Analytical Bias in a Quality Control Scheme - Clinical Chemistry

Analytical Bias in a Quality Control Scheme
John R. Allen,* Rachel Earp, E. Christis Farrell, Jr.,
and H. D. GrUmer
A quality control program utilizing both “known” and “blind” control specimens
was analyzed in the routine clinical chemistry laboratory. The results obtained with
the control samples of 18 automated and nonautomated procedures demonstrated
the presenceof analytical bias. Only through the evaluation of blind control samples
tested at random times can a reliable measure of the proficiency of the laboratory
be achieved.
‘THE CLINICAL CHEusmY Laboratories of The Ohio State University
Hospitals perform approximately 500,000 analyses per year, including
24,000 control specimens. A considerable number of these control specimens
are sent through the laboratory as “blind” samples. When
reviewing our control data, we wondered whether a systematic difference
appears between the data derived from known and unknown
control samples. If no significant divergence between the 2 groups
could be found in our laboratory, it would permit us to reduce the
number of “blind” samples. When reviewing the literature we found
to our surprise only one reference dealing with this subject matter in
a clinical laboratory, and this one reported tile absence of bias between
known and unknown control results (i).
The purpose of this h-t’portis to make available our data on known
and unknown control samples and to report the presence of bias. In
addition, we would like to present our standard deviations and coefficient
of variation of a variety of tests from a single identified source.
From the Division of Clinical Cheuiiistu-y, Departmont of Pathology, The Ohio State Uni.
versity, Columbus, Ohio 43210.
Presented at the meeting of the American Association of Clinical Chemists, Washington,
DC, August 1968.
Thanks are due Dr. Ransom Whitney, Director of the Statistics Laboratory, and Dr. Martin
Keller’s Biometric Group of tIn’ Department of Preventive Medicine, The Ohio State
U niversity.
Received for publication %OV 29, 1968; accepted for publication July 15, 1969.
Present address: Clinical Chemistry Section, Department of Pathology, Decatur Memorial
hospital, 2300 N Edwod St, Decatur, Ill 62526. -
1039

1040 ALLEN fT AL Clinical Chemistry
Methods
The presently used quality control method was instituted 2 years
ago with assistance from The Ohio State I. lilversity Statistics l)ivisioii
(2-4).
Control samples analyzed in the laboratory are of three types. There
are samples the technologist knows are controls: These are of pooled
serum processed and stored frozen. After the test has beeii standardized,
only the first daily result of the 1)001 for every constituent is
used for statistical analysis. Tilell there are coniinerical control
samples, used at the discretion of the technologist; these results are
not kept on record. Finally, there are control samples unknown to the
technologist, consisting of commercial control samples at two or three
levels of concentration. Oniy the results of the control having a concentration
approximating that of the pool have been tabulated (Tables
land 2).
Control samples, not known as controls to tile technologist, were
submitted at randoni times during the work day with a sillluiated
patient ielentilication. The time was determined by a table of ratidom
numbers. Results of the pool as well as the unknown commercial controls
were plotted daily on conlputer-prepared control charts indicating
the mean and tile permissible deviation. Values outside the prescribed
limits were brought to the attention of the supervisor. The limits used
to evaluate daily performance were usually the 2 standard deviations
calculated over a 3-month period. The number of analyses is about fit)
for each test. No changes iii methodology occurred during this -niontii
period for any one test.
Results
Results of control analysis over a period of 6 months showed differences
between sonic of tile known and unknown controls (Tal)le 1).
rflle mean values of the known calcium pool and tile blind control for
calcium, for instance,differ by about 1% while the standard deviation
differsby a factor of about 2.
1xanumation of time results of 18 tests in Table 1 shows that the
known control samples for 16 of the 18 tests had a smaller standard
deviation and 15 of tile 18 tests had a smaller coefficient of variation
than the unknown controls. The prohalility of this occurring by cilance
alone is less than 0.001 for the standard deviation and less than 0.01
for the coefficient of variation in a one-tailed test. Since the data did
not follow a parametric or normal distribution, a nonparametric statistical
analysis, that is, a binomial distribution, was employed for the
calculations.

