CE Update—Clinical Chemistry IV

Validation Protocol for New Methods or

Instruments in the Clinical Laboratory
Alfred E. Hartmann, MD

T he introduction of new instru-

ments and methodologies in the
clinical laboratory necessitates vali-
method is a considerable undertaking
and a time-consuming task that may
well be beyond the resources of many
turers with the resources to establish
accuracy and precision claims for their
products. They were never intended
dation studies of accuracy and preci- laboratories in regard to the afore- to be used in clinical laboratories for
sion to ensure that the new method mentioned factors. t h e v a l i d a t i o n of m a n u f a c t u r e r s '
will meet acceptable standards of per- Numerous articles have appeared claims.
formance. The e v a l u a t i o n of new in the literature concerning product Recently, NCCLS has concentrated
methods is approached in various ways evaluation. 1 8 Evaluation protocols on writing protocols designed to be
in different laboratories. All too often, have been presented, but generally used by the enduser (laboratorian) for
many laboratories begin routinely us- these are extensive and suitable only verification of accuracy and precision
ing new methods for analysis of pa- for the largest laboratories with plen- in the clinical laboratory. These pro-
tient specimens after only a short tiful resources. For example, almost tocols concern precision, linearity, in-
familiarization period and without any all the protocols involve a method terference, recovery, comparison of
formal evaluation or documentation comparison experiment. 9 1 3 These pro- methods, and method validation
of precision or accuracy. Other labo- tocols are worthwhile to assess accu- schema. The first of these new pro-
ratories complete an extensive eval- racy, but are of limited value for the tocols has been published as EP5T,
uation of each new method consuming majority of laboratories in the United Tentative Guidelines for User Evalu-
several months and evaluate in great States, where more t h a n one half of ation of Precision Performance of
depth precision, accuracy, interfer- hospital laboratories are in hospitals Clinical Chemistry Devices." This ar-
ence, linearity, recovery, and compar- less than 100 beds in size, and where ticle summarizes some of the princi-
ison with a reference method. In any the new method is frequently much ples discussed in the preparation of
given laboratory, the nature of the better t h a n the available existing these protocols and presents the crit-
method evaluation depends on the ex- methods. Other protocols have been ical factors t h a t need to be evaluated
pertise and interest of the personnel, utilized that are even too burdensome before a method is used to generate
time constraints, laboratory size, re- for larger laboratories. 1415 The Na- clinical results. It is not intended as
sources available in terms of equip- tional Committee for Clinical Labo- a definitive e v a l u a t i o n of a new
ment, reagents, previous methodology, ratory Standards (NCCLS) protocols method, but rather an assessment of
and financial factors such as cost per (EP2T, EP3T, EP4T) 16 have been used a method relative to the claims of the
test for the new method and budget to evaluate several methodologies in manufacturer. It is assumed t h a t the
committed for product evaluation. A the literature with excellent, statis- laboratory has accepted as reasonable
laboratory should do more than sim- tically valid results with high confi- the manufacturer's claims of preci-
ply familiarize itself with a method dence limits; however, these protocols sion and accuracy listed in the prod-
before clinical i n t r o d u c t i o n even generate voluminous data. They are uct labeling or as reported in the
though the full evaluation of a new time-consuming, expensive, and re- literature by more extensive method
quire computer handling of data. These e v a l u a t i o n s and h a s selected t h e
From McKennan Hosp. Sioux Falls, SD 57101. protocols were designed for manufac- m e t h o d w i t h a k n o w l e d g e of t h e
LABORATORY MEDICINE • VOL. 14, NO. 7, JULY 1983 4 1 1
Downloaded from https://academic.oup.com/labmed/article-abstract/14/7/411/2642914
by guest
on 29 March 2018
 claimed precision and accuracy that tal imprecision, such as fluctuations (2) to assign quality control limits for
will satisfy the clinical needs of that of line voltage, wear and tear of the the daily quality control material that
institution. 1 8 2 0 Details of fairly sim-
instrument, reagent lot variability, will be used to assess random error.
ple experimental protocols are pre- multiple calibrations, technologist in- The total imprecision protocol in-
t e r a c t i o n , and other physical and
sented, t o g e t h e r with a m e a n s of volves analyzing a stable specimen in
assessing whether the results indi- chemical changes. 21 Before commenc- duplicate each day over at least a 20-
cate that acceptable performance has ing with these reproducibility stud- day period. The data are tabulated as
been achieved in the user's hands. The ies, samples need to be obtained that in Table I of the Appendix. ANOVA
protocols described in this article may will be stable for the time period of statistics are utilized to determine the
not be universally applicable or rele- the study and be similar in character w i t h i n - r u n t o t a l imprecision esti-
vant to all methods, and some parts to the patient samples to be analyzed mates. 2 2 The within-run imprecision
may require modification before they (ie, serum for serum, urine for urine, estimate (Swr) is determined by cal-
