Dllena, penberthy, Fraser compare six commonly used methods for protein determination. The sulfosalicylic acid / biuret technique is not linear at low concentrations of urine protein. The Coomassie Brilliant Blue technique has a narrow range of linearity and poor precision.
Original Description:
Original Title
Six Methods for Determining Urinary Protein Compared
Dllena, penberthy, Fraser compare six commonly used methods for protein determination. The sulfosalicylic acid / biuret technique is not linear at low concentrations of urine protein. The Coomassie Brilliant Blue technique has a narrow range of linearity and poor precision.
Dllena, penberthy, Fraser compare six commonly used methods for protein determination. The sulfosalicylic acid / biuret technique is not linear at low concentrations of urine protein. The Coomassie Brilliant Blue technique has a narrow range of linearity and poor precision.
CLINICALCHEMISTRY, Vol. 29, No. 3, 1983 553 Six Methods for Determining Urinary Protein Compared Beverly A. Dllena, Lloyd A. Penberthy, and Callum G. Fraser Inter-laboratory surveys have shown that routine methods of urinary protein determination are often unsatisfactory. Therefore, we compared six frequently used methods for determination of protein in urine with respect to linearity, within-batch and between-batch precision, comparative bias, and practicability. We assayed dilutions of human and bovine albumin and serum, and fresh and lyophilized human urine. We find that the AACC Selected Method has poor practicabil- ity and poor precision under routine conditions, but good linearity. The sulfosalicylic acid/biuret technique is also im- practicable, requires a large sample, and is not linear at low concentrations of urine protein. The Coomassie Brilliant Blue technique has a narrow range of linearity and poor precision. The sulfosalicylic acid/sodium sulfate turbidimetric method is not precise and cannot be standardized with bovine materi- als. The Ponceau-S technique has good performance char- acteristics and practicability, and we recommend it for routine laboratory use. Inter-laboratory quality-assurance surveys of the stan- dard of performance of urinary protein analyses in the United States (1-4) and in Australia (5-7) have shown that the biases within and between methods are high; the range between low and high values for the same specimen varied from normal to distinctly abnormal. In addition, many techniques have poor intra-laboratory precision. It has been suggested (1) that disease states may well go undetected because of this poor performance; in addition, healthy individuals may also be labeled as diseased. Even though a Selected Method for the quantitation of urinary protein (8) and an Australian recommended method (5) have been published, no single method has been widely adopted by laboratories. Recently, McElderry et al. (9) compared the linearity, accuracy, precision, sensitivity, and specificity of six methods: the Coomassie Brilliant Blue method in two commercial forms, Ponceau-S dye binding, ferric chloride/tannic acid turbidimetry, turbidimetry with benzethonium chloride in alkali, and trichloroacetic acid! biuret colorimetry. However, many other methods are wide- ly used by laboratories. We therefore report here our studies in which we have compared the Selected Method of the American Association for Clinical Chemistry (8), the recom- mended method of the Royal College of Pathologists of Australia (5), the manual Coomassie Brilliant Blue tech- nique of Heick et al. (10), sulfosalicylic acidlsodium sulfate turbidimetry (11), trichloroacetic acid turbidimetry (12), and Ponceau-S dye binding (13). For each method we assessed linearity, within-batch and between-batch preci- sion, comparative bias, and practicability. Materials of animal origin, materials primarily designed for use for serum analyses, and lyophilized materials are widely used both as calibration materials and in intra- and inter-laboratory quality-control and assurance programs. Because such materials may not be truly representative of Department of Clinical Biochemistry, Flinders Medical Centre, Bedford Park, South Australia, 5042. Received Aug. 9, 1982; accepted Dec. 23, 1982. specimens from patients (14), we therefore examined the precision and bias of the six