Special Report
Desirable Standards for Hematology Tests: A Proposal
CALLUM G. FRASER, PH.D., F.A.A.C.B.
Few analytic goals have been proposed for hematology tests,
and these are based upon subjective clinical opinons. It is proposed that the current consensus strategies for delineating analytic goals in clinical chemistry can be validly used in the other
disciplines of clinical pathology. Goals for imprecision are best
based upon average intraindividual biologic variation; such
goals are more stringent than those previously published.
There are many cogent reasons that justify the goal for inaccuracy of no bias. In consequence, goals for imprecision should be
viewed as goals for total laboratory error. Further experimental work on biologic variation is required to develop a wider
range of goals for hematology and coagulation tests, if these
strategies are considered worthy of adoption. (Key words: Accuracy; Precision; Allowable error; Biologic variation; Hematology tests; Method evaluation) Am J Clin Pathol 1987; 88:
667-669
RECENTLY, Hackney and Cembrowski 14 confirmed
our published findings13 that method comparison studies in the literature of clinical chemistry either omit useful statistics or rely too heavily on inappropriate tests. A
new finding was that evaluations of hematology and
coagulation tests were generally even less rigorous than
clinical chemistry evaluations; of particular interest,
only one evaluation compared error with any type of
desirable performance standard or analytic goal, and
that was in the field of clinical chemistry.
It was stated that, with the exception of hemoglobin,
there are no accepted estimates of allowable error for
hematology and coagulation tests. There are undoubtedly a number of plausible reasons for this situation.
However, because many of these tests are performed on
multichannel automated equipment and are reported as
single numeric results, there appears to be little reason
why the considerable body of work on analytic goal-setting in clinical chemistry could not be straightforwardly
adapted for the other disciplines that make up clinical
pathology.
In this article, therefore, strategies for defining analytic goals are summarized and numeric desirable per-
Received November 5, 1986; accepted for publication February 27,
1987.
Address reprint requests to Dr. Fraser: Department of Biochemical
Medicine, Ninewells Hospital and Medical School, Dundee DD19SY,
Scotland.
Department of Biochemical Medicine, Ninewells Hospital and
Medical School, Dundee, Scotland
formance standards for common hematology tests delineated.
Imprecision
Background
Analytic goals for the imprecision of clinical chemistry tests have been developed with the use of (1) reference values; (2) the state of the art, usually evidenced
from interlaboratory quality assessment schemes; (3) the
opinions of clinicians, most often assessed by questionnaire; (4) the views of expert persons or groups; and (5)
biologic variation data. The details of these strategies,
and their advantages and disadvantages, have been previously discussed in depth. 7,9 A further review of these
approaches, and of analytic goals for other performance
characteristics and matrices other than serum or plasma,
has been recently published, using glucose as an
example.''
Until now, goals for hematology tests have been determined only by seeking the opinions of clinicians. '•6-20-21 Most recently, Skendzel and colleagues21
proposed the goals for hematology tests shown in Table
1. However, this study has been criticized on a number
of seemingly cogent grounds, 3,10,18 and it should be realized that, as stated by Batsakis,2 recognizing the very
subjective quality of clinical opinions, formulation of
analytic goals by collation and quantitation of such surveys is not very productive and is often misleading.
In clinical chemistry, the current consensus view regarding the setting of analytic goals for imprecision is
based upon the statistical postulates of Harris. 15 These
state that, for comparison of a single result with reference values, or for assessment of serial results obtained
on a person, the analytic coefficient of variation (CV)
should be equal to or less than half of the average intraindividual biologic variation (which is the average
fluctuation around the homeostatic setting points of
persons, assuming that the fluctuation is truly random).