can be applied to a particular method. aqueous for CSF). These samples can culating the square root of the vari-
be lyophilized control samples, stabi-
This is p a r t i c u l a r l y t r u e for new ance of the duplicates with 20 degrees
methods t h a t are poorly documented lized liquid control samples, or pooled of freedom (number of duplicates ana-
or extensively studied by colleagues. patient samples. If only one specimen lyzed). The total imprecision (ST) is
Nonetheless, the laboratory should at is used, the concentration of analyte determined by the square root of the
a m i n i m u m assess reproducibility, should be near a medical decision limit variance of the daily mean (B) and
accuracy, linearity, and reference in- and near the manufacturer's level of within-run imprecision estimate (Swr)
tervals for all new methods. determined precision. The reproduci- using the formula in the Appendix.
bility of m a n y a n a l y t i c a l methods Bauer and Kennedy 22 have written a
Familiarization varies with concentration, sometimes computer program in BASIC for this
by more than threefold over the range data analysis procedure. The deter-
The first step in any method eval- encountered in clinical practice. 5 A mination of the degrees of freedom (T)
uation is for the evaluators to be more complete protocol therefore for the total imprecision estimate is
trained and familiar with the method. would use three specimens with an- complex and relies upon computation
This should include studying the op- alyte concentrations in the low, nor- by S a t t e r t h w a i t e ' s formula. 2 2 The
erating instructions, writing up the mal, and high ranges. A quantity of within-run imprecision and total im-
method procedure in a logical oper- each specimen sufficient for at least precision e s t i m a t e s are t h e n com-
ating protocol, receiving training from 60 analyses and several practice runs pared to the manufacturer's claims. If
the manufacturer's representatives, should be obtained and stored appro- the calculated standard deviations are
and a n a l y z i n g a few samples and priately to ensure stability. A com- less t h a n or equal to the claimed im-
q u a l i t y control samples to become mon error is to run out of the required precision, the manufacturer's claims
aware of any idiosyncrasies in the samples before completion of the ex- are validated. If the calculated stand-
method. 12 At this time, quality con- periment. Several protocols have been ard deviation is g r e a t e r t h a n t h e
trol specimens should be selected that designed to assess precision.5-91213 The claimed standard deviation, chi-square
will be used subsequently to assess protocol presented here is an adap- (x2) statistics are utilized to deter-
imprecision when clinical samples are tation of the NCCLS draft protocol for mine if there is a statistically signif-
analyzed. user evaluation of precision perform- icant difference. 22 The x2 statistic is
ance (EP5T). 17 An experiment assess- determined by dividing the variance
Reproducibility ing the within-run imprecision is first of the determined estimate by the
accomplished by analyzing at least 20 variance of the claimed estimate and
The first operational item to be val-
aliquots of a single specimen and de- multiplying by the degrees of freedom
idated is reproducibility. Good preci-
termining the standard deviation and for the determined estimate. The cal-
sion is perhaps the most important
attribute of an analytical method. If
coefficient of variation of the results. culated x2 should be compared with a
the method fails the reproducibility These results are then compared to statistical table of x2 values using the
claims, any other data that are gen- the manufacturer's claim of within- upper 95% critical value with the ap-
erated probably are invalid and there run imprecision and any considerable propriate degrees of freedom. If the
is no point in proceeding to other de- discrepancy should be addressed and calculated value is less than this tab-
terminations. Reproducibility is gen- solved before c o n t i n u i n g with t h e ulated value, then the manufacturer's
erally divided into within-run and evaluation. If this preliminary run is precision claims are validated. 17 These
between-day or t o t a l imprecision. satisfactory, one can proceed with the precision estimates can also be uti-
Within-run imprecision is the ran- total imprecision protocol. The total lized as preliminary control values for
dom error inherent in the method and imprecision study takes approxi- quality control specimens when the
is the optimum reproducibility that mately 20 days to run. Although the test is introduced for analyzing clin-
should be expected by the method. To- time commitment is great, the bene- ical samples.
tal imprecision differs from within-run fits of additional knowledge over and
imprecision in that assays are per- above the preliminary imprecision Linearity
formed over multiple runs on differ- study are worthwhile in assessing the
expected imprecision of the method. If the within-run precision study was
ent days and total imprecision is the
The two major benefits are (1) to val- satisfactory, one can proceed to as-
reproducibility t h a t should be ex-
pected from the method over a period idate the manufacturer's claims for s e s s i n g o t h e r p a r a m e t e r s while
of time. Many factors contribute to to- total and within-run imprecision, and simultaneously performing the total
4 1 2 LABORATORY MEDICINE • VOL. 14, NO. 7, JULY 1983 precision study. The second parame-