methods with a variety of such materials. Our aims were to investigate the possible reasons for the documented poor performance of six frequently used meth- ods and to recommend a single method that would be suitable for wide use in the clinical laboratory. Materials and Methods Materials Bovine serum albumin (Cohn Fraction V powder) and human serum albumin (crystallized and lyophilized) were obtained from Sigma Chemical Co., St. Louis, MO 63178. Stock 35 g/L solutions were prepared and, after determina- tion of total protein by the biuret method with a Technicon 6-Plus AutoAnalyzer (Technicon Instruments Corp., Tarry- town, NY 10591), with use of Ortho Automated Reference Serum (Ortho Diagnostics Inc., Raritan, NJ 08869) as standard, solutions in saline (9 g of NaC1 per liter) of approximately 0.25 and 0.50 g/L were prepared. Aliquots were stored at -20 #{176}C. Fresh bovine serum was obtained by centriftigation of freshly clotted blood obtained at slaughter; fresh human serum was obtained by centrifugation of freshly clotted blood obtained from apparently healthy laboratory person- nel. After determination of total protein by the biuret method, both sera were diluted with saline to prepare solutions of approximately 0.25 and 0.50 g/L. Aliquots were stored at -20 #{176}C. Monitrol Ii was supplied by Dade Division, American Hospital Supply Corp., Miami, FL 33152; and Weilcome Unassayed bovine serum material from Wellcome Reagents Ltd., Beckenham, Kent, U.K. Vials of each were reconstitut- ed each day and dilutions to approximately 0.25 and 0.50 g of protein per liter were prepared in saline. The primary clinical standard used throughout this study was a human urine to which a known amount of human serum albumin had been added. This material was specially supplied by Commonwealth Serum Laboratories, Parkville, Victoria, 3052, Australia, and had been prepared from a pool of random urine samples collected from apparently healthy males. Except for protein, the base pool and the standard did not have any tests positive on qualitative urinalysis, and ascorbate concentrations were undetectable. The concentration of protein in the base pool was not significantly different from the detection limits of the meth- ods used, and the relative density was 1.015. The standard had an assigned value (by weight) of 0.50 g of protein per liter. We assigned concentrations of protein in each of the materials used in this study by using this standard and the Selected Method (8); analyses were performed in 10 separate analytical batches. Ortho Control Urine II was supplied by Ortho Diagnostics Inc., Raritan, NJ 08869. Several vials were reconstituted, the material was pooled, and aliquots of it were stored frozen at -20#{176}C. Fresh human urine samples with low, medium, and high protein concentrations were obtained from patients in the Flinders Medical Centre. These were analyzed by use of the Selected Method (8) in duplicate on three occasions. Values of 0.22, 2.48, and 20.40 g/L were obtained. The latter two urines were diluted fourfold and 20-fold, respectively, with saline. Aliquots of these dilutions and of the undiluted urine were stored at -20 #{176}C. All other reagents used were as stated in the published methods or were equivalent Analar grade. The performance characteristics of the methods with materials stored at -20 #{176}C were not significantly different from those achieved with similar fresh materials. Methods All absorbances were measured in 1-cm cuvettes with a Model 300-Ti spectrophotometer (Gilford Instrument Labs. Inc., Oberlin, OH 44074). Samples were transferred with Eppendorf pipettes. In addition, in separate experiments, the Selected Method, as recommended (8), was performed with use of glass volumetric pipettes throughout. The six procedures examined were done exactly as de- scribed in the references except for minor modifications to the Selected Method (8). Disposable syringes (Terumo Corp., Tokyo, Japan) of 5-mL capacity were used as chromatogra- phy columns. A single glass bead was used as the bed support. In this study, the Selected Method when performed with glass pipettes is termed the Selected Method, and when performed with air-displacement pipettes is termed the Selected Method (routine). Results and Discussion Analytical Variables