Vol. 15, No. II. 1969 ANALYTICAL BIAS 1041
Table 1. ItEsuhis wirti BLIND (\ERsroI.) ANn KNOWN (Pool.) C sraoi.s .r
Onio STATE UNIVERSITY HOSPITALS, 0’T. 1967-MARcH 1968
Specimen .Wean
Versatol
SI) CV .Vo. Mean
Pool
SD CV No.
Albumin 3.4 0.16 2.9 127 4.3 0.13 3.0 133
Calcium 4.8 0.10 40 127 4.2 0.09 2.2 164
Chb)ridet 102 1 .6 1 .6 127 104 0.8 0.8 182
Creatituiuuet 1.7 0.11 6.3 123 1.9 0.08 4.4 183
(iIobiJin 2.2 0.12 5.7 129 2.6 0.13 5.6 137
(;lucoset 82 :1.4 4.2 126 116 2.7 2.3 183
r.P. 4.3 0.30 6.7 125 4.0 0.21) 4.9 126
Kt 3.0 0.13 3.0 126 4.7 0.12 2.5 183
Nat 140 2.4 1.7 127 143 1.2 0.8 l83
Tot, protein 7.6 0.19 2.5 12$ 73) 0.17 2.4 136
Urea-Nt 12.5 0.8 6.6 125 25.2 0.8 3.0 183
Uric Acidt 3.9 0.27 4.5 173 5.6 0.22 4.0 182
Cholesterol 224 13 6.5 127 200 7.0 :1.3 124
Alk. p’tase 6.9 0.4:1 6.2 126 2.6 0.15 3.6 127
Amylase 230 17 6.9 127 103 10 9.7 173
LDH 303 52 10.4 12(1 261 24 9.3 131
SOOT 133 12 9.3 121 23.7 3.2 13 129
SGPT 10.7 3.0 28 111 18.9 3.1 16 121
* Ntunber of analyses for each constituent is equal to the number of days on which the given
procedure ha.s beemi reported.
f Automated procedures.
Examination of each of the seven automated tests shows that the
average of the monthly standard deviations and of the monthly coefficients
of variation of the unknown control samples was greater
than the known samples in all cases (7+). The probability of this
occurring by chance is less than 0.01.
Of the 11 nonautomated tests, tile standard deviation of tile unknown
controls was greater than that of known controls in 9 instances, and
tue coefficient of variation was greater in eight tests. The probability
of this occurring by chance is 0.027 for the standard deviation arid
0.081 for the coefficient of variation.
Discussion
The data in Table 1 illustrate tile variation between “known” controls
and ‘‘blind’’ controls in several routine clinical chemistry tests.
The “blind” sample, more like that of a patient’s specimen, measures
Ilie precision actually achieved Ill the laboratory, in contrast to
“known’’ controls which measure tile precision the laboratory is
capable of achieving. It should 1)e emphasized that these data were
collected by doing just one set of control samples a day, and except
for the known control samples chosen by the technologist for checking

1042 ALLEN fT AL Clinical Chemistry
himself, he was not aware that lie was using a control sample until
after the work was completed.
Table 2 presents the coefficient of variation of’ some constituents
that appear on Table 1 with those data reported by Straumfjord and
Copeland (5). These authors made no mention as to whether their data
were derived from known or unknown samples or a combination
thereof. Thus, for instance, we might have a problem with our manual
phosphorus procedure if Straumfjord and Copeland’s data (5) are
derived from unknown samples. The calcium shows a difference between
the known and the unknown specimen; the known value compares
favoi’ably with the coefficient of variation of these authors. The
amylase results are contrary to what we expect with regard to coefficient
of variation of the known and blind specimens, and are probably due
to the higher mean of the unknown sample (250 versus l0 Somogyi
units for the known sample).
For the remainder of the procedures our coefficient of variation for
the known samples is in the same range as those reported by
Straumfjord and Copeland. Although the occurrence of bias in the
imonautomated teclinics group, as a whole, was not significant at the 5%
level, this does not exclude the possibility that bias may occur in any
one method. The ease with which a binomial sign test can be carried
out recommends this method as a useful tool for recognizing differences
in handling of control samples in a group of procedures.
