By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Due to essential maintenance the subscribe/renew pages will be unavailable on Wednesday 26 October between 02:00- 08:00 BST/ 09:00 – 15:00 SGT/ 21:00- 03:00 EDT. Apologies for the inconvenience.
Total protein measurement in canine cerebrospinal fluid: agreement between a turbidimetric assay and 2 dye-binding methods and determination of reference intervals using an indirect a posteriori method
Clinical Laboratory, Vetsuisse Faculty, University of Zurich, Zurich, Switzerland
Barbara Riond, Clinical Laboratory, Vetsuisse-Faculty, University of Zurich, Winterthurerstrasse 260, 8057 Zürich, Switzerland
In veterinary clinical laboratories, qualitative tests for total protein measurement in canine cerebrospinal fluid (CSF) have been replaced by quantitative methods, which can be divided into dye-binding assays and turbidimetric methods. There is a lack of validation data and reference intervals (RIs) for these assays.
The aim of the present study was to assess agreement between the turbidimetric benzethonium chloride method and 2 dye-binding methods (Pyrogallol Red-Molybdate method [PRM], Coomassie Brilliant Blue [CBB] technique) for measurement of total protein concentration in canine CSF. Furthermore, RIs were determined for all 3 methods using an indirect a posteriori method.
For assay comparison, a total of 118 canine CSF specimens were analyzed. For RIs calculation, clinical records of 401 canine patients with normal CSF analysis were studied and classified according to their final diagnosis in pathologic and nonpathologic values.
The turbidimetric assay showed excellent agreement with the PRM assay (mean bias 0.003 g/L [−0.26–0.27]). The CBB method generally showed higher total protein values than the turbidimetric assay and the PRM assay (mean bias −0.14 g/L for turbidimetric and PRM assay). From 90 of 401 canine patients, nonparametric reference intervals (2.5%, 97.5% quantile) were calculated (turbidimetric assay and PRM method: 0.08–0.35 g/L (90% CI: 0.07–0.08/0.33–0.39); CBB method: 0.17–0.55g/L (90% CI: 0.16–0.18/0.52–0.61). Total protein concentration in canine CSF specimens remained stable for up to 6 months of storage at −80°C.
Due to variations among methods, RIs for total protein concentration in canine CSF have to be calculated for each method. The a posteriori method of RIs calculation described here should encourage other veterinary laboratories to establish RIs that are laboratory-specific.
Determination of total protein concentration is an integral component of canine cerebrospinal fluid (CSF) analysis. The total CSF protein concentration increases during inflammatory, neoplastic, traumatic, or other disorders that alter the blood brain barrier, or cause changes in the synthesis of immunoglobulins within the central nervous system (CNS).[1, 2]
Several tests were developed to assess qualitative and semiquantitative changes in CSF protein. The Pandy test and Nonne Apelt test detect an increase in globulins. Urinary reagent strips can be used to estimate, but not quantify, CSF protein concentrations between 0.3 g/L and 1 g/L. The test strips are highly efficient for albumin detection, but not for globulins. All 3 tests can be easily and quickly performed in veterinary practice; however, the obtained results are only semiquantitative approximations. Therefore, these assays should only be used for screening, as qualitative tests have been replaced by quantitative methods. Quantitative methods can be broadly divided into 2 main groups: dye-binding and turbidimetric. Dye-binding methods are rapid, simple, economically advantageous, and have been utilized for many years in human and veterinary clinical laboratories. Disadvantages of these methods include the associated variations in binding to different proteins as well as the frequent measurement of nonprotein nitrogenous substances. The Coomassie Brilliant Blue (CBB) technique was established by Marion Bradford for use in research laboratories and was later adapted for clinical laboratories. The Pyrogallol Red-Molybdate (PRM) method has been developed for protein determination in urine specimens; this assay yields a more unique response to albumin and globulin. In addition, this method has a lower tendency for dye precipitation in cuvettes and automated systems than the CBB technique, and is therefore preferable for total protein determination in CSF. Precipitating agents are also utilized for CSF protein detection; these latter ones form protein precipitates that can be quantified by turbidimetry. In human CSF specimens, benzethonium chloride forms stable precipitates with slow formation kinetics. The resulting turbidity lasted longer with this method than with the trichloroacetic acid or sulfosalicylic acid techniques.[8, 9] Turbidimetric assays are commonly used in human medicine. Consequently, many veterinary clinical laboratories use the turbidimetric assay for total protein determination in CSF.
