAT CV%	EPOC CV%	Lactate Pro CV%
pH	0.5*	0.1	0.2	—
PCO2	4*	4.1	1.7	—
Bicarbonate	—	3.2	1.8	—
Sodium	2*	0.7	0.6	—
Potassium	4*	1.1	0	—
Glucose	2.3†	0	0	—
Lactate	13.6†	—	4.2	5.1

Published target values are not available for bicarbonate.

PCO₂ — partial pressure of carbon dioxide.

^*Reference 16.

^†Reference 17.

The calculated reference values (RVs) and 90% CIs around the reference upper and lower reference value for the EPOC (Element POC) and the handheld lactate analyzer (Lactate Pro) are shown in Table II. Similar calculations were carried out for the i-STAT analyzer (data not shown). It was not possible to establish RVs for chloride and hematocrit as they were not within the measurement range of the POC analyzers. The glucose concentration was beneath the lower limit of detection for the handheld glucometer (AlphaTrak) (< 1.1 mmol/L) for all healthy cats with the exception of 2 cats with a glucose concentration of 1.2 mmol/L and 1.3 mmol/L, respectively. The WCI/WRI ratio was 1.3 to 1.6 for all analytes on the EPOC machine.

Table II.

Reference values (RVs) for bile in healthy cats as determined by EPOC and Lactate Pro analyzers.

Analytes	Units	N	Mean	Med	SD	Min	Max	RV	Lower 90% CI*	Upper 90% CI*	Method	Distribution
pH	—	20	7.61	7.64	0.24	7.16	7.97	7.09 to 8.12	6.95 to 7.27	7.95 to 8.29	P	G
PCO2	mmHg	20	—	16.8	7.3	10.4	33.3	1.5 to 32.1	−3.9 to 6.6	25.9 to 37.7	R	NG
PO2	mmHg	20	128.1	124.2	24.9	92.5	179.7	74.7 to 181.5	59.3 to 92.8	164.4 to 199.2	P	G
HCO3−	mmol/L	19	18.4	17.2	9.1	5.1	35.0	0 to 37.9	−7.0 to 5.6	31.5 to 44.5	P	G
BE	mmol/L	20	−1.7	−2.5	13.9	−23.0	30.0	−31.5 to 28.0	−40.1 to −21.4	18.5 to 37.8	P	G
Na	mmol/L	20	—	99	14	85	131	70 to 128	61 to 79	117 to 140	R	NG
K	mmol/L	20	3.5	3.4	0.4	2.8	4.4	2.5 to 4.4	2.3 to 2.8	4.1 to 4.7	P	G
iCa	mmol/L	20	0.67	0.67	0.16	0.38	0.95	0.32 to 1.02	0.22 to 0.44	0.91 to 1.14	P	G
Lactate§	mmol/L	20	1.4	1.4	0.6	0.7	2.7	0.2 to 2.6	−0.1 to 0.6	2.2 to 3.0	P	G
Creatinine	μmol/L	20	406	422	152	139	672	80 to 732	−14 to 190	628 to 840	P	G
Lactate†	mmol/L	21	5.4	5.4	1.8	2.2	8.4	1.6 to 9.2	0.6 to 2.9	8.0 to 10.4	P	G

N — Number of data available to calculate the reference interval; Med — median; SD — standard deviation; Min — minimum; Max — maximum; PCO₂ — partial pressure of carbon dioxide; PO₂ — partial pressure of oxygen; HCO₃− — bicarbonate; BE — base excess; Na — sodium; K — potassium; iCa — ionized calcium; G — Gaussian; NG — non-Gaussian.

Reference values were established and confidence intervals (CIs) around the upper and lower reference limits were calculated following guidelines of the American Society for Veterinary Clinical Pathology (ASVCP).

^*90% confidence intervals (CIs) of the lower and upper limits of the reference values.

^§Measured with the EPOC analyzer.

^†Measured with the Lactate Pro analyzer.

Correlation statistics for the i-STAT and EPOC analyzers and, the Lactate Pro and EPOC analyzers, are listed in Table III. As the Pearson’s correlation coefficient (r) was not statistically significant for the lactate analyte, no regression was done for this analyte. The Passing-Bablok regression analysis revealed the presence of proportional and constant errors in pH and Na⁺ and only a constant error in HCO₃⁻ and K⁺ analytes. Agreement using Bland-Altman graphs and preset clinically acceptable levels of agreement was assessed for pH (Figure 1), PCO₂, HCO₃⁻, Na⁺, and K⁺ and there was no acceptable level of agreement between the machines for any of these analytes (Table IV).