Since we are also reporting the presence of bias in fully automated
tests, where, at least in theory, bias should be negligible, one wonders
how this bias is introduced. Instrumental inexactness of the sampler
phasing in the AutoAnalyzer system has been reported (6, 7). If known
control samples are introduced directly after the standards, as in our
case, arid only then reported by the medical technologist after he has
Table 2. CoMPARISON OF COEFFICIENT OF \ARIATION:
OHIO STATE U NIVERSITY IIosPcrus
(OcT 1967-MARCH 1968) AND STRAUMFJO RI) & Cul’EJ.AND
Specimen
O.S.U.I!. Slraumfjord&
uaknown - Copeland
Cholesterol 6.5 .5. 1 3.5
Alkaline p’tase 6.2 10.6 5.6
Amylase
6.9 8.1
9.7
Calcium
4.0 2.3
2.2
Chloride’
1.6 1.0
0.8
Glucose
4.2 3.1
2.3
I.P.
6.7 4.9
49
Potassium 3.0 2.5 2.5
Sodium 1.7 1.1 - 0.8
().S.U.II.
known

Vol. 15, No. II, 1969 ANALYTICAL BIAS 1043
convinced himself of initialsatisfactoryperformance of the procedure,
these samples failto unmask instrument driftsand specimen interaction
(8). Amenta (8) also reported twofold differences in standard errors
between samples which in each run were placed in the same position,
usually directly after tile standard and duplicate samples which were
placed at random. It is generally acknowledged that drifts occur, as
evidenced by shifts of known control results. Unfortunately, under the
pressure of daily routine and the boredom of reading peak after peak
from the recorder, it is virtually impossible to correct for drifts unless
the error becomes of such magnitude that the elimination of the
underlying malfunction becomes mandatory, or if the error is of lesser
magnitude, it call be compensated, for instance, with a computer-aided
program. We feel that, at least ill part, a similar situation can occur
for automated procedures as that described for manual technics by
Robinson (9).
Addendum
The methods used in this laboratory and listed in Table 1 are
itemized below. Some modifications have been made in some of the
technics; therefore, reference to the original literature only implies
the underlying principle of the technic used.
Albumin Kingsley, C. B., J, Biol. Cheni. 133, 731 (1940).
I1ove, P. E., J. Biol, Chem. 49, 109 (1921) ; 57, 235 (1923).
Calcium Dielul, H., and Ellingboe, J. L., Anal. Chem. 28, 811 (1946).
Schwarzenbach, C., Helu. C/jim. Acta 29, 811 (1946) ; 30, 1798 (1947)
32, 1314 (1949); 37, 113 (1954).
Chloride AutoAnalyzer modification (N-5a) of the method of Zall, D. M., Fisher, 1).,
:uuid Garner, M. 0., Anal. Chenu. 28, 1665 (1956).
Creatinine AutoAnalyzer modification (N-ha) of the method of Folin and Wu, taken
fu-oun Hawk, Oser, and Sumnmerson, Practical Ph ysiological Chemistry
(ed. 12). Blakistone, p 506.
Globulin Same as albumin.
Glucose AutoAnalyzer modification (N-2a) of the method of Hoffman, W. S., J.
Biol. Chem. 120, 51 (1937).
1.1’, Shuiuiow:uu-a, C., Jones, L., ajud Reinluart, 11., .1. L’iol. Chum. 142, 921 (1942).
K and Na Auto,\uialyzer technic N-20a, Technicon Coup, Ardsley, NY.
Protein Iliuret reaction, total protein and albumin ; Gornall, A. C., Bardewell, C. J.,
and 1)avid, M. M., J. Biol, Chem. 177, 751 (1949).
(le Ia Huerga, J., Smatten, G. W., and Sherrick, J. C., Assoc. Clin. Scientists
App1. Seminar on Proteins, Oct 1963, Washiiiigtouu, DC.