667
AJ.C.P. • November 1987
FRASER
668
Table 1. Analytic Goals Derived from a Questionnaire
on the Views of Clinicians
Analyte
Goal (CV, %)
Hemoglobin
Hematocrit
Leukocytes
MCV
Prothrombin time
3.6
5.4
16.4
3.2
15.2
The strategy, which can be expressed as follows,
CV
< J- . CV
^ * analytical ^ 2
^
intraindividual s
has been widely suported by clinical pathologists worldwide, the Aspen Conference of the College of American
Pathologists,4 and the World Association of Societies of
Pathology.23 Thus, although possibly not the theoretically ultimate strategy, and inapplicable to certain tests
such as assays used in therapeutic drug monitoring for
which goals should be set with the use of pharmacokinetic data12 (the elimination half-life and the time interval between doses), it is proposed that this strategy be
adopted for hematology and coagulation tests.
Biologic Variation Data
There are much data on the components of biologic
variation of analytes of interest to the clinical chemist,
and these have been superbly collated and tabulated by
Ross.17 In contrast, few numeric data are available for
hematology and coagulation tests; these have been summarized by Statland and Winkel.19 Although some parameters have diurnal or other biologic rhythms, disection of the components of variation according to the
concept that they vary around homeostatic setting
points has led to the average intraindividual biologic
variations shown in Table 2. In accordance with the
Aspen Conference of the College of American Pathologists,4 these data from Statland and Winkel19 can be
used to delineate analytic goals for imprecision.
It could be argued that the relevance of average intraindividual biologic variation to the presence or absence
Table 2. Average Intraindividual Biologic Variation
and Derived Objective Analytic Goal
Biologic Variation
Analyte
(CV, %)
Goal (CV, %)
Hemoglobin
Hematocrit
Platelets
Leukocytes
Neutrophils
Lymphocytes
Monocytes
Eosinophils
Basophils
2.4
2.5
6.6
15.6
24.6
11.0
16.2
21.1
13.2
1.2
1.3
3.3
7.8
12.3
5.5
8.1
10.6
6.6
of disease on health is tenuous. However, studies on the
biologic variation of clinical chemistry tests results in
chronic disease states,9 and certain haematology test results in multiple sclerosis and chronic pyelonephritis,16
have suggested that the data derived from the healthy
patients are valid in such conditions. Moreover, more
recent data published by Costongs5 are in good agreement—as far as comparison is possible—with those of
Statland and Winkel.19
Analytic Goals
Analytic goals for the imprecision of hematology tests,
derived with the use of the theories of Harris15 and the
data of Statland and Winkel,19 are shown in Table 2.
These goals, as in clinical chemistry, are more stringent
than those derived from the opinions of clinicians7,9
(Table 1), possibly because the clinician incorporates
preanalytic, biologic, and other sources of variation into
estimates of test result variability, not only analytic variation.918
Although some would undoubtedly argue that such
goals are unrealistic and of little relevance to the ways in
which clinicians use tests, they are required to ensure
that (1) the single numeric result has little random error,
(2) changes in serial results from a person do reflect real
changes in homeostasis, (3) reference values are not unduly widened by analytic variation, and (4) other performance characteristics can be better defined.9
Such goals can be used9 to (1) assess current laboratory performance and then direct available resources to
improvement of those methods that do not attain or
surpass goals, (2) set criteria for acceptance or rejection
of analytic batches instead of traditional statistical quality control rules used with the reference sample technic;
(3) judge whether new methods, instruments, or reagent
kit sets are suitable for adoption in the laboratory8; and
(4) facilitate various aspects of laboratory management.
Inaccuracy
As stated previously,7'9'23 the goal for the inaccuracy
of all clinical laboratory tests is that no bias should exist.
This simple desirable standard of performance is also
required here so that hematology test results are comparable over time (as methods change) and over geography
(because tests may be done in different hospital laboratories, clinics, and physicians' offices). Moreover, comparison of test results with reference values and fixed
agreed interpretative criteria can only be done correctly
if no bias exists.
Total Analytic Error
As advocated by many, and recently reemphasized
byHackney and Cembrowski,14 the most important goal
Vol. 88 • No. 5
AWARD LECTURES AND SPECIAL REPORTS
22
to consider is that for total analytic error (imprecision
plus inaccuracy). Because the goal for inaccuracy is that
of no bias, analytic goals for imprecision should be used
as goals for total analytic error.
