Clinical Pathology: Preanalytical Variation in Preclinical Safety Assessment Studies-Effect on Predictive Value of Analyte Tests * JULIA H. RILEY Investigative Toxicology, Department of Toxicology and Pathology, Hoffmann-La Roche, Inc., Nutley, New Jersey 07110

Significant differences in concentrations of analytes in samples may be introduced before samples enter analyzers. These differences are known as preanalytical variation and are part of the overall variation in analytical data. Preanalytical variation is caused by factors that operate during animal preparation prior to sampling, sample collection, sample processing, and sample storage prior to measurement. Preanalytical variation is important because it detracts from the predictive value of analyte measurements. Preanalytical variation may permanently damage data. Because its effects are difficult to quantitate it should be minimized in safety assessment studies. Sources of preanalytical variation are actions performed on animals prior to sample collection and actions performed on the specimen prior to analysis. Preanalytical variation produces a range of artefacts in experimental data. Consequences of preanalytical variation are loss of confidence in the data, obfuscation of real test article effects, false effects, and possibly the expense of repeating a study. To limit preanalytical variation, its sources must be identified, the effects documented, and measures devised to eliminate its sources. Predictive value (likelihood of actual disease) of appropriate clinical pathology tests in toxicology is inversely dependent on preanalytical variation: uncontrolled variation produces data with low predictive values, and controlled variation produces data with high predictive values. Keywords. Artefacts; animals; sample collection and handling; toxicology

INTRODUCTION

troduced before samples enter an analyzer. Preanalytical variation is part, along with interindividual, intraindividual, and analytical variation, of the overall variation in analytical data. Preanalytical variation or preinstrumental variation in toxicology studies is caused by factors that operate during animal preparation, sample collection, sample transport, sample processing, and sample storage prior

The objective of this article is to discuss the effects of preanalytical variation on the predictive value of analyte tests performed by the clinical pathology laboratory in a toxicology testing facility. Based on experience, it is the author’s contention that standards similar to those used in human medicine are necessary to assure the quality and uniformity of our data before we attempt to assess its predictive value. Data from clinical pathology studies in preclinical safety assessment testing serves two predictive functions: It is used to predict a disease in the test animal and, by extrapolation, in the ultimate

to measurement.

IMPORTANCE OF PREANALYTICAL VARIATION

Preanalytical variation in clinical pathology analyses is of critical importance because it affects the predictive value of all sample measurements (23, 24, 71, 79, 84). An artefact introduced prior to sample analysis is going to affect these analyses. Like other forms of experimental variation, preanalytical variation detracts from the accuracy or specificity of analyte measurements. Preanalytical variation is important to us, as clinical pathologists, because we make interpretative judgments about the effects of drugs, chemicals, and other products in test animal

(humans). Preanalytical variation or uncertainty refers to differences in analyte concentrations in specimens inuser

* Address correspondence to: Dr. Julia H. Riley, Investigative Toxicology, Department of Toxicology and Pathology, Hoffmann-La Roche, Inc., 340 Kingsland Street, Nutley, New Jersey

07110. Presented at the International Workshop on Clinical Pathology Testing in Preclinical Safety Assessment, George Washington University, Washington, D.C., July 27, 1991.

490

491

populations in preclinical safety assessment studies (23, 71, 79). We base these professional opinions on clinical pathology data obtained from preclinical studies (79). CHARACTERISTICS OF PREANALYTICAL VARIATION Medical opinions of the clinical pathologist are based on analyte measurements (23, 71, 79, 84). The confidence limit, or range around an experimentally determined statistic that has a known probability of including the true value, will directly influence how decisively the clinical pathologist’s medical opinion is expressed. Large preanalytical variations broaden the confidence interval and limit our trust in the test result. For example, test results from hemolyzed specimens are always viewed with circumspection because hemolysis interferes with the accuracy and specificity of most analyte tests. No definitive statements concerning analyte concentration or, in other words, test article effects should be made when the specimen is hemolyzed. A hemolyzed sample should be considered a waste of time and effort. Preanalytical variation is embodied in all sample measurements (23, 71, 79). Some inherent variability in animal preparation, collection technique, transport, handling, and storage of samples is unavoidable because of human, animal, and other factors that are beyond our control. No matter how good our intentions, no 2 animals will be sampled in an identical manner. When blood samples are collected from the orbital sinus plexus, for example, the amount of interference from tissue damaged during the procedure will vary from animal to animal even when the same operator is employed. The inherent or unavoidable variation may be introduced by such factors as movement of the animal during the procedure, sharpness of the collection tube, or skill of the operator. In addition, the animal may introduce variability into the sample simply because some animals have a quiet disposition and manifest a minimal stress response while others strongly resist any attempt to collect samples (47,

49, 55). The factors that produce preanalytical variation difficult to quantitate on an individual basis or, for that matter, on a sampling or study-by-study basis (2, 23, 76). Thus, it becomes necessary either to avoid these factors or to standardize them, whichever is more appropriate. Artefacts produced by preanalytical variation may be introduced into data from single animals, groups of animals, or study intervals by varying, for example, the sampling site (2, 23, 71, 79), sample collection technique, operator, or method of animal

are

handling.