Linearity. Linearity was assessed by duplicate analysis, in each of two batches, of at least nine samples of different concentrations prepared by dilution of the materials recom- mended as standard with saline. In addition, linearity was Table 1. Maximal Useful Limit of Linearity with Samples Used for Assessment of Linearity and with Urine Maximal limit of linearity, g/L Method Selected Method SSA/biuret Coomassie Brilliant Blue SSA/sodium sulfate TCA turbidimetry Ponceau-S Sample Human albumin Human serum Human albumin Human serum Human serum Ortho serum Tables1-5: SSA, sulfosalicylic acid;TCA, trichloroacetic acid. 554 CLINICALCHEMISTRY, Vol. 29, No. 3, 1983 assessed by duplicate analysis, in each of two batches, of nine dilutions with normal human urine of a human urine with a high protein concentration. Table 1 shows the samples and the maximal useful limits of linearity. The Selected Method showed the widest range of linear- ity. The trichloroacetic acid turbidimetry technique had the narrowest useful range of linearity; use of this method would entail dilution and reanalysis for a substantial pro- portion of specimens from patients. We assayed, in duplicate, six samples of urine to which albumin was added in various amounts up to a concentra- tion of 2.50 g!L. The maximum limit of linearity with these samples was identical to that found by the corresponding method for urines containing protein. Within-batch precision. For each method, within-batch precision was assessed by analysis in duplicate of thawed aliquots of the three human urine specimens, with low, medium, and high protein concentrations, in 10 batches. The standards recommended for each method were used. These had values assigned by the Selected Method. Values for precision calculated by use of the paired results, are shown in Table 2, both as standard deviation (SD) and coefficients of variation (CV). When the precision of the Selected Method was used as the comparison point for application of the F-test, all meth- ods had similar precision at a low urinary protein concentra- tion (0.22 gIL). At a medium (0.62 gIL) concentration of urinary protein, only the sulfosalicylic acid/sodium sulfate method was significantly more precise than the Selected Method, whereas the Selected Method (routine) had signifi- cantly poorer precision than the Selected Method. At a high (1.02 g/L) concentration of urinary protein, all methods except the Coomassie Brilliant Blue technique had signifi- cantly better precision than the Selected Method and, again, the Selected Method (routine) had the poorest precision. The sulfosalicylic acid/biuret, sulfosalicylic acid/sodium sulfate turbidimetry, and trichloroacetic acid turbidimetry tech- niques were all precise for all concentrations examined. Between-batch precision. Between-batch precision was assessed by 10 replicate analyses, in separate batches, of each of the eight calibration materials; each material was studied at concentrations of approximately 0.25 and 0.50 g/ L. In addition, two urine samples from patients, with protein concentrations of 0.22 and 0.62 g/L, were analyzed 10 times. Sample Urine The precision for each method, as SD and CV, calculated 4.0 4.0 with the primary standard used in this study, is shown in 3.5 3.5 Table 3. 1.0 0.7 The conclusions we drew from this study were: 1 6 1 6 #{149} The poorest overall precision was found with Coomassie 0:6 0.6 Brilliant Blue technique and sulfosalicylic acid/sodium sul- 1.6 1.6 fate turbidimetry. #{149} There did not appear to be any direct relationship be- tween the precision achieved and the type of material Method Table 2. WIthin-Batch Rd Imprecision ative protein concn of sample Low Medium High SD, g/L CV, % SD, g/L CV, % SD, g/L CV, % Selected Method 0.013 5.9 0.023 3.7 0.035 3.4 Selected Method (routine) 0.015 6.8 0.043 6.9 0.064 6.3 SSA/biuret 0.011 5.0 0.0118 1.8 0.0198 1.9 Coomassie Brilliant Blue 0.015 6.8 0.022 3.5 0.030 2.9 SSA/sodium sulfate 0.008 3.6 0.016 2.6 0.01 48 1.4 TCAturbidimetry 0.014 6.4 0.016 2.6 0.0148 1.4 Ponceau-S 0.011 5.0 0.020 3.2 0.0148 1.4 8Precision significantly better than that obtainedby the Selected Method(p 0.05). Method Material 0.089(17.8) 0.080(15.4) 0.064(12.8) 0.038(8.3) 0.065(13.8) 0.065(13.8) 0.071(15.1) 0.065(18.6) 0.063(10.2) CLINICAL CHEMISTRY, Vol. 