Downloaded from https://academic.oup.com/labmed/article-abstract/14/7/411/2642914

by guest
on 29 March 2018
 ter to evaluate is linearity. This is to most method evaluations, inaccuracy Comparison of Quantitative Clinical
identify the range of analyte concen- is assessed by a comparison of meth- Laboratory Methods Using Patient
tration over which the assay is linear ods in which patient samples are ana- Samples should be utilized. 26 This
or usable. 8 Frequently, one of the first lyzed w i t h t h e new m e t h o d and protocol entails at least 40 patient
indications of a deterioration of re- compared to results of a reference comparisons having a broad range of
agents or equipment is a decrease in method or a well-studied method of concentrations with a distribution of
the sensitivity of the method, with a known accuracy utilizing linear 20% low, 50% normal, and 30% high
corresponding decrease in linearity. regression statistics. Such a reference results. The patient comparisons are
Linearity can be assessed by analyz- method is not available for method evaluated visually on an x/y coordi-
ing a number of dilutions of a speci- comparison in the majority of labo- n a t e g r a p h a n d by u s i n g l i n e a r
men with a markedly elevated analyte ratories. An alternative method to regression statistics. A method com-
concentration. Obtaining a specimen validate the new method's accuracy is parison can help to relate one method
with an adequate concentration of an- to analyze several different speci- to another in terms of patient results
alyte can be the major problem in as- mens with known analyte values. 6 ' 7 and reference intervals when more
sessing linearity. Suggested sources These specimens can be obtained from t h a n one method will be used at var-
are standard reference materials from several sources, such as in-house ious times for the same analyte.
the National Bureau of Standards preparation from analytical grade re-
(NBS), reference standards from the agents, NBS certified standard ref- Reference Ranges
College of American Pathologists, as- erence materials, College of American Reliably obtained reference inter-
sayed commercial sera, and patient Pathologists reference materials, as- vals are as important as accuracy,
samples. Spiking patient pools with sayed commercially prepared mate- precision, linearity, and other para-
tissue extracts obtained from autopsy r i a l s , e x c h a n g e of s a m p l e s w i t h meters generally investigated for an-
or surgical material is particularly reputable laboratories having exper- alytical procedures. The reference
useful for enzyme studies. Dilutions tise in the area of interest, and per- interval should be derived from data
can be made with saline, water, or in- h a p s m o s t useful for s m a l l e r generated by using specimens from the
activated serum consistent with the laboratories, interlaboratory profi- a p p r o p r i a t e reference population
manufacturer's recommendations. ciency test samples, which can be (outpatient samples, hospitalized pa-
Enzymes should be diluted with low p u r c h a s e d t h r o u g h the College of tients, medical students, etc). The use
level serum to prevent matrix effects. American Pathologists Surveys pro- of blood bank donors for establishing
Appropriate dilutions are: 1:1, 7:8, 3:4, gram. These materials have consen- reference i n t e r v a l s h a s p a r t i c u l a r
1:2, 1:4, and 1:8. At least five concen- s u s v a l u e s o b t a i n e d by v a r i o u s m e r i t ; t h e donors a r e p r e s u m a b l y
t r a t i o n points should be analyzed methodologies, usually including the healthy and specimens are easily col-
throughout the analytical range and methodology u n d e r i n v e s t i g a t i o n . lected. 8 A reference population is not
each point should be analyzed in tri- Some analyte values have also been necessarily a normal population and
plicate to avoid spurious data gener- verified by definitive methods. the reference range may vary from
ation. The first step toward assessment The test protocol consists of analyz- population to population. To estimate
of the results involves graphic pre- ing at least 20 aliquots of one of the the reference interval with 90% con-
sentation on an x/y coordinate graph, above specimens and determining the fidence, at least 120 patient samples
should be analyzed and results plot-
plotting the experimental results along mean value. 2225 The determined mean
the ordinate (y axis) against expected ted using a histogram. 27 The shape of
value is then compared to the known
values on the abscissa (x axis). The the histogram will indicate whether
value. To assess whether the deter-
the underlying population is Gaus-
graph should show whether the mea- m i n e d m e a n is e q u i v a l e n t to t h e
sian. Of the methods devised to assess
surements follow a linear relation- known analyte value, we test the null
reference intervals, the use of non-
ship. If a quantitative measure of hypothesis t h a t Ho: determined mean
parametric statistics estimates the
nonlinearity is desired, Burnett 2 3 has (pC) = known mean (|xo). A signifi-
reference interval most accurately,
described a method using second-de- cance test is based on the quantity t
especially with non-Gaussian data. 28
gree polynomial regression that tests = (pi-(xo) H- (s/n1/2), which is simply the This protocol uses the nonparametric
the significance of the coefficient of bias observed divided by the standard p e r c e n t i l e e s t i m a t e procedure de-
the second-degree term. For accepta- error of the mean. We accept Ho if scribed by Reed and colleagues. 29 The
ble performance, the linearity range this determined t value is within the data is first ranked in order of mag-
determined should be comparable to t-distribution tabulated for n-1 de- nitude. The central 95% of all values
the manufacturer's claims. grees of freedom at a significance level ranges between the 2.5 and 97.5 per-
ofP = .05. If applicable to the method centile ranks. The estimate of the 2.5
Accuracy being studied, one can combine the percentile rank is the r th sample from
Inaccuracy is defined as a numeri- linearity experiment and accuracy the low end of values where r =
cal difference between the mean of a experiment together by using the same 0.025(N + 1) and N = total number
set of replicate measurements and the specimen data to assess both areas of of values. The 97.5 percentile esti-
true value. 24 Accuracy is probably the interest. mate is the r th rated sample from the
most elusive quality to determine in If the new method is to be compared high end of values. For most values
a method, due mainly to the difficulty to another method, a protocol such as of N, r will not be a whole number
of determining the true value for a that to be published by NCCLS as and interpolation between the two
component in a biologic specimen. In EP9P, Proposed Guidelines for User