According to the recommendations of the International Federation of Clinical Chemistry (IFCC), each laboratory should establish reference intervals (RIs). However, only a few clinical laboratories produce such RIs, and typically only for a few analytes. The difficulty in selecting reference individuals as well as the frequent change in methods and analyzers renders the determination of RIs difficult, time-consuming, and expensive. Therefore, in many laboratories, RIs are used from scientific or commercial literature, even if these values are not readily transferable. The situation in veterinary clinical laboratories might be comparable or even worse due to the wide variety of species and breeds. The latest CLSI (Clinical and Laboratory Standards Institute) guideline addressed this issue and stated that it is unrealistic to expect each laboratory to develop its own RIs. Generally, RIs can be obtained by several approaches. They can be determined de novo from measurements made in reference individuals. When it is too difficult to apply a de novo procedure, preexisting RIs can be transferred, or previously established or transferred RIs can be validated. The most often performed de novo determination of RIs can be achieved by using an a priori approach in which reference individuals are selected according to predefined criteria. As an alternative to the a priori determination of RIs, an a posteriori approach can be applied in which preexisting laboratory and clinical data are exploited to establish RIs.[13, 15] The same preanalytical, analytical, and selection factors should be applied as for a de novo determination of RIs. The only difference is that the selection of reference individuals is made after analysis has been performed.
In this regard, the determination of RIs for CSF represents a unique problem. Collection of CSF by lumbar or atlanto-occipital puncture from healthy animals is generally not justified for animal welfare reasons. However, RIs for canine total protein concentration in the literature are relatively variable and reflect 2 major problems. These values are method-dependent and the population used for RIs determination is often poorly defined. In addition, RIs used were often reported without information regarding the method and the number of dogs studied.[16, 17]
The present study pursued 2 aims. First, agreement between the turbidimetric benzalkonium chloride method and 2 dye-binding methods (PRM method and CBB technique) for measurement of total protein in canine CSF was determined. Furthermore, imprecision of the turbidimetric assay was assessed. Second, RI were determined for total protein concentration in canine CSF for the turbidimetric method, the PRM method, and the CBB technique using an indirect a posteriori method, according to the IFCC/CLSI guidelines[12, 13] and the ASVCP Quality Assurance and Laboratory Standards Committee (QALS) Guidelines for the Determination of Reference Intervals in Veterinary Species.
Materials and Methods
Materials for assay comparison
A total of 118 canine CSF specimens were analyzed. Total protein concentrations of the 118 canine CSF specimens ranged between 0.02 g/L and 2.35 g/L. CSF specimens were collected by atlanto-occipital puncture from dogs presented to the Clinic for Small Animal Surgery at the Vetsuisse Faculty, University of Zurich (2006–2007). CSF was collected into sterile plastic tubes without anticoagulants (Milian SA, Geneva, Switzerland). The basic diagnostic CSF analysis for each sample included macroscopic evaluation, WBC and RBC counts, as well as cytologic evaluation of a cytospin specimen. Macroscopic evaluation of the CSF consisted of the color of the CSF and the color of the supernatant as well as the transparency. A Fuchs-Rosenthal hemocytometer was used to determine total WBC count by diluting the native CSF with Samson's Reagent (Kantonsapotheke Zürich, Zurich, Switzerland) 1 : 10. RBC counting was performed with undiluted CSF in a Neubauer-improved hemocytometer. Cytological evaluation was performed after cytocentrifugation (Shandon Cytospin 4, Thermo Electron Corporation, Allschwil, Switzerland) of the sample, followed by Wright staining (Hema Tek 1000; Siemens Healthcare, Zurich, Switzerland). Total protein was measured after centrifugation (5 minutes at 390g in a Hettich Rotina 35R centrifuge) in the CSF supernatant. The turbidimetric assay was performed within one hour after CSF collection. The remaining CSF supernatants were stored at −80°C for a maximum of 6 months for the PRM and CBB analyses. At the time of analysis, CSF specimens were gently thawed, homogenized by vortexing, batched, and assayed. Furthermore, the influence of storage time on total protein concentration in canine CSF was assessed. In total, 52 canine CSF supernatants, stored at −80°C after the day of collection, were gently thawed and total protein concentrations were determined a second time using the turbidimetric assay on a Cobas Integra 800 instrument (Roche Diagnostics, Rotkreuz, Switzerland). Eighteen of the 52 canine CSF supernatants were stored for 2 months, 15 specimens for 4 months, and 19 specimens for 6 months at −80°C.