29, No. 3, 1983 555 Table 3. Between-Batch Imprecision (SD in g/L, CV % in Parentheses) for Eight Calibration Materials at Two Different Concentrations and for Two Urine Samples from Patients Bovine Human Bovine Human Dade Ortho Wellcome Ortho Human albumin albumin serum serum serum serum serum urine urine Selected method 0.039(13.9) 0.027(10.8) 0.045(20.5) 0.036(16.4) 0.024(10.9) 0.037(16.1) 0.041(18.6) 0.026(15.3) 0.028(12.7) (routine) SSA/biuret 0.021(9.1) 0.021(8.4) 0.035(16.4) 0.026(11.8) 0.044(20.0) 0.028(12.2) 0.030(13.0) 0.036(21.2) 0.051(23.2) 0.041(8.2) 0.043(8.3) 0.058( 11.6) 0.055( 12.0) 0.051(10.9) 0.041(8.7) 0.036(7.7) 0.025(7.1) 0.061(9.8) Coomassiebrilliant 0.023(10.0) 0.041(16.4) 0.050(22.7) 0.047(21.4) 0.024(10.9) 0.042(18.3) 0.034(15.5) 0.048(28.2) 0.037(16.8) blue 0.049(9.8) 0.046(8.8) 0.046(9.2) 0.052(11.3) 0.030(6.4) 0.030(6.4) 0.032(6.8) 0.049(14.0) 0.087(14.0) SSA/sodium sulfate 0.047(20.4) 0.028(11.2) 0.043(19.5) 0.038(17.3) 0.028(12.7) 0.027(11.7) 0.026(11.8) 0.044(25.9) 0.040(18.2) 0.100(20.0) 0.047(9.0) 0.085(17.0) 0.060(13.0) 0.077(16.4) 0.047(10.0) 0.070(14.9) 0.060(17/1) 0.070(11.3) TCA turbidimetry 0.022(9.6) 0.018(7.2) 0.043(19.5) 0.025(11.4) 0.019(8.6) 0.012(5.2) 0.031(14.1) 0.012(7.1) 0.024(10.9) 0.050(10.0) 0.026(5.0) 0.058(11.6) 0.047(10.2) 0.042(8.2) 0.029(6.9) 0.046(9.8) 0.047(13.4) 0.046(7.4) Ponceau-S 0.026(11.3) 0.022(8.8) 0.039(17.7) 0.012(5.5) 0.015(6.8) 0.011(4.8) 0.022(10.0) 0.025(14.7) 0.027(12.3) 0.052(10.4) 0.020(3.8) 0.057(11.4) 0.024(5.2) 0.016(3.4) 0.024(5.1) 0.027(5.7) 0.030(8.6) 0.036(5.8) analyzed; this implies that use of frozen or lyophilized, bovine or human albumin, serum or urine would be, in general, satisfactor.y as quality-control materials to monitor precision. #{149} The sulfosalicylic acid/biuret, trichloroacetic acid turbi- dimetry, and Ponceau-S methods had the best overall preci- sion. #{149} The Selected Method had significantly better between- batch precision than the Selected Method (routine). We calculated the precision found when using the stan- dards recommended for the methods. The Coomassie Bril- liant Blue technique had better precision when human albumin diluted in saline was the standard, but the preci- sion of the sulfosalicylic acid/sodium sulfate turbidimetry technique was poorer with human serum diluted with saline as the standard. We consider it likely that, as shown previously in our laboratory (17, 18), detailed examination of mode of standardization and type and concentration of standard by the individual laboratory could significantly improve the precision of the method chosen by that labora- tory. Comparatiue bias. Peters (19) has suggested that the method of choice for measurement of protein should be a version of the biuret reaction and that bovine or human serum albumin, assayed by Kjeldahl analysis, should be used for standardization. However, the composition of uri- nary proteins can vary very widely (20), a primary standard for such a complex and variable mixture of components cannot truly be defined, and neither definitive nor reference methods exist. It is therefore difficult at this time to define the accuracy of urinary protein methods in strict numerical terms as deviation from the true value. In this study, we assigned values to all albumin and serum-based materials and urine samples by using the Selected Method because such a method has been thorough- ly evaluated and verified by independent workers before being accepted. Human albumin weighed in to a normal urine obtained from healthy men was used as the standard. The means of 10 replicate analyses on each of eight calibra- tion materials at concentrations of approximately 0.25 and 0.50 g/L and on two urine samples are shown in Table 4. The same technique and standard having been used to assign values, we believe that the data obtained are valid for assessing comparative bias among methods and materials. The results obtained for the different proteins showed, in general, little difference between methods. This implies that, for most methods, human or bovine albumin, serum, or urine could be used for standardization and quality control, provided that the protein concentration was adequately assessed. However, for the sulfosalicylic acid/sodium sulfate meth- od, results for bovine serum albumin