LABORATORY MEDICINE • VOL. 14, NO. 7, JULY 1983 4 1 3

Downloaded from https://academic.oup.com/labmed/article-abstract/14/7/411/2642914

by guest
on 29 March 2018
 nearest samples is done. This pro- required to complete the evaluation methods in routine laboratory service. Clin Chem
1976;22:489-496.
vides the central 95% intervals re- protocol. After completion of the eval- 11. Brooks RJ: Clinical chemistry method documen-
gardless of the shape of distribution. uation and introduction of the method tation. Lab World May 1979, 30-33.
This method places great importance for clinical testing, an on-going eval- 12. Buttner J, Borth R, Boutwell JH, et al: IFCC
provisional recommendation on quality control
on data points near the reference lim- uation program should be instituted. in clinical chemistry. Part 2. Assessment of an-
its and may raise questions as to The method should be observed by the alytic methods for routine use. Clin Chem
1976;22:1922-1932.
whether extreme values are in fact usual quality control practices and 13. Westgard J, deVos DJ, Hunt MR, et al: Concepts
outliers that should be rejected. Out- particular emphasis placed on results and practices in the evaluation of clinical chem-
istry methods. Parts I-V. Am J Med Technol
liers can be determined by a modifi- of proficiency testing and survey pro- 1978;44:290-300,420-430,522-571,727-742,803-
cation of the rlO statistic proposed by grams as to biases between the new 813.
14. Siegel SR, Chesler R: Using proposed NCCLS
Dixon. 30 According to this method, a methodology and other laboratories protocol for evaluation of automated instru-
value can be rejected as an outlier if using the same method. The assess- ments. Am J Med Technol 1982;48:595-600.
15. Garber CC, Westgard JO, Milz L, et al: Dupont
the distance between it, X(N), and the ment of an analytical method is a con- aca III performance as tested according to NCCLS
next value, X(N-l) is greater than tinuing process and information about guidelines. Clin Chem 1979;25:1730-1738.
1/3 of the range of values. (See Ap- its characteristics accumulates grad- 16. National Committee for Clinical Laboratory
Standards. Tentative guidelines for manufac-
pendix for an example of such a cal- ually. Although the above protocol turers for establishing performance claims for
culation.) The reference interval that should not be considered an extensive clinical chemical methods. EP2T Introduction
and performance check. EP3T Replication ex-
is determined should agree closely evaluation of a method's accuracy or periment. EP4T Comparison of methods exper-
(within 10%) with what is in the lit- precision, it should prove worthwhile iment. Villanova, PA, 1982.
17. National Committee for Clinical Laboratory
erature or stated by the manufac- in assessing the manufacturer's claims Standards. EP5T Tentative guidelines for user
turer. and ensuring the method's reliability. evaluation of precision performance of clinical
chemistry devices. Villanova, PA, 1983.
With well-studied analytes, such as 18. James GP: Evaluation of clinical laboratory in-
sodium, glucose, etc, the determina- struments. Part 1. Defining laboratory needs.
Am J Med Technol 1981;47:225-228.
tion of the reference interval in-house 19. Barnett RN: Analytic goals in clinical chemis-
References
is not as critical as with analytes try: the pathologist's viewpoint. Pathologist
1. Barnett RN: Scheme for comparison of quanti- 1977;31:319.
known to have a high variability. In tative methods. Am J Clin Pathol 1965;43:562- 20. Ross JW, Fraser MD: Analytic clinical chemis-