Preanalytical variation is difficult to amend or repair once introduced into samples. The reason for this being so is that quantitation, with the intent of correcting analyte values for the effect of factors producing preanalytical variation, is essentially another experiment. Even correction for a known preanalytical variation is not an easy task in a toxicology study, because of the documentation necessary, and may not be wise because of the

possi-

bility of interactions of the test article with the variable! It is certainly preferable not to introduce artefacts into data in the first place rather than ponder upon the relevance of correction factors. The statement by Statland and Winkle (71 )- &dquo;The preparation of the specimen is not as rigidly controlled as is the measurent of analytes; thus, it is during the preinstrumental phase that numerous sources of variation are in operation&dquo;-is not strictly applicable to clinical pathology in our setting. Preanalytical variation should be minimal in toxicologic clinical pathology testing. The animal environment is by necessity stabilized, and uniform specimen collection and handling methods can be specified in standard operating procedures. Toxicology studies should be designed so that factors producing noisy data are eliminated and the only variable is the test article. To achieve this end, sampling should be controlled and closely supervised by the clinical pathologist, and standard operating procedures described in enough detail to limit

preanalytical variation. SOURCES OF PREANALYTICAL VARIATION the sources of preanalytical variation be divided into actions performed on the animal prior to sample collection, actions performed on the animal at sample collection, and actions performed on the specimen prior to analysis. Sources of preanalytical variation and their effects on the data are listed in Table I.

Generally,

can

ARTEFACTS CAUSED BY PREANALYTICAL VARIATION

Discussions of sample collection procedures for various analytes are found in most medical clinical pathology textbooks. For the most part, these procedures can be directly applied to animal sample collections. Collection Technique

Artefacts

analytes the collection method has direct bearing on values obtained for the analytes. For example, when a blood sample contacts activators (membrane phospholipids, connective tissue elements of the subendothelium, and Factor VIII/von For most

492

FIG. 1.-Effect of poor collection technique is shown in this blood smear. Blood was collected into EDTA by cutting the rat’s tail. Platelet clumps are indicated in the tail of this smear by arrows. Modified Wright’s stain. Oil immersion. x 400. FIG. 2.-Effect of good collection technique is shown in this blood smear. Blood was collected from the orbital sinus plexus with an EDTA-coated capillary tube and was placed into an EDTA tube prior to smear preparation. Individual platelets are indicated in the tail of this smear by arrows. Modified Wright’s stain. Oil immersion. x 1,000. FIG. 4. -Specimen handling and storage. Normal primate blood smear. 1 ) Lymphocyte. 2) Eosinophil. 3) Platelet. Modified Wright’s stain. Oil immersion. x 630. FIG. 5. - Specimen handling and storage artefact. Blood smear from a primate made after its blood was refrigerated overnight. 1 ) Spurious spheroechinocyte. 2) Spurious autoagglutination. 3) Variant lymphocyte. 4) Abnormal neutrophil morphology. In addition, spurious thrombocytopenia and anisocytosis are seen. Modified Wright’s stain. Oil immersion. x

1,000.

FIG. 6.-Hemolysis artefact. Plasmas from 4 blood samples obtained by venipuncture from a dog and anticoagulated with heparin. 1) Most hemolysis (+++) is seen in a 3-mL Vacutainer. 2) Hemolysis (++) in a 7-mL Vacutainer. 3) Hemolysis (+) in a 5-mL syringe-drawn sample. 4) No visible hemolysis in a carefully drawn 5-mL syringe sample.

493 TABLE L-Sources and effects

of preanalytical

variation.

,-

-

-

I

FIG. 3.-Effect of poor technique. A leukocyte histogram from a Coulter instrument (permission obtained from Coulter). Platelet interference.

Willibrand’s factor) platelet aggregation occurs. Figure 1 is a blood smear that illustrates one of the consequences of platelet activation in data. In this example, platelets were exposed to a cut tissue surface when the blood sample was obtained from a rat by cutting the tail with a scalpel. The blood smear from this sampling contains numerous clumps of platelets. Clumped platelets produce artefacts in data because they are falsely enumerated in the leukocyte count (those the size of leukocytes) or erythrocyte count (those the size of erythrocytes) by instruments using the Coulter technology ( 11 ). By contrast, Fig. 2 is a blood smear made from a blood sample obtained from the orbital sinus plexus of a rat using

ethylenediaminetetraacetate (EDTA)-coated capillary tube. There are no clumped platelets in the tail of this smear. Figure 3 is a histogram from a Coulter S4 + indicating platelet noise in the leukocyte count. Microclots, as demonstrated in Fig. 1, invalidate the platelet count and produce inaccuracies in the erythrocyte and leukocyte counts from Coulter-type instruments. The effect of platelet clumps is more obvious in the leukocyte count than in the erythrocyte count because of the relative num-

an

bers involved. Microclots also cause alterations in erythrocyte morphology on blood smears because sticky platelet clumps tear cells apart as the smear is made. Improper collection techniques are reflected in our data (29).

Specimen Handling and Storage Artefacts Blood and tissue samples deteriorate after collection. This deterioration, reflected in the analyte val-

-

FIG. 7.-Deterioration artefact. A bone marrow smear from a mouse femur made immediately after the mouse was sacrificed. (Note this part of the smear is used for illustration-It is too thick for normal evaluation.) 1) Macrophage with cellular debris. 2) Granulocyte. 3) Intact cell membrane. 4) Modified Wright’s stain. Oil immersion. x 1,000.