Table III.

Correlation statistics for comparison of the i-STAT and EPOC analyzers and the Lactate Pro and EPOC analyzers, including the number of comparisons (N), Pearson’s correlation, correlation regression equations (by Passing-Bablok regression technique), slope, and intercept, each with 95% confidence intervals (CIs).

	N	Pearson’s (r)	Regression	Slope	95% CI of slope	Intercept	95% CI of intercept
pH	30	0.943a	1.51 + 0.81x	0.81	0.7 to 0.95	1.51	0.44 to 2.3
PCO2 (mmHg)	29	0.823a	−3.31 + 1.07x	1.1	0.83 to 1.4	−3.31	−10.3 to 1.0
HCO3− (mmol/L)	25	0.96a	0.57 + 1.0x	1.03	0.95 to 1.15	0.57	−0.86 to −1.6
Na+ (mmol/L)	25	0.96a	−70.6 + 1.48x	1.48	1.29 to 1.77	−70.6	−103.8 to −48.6
K+ (mmol/L)	29	0.64b	−1.94 + 1.24x	1.24	0.83 to 2.0	−1.94	−5 to −0.25
Lactate (mmol/L)	31	0.01

^a(P < 0.001).

^b(P = 0.005).

PCO₂ — partial pressure of carbon dioxide; HCO₃⁻ — bicarbonate; Na⁺ — sodium; K⁺ — potassium.

Figure 1.

Bland-Altman percentage difference plots for evaluating the clinical interchangeability of 2 point-of-care (POC) analyzers with the clinical limits of agreement determined by subjective preset limits based on whole blood values for bile pH. The subjective clinically acceptable limits of agreement are graphed around the mean difference of zero (fine dotted line) and are represented by the dashed-dot-dashed black lines.

Table IV.

Method of comparison study between the i-STAT and the EPOC analyzers, including number of comparisons (N), percentage mean difference, standard deviation (SD) of the percentage mean difference, calculated percentage limits of agreement (LOA), subjective a priori clinically acceptable limits of agreement based on published whole blood results, and percentage of results within the clinically acceptable values.

	N	% mean difference	SD mean difference	LOA of %differences	Clinically acceptable target values for agreement (%)	Results within clinically acceptable agreement values (%)
pH	30	−0.84	2	−4.7 to 3.0	1	36.7
PCO2 (mmHg)	29	9.4	24.63	−38.8 to 57.8	15	44.8
HCO3− (mmol/L)	25	−5.6	14.28	−33.69 to 22.30	15	68.0
Na+ (mmol/L)	25	13	6.6	−0.22 to 26	5	20.0
K+ (mmol/L)	25	23	10	2.2 to 43	15	20.0

PCO₂ — partial pressure of carbon dioxide; HCO₃⁻ — bicarbonate; Na⁺ — sodium; K⁺ — potassium.

In group 3, 5 cats met the inclusion criteria: 1 cat showed bactibilia on cytology, 1 cat had a positive culture but no bactibilia on cytology, and the remaining 3 cats had both cytologic bactibilia and a positive culture. Of the latter 3 cats, 2 had degenerated neutrophils present on cytology. Positive culture results identified the following bacteria: Escherichia coli (n = 3), Streptococcus alpha-hemolytic (n = 1), and Enterococcus faecium (n = 1). One cat had multiple bacteria present (E. coli and E. faecium).

While all analyzers yielded results among the healthy cats, this was not the case in the affected cats. For example, it was not possible to measure pH with the i-STAT machine for 4 affected cats (3 in group 2 and 1 in group 3) and with the EPOC machine for 1 cat in group 2. The authors therefore chose to make comparisons among the 3 groups using only the EPOC and lactate analyzers as they subjectively performed better.

On the EPOC analyzer, there was no significant difference among the 3 groups for the pH analytes (P = 0.88) and lactate (P = 0.85) analytes (Figures 2A and 2B, respectively). Furthermore, there were no statistically significant differences among the groups when lactate was analyzed with the handheld lactate analyzer (P = 0.72). Sodium (Na⁺) and PCO₂ were significantly higher in group 3 than in group 1 (P < 0.01) and Na⁺ was significantly higher in group 3 than in group 2 (P < 0.05) (Figures 3A and 3B). Potassium (K⁺) was significantly lower in group 2 than in group 1 cats (P < 0.01) (Figure 3C). The Na/K ratio was significantly lower in group 1 than in group 3 (P < 0.01). Glucose concentration in the bile of affected cats was always < 1.1 mmol/L with the exception of 1 cat in group 3 (3.5 mmol/L) and 1 cat in group 2 (23.7 mmol/L), which were both diagnosed with diabetes mellitus. The diagnosis was based on consistent clinical signs and the presence of persistent hyperglycemia and glucosuria.