Discussion
Few analytic goals have been proposed for hematology tests. It is considered that the models currently
deemed to be of most value in clinical chemistry can be
validly transferred to other disciplines of clinical pathology. Goals for imprecision are best based upon intraindividual biologic variation data, and it is hoped that
further experimental work will be undertaken to increase our quantitative knowledge of these. The only
logical goal for inaccuracy is that methods should have
no bias and, in consequence, goals for imprecision are
the goals for total analytic error.
It is hoped that this proposal will stimulate discussion
on the question of deciding what the desirable standards
of performance are for hematology tests.
References
1. Barnett RN: Medical significance of laboratory results. Am J Clin
Pathol 1968;50:671-676.
2. Batsakis JG: Analytical goals and the College of American Pathologists. Am J Clin Pathol 1982; 78:678-680.
3. Boerma GJM: Questionnaire highly interesting: Conclusions
make little sense. Am J Clin Pathol 1986; 85:392.
4. CAP Aspen Conference 1976: Analytical goals in clinical chemistry. Edited by FR Elevitch. Skokie, College of American Pathologists, 1977.
5. Costongs GMPJ: Intra-individual variations and critical differences of clinical laboratory parameters. PhD thesis. The Netherlands, Rijksuniversiteit Limburg te Maastricht, 1984. Published in J Clin Chem Clin Biochem 1985; 23:69-76 and J Clin
Chem Clin Biochem 1985; 23:405-410.
669
6. Elion-Gerritzen WE: Analytic precision in clinical chemistry and
medical decisions. Am J Clin Pathol 1980; 73:183-195.
7. Fraser CG: Analytical goals in clinical biochemistry. Prog Clin
Pathol 1980;8:101-121.
8. Fraser CG: Acceptable performance standards for clinical laboratory methods. Journal of Automatic Chemistry 1982; 4:1-2.
9. Fraser CG: Desirable performance standards for clinical chemistry
tests. Adv Clin Chem 1983; 23:299-339.
10. Fraser CG: Better criteria for desirable laboratory performance
exist. Am J Clin Pathol 1986; 85:251.
11. Fraser CG: Analytical goals for glucose analyses. Ann Clin Biochem 1986; 79:379-389.
12. Fraser CG: Desirable standards of performance for therapeutic
drug monitoring. Clin Chem 1987; 33:387-389.
13. Fraser CG, Singer R: Better laboratory evaluations of instruments
and kits are required. Clin Chem 1985; 31:667-670.
14. Hackney JR, Cembrowski GS: Need for improved instrument and
kit evaluations. Am J Clin Pathol 1986; 38:391-393.
15. Harris EK: Statistical principles underlying analytic goal-setting in
clinical chemistry. Am J Clin Pathol 1979; 72:374-382.
16. Raun RE, Moller BB, Back U, Gad I: On individual reference
intervals based on a longitudinal study of plasma proteins and
lipids in healthy subjects, and their possible clinical application.
Clin Chem 1982;28:294-300.
17. Ross JW: Evaluation of precision, CRC handbook of clinical
chemistry, vol 1. Edited by M Werner. Boca Raton, CRC Press,
1982, pp 391-422.
18. Ross JW: Goals for allowable analytic error better based on medical usefulness criteria. Am J Clin Pathol 1986; «5:391-392.
19. Statland BE, Winkel P: Effects of preanalytical factors on the
intraindividual variation of analytes in the blood of healthy
subjects: Consideration of preparation of the subject and time
of venepuncture. CRC Crit Rev Clin Lab Sci 1977; 10:105144.
20. Skendzel LP: How physicians use laboratory tests. JAMA 1978;
293; 1077-1080.
21. Skendzel LP, Barnett RN, Piatt R: Medically useful criteria for
analytic performance of laboratory tests. Am J Clin Pathol
1985;83:200-205.
22. Westgard JO, Carey RN, Wold S: Criteria for judging precision
and accuracy in method development and evaluation. Clin
Chem 1974;20:825-833.
23. World Association of Societies of Pathology: Proceedings of the
Subcommittee on Analytical Goals in Clinical Chemistry. Analytical goals in clinical chemistry: Their relationship to medical
care. Am J Clin Pathol 1979; 71:624-630.