494 ues, may be hastened

retarded by handling and storage techniques (51). A simple example of the effect of inappropriate sample processing in hematology is illustrated in Fig. 4. This is a blood smear made from portion of an EDTA blood sample obtained from a primate shortly after the animal was bled. By contrast, Fig. 5 is a blood smear from the same animal made after the blood sample was refrigerated overnight. In the refrigerated specimen, the central pallor or biconcave shape of many normal erythrocytes has been lost, and there are spurious spheroechinocytes, spurious agglutination, and a variant lymphocyte morphology. Analyte values in biochemistry, too, reflect sample handling and storage methods. Serum potassium and inorganic phosphate are spuriously elevated when blood samples are allowed to clot at refrigerator temperature (20). Many other examples of processing artefacts can

or

be found in the medical literature. Specimens be handled in a timely manner and stored ap-

must

propriately.

Figure 6 illustrates 2 means of introducing hemolysis into samples. The 4 (Li) heparinized blood samples were collected from a dog via jugular punctures and all were centrifuged simultaneously. Comparison of the plasma color in the 3- and 7-mL Vacutainer tubes shows that hemolysis is more severe in the smaller tube. Comparison of both Vacutainer samples to those drawn via syringe shows that both Vacutainer plasmas have more hemolysis than syringe-drawn samples. When the two 5-mL syringedrawn plasmas are compared, it is seen that one contains more hemoglobin than the other. The sample with no visible hemolysis was collected carefully, using minimum negative pressure; the other sample, obtained more rapidly, had obvious hemolysis. Glick et al (27) illustrate by means of interferographs the effect of hemolysis on some common analytes with the then current analyzers. The spurious effect of hemolysis (1 g spiked sample) is especially obvious in distorting bilirubin values measured on the ACA analyzer. Almost all analytes, even pH measurements, are affected to some degree by hemolysis. Hemolysis may cause a positive or negative interference in analyte measurements. Hemoglobin directly interferes with some analyte measurements by virtue of its red color. Its presence in a sample indicates that intracellular material is now in the specimen, and one expects biochemical analytes to be affected and the erythrocyte count to disagree with the hemoglobin value because red cells have been ruptured. The presence (or absence) of hemolysis must always be documented for all samples because in vivo hemolysis signifies a serious disease process, whereas in vitro hemolysis is a sig-

The deleterious effects of storage are the result of multiple factors operating on the sample (7, 20, 25, 49, 51, 52, 61-63, 65, 78, 80). A short list of factors includes evaporation, interaction of the sample with the container, temperature, light, and humidity. The effect of these factors may be direct on the analyte or indirect where change in one analyte leads to a secondary change in another. A direct effect is the pH change that occurs when gaseous carbon dioxide is evolved from a blood sample, and the secondary change is the decrease in values for ionized calcium in the sample (61 ) or increase in hemolysis of pHsensitive erythrocytes (26). Storage conditions for one analyte (serum creatine kinase) are not ideal for others (isoenzymes of lactate dehydrogenase) (68). Most analyte values change when the sample is stored.

gist’s albatross (11, 12, 13, 14, 16, 23, 26, 27, 32, 33, 36, 42, 66, 71, 73, 79).

Sampling Site Artefact

Deterioration

Samples collected from different sites in the animal do not have comparable analyte values (2, 18, 23, 60, 71, 76, 79, 84). Archer and Riley (2) illustrate this point by examining the effect of different sampling sites on blood leukocyte and erythrocyte numbers. Their data also show that analyte values from some sites are more uniform than those from other sites. Leukocyte counts in blood samples from unanesthetized rats bled from the tail and jugular vein were characterized by an extremely large standard deviation when compared to counts from the orbital sinus plexus. There are significant differences in analyte concentrations in specimens collected from dif-

The quality of a specimen depends on the collection technique and timely processing. The importance of both these factors is best illustrated using a bone marrow smear as an example. Figure 7 is a mouse bone marrow smear made from the femur, stained with a modified Wright’s stain. The mouse was sacrificed and the smear made immediately. Notice in this smear that the cell membranes appear intact. A centrally located macrophage is easily identified, the charactistic chromatin pattern seen in the neutrophils is obvious, and the nuclear material is contained in the nucleus. The smear in Fig. 8 is identical in every respect to that in Fig. 7 except it was made from the opposite femur 15 min after the first smear. The cell indicated by the arrow is probably a macrophage. Its cell membrane and nuclear characteristics are blurred, thus making positive identification of the cell difficult. Nuclei of other

ferent sites. Hemolysis Artefact Hemolysis in serum blood) is

a

and plasma (anticoagulated serious collection artefact best avoided.

nificant artefact.

Hemolysis

is the clinical

patholo-

Artefact

495 cells are pyknotic or have clumped chromatin, or the nuclear material is strung out across the face of other marrow cells obscuring their identity. In other fields of this smear, the nucleoli are spuriously enlarged and falsely present in more mature cells than one would expect to find normally. Morphologic bone marrow interpretation is subjective, and difficult, and even more difficult when spurious changes are introduced by poor technique. In Fig. 8 the nucleolar enlargement would suggest a neoplastic change were it not for the fact that this pathologist knows the sample was obtained from a normal mouse.

right side of Fig. 9 is an example of a lymphocyte and erythrocytes from a normal rat. By contrast, the left side of Fig. 9 is a cell that by a process of elimination is probably a lymphocyte. It is surrounded by erythrocytes with very abnormal morphology. The smear was made from a normal rat. The cause of the changes was not determined. Figure 10 is a rat basophil. Both mice and rats have basophils in the peripheral blood, but they are mostly overlooked or the granules are washed out during staining. The quality of bone marrow smears and blood smears depends on both good technique and timely collection (43, 81). The

centrations stimulate the heart and respiration and higher concentrations cause asphyxia and respiratory failure (59). Anesthetic agents have multiple effects on most analyte values (74).