Materials for RI calculation for the turbidimetric assay
Reference sample group selection for RIs calculation was performed in 2 steps. First, 401 results for total protein in CSF from dogs having normal CSF analysis were picked out from the databank of the Laboratory Information System. Normal CSF analysis was defined as CSF with RBC counts < 10/μL, WBC count < 5/μL, no abnormalities in macroscopic (crystal clear, colorless) and cytologic examination (only normal lymphocytes and normal monocytes present). All 401 CSF specimens were collected from the same site (atlanto-occipital puncture), and underwent the same handling procedure. CSF analysis was performed as described in materials for assay comparison. Total protein concentration was determined with the turbidimetric assay in the supernatant within one hour after collection. These dogs underwent complete clinical and neurologic examination and CSF puncture at the Clinic for Small Animal Surgery at the Vetsuisse Faculty, University of Zurich (2006–2007); in addition, a CBC and clinical chemistry analyses were performed. In a second step, clinical records of these 401 patients were studied, and dogs were classified according to their final diagnosis into 2 groups. A detailed list of conditions that excluded certain canine protein values from the reference calculation was determined to accurately describe the reference population as well as to differentiate between pathologic and nonpathologic results. Dogs with neurologic disorders, an abnormal myelogram, or computed tomography findings (conditions listed in Table 1) were excluded. The exclusion criteria were established with a board-certified veterinary neurologist (ECVN). Reference intervals for the PRM and CBB methods were determined by transferring the RIs determined for the turbidimetric assay using Passing–Bablok equations.
Table 1. List of CNS conditions that excluded dogs from reference sample group.
Benign tumors or masses
Inflammatory disorders (eg, Meningitis)
Intrathecal drug or radiocontrast administration within the last 48 h
Malignancies, primary or secondary
Trauma of any cause
A commercial kit that used Benzethonium chloride as the precipitating agent was utilized (Total Protein Urine/CSF Gen.3; Roche Diagnostics) on an Integra 800 instrument (Roche Diagnostics). A human calibrator for CSF proteins consisting of albumin and immunoglobulin G (C.F.a.s. PUC; Roche diagnostics) was used for determination of standard concentrations. To monitor accuracy and precision, commercially available controls with low and high protein concentration were included on a daily basis (Precinorm PUC, Precipath PUC; Roche Diagnostics).
Within-series precision was determined by performing 12 tests on 3 canine CSF specimens with low (mean 0.13 g/L), intermediate (mean 0.34 g/L), and high (mean 2.55 g/L) protein concentrations. Day-to-day precision was determined by testing total protein concentration in 2 canine CSF specimens with low (mean 0.17 g/L) and high (mean 1.37 g/L) protein concentrations on 5 consecutive days. CSF specimens were stored at 4°C between measurements. Maximal allowable imprecision for total protein concentration in canine serum samples is indicated as 1.3%. For internal quality control, 2 commercially available specimens with low (Precinorm PUC) and high (Precipath PUC) protein concentrations were tested on a daily basis.