solutions were falsely increased; therefore, bovine serum albumin should not be used as a calibration material for this method, and material supplemented with bovine serum albumin should not be used in inter-laboratory quality-assurance surveys. Human albumin has also been reported to give increased results in this method, but the composition of the precipitating rea- gent appears to be important, the absorbance varying with the amount of sodium sulfate added (21). In addition, higher results were found with the Ortho urine material. This material is stated to be human urine supplemented by material of nonhuman origin; bovine albumin has been shown (22) to be present. In contrast, bovine serum and Wellcome serum, a bovine serum-based material, appeared to be satisfactory as calibration materials. The sulfosalicylic acid/biuret technique gave increased results for all albumin- and serum-based materials but gave correct results for urine materials. This is because this method, recommended for Australian laboratories, requires that a blank be performed with urine samples submitted for analysis but does not recommend using a blank for the saline dilutions of human serum to be used as standards. Although McElderry et al. (9) showed considerable differ- ences in response between albumin and globulin in the Coomassie Brilliant Blue technique, we did not find, in this study, any particular problems of bias with serum-based materials. albumin albumin serum urine urine Table 5. Practicability Scores for the Methods Evaluated Turnaround Reagent Technical Sample Method time Equipment preparation skill size Costs Total Selected Method 1 3 2 1 2 1 10 SSA/biuret 4 4 2 4 1 3 18 CoomassieBrilliantBlue 5 5 5 5 5 4 29 SSAlsodium sulfate 5 5 5 4 2 4 25 TCA turbidimetry 5 5 5 4 1 5 25 Ponceau-S 5 4 4 3 4 4 24 556 CLINICAL CHEMISTRY, Vol. 29, No. 3, 1983 Method Table 4. Mean Protein Concentrations (g/L) for Albumin- and Serum-Based Materials at Two Concentrations and for Two Urine Samples from Patients Material Bovine Human Bovine Human Dade Ortho Wellcome Ortho Human serum serum serum serum Selected method (routine) 0.23 0.25 0.22 0.22 0.22 0.23 0.22 0.17 0.22 SSA/biuret 0.50 0.52 0.50 0.46 0.47 0.47 0.47 0.35 0.62 0.27 0.31 0.31 0.30 0.30 0.29 0.30 0.19 0.17 0.54 0.58 0.62 0.59 0.63 0.59 0.59 0.34 0.60 Coomassie brilliant blue 0.31 0.28 0.19 0.22 0.21 0.25 0.23 0.21 0.23 0.54 0.52 0.36 0.42 0.39 0.45 0.41 0.35 0.51 SSA/sodium sulfate 0.42 0.25 0.17 0.23 0.24 0.25 0.21 0.22 0.20 1.05 0.52 0.51 0.48 0.51 0.54 0.55 0.61 0.57 TCA turbidimetry 0.22 0.22 0.28 0.26 0.29 0.28 0.31 0.17 0.23 Ponceau-S 0.47 0.43 0.58 0.51 0.56 0.56 0.59 0.43 0.50 0.28 0.29 0.21 0.25 0.25 0.26 0.25 0.18 0.24 054 0.55 0.43 0.49 0.47 0.50 046 0.34 0.55 For turbidimetnc methods, there were no apparent prob- lems of comparative bias between human albumin and serum-based materials and urines used in this study. Per- haps this was because all materials were diluted in 9 g/L saline: turbidimetric methods generally suffer from failure of standards and samples to form precipitates identically, and precipitation may not occur at low protein concentra- tions in urines of high ionic strength. It is therefore possible that results on urine samples from patients do have consid- erable bias. Practwability. Turnaround time, equipment required, dif- ficulty of reagent preparation, technical skill required, sample size, and costs were judged on a scale of 1 to 5, 5 being the best possible score (Table 5). The Selected Method had the worst score and the method recommended for Australian laboratories the second worst score; the other four methods had comparable high scores, with the Coomas- sie Brilliant Blue technique being the most practicable. Conclusions The Selected Method was the least practicable of the methods studied. The precision of the method was poor when performed under routine laboratory conditions. The method had the widest linear range and little apparent bias. The best use of this method would perhaps be as a reference method. The sulfosalicylic acidibiuret technique was also some- what impracticable and, specifically, requires a large sam- ple volume. The method had good precision and linear range, and little apparent bias with urine samples; however, it overestimated the protein concentration of