Figure 2.

Scatter plots for bile pH (A) and lactate (B) among the 3 groups. These analytes were measured on the EPOC analyzer. Horizontal black line represents the mean and the error bars represent the standard deviation of the mean.

Figure 3.

Scatter plots for bile sodium (Na⁺) (A), partial pressure of carbon dioxide (PCO₂) (B), and potassium (K⁺) (C) among the 3 groups. These analytes were measured on the EPOC analyzer. Horizontal black line represents the mean and the error bars represent the standard deviation of the mean.

* P < 0.01; ^# P < 0.05.

Discussion

The primary objective of this study was to measure several analytes of bile in healthy cats and to investigate whether these analytes had any predictive properties for the early detection of biliary bacterial infection in cats with suspected cholangitis. In order to accomplish this goal, we investigated the use of POC analyzers and handheld lactate and glucose analyzers to assess if they could provide rapid and reliable cage-side results on bile in healthy cats and cats with suspected hepatobiliary disease. All analyzers were deemed repeatable and reliable, with the exception of the glucometer as the measured values for biliary glucose concentration in the majority of cats were below the detection threshold. Subjectively, the EPOC analyzer performed better than the i-STAT analyzer in the affected cat group and was therefore used for the clinical comparisons.

Reference values (RVs) were calculated for all analytes on the EPOC and i-STAT machines; however, the WCI/WRI ratios were higher than those recommended by the ASVCP guidelines. This was expected as it has been demonstrated that, under real conditions, the WCI/WRI < 0.2 requirement cannot be fulfilled for reference sample groups with less than 55 participants (22). Although clinically useful, caution should be taken when applying the RVs presented in this preliminary study to a more generalized population as there is likely significant variability due to the small reference sampling group. Furthermore, when the imprecision of the reference limits is high, it is recommended that the point at which a result may be considered abnormal should not be the reference limit, but instead should be the upper and lower limit of the CI (22).

There was no agreement between the POC analyzers for any of the analytes evaluated and there was no significant correlation between the EPOC and handheld lactate analyzers for the lactate analyte. This lack of agreement is also present in canine blood and therefore it may not be surprising that there were similar results in bile (17). This implies that if bile acid-base balance, electrolytes, or lactate variables are to be monitored, the same analyzer must be used for all these measurements.

Point-of-care (POC) analyzers did not always yield results among affected cats. Clinically, a change in the viscosity of bile was observed in several of the analyzers and we hypothesized that this was the main reason for the inability to obtain results. Indeed, when the viscosity of the bile increases, it may not flow within the cartridge and thereby gain access to the measurement electrodes.

A specific clinical goal of this study was to evaluate pH, glucose, and lactate analytes for their ability to predict the presence of bacteria in the bile of affected cats. There was no difference in lactate concentration among the groups with either the EPOC or handheld lactate analyzer. This could be because these analyzers only detect the L-lactate, whereas bacteria may synthesize both the L-lactate and D-lactate stereoisoforms (23). Considering that D-lactate can only be measured in specialized laboratories (23) and that the clinical purpose of this study was to evaluate bedside tests, this measurement was not taken. The small number of affected cats is also an important limitation in this preliminary study to detect a difference among the groups.

The bile of healthy cats contained little or no glucose, which corresponds to what has been observed in bile from humans, rabbits, and rats. In rats, the intracellular concentration of free glucose in the liver has been found to parallel the plasma concentration (24), but cholangiocytes possess sodium-glucose transport proteins on the apical membrane and glucose transporters on the basolateral membrane that allow glucose from the bile to be reabsorbed to the liver (25). It is therefore possible that cholangiocytes in cats contain comparable channels for glucose reabsorption, although further studies are required to confirm this hypothesis.

An interesting finding in the present study was that glucose was detected in the bile of 2 diabetic cats. In rats and humans, glucose may be present in the bile when there is a supraphysiologic level of glucose in the blood that exceeds the threshold for biliary glucose reabsorption (24,26). There is probably a similar mechanism for the reabsorption of glucose from the bile in cats, although this remains to be determined. Given the preliminary results reported in this study, measuring glucose, pH or lactate in the bile is not useful for predicting the presence of bacteria in the bile of cats.