Muscle Activity

Artefact

Struggling increased muscle activity produces changes in many analytes: Potassium is released from the liver, serum creatine kinase is increased, plasma lactate is increased, and blood ieukocytes and, in some species, erythrocytes are increased (23, 41, 66, 79). Glover and Jacobs (28) monitored the activity pattern of normal rats over a 24-hr period. It is interesting to note that the period of inactivity occurs during the time most sample collections are scheduled. Muscle activity above resting levels produces significant effects on many analytes (23, 71, 79). or

Seasonal Artefacts

Berger (5) demonstrated a seasonal effect on blood leukocyte numbers in male rats reared under artificial light. Female dogs housed in laboratory environments demonstrate seasonal hormonal fluctuations (personal experience). Seasonal effects alter analyte values in most species.

Anesthetic Artefacts

Dilution Errors

Anesthetic agents induce 2 types of artefact into data values: physiological and biochemical. Physiological effects are produced by alterations in blood pressure, respiration, recumbency, and excitement, whereas biochemical effects are the result of enzyme induction, tissue hypoxia, and toxic damage (17, 37). The data shown in Table II are an example of a biochemical interference in an investigative study where the animals (10 of each sex/group) were given xylazine and a combination of xylazine and ketamine. In this study, the effect of xylazine on biochemistry values is illustrated by statistically significant differences (Dunnett’s) in many of the commonly measured analytes. Xylazine, a sedative/ hypnotic/analgesic/muscle relaxant has been reported to interfere with growth hormone, glucagon, insulin, glucose, follicle-stimulating hormone, and possibly other analytes (77). In a study by Archer and Riley (2), different anesthetics were compared for their effect on leukocytes. The agents used in this study in addition to altering total leukocyte counts appeared to increase the interanimal variation (large standard deviations). Such a finding is not unexpected and is probably the result of an exaggeration of interanimal variation caused by the varied response of individuals to anesthetic agents. Some laboratories use carbon dioxide as an anesthetic agent. It should be remembered that 30% carbon dioxide is the anesthetic concentration; lower con-

Dilution of samples alters analyte values (23, 34, The effect of dilution is illustrated by Ho (34), who demonstrated that significant errors (17%) in the concentration of serum amylase were produced by the first 2-fold dilution. Dilution of samples produces errors in analyte measurements.

79).

Evaporation Artefact Many of our samples

are obtained from rodents and small primates, where sample volumes obtained are small. The documented effect of evaporation is pronounced on small samples (especially if left uncapped or placed in containers with a large airspace above the sample) (23, 71, 79). Evaporation may produce a 50% increase in analyte concentrations after 4 hr of exposure (73).

Collection Device Artefacts The sample collection device, syringe, curette, needle, Vacutainer, brush, and tube can have significant effects on analyte values measured in that sample (71). For example, samples collected from the orbital sinus plexus (Fig. 11) with a Pasteur pipette are inappropriate-Blood contact with the glass pipette initiates the coagulation cascade when Hageman factor is activated by the glass. Samples obtained by this method are likely to contain clots and activated cascade proteins. The introduced artefacts

invalidate blood cell counts and disturb the coag-

496

FIG. 8. -Deterioration artefact. A bone marrow

smear

from a mouse femur made 15 min after the mouse was sacrificed.

(Note this part of the smear is used for illustration-It is too thick for normal evaluation.) 1) Possible macrophage(?) with cellular debris. 2) Granulocyte with clumped chromatin artefact. 3) Rupture of cell membrane. 4) Strand of chromatin. 5) Pyknotic nuclei of unidentifiable cells. Modified Wright’s stain. Oil immersion. x 1,000. FIG. 9.-Deterioration artefact. Blood smears from normal rats. Smear on the left shows normal lymphocyte, platelet, and erythrocyte morphology. Smear on the right shows spurious erythrocyte and lymphocyte morphology and a spurious thrombocytopenia. The cause of the artefact was not determined. Modified Wright’s stain. Oil immersion. x 630. FIG. 10.-Deterioration artefact. Basophil from a normal rat. Modified Wright’s stain. Oil immersion. x 1,000. FIG. 11. -Collection device artefact. Blood samples collected from rodents. 1) Blood collection using an inappropriate sampling device. 2) Blood collection for hematology using an EDTA-coated capillary tube.

ulation screening tests. The device used to collect a sample has significant effects on the subsequent an-

alyte

measurements

Serum

versus

(78).

Plasma

As illustrated by Korsrud and Trick (50) and others (39, 75), most analytes in serum samples have different numerical values from those obtained in plasma samples. These differences are the result of removal of coagulation proteins from the serum when the clot is removed, release of intracellular material (hemoglobin, inorganic phosphates, potassium) into serum during the clotting process, and dilution and contamination of the plasma with the chosen anticoagulant. The concentration of an analyte is not the same in serum as it is in plasma.