albumin and serum-based materials by not requiring blanks to be per- formed with such samples. The Coomassie Brilliant Blue technique was most practi- cable, but the relatively low linear range and poor precision were disadvantages. The sulfosalicylic acid/sodium sulfate turbidimetry tech- nique was simple to perform but a large sample size is required. The method had poor precision, and bovine mate- rials could not be used as calibration material. The trichioroacetic turbidimetry technique had generally good practicability but required a large sample and had a narrow linear range. The precision was good, and there was little apparent bias with the samples used in this study. The Ponceau-S technique had good practicability, re- quired little sample, had good precision, and little apparent CLINICAL CHEMISTRY, Vol. 29, No. 3, 1983 557 bias. We recommend this technique for routine clinical chemistry laboratory use, particularly because human or bovine albumin, serum, or urine appeared to be satisfactory for calibration and quality-control or assurance. McElderry et al. (9) also recommend this method as one that has considerable advantages. We thank Miss M. E. Coles, Division of Clinical Chemistry, Institute of Medical and Veterinary Science, Adelaide, for her advice, and the referees of the manuscript for their detailed helpful critiques. References 1. Glenn CC, Hathaway TK. Urine chemistry. A new CAP pro- grain. Am J Clin Pathol 68, 153-158 (1977). 2. Glenn GC. The CAP urine chemistry program for 1977. Am J Clin Pat/wi 70, 513-515 (1978). 3. Glenn CC. Evolution of the urine chemistry survey program of the CAP. Am J Clin Pathol 72, 299-305 (1979). 4. Glenn CC. The results of analyte enhancement and use of supplied urine protein standard in the CAP urine chemistry survey program. Am J Clin Pat/wi 74, 531-534 (1980). 5. The Royal College of Pathologists of Australia. Test and Teach Programme 1973. Protein in Urine. RCPA, Sydney, New South Wales, Australia. 6. Shephard MDS, Penberthy LA, Fraser CG. Analytical goals for quantitative urine analysis. Pathology 13, 543-546 (1981). 7. Shephard MDS, Penberthy LA, Fraser CG. A quality assurance programme for quantitative urine analyses. Pathology 14, 327-331 (1982). 8. Doetsch K, Gadsden RH. Determination of urinary total protein by use of gel filtration and a modified biuret method. Selected Methods Clin Chem 8, 179-182 (1977). 9. McElderry LA, Tarbit IF, Cassells-Smith AJ. Six methods for urinary proteins compared. Clin Chem 28, 356-360 (1982). 10. Heick HMC, Begin-Heick N, Acharya C, Mohammed A. Auto- mated determination of urine and cerebrospinal fluid proteins with Coomassie Brilliant Blue and the Abbott ABA-100. Clin Biochem 13, 81-83 (1980). 11. Tietz N. Fundamentals of Clinical Chemistry, 2nd ed., Saunders, Toronto, 1970, pp 360-363. 12. Henry RJ, Cannon DC, Winkelman JW. Clinical Chemistry: Principles and Technics, 2nd ed., Harper and Row, New York, NY, 1974, pp 430-431. 13. Pesce MA, Strande CS. A new micromethod for determination of protein in cerebrospinal fluid and urine. Clin Chem 19, 1265- 1267 (1973). 14. Fraser CG, Peake MJ. Problems associated with clinical chem- istry quality control materials. Crit Rev Clin Lab Sci 12, 59-86 (1980). 15. Hiller A, Greif RL, Beckman WW. Determination of protein in urine by the biuret method. J Bail Chem 176, 1421-1429 (1948). 16. Bradford MM. A rapid and sensitive method for the quantita- tion of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72, 248-254 (1976). 17. Peake MJ, Duncan BM, Fraser CG. Comparison of standardiza- tion techniques for manual colorimetric analyses. Clin Biochem 13, 12-16 (1980). 18. Hearne CR, Peake MJ, Duncan BM, Fraser CG. Comparison of standardization techniques for colorimetnc analyses on a centrifu- gal analyzer. Clin Biochem 15, 173-178 (1982). 19. Peters T Jr. Proposals for standardization of total protein assays. Ciin Chem 14, 1147-1159 (1968). 20. Manuel Y, Revillard JP, Betnal H. Proteins in Normal and Pathological Urine, University Park Press, Baltimore, MD, 1970, p 355. 21. Meulemans 0. Determination of total protein in spinal fluid with sulphosalicylic acid and trichloroacetic acid. Clin Chim Acta 5, 757-761 (1960). 22. Balazs NDH, Dennis PM. The curse of the cow in clinical chemistry. Clin Biochem News 62. 31-33 (1981).