A commercial test kit for the determination of urinary proteins was used (ABX Pentra Urinary Proteins CP, Horiba Group, Montpellier, France). The test was performed with a Cobas Mira instrument, type S (Roche diagnostics). A bovine serum albumin protein standard served as the calibrator (Sigma-Aldrich Inc., St. Louis, MO, USA), and a urine control was used with high and low protein concentrations (ABX Pentra Urinary Proteins CP, Horiba Group).
The CBB technique was performed according to the Bradford protocol with Bradford Reagent (Sigma-Aldrich Inc.). The test was performed with a Cobas Mira instrument, type S (Roche diagnostics). Bovine serum albumin was utilized as a protein standard (Sigma-Aldrich Inc.), and a urine control was also used with high and low protein concentrations (ABX Pentra Urinary Proteins CP, Horiba Group).
All results were compiled in a table-calculation program (Analyse-it for Microsoft Excel, v. 2.09; Analyse-it-Software, Ltd., http://www.analyse-it.com). Spearman coefficient of correlation (r), the intercept and slope with 95% confidence intervals (CI) calculated by Passing–Bablok regression, and Bland–Altman biases with 95% limits of agreement were reported. Correlation was considered excellent if r ≥ .95, very good if r = .90–.94, good if r = .80–.89, fair if r = .59–.79, and poor if r < .59. Normality was assessed by visual inspection of the normality plot and the Shapiro–Wilk test. Imprecision was calculated as the coefficient of variation (CV = standard deviations [SD]/mean). RIs were determined nonparametrically for the 2.5%, and 97.5% quantiles. The CIs of the limits of the RI were determined using a bootstrap method (Reference Value Advisor V 1.3 [available as freeware at http://www.biostat.envt.fr/spip/spip.php?article63]). A Mann–Whitney U-test was used to assess gender- and age-related differences in the reference population. A multiple t-test was used to assess the effect of storage on total protein concentration in canine CSF. P values < .05 were considered significant.
Storage of canine CSF supernatants for up to 6 months at −80°C had no statistically significant influence on total protein concentration. Mean differences between measurements before and after storage were only 0.7%, 0.8%, and 2.6%. After 2 months of storage, mean total protein concentration decreased from 0.384 g/L to 0.374 g/L, after 4 months from 0.389 g/L to 0.374 g/L, and after 6 months from 0.328 g/L to 0.326 g/L (Table S1). Spearman's coefficient of correlation r, intercept and slope with 95% confidence intervals calculated by Passing–Bablok regression analysis, and Bland–Altman biases with their 95% limits of agreement for canine total protein in CSF are tabulated in Table 2. Bland–Altman difference plots and linear regression analysis by Passing–Bablok are presented in Figures 1-3. The turbidimetric assay and the PRM method showed excellent correlation (r = .97), excellent regression analysis (y = 1x + 0) and a small positive bias (0.003 g/L). Despite the slightly better correlation between the turbidimetric assay and the CBB (r = .98), the turbidimetric assay underestimated CSF protein concentration due to a constant systematic error (mean bias −0.139 g/L). Also the PRM method underestimated protein concentration compared with the CBB assay due to a constant systematic error (mean bias −0.142 g/L). Two outliers could be identified, which were not excluded from statistical analysis. In one case, the turbidimetric method showed an extremely lower value than the dye-binding methods. In the other case, the turbidimetric assay revealed a higher result than the PRM method and the CBB assay. Results from the precision study are shown in Table 3. The turbidimetric assay demonstrated in the middle and low range imprecision data below the maximal acceptable imprecision for measurement of total protein in canine serum specimens, whereas in the canine CSF specimen with a low protein concentration (mean 0.17 g/L), imprecision was higher than 1.3%.
Table 2. Agreement between the turbidimetric assay and the Pyrogallol Red-Molybdate (PRM) method, between the turbidimetric assay and the Coomassie Brilliant Blue (CBB) method, and between the PRM method and the CBB assay for total protein concentrations in canine cerebrospinal fluid (CSF).