Circadian

Effect Most analytes are affected to some degree by periodicity. The circadian periodicity in hormone concentrations in male rats (64, 83) and the well-documented changes in serum iron, corticosteroids, and blood carbon dioxide content are further examples of periodicity in analyte concentrations (42, 73). Test Article

Effects

Drugs and their metabolites may interfere with analyte test methods (74). Samples collected during the period of maximum drug or metabolite concentration may have significant drug-related effects in analyte values because of drug test method interactions. The type of interference ranges from lack

497 TABLE

c

rats

(pooled sexes).a

20 (10 of each sex). Statistical difference (p < 0.05). Statistical difference (p < 0.01 ).

° n b

IL-Drug interference with chemistry values in

=

of discrimination between two ions (bromide and chloride, lithium and sodium), to pigment or color effects of the drug, to spurious increases in electronic cell counts (lipid infusion), to false-negative values for fecal blood (ascorbic acid) (74). GOOD DATA/POOR DATA

There are good data and poor data-the distinction is in part due to the amount of preanalytical variation that the experimenter tolerates. In good data from a toxicology study, the standard deviations are usually small and precision tight. In a laboratory performing poorly, the data values are widely scattered, individual animal values from different samplings are spread apart, the reference range is very large, and some values are unrealistic. It is not easy to interpret data from studies where the data are noisy, where, for example, normal rats have bilirubin values of 12 mg/dL and normal squirrel monkeys have glucose values of - 20 mg/dL (personal

experience).

by ignoring the necessity to minimize preanalytical variation. A large or uncontrolled preanalytical variation may introduce false treatment effects. In hematology analysis, platelet clumping, as described previously, may produce a false thrombocytopenia and leukocytosis in those animals in which this artefact is prominent. There is usually minimal effect of platelet clumps on erythrocyte counts because of the relative numbers involved. Poor technique in making blood smears produces many artefacts that include inaccurate enumeration of leukocytes when the smear is uneven, too long (for automated staining) or too thick, spherocytes, and red cell fragments. In chemistry analysis, falsely low (light-exposed) or falsely high (hemolyzed) serum bilirubin values, falsely low (water contamination) or falsely high (evaporation) urine sodium values, and hyperalbuminemia (tourniquet effect) are examples of false

data

treatment effects that may creep into data if too little attention is paid to controlling preanalytical varia-

tion.

CONSEQUENCES

OF

LARGE OR UNCONTROLLED PREANALYTICAL VARIATION One consequence of a large or uncontrolled preanvariation in preclinical safety assessment studies is loss of confidence in the test result: accurate results from clinical laboratory tests ultimately depend on the collection of appropriate spec-

The ultimate consequence of a large preanalytical variation is the expense of having to repeat a study when there is loss of confidence in the test results because of poor quality data.

CONTROL

alytical

imens. A large

Control of preanalytical variation is complex. It approached by several different means be-

must be

simple system will meet the requirements of all analytes and there are numerous variables. The most important step toward controlling sources of preanalytical variation is to identify those sources and document their effects for your laboratory. Know how measured amounts of hemoglobin affect analyte values, how long is too long to hold off a vessel before the effect of the tourniquet is reflected in analyte values, and what effect more than one venipuncture attempt has on subsequent

cause no

uncontrolled preanalytical variation in analyte tests may obscure or obfuscate interpretation of analyte results. In the bilirubin example cited previously, where both control and treated rats had bilirubin values of 12 mg/dL, 12 times that of a normal rat, some rats may truly have had elevated values. In serum samples with in vitro or introduced hemolysis, real or in vivo hemolysis will most likely be obscured by the artefactual hemolysis. There are many other artefacts that can be introduced into or

OF

PREANALYTICAL VARIATION

data values.

498 All standard operating procedures should be writdescriptive detail to eliminate variations in sample collection, transport, storage, and other preprocessing procedures and between operators or between samplings. The guidelines of the National Committee for Clinical Laboratory Standards (NCCLS; Villanova, Pennsylvania) have been developed to aid laboratories in establishing uniformity and good technique in these areas. Special attention is necessary in training those whose job it is to collect quality specimens (70). Medical technologists, by virtue of their training, are alert to the artefacts produced by poor or inconsistent sample collection techniques, but samples are also collected by others who do need additional training. In my experience, when sufficient training is provided to these people and the sampling artefacts produced by poor technique are demten with sufficient

onstrated, they can perform technically difficult sampling procedures extremely well. CONCLUSION

Preanalytical variation has largely been ignored by clinical pathology, and yet simple steps can be taken to eliminate many of its sources. Contrast the enormous volume of literature on analytical variation in journals to the paucity of literature on preanalytical variation. It is now time to close this gap. Veterinary clinical pathologists involved in toxicologic pathology are in the position to do just that! To conclude, in clinical pathology, from preclinical safety assessment studies, and especially with the pharmaceutical products of today, we must be aware of the deleterious effects that preanalytical variation has on our data and take steps to limit these effects. It is recommended as a first step that standards similar to those of the NCCLS be developed and implemented prior to evaluation of the predictive value of test profiles in preclinical testing

(3, 9). ACKNOWLEDGMENTS The author wishes to express her gratitude to the following people for technical advice and help in preparing this presentation: Julia Symington, MT, and Diane Schmoltze, MT, for sample analysis; Janet Becker, BS, MS, for photography (Pathology Dept.); Bob Pitler, Laboratory Technician; Rudy Wolf, Laboratory Technician (Toxicology Dept.); and Nicholas Riley, for slide preparation; and Roche photographic services. REFERENCES 1. Annino JS and Reiman AS