laboratories with limited resources, the 569. try precision. Am J Clin Pathol 1977;68:130-
2. Westgard JO. Hunt M: Use and interpretation 141.
manufacturer's normal range can be of common statistical tests in method compari- 21. Bauer S, Kennedy JW: Applied statistics for the
utilized until data on these well-stud- son studies. Clin Chem 1973;19:49-57. clinical laboratory. Part I. J Clin Lab Autom
3. Taylor DW, Baird HW: Design for evaluation 1981;1:128-132.
ied analytes can be collected on an on- and justification of laboratory instruments. Am 22. Bauer S, Kennedy JW: Applied statistics for the
going basis and reference intervals J Med Technol 1979;43:139-142. clinical laboratory. Parts III-V, VII. J Clin Lab
determined. In larger laboratories, an 4. Nielsen LG, Ash KO: Protocol for adoption of Autom 1982;2:35-40,129-133,209-215,354-357.
analytic methods in the clinical chemistry lab- 23. Burnett RW: Quantitative evaluation of line-
in-house computer can easily compile oratory. Am J Med Technol 1978;44:30-37. arity. Clin Chem 1980;26:644-646.
these statistics. 5. Kim EK, Logan JE: Scheme for evaluation of 24. Buttner J, Borth R, Boutwell JH. et al: IFCC
methods in clinical chemistry with particular provisional recommendation on quality control
application to those measuring enzyme activi- in clinical chemistry. Part 1. General principles
ties. Part 1. General considerations. Clin Biochem and terminology. Clin Chem 1976;22:532-540.
Foliow-up 1978;11:238-243. 25. Bauer S, Kennedy JW. Applied statistics for the
6. Percy Robb IW, Broughton PMG, Jennings RD, clinical laboratory. VIII. Measure of inaccuracy.
et al: Recommended scheme for evaluation of J Clin Lab Autom 1983;3:46-48.
During the evaluation period, a log kits in the clinical chemistry laboratory. Ann 26. National Committee for Clinical Laboratory
of all steps taken and problems en- Clin Biochem 1980;17:217-226. Standards. Draft document EP9P Proposed
7. Broughton PMG, Gowenlock AH, McCormack guidelines for user comparison of quantitative
countered should be kept along with JJ, et al: Revised scheme for evaluation of au- clinical laboratory methods using patient sam-
all data generated. The laboratory's tomatic instruments in clinical chemistry. Ann ples. Villanova, PA, 1983.
Clin Biochem 1974;11:207-218. 27. Massod MF: Nonparametric percentile esti-
needs for validation of a new method's 8. James GP, Bee DE: Evaluation of clinical lab- mates of clinical normal ranges. Am J Med
accuracy and precision should be met oratory instruments. Part 2. Use and abuse of Technol 1977;43:243-252.
statistical tools. Am J Med Technol 1981;47:303- 28. Butts WC, Lilje DJ: Clinically significant ref-
by the above evaluations. However, 310. erence intervals. Am J Med Technol 1982;48:587-
on some occasions it is necessary to 9. Westgard JO, Carey RN, Wold S: Criteria for 594.
judging precision and accuracy in method de- 29. Reed AH, Henry RJ, Mason WB: Influence of
proceed further and obtain additional velopment and e v a l u a t i o n . Clin Chem statistical method used on the resulting esti-
data. It is estimated that a time com- 1974;20:825-833. mate of normal range. Clin Chem 1971;17:275-
mitment of approximately two hours 10. Westgard JO, Carey RN, Feldbruegge DH, et al: 284.
Performance studies on the Technicon SMAC 30. Dixon WJ: Processing data for outliers. Biomet-
daily over a period of one month is analyzer: precision and comparison of values with rics 1953;9:74-76.