Spearman Coefficient of Correlation r
Intercept with 95% CI
Slope with 95% CI
Bias (g/L) with 95% Limits of Agreement
Turbidimetric Assay/PRM Method
Turbidimetric Assay/CBB Assay
−0.06 (−0.07 to–0.04)
PRM Method/CBB Assay
−0.06 (−0.07 to–0.04)
Table 3. Day-to-day and within-series precision of the turbidimetric assay for canine cerebrospinal fluid (CSF) total protein, tested in canine CSF specimens with low, middle (at medical decision limit), and high total protein concentration (g/L).
Day-to-Day Precision (n = 5)
Within-Series Precision (n = 12)
Mean ± SD (g/L)
Mean ± SD (g/L)
0.17 ± 0.005
0.13 ± 0.004
0.34 ± 0.003
1.37 ± 0.015
2.55 ± 0.014
Of 401 dogs with normal CSF analysis, 311 were excluded from the reference calculation due to the presence of CNS conditions listed in Table 1. Thus, reference population for RIs calculation consists of 90 canine CSF specimens. Ages of the 90 dogs ranged from 2 months - 15 years with a median of 5.0 years. Seventy-one percent of the dogs were between 1 and 9 years – age, 10 dogs were < 1 year, and 16 dogs > 10 years. Gender distribution was 60 (67%) male reference individuals and 30 (33%) female reference individuals. No significant sex- or age-related differences were found among the 90 dogs included in the RIs calculation. The reference population consisted of 43 different canine breeds. A detailed breed distribution is shown in Table S2. The Shapiro–Wilk test indicated that canine CSF total protein concentration is not normally distributed in the reference population; a frequency histogram is shown in Figure 4. Therefore, the 2.5%, and 97.5% quantiles for the turbidimetric assay, PRM method, and CBB method were calculated instead of using the mean and SD to characterize the 95% reference interval (Table 4).
Table 4. Nonparametric reference interval (RI) for total protein concentration (g/L) in canine cerebrospinal fluid (CSF), determined from 90 dogs.
2.5%-Quantile (90% CI)
97.5%-Quantile (90% CI)
Total CSF protein determination in veterinary clinical laboratories is usually performed in the context of a daily routine diagnostic on an automated chemistry analyzer using assays derived from human medicine. Several automated qualitative methods for measurement of total protein concentration in CSF are available. For adaption to veterinary use, methods have to be validated and RIs to be established for each individual species. In the present study, the turbidimetric assay using benzethonium chloride as the precipitating agent showed high agreement with the PRM assay. The CBB method generally yielded higher total protein values in the analyzed canine CSF specimens compared with the turbidimetric assay and PRM method, particularly in the lower concentration range. Disagreement between the CBB method and the turbidimetric assay as well as between the CBB method and the PRM method have been already shown in a study with human urine samples, but not CSF. The bias found in the present study might result from the fact that dye-binding methods are affected by variations in binding to different proteins. Coomassie Brillant Blue might have a higher affinity to aromatic amino acids in the sample than Pyrrogallol Red, thus yielding colored products with higher absorbances. Dye-binding methods have also been demonstrated to measure nonprotein nitrogenous substances, resulting in an overestimation of total protein concentration in the CSF.
The choice of standard reference material has been reported to be crucial for avoiding variations in results.[4, 23] One limitation of this study is the use of different calibrators. A human albumin/globulin calibrator was used for the turbidimetric assay, while only an albumin standard was used for the PRM assay. Nevertheless, both tests demonstrated excellent agreement. Although the same calibrator (bovine albumin) was used for the PRM and CBB methods, these methods exhibited a low correlation and a negative bias. Despite lack of proof, it is likely that the biases were largely due to differences in the reactivity of the methods with different proteins rather than the standard and equipment used.