(1959). The effect ofeating

chemical constituents of the blood. Am. J. Clin. Pathol. 31(2): 155-159. on some

clinically important

2. Archer RK and Riley J (1981). Standardized method for bleeding rats. Lab. Anim. 14: 25-28. 3. Astaldi A, Bricarelli D, Palomba R, Tassi GC, Ugazio AG, and Astaldi G (1975). Standardized procedures for specimen collection in cellular immunology. In: Current Studies on Standardization Problems in Clinical Pathology, Hematology and Radiotherapy in Hodgkin’s Disease, G Astaldi (ed). Excerpta Medica, pp. 14-19. 4. Bailey CJ, Day C, Flatt PR, and Atkins TW (1980). Effect of repeated sodium pentobarbitone administration on glucose homeostasis in mice. Gen. Pharmacol. 11: 303-305. 5. Berger J (1983). Seasonal influences on circadian rhythms in the blood picture of SPF rats housed under artificial illumination. Folia Hematol. 110: 55-70. 6. Blank DW, Kroll MH, Ruddel ME, and Elin RJ (1985). Clin. Chem. 31(9): 1566-1569. 7. Boink BTJ, Bijster P, Kees L, Vink J, and Maas AHJ

(1985). Direct potentiometric determination of sodium ion in blood. III. Influence of (Bi) carbonate. Clin. Chem. 31(4): 523-526. 8. Brady AG and Koritnik DR (1985). The effects of ketamine anesthesia on glucose clearance in African green monkeys. J. Med. Primatol. 14: 99-107. 9. Burlina A (1975). Specimen collection and preservation in clinical enzymology. In: Current Studies on Standardization Problems in Clinical Pathology, Hematology and Radiotherapy in Hodgkin’s Disease, G Astaldi (ed). Excerpta Medica, pp. 8-13. 10. Chin BH, Tipton TR, and Kozbelt SJ (1979). The interfering effects of hemolyzed blood rat serum chemistry. Toxicol. Pathol. 7(1): 19-22. 11. Cornbleet J (1983). Spurious results from automated hematology cell counters. Lab. Med. 14(8): 509-514. 12. Craft NE, Brown ED, and Smith JC (1988). Effects of storage and handling conditions on concentrations of individual carotenoids, retinal and tocopherol in plasma. Clin. Chem. 34(7): 1505. 13. Czerwek BH and Bleuel H (1981). Normal values of alkaline phosphatase, glutamic oxaloacetic transaminase and pyruvic transaminase in the serum of experimental animals using optimised methods, and the effect of hemolysis on these values. Exp. Pathol. 19: 161-163. 14. D’Alesandro MM, Gruber DF, Reed HL, O’Halloran KP, and Robertson R (1990). Effects of collection methods and storage on the in vitro stability of canine plasma catecholamines. Am. J. Vet. Res. 51(2): 257259. 15. D’Alleinne CP and Mann DD (1982). Evaluation of ketamine/xylazine anesthesia in the guinea pig: Toxicological parameters. Vet. Hum. Toxicol. 24(6): 410412. 16. Davey CW, Jackson MR, and Walker JM (1984). Stabilities of marmoset ) Callithrix jacchus plasma ( under conditions of storage. Clin. Chem. 30(1): 101104. 17. Dundee JW, Fee JPH, Moore PDA, Mcllroy, and Wilson DB (1980). Changes in serum enzyme levels following ketamine infusions. Anesthesia 35: 12-16.

499 18. Doell BH and Hegarty PV (1970). The hemoglobin concentration in peripheral and central blood of the laboratory rat. Br. J. Hematol. 18: 503-509. 19. Dorner JL, Hoffmann WE, and Lock TF (1981). Ef fects of in vitro hemolysis on equine serum chemical values. Am. J. Vet. Res. 42(9): 1519-1521. 20. Ezigbo JC and Storey DM (1980). Effect of storage at various temperatures on the activity of cotton rat serum enzymes. Indian J. Exp. Biol. 18: 96-97. 21. Finbarr P, Hurley T, and Martin N (1985). Centrif ugation speeds and the removal of platelets from heparinized plasma. Clin. Chem. 31(6): 1082-1083. 22. Fowler JSL, Brown JS, and Fowler EW (1980). Comparison between ether and carbon dioxide anesthesia for removal of small blood samples from rats. Lab. Anim. 14: 275-278. 23. Garber CC and Carey RN (1984). In: Clinical Chemistry- Theory, Analysis and Correlation, LA Kaplan and AJ Pesce (eds). CV Mosby, St. Louis. 24. Gerstman BB and Cappucci DT (1986). Evaluating the reliability of diagnostic test results. JA VMA 188(3): 248-251. 25. Giampietro O, Navalesi R, Buzzignoli G, Boni C, and Benzi L (1980). Decrease in plasma glucose concentration during storage at —20°C. Clin. Chem. 26(12): 1710-1712. 26. Giger U, Harvey JW, Yamaguchi, McNulty PK, Chiapella, and Beutler E (1985). Inherited phosphofructokinase deficiency in dogs with hyperventilation-induced hemolysis: Increased in vitro and in vivo alkaline fragility of erythrocytes. Blood 65(2): 345-351. 27. Glick MR, Ryder KW, and Jackson SA (1986). Graphical comparisons of interferences in clinical chemistry instrumentation. Clin. Chem. 32(3): 470475. 28. Glover J and Jacobs A (1972). Activity pattern of iron-deficient rats. Br. Med. J. 2: 627-628. 29. Goldberg DM (1975). Quality control of specimen collection in clinical chemistry. In: Current Studies on Standardization Problems in Clinical Pathology, Hematology and Radiotherapy in Hodgkin’s Disease, G Astaldi (ed). Excerpta Medica, pp. 24-31. 30. Goldberg DM, Waheed M, Alexander YU, and Roncari AK (1980). The effects of phenobarbital upon triacyglycerol metabolism in the rabbit. Biochem. J. 192: 165-175. 31. Green JE, Weintraub HA, Donnelly BS, and Mordecai BG (1979). Sample preparation variation and its effects on automated blood cell differential analysis. Anal. Quantit. Cytol. 1(3): 187-201. 32. Harrison SP and Barlow IM (1990). Evaporative loss from small samples. Clin. Chem. 36(8, pt. 1): 1520. 33. Harsten A, Berg B, Inerot S, and Muth L (1988). Importance of correct handling of samples for the results of blood gas analysis. Acta Anesthesiol. Scand. 32(5): 365-368. 34. Ho CS (1985). Effect of dilution on amylase measurement with Astra-8. Clin. Chem. 31(6): 1098-1099. 35. Hoffman WE, Miletich D, and Albrecht RF (1982). Cardiovascular and regional blood flow changes dur-