Urea-N - AutoAnalyzer method for the determination of urea nitrogen in serum
(N-la), Skeggs, L. T., Am. J. Clin. Path. 28, 311 (1957).
Marsh, W. H., Fingerhut, B., and Kirsch, B., Amer. J. Clin. Path. 28, 681
(1957).
Uric acid - AutoAnalyzer method N-13a. Technicout Corp, Ardsley, NY.
Cholesterol Leffler, H. H., liner. J. Clin. Path. 31, 310 (1959).
Ziathis, A., Zak, B., and Boyle, A. J., J. Lab. CUn. Med. 41, 486 (1953).

1044 ALLEN FT AL Clinical Chemistry
Alk. phosphatase hiabson, A. L., Guculey, S. .T., Coleman, C. M.,.utiul I’luihipps, C. B., Clin.
(‘hem. 12, 484 (1966).
Amylase Somogyi, 0., J. Biol. Chein. 125, 399 (1938).
henry, B. ,I.,Clinical Chemistry, Principles and Technics, hloeber, New
York, 1964, p. 471.
LDH ltergmeyer, II. U., Methods of Enzymatic Analysis. Acad Press, New York,
1963, p. 738.
LDH Wr#{243}hheski,F., and LaDue, J. S., Proc. Soc. Exp. Thai. Med. 90, 210 (1955).
GOT . GPT Karmen,
(1955).
A., WrThleski, F., and LaDue, J. S., J. Clin. Invest. 34, 126
Karmeuu, A., .1. Cliv. Invest. 34, 131 (1955).
Gilford recording spectropliotometer with automatic cuvet positioner and
constant temperature at 25#{176} (±0.5n).
Labeled values of control serum prepared by Warner-Cliilcott Laboratories:
calcium, 5.1 mEq/L; chloride, 101.3 mEq/L; creatiriine, 1.6
mg/100 ml; glucose, 85 mg/100 ml; inorganic phosphorus, 4.0 mg/
100 ml; potassium, 4.9 mEq/L; sodium, 140 mEq/L; total protein,
7.0 g/100 ml; urea nitrogen in serum, 12.1 mg/100 ml; uric acid, 6.2
mg/100 ml; cholesterol, 205 mg/100 ml; alkaline phosphatase, 66 TJ or
6.6 OSU-units; amylase, 262 Somogyi units: labeled value for LDII,
not available in Wr#{243}hleskiunits; SGOT, 157 Karmen units: 110 data
given for SGPT.
References
1. Weinberg, M. S., auni Tlau-uiett, 11. N., Absence of analytical bias in a cinahity control
program. Amer. ,I. Cliv. Path. 38, 468 (1962).
2. Tonks, D. B., A study of the accuracy and precision of chinienl chemistry dete,-uuuiii:utioius
iii 170 Canadian laboratories. Cliv. Chem. 9, 217 (1963).
3. Sax, S. M., Dorinan, L., Lihensen, 1). 1)., and Moore, J. J., Design and operation of an
expanded system of quality couitrol. Cli,u. Chem, 13, 825 (1967).
4. Conn, B. B., Jr., Best, M. L., Anido, V., and Cram, F. D., Computer analysis of quality
control data. Bnii. Req. Med. Technol. 37, 243 (1967).
5. Straumfjord, .1. V., .Jr., and Copeland, B. B., Clinical chemistry quality control values in
thirty-three university medical school hospitals. Amer. J. Cliii. Path. 44, 252 (1965’).
(1. Thiers, B. B., Cole, B. B., and Kirsch, W. J., Kinetic parameters of continuous finnanalysis.
Clin. Chem. 13, 451 (1967).
7. Young, D. S., Montague, B. M., and Snider, B. B., Studies of sampling tiunes in a
continuous flow analytic system. Cliii. Chenu. 14, 993 (1968).
8. Amenta, J. S., Analysis of variance, control charts, and the clinical laboratory. Anuer, ,J.
Clin. Path. 49, 842 (1968).
9. Robinson, R., Tedium and the law of dinuimuishingreturns in clinical biochemistry hiul,ou;utories.ASCP
Snvu?ulary Report, June 1967.