Inconsistency in preanalytical treatment of the specimens used for assessing agreement among the 3 methods could be the cause of a bias in the present study. Results obtained by the turbidimetric assay were determined in fresh CSF specimens, whereas results with the PRM assay and CBB assay were measured after storage for up to 6 months at −80°C. However, the results of the stability study clearly indicated that storage of canine CSF for up to 6 months at −80°C had no statistically significant influence on total protein concentration in canine CSF.
The turbidimetric assay yielded low imprecision at a low concentration of proteins, suggesting that this assay is highly reproducible and could allow clear discrimination between normal and pathologic specimens. Although the maximal allowable imprecision was exceeded at the lowest range (mean 0.17), imprecision at the clinical decision limit (mean 0.34) and in the higher range was within the acceptable limit. The turbidimetric assay provided results within 5 minutes, resulting in a short turnaround time that could avoid delays in the diagnosis and treatment of CNS disorders.
The establishment of RIs for total protein in canine CSF is particularly difficult, time-consuming, and costly. Moreover, collecting CSF specimens from healthy dogs raises ethical concerns. Consequently, few laboratories collect such data, and most laboratories use RIs reported in the literature. These RIs for CSF total protein vary greatly depending on the site of collection, the method used, and the dog population used as reference individuals. In addition, a small reference sample group widens the confidence limits. Therefore, the present study aimed to determine RIs for total protein in canine CSF using the indirect a posteriori approach. Preexisting data from a databank were used; inclusion and exclusion criteria were applied retrospectively to separate, as efficiently as possible, healthy from unhealthy dogs. Preanalytical, analytical, and biological factors causing variations for total protein concentration in canine CSF were taken into account. All CSF specimens were collected from the same site (atlanto-occipital puncture), and underwent the same handling procedure. The a posteriori approach for RIs determination has already been applied in human medicine[11, 25] and also in veterinary clinical pathology.[15, 26] It has to be considered that use of an indirect a posteriori approach for determination of RIs can introduce several biases. The use of retrospective data is prone to the introduction of verification bias, as only verified cases are excluded. In addition, diagnostic modes as CT-scanning and magnetic resonance imaging were not applied in all 401 investigated cases. It has been reported that the indirect a posteriori approach is less reliable when distributions have skewness of 13 and higher and/or coefficients of kurtosis of 240 and higher. For total protein concentrations in canine CSF, these values were far below these cutoff values (skewness 1.5; kurtosis 2). For the analyte investigated in the present study, the indirect a posteriori approach provides many advantages over the conventional de novo determination of RIs. The procedure is much easier and less expensive than collection of CSF from healthy individuals. Extensive digitized data are available in the Laboratory Information System; thus, this method is practical for a clinical laboratory. Furthermore, these RIs are derived from values obtained using the same method and the same preanalytical conditions to allow comparison with patient results. The calculated RIs for the turbidimetric assay in this study were slightly lower than those established in a previous study of 20 dogs. To our knowledge, such RIs have not been previously reported for the PRM method. The RIs were nearly in the same range due to high agreement between the turbidimetric assay and the PRM method. RIs for the CBB method reported in previous studies[28, 29] yielded lower values than the values reported in this study.
Due to variations among methods, RIs for total protein in canine CSF have to be calculated for each method.
RIs for total protein concentrations in canine CSF were established for the first time using an indirect a posteriori approach for the turbidimetric assay, the PRM method, and the CBB method. The technique of RIs calculation described herein should encourage other veterinary laboratories to establish RIs that are laboratory specific.
We thank B. König, E. Rogg, E. Schuler, Y. Bosshart, U. Egger, E. Grässli, M. Huder, B. Lange, J. Wälchli, M. Maurer, and Y. Gahler for excellent laboratory assistance. R.H.-L. is the recipient of a professorship from the Swiss National Science Foundation (PP00B-102866 and PPOOP3-119136).
Disclosure: The authors have indicated that they have no affiliations or financial involvement with any organization or entity with a financial interest in, or in financial competition with, the subject matter or materials discussed in this article.