halothane anesthesia in the aged rat. Anesthesi444-448. 36. Hollaar L and van der Laarse A (1979). Interference of the measurement of lactate dehydrogenase (LDH) activity in human serum and plasma by LDH from blood cells. Clin. Chem. Acta 99: 135-142. 37. Hsu W and Hembrough FB (1982). Intravenous glucose tolerance test in cats: Influenced by acetylpromazine, ketamine, morphine, thiopental and xylazine. Am. J. Vet. Res. 43(11): 2060-2061. 38. Hubbard RW, Matthew WT, Horstman D, Francesconi R, Mager M, and Sawka MN (1984). Albumininduced plasma volume expansion: Diurnal and temperature effects. J. Appl. Physiol. 56(5): 1361-1368. 39. Hyman D and Kaplan NM (1985). The difference between serum and plasma potassium. N. Engl. J. Med. 313(10): 642. 40. Ishida M, Sieno Y, Yamaoka K, Tanaka Y, Satomura K, Kurose Y, and Yabunchi H (1983). The circadian rhythms of blood ionized calcium in humans. Scand. J. Clin. Lab. Invest. 43(suppl. 165): 83-86. 41. Jain N (ed) (1986). Schalm’s Veterinary Hematology. Lea & Febiger, Philadelphia. 42. Kaplan LA and Pesce AJ (eds) (1984). Clinical Chemistry- Theory Analysis and Correlation. CV Mosby, St. Louis. 43. Kass J (ed) (1979). Bone Marrow Interpretation. J. B.

ing

ology 56:

Lippincott, Philadelphia. Baily DN (1985). Mismeasurement of methemoglobin ("methemoglobin revisited"). Clin. Chem. 31(1): 169. 45. Kimura F, Okano H, and Kawakami M (1981). Development of circadian rhythms in serum hormone

44. Kelner MJ and

46.

47.

48.

49.

50.

51.

52.

53.

levels in the immature female rat. Neuroendocrinology 32: 19-23. Koepke JA, Rodgers JL, and Ollivier MJ (1975). Preinstrumental variables in coagulation testing. Am. J. Clin. Pathol. 64: 592-596. Kolata RJ and Rawlings CA (1982). Cardiopulmonary effects of intravenous xylazine, ketamine, and atropine in the dog. Am. J. Vet. Res. 42(12): 21-96. Komearek J (1984). The effect of ether, pentobarbitone sodium and fentyl on blood gases, acid-base balance and hematological parameters in the rat. Experimentia 40: 476-478. Koo HY (1985). Evaporative effect of Ektachem pierceable caps. Clin. Chem. 31(2): 347-348. Korsrud GO and Trick KD (1973). Activities of several enzymes in serum and heparinized plasma from rats. Clin. Chem. Acta 48: 311-315. Kubasik NP, Ricotta M, Hunter T, and Sine HE (1982). Effect of duration and temperature of storage on serum analyte stability: Examination of 14 selected radioimmunoassay procedures. Clin. Chem. 28(1): 164-165. Larsson L and Ohman S (1985). Effect of silicone separator tubes and storage time on ionized calcium in serum. Clin. Chem. 31(1): 169-170. Lenahan JG and Philips GE (1966). Some variables which influence the activated partial thromboplastin time assay. Clin. Chem. 12(5): 269-273.