The pH of bile in healthy cats in this study was 7.6 ± 0.2 (mean ± standard error of mean) and was higher than previously described in other studies: 6.61 ± 0.29 (27) and 6.70 ± 0.14 (28). These differences may be attributed to the different analyzers used or to the duration of pre-sampling fasting time, which was 12 h in our study compared to 24 h in the other studies, as it has been demonstrated that the mucosa of the gallbladder continuously secretes H⁺ (27). Also, the influence of diet on biliary composition in the cat has not been investigated. Before enrolling in this study, healthy cats were all fed the same commercial cat food for 7 d, a time that was chosen arbitrarily. After this study had begun, however, it was demonstrated that either a high-fat/high-cholesterol diet or a low-fat diet for 2 wk induced changes in the composition of gallbladder bile acid and gallbladder motility in dogs (29). Given these results in the dog, a change in the diet for 7 d may not have been sufficient to account for any changes in the bile of cats as a result of diet.

In general, the lower pH in septic effusions is partially due to the production of lactate from neutrophilic glycolysis and bacterial metabolites (10). The fact that the pH was not significantly different among the groups in this study may partially be explained by the lack of significant difference in lactate among the groups or the small number of cats with neutrophils present on cytology.

The partial pressure of carbon dioxide (PCO₂) was significantly higher in group 3 than in group 1. This is in contrast to a study in which a buffer solution was instilled into a cannulated gallbladder and demonstrated a decrease in pH and an increase in PCO₂ in normal feline bile, but no change in pH or PCO₂ in the presence of inflammation (30). These divergent results could be attributed to the fact that the inflammation was not septic in the latter study. The increase of PCO₂ may result from the reaction between H⁺ and bicarbonate ions, yielding CO₂ and water (31), but further studies are warranted to confirm this hypothesis.

Sodium (Na⁺) was significantly higher in group 3 than in groups 1 and 2. There may be a net loss of free water in the bile in the presence of bacterial infection, although this seems unlikely considering that there is a net secretion of water in an inflamed feline gallbladder, whereas there is a net absorption in the normal gallbladder (30). As the Na/K ratio was significantly lower in group 1 than in group 3 in this study, there may have been a change in the function of the Na⁺/K⁺-ATPase pump (31) in the gallbladder wall in the presence of inflammation and/or infection. Some factors that could have contributed to Na⁺/K⁺ pump dysfunction include membrane damage, impaired adenosine triphosphate (ATP) production, or impaired hormone receptors involved in transport control, such as aldosterone. Indeed, it has been demonstrated in rabbits that aldosterone influences sodium excretion in the bile (32,33). The effect of aldosterone on bile composition in cats is unknown and further investigation is recommended.

There are several limitations to this study. Firstly, the diagnosis of hepatobiliary disease was a presumptive diagnosis as liver biopsies were not carried out to definitively confirm the diagnosis histologically. Since only 2 of the 5 cats in the bacterial infected group had neutrophils observed on cytology, a transient bactibilia instead of an infection cannot be totally excluded. However, bile of healthy cats is deemed sterile (2) and none of the healthy cats in the present study had evidence of transient bactibilia. No bacteria were found in the gallbladder of 10 healthy cats in a previous study (34) and only 1 study has demonstrated transient bactibilia in cats after polyethylene tubes and human cholesterol stones were implanted into the gallbladder (35). Furthermore, as point-of-care (POC) analyzers have not been validated for analysis of bile, the influence of bilirubin on the analyte measurements cannot be ruled out. Incidentally, we were not able to obtain a significant number of data points in affected cats, particularly on the i-STAT machine, which further compromised our results and could limit the use of POC analyzers for analyzing bile. In addition, as only 5 cats were recruited into group 3, the absence of statistical differences among the groups could be because the study was underpowered.

In conclusion, this prospective observational preliminary study showed that analytes can be measured in bile with point-of-care (POC) blood gas analyzers and yielded reference values for bile analytes in healthy cats. The POC analyzers used were not interchangeable, but each machine demonstrated acceptable reliability for the measured analytes. The bile of healthy cats contains little to no glucose. There was no difference in pH or lactate concentration among the groups. Sodium was the only analyte that demonstrated significant differences between group 2 and 3. Further studies evaluating the biliary composition of cats are required before sodium concentration can be used to predict bacteria in the bile of cats.