500 54. Lindena J, Buttner D, and Trautschold I (1964). Biological, analytical and experimental components of variance in a long-term study of plasma constituents in rat. Clin. Chem. Clin. Biochem. 22(1): 97-104. 55. Little J and Anderson TD. A study to evaluate the effects of the anesthetic ketamine hydrochloride and xylazine combination on the clinical chemistry parameters in rats. Study No. 05435. Data on file, Hoffmann-La Roche Inc. 56. Lunam CA, Cousins MJ, and Hall P (1985). Guineapig model of halothane-hepatotoxicity in the absence of enzyme induction and hypoxia. J. Pharmacol. Exper. Therapeut. 232(3): 802-809. 57. Massarrat S and Massarrat S (1979). Transient liver deterioration induced by general anesthesia. Acta Hepato-Gastroenterol. 26: 106-111. 58. McGuill MW and Rowan AN (1989). Biological ef fects of blood loss: Implications for sampling volumes and techniques. ILAR News 31(4): 5-18. 59. Miller RD (ed) (1990). Anesthesia. Churchill Living-

Shapiro WF, Barone J, and Anagnostou A (1978). Effect of protein deprivation on hematopoietic stem cells and on peripheral blood counts. J. Lab. Clin. Med. 92(2): 300-310. 70. Solberg HE and PetitClerc C (1988). Approved recommendation (1988) on the theory of reference values. Part 3. Preparation of individuals and collection of specimens for the production of reference values. Clin. Chem. Clin. Biochem. 26(9): 593-598. 71. Statland BE and Winkle P (1984). Pre-instrumental sources of variation. In: Clinical Diagnosis and Management by Laboratory Methods, JB Henry (ed). W. B. Saunders, Philadelphia. 72. Thompson DS and Friday CD (1978). Changes in 69.

liver enzyme values after halothane and enflurane for surgical anesthesia. S. Med. J. 71(7): 779-782. 73. Tietz NW (ed) (1986). Textbook of Clinical Chem74.

istry. Saunders, Philadelphia. Tryding N and Roos K-A (eds) (1986). Drug Interferences and Drug Effects in Clinical Chemistry. Apo-

75.

teksbolaget, Stockholm. Ungara B, Corso G, and Russo AD (1985). Cholesterol concentration in serum-plasma pairs differs be-

ston.

60. Neptun DA, Smith CN, and Irons RD (1985). Effects of sampling site and collection method on variations in baseline clinical pathology parameters in Fischer344 rats. Fund. Appl. Toxicol. 5: 1180-1185. 61. Nikolakakis NI, de Franscisco AM, Rodger RSC, Gaiger E, Goodship THJ, and Ward MK (1985). Ef fect of storage on measurement of ionized calcium in serum of uremic patients. Clin. Chem. 31(2): 287289. 62. Olson PN, Bowen RA, Husted PW, and Nett TM (1981). Transport of specimens for clinical chemistry analysis. Ann. Clin. Biochem. 14: 301-306. 63. Ono T, Kitaguchi K, Takehara M, Shiba M, and Hayami K (1981). Serum-constituents analysis: Effect of duration and temperature of storage of clotted blood. Clin. Chem. 27(1): 35-38. 64. Ooka-souda S (1981). Changes in certain diurnal hormonal rhythms in developing male rats exposed to adiumal environmental cues. Mech. Aging Dev. 15: 227-233. 65. Reimers TJ, McCann JP, and Gowan RG (1983). Effects of storage times and temperatures on T3, T4, LH prolactin, insulin, cortisol and progesterone concentrations in blood samples from cows. J. Anim. Sci. 57(3): 683-691. 66. Riley JH and Cornelius LM (1989). In: Clinical Chemistry of Laboratory Animals, WF Loeb and FW Quimby (eds). Pergamon Press, New York. 67. Scheving LE and Pauly JE (1967). Daily rhythmic variations in blood coagulation times in rats. Anat. Rec. 157: 657-666. 68. Shain SA, Boesel RW, Klipper RW, and Lancaster CM (1983). Creatinine kinase and lactate dehydrogenase : Stability of isoenzymes and their activity in stored human plasma and prostatic tissue extracts and effect of sample dilution. Clin. Chem. 29(5): 832835.

cause

of water shifts from blood to cells. Clin. Chem.

31(6): 1096-1097. 76. Upton PK and Morgan DJ (1975). The effect of sampling technique on some blood parameters in the rat. Lab. Anim. 9: 85-91. 77. USP DI (1990). Drug Information

for the Health Care Professional. Arcata Graphics, Kingport, Tennessee

37550. 78. Verresen L, Lins RL, Neels H, and De Broe ME (1986), Effect of needle size and storage temperature on measurements of serum potassium. Clin. Chem. 32 (4): 698-699. 79. Westgard JO and Klee G (1986). Quality assurance. In: Textbook of Clinical Chemistry, NW Tietz (ed). 80.

W. B. Saunders, Philadelphia. Wilding P, Zilva JF, and Wilde CE (1977). Transport of specimens for clinical chemistry analysis. Ann. Clin.

Biochem. 14: 301-306. 81. Wintrobe MW (1981). Clinical Hematology. Lea and

Febiger, Philadelphia. 82. Wisser H and Breuer H (1981). Circadian changes of clinical chemical and endocrinological parameters. J. Clin. Chem. Clin. Biochem. 19: 323-337. 83. Wong C-C, Dohler K-D, Geerings H, and Muhlen (1983). A. Influence of age, strain and season on circadian periodicity of pituitary, gonadal and adrenal hormones in the serum of male rats. Hormone Res. 17:202-215. 84. Yoshida and Kanno T (1991). Present situation and problems of sampling and transportation. Rinso By85.

ori. 39(1): 41-42. Young GP, Friedman S, Yedlin ST, and Alpers DH (1981). Effect of fat feeding on intestinal alkaline phosphatase activity in tissue and serum. Am. J.

Physiol. 241: G461-G468.