4 1 4 LABORATORY MEDICINE • VOL. 14, NO. 7, JULY 1983

Downloaded from https://academic.oup.com/labmed/article-abstract/14/7/411/2642914

by guest
on 29 March 2018
 Appendix
Example of typical evaluation—glucose

1. Reproducibility Study

Day Result Result Run (Rep1-Rep2) 2

1 2 Mean
1 242 246 244 16
2 243 242 242.5 1
3 247 239 243 64
4 249 241 245 64
5 246 242 244 16
6 244 245 244.5 1
7 241 246 243.5 25
8 245 245 245 0
9 243 239 241 16
10 244 246 245 4
11 252 251 251.5 1
12 249 248 248.5 1
13 242 240 241 4
14 246 249 247.5 9
15 247 248 247.5 1
16 240 238 239 4
17 241 244 242.5 9
18 244 244 244 0
19 241 239 240 4
2p 247 240 243.5 49
Sum 289
daily mean 241.1
2(Rep1-Rep2) 2 "
SD (B) 2.98 Swr = 1/2
2(# days)
variance (B2) 8.88
289 " 1/2 = 2.69 = within-run imprecision
Swr =
2(20) _
variance S2wr = 7.225

N-1 „ , 2-1
ST = B2 + 1/2 = 8.88 + — (7.225) 1/2 = 3.53 = total imprecision
•ws wr
_ 0
where B2 = variance of daily means
S2wr = variance of within-run precision
N = number of replicates per run

Satterthwaite's formula to determine degrees of free dom (T) for total imprecision

[(N-1)S2wr + (N)(B2)]2 (7.225 + 2(8.88))2

[(N-1)(S2wr)]2 + [(N)(B2)]2 (7.225)2 + (17.76)2
I 1-1 20 19

where N = # replicates per run

# of days

Comparison to reproducibility claims:

Within-run imprecision: To al imprecision:
determined Swr= 2.69 de ermined ST= 3.53
variance S 2 wr= 7.225 ariance ST 2 = 12.46
claimed SD= 2.5 claimed SD= 3.4
2
variance SD 2 = 6.25 variance SD = 11.56
degrees of freedom = 20 degree;> of freedom = 32

Y2 S2wr(degrees of freedom) 7.225(20)

2
SD2 6.25

critical x2 at 95% level for 20 DF = 31.4

if determi ned x 2 < critical x2 = c l a i m
validated = 23.12 < 31.4 acceptable

continued on next page

LABORATORY MEDICINE • VOL. 14, NO. 7, JULY 1983 4 1 5

Downloaded from https://academic.oup.com/labmed/article-abstract/14/7/411/2642914
by guest
on 29 March 2018
 2 ST2(degrees of freedom) 12.46(32)
34.49
SD2 11.56

critical X at 95% level for 32 DF = 46.2

if determined X < critical X = claim
validated = 34.49 < 46.2 acceptable

2
Abbreviated X Table (95% level)
DF X critical value
5 11.1
10 18.3
15 25.0
20 31.4
25 37.7
30 43.8
35 49.8
40 55.8
50 67.5
60 79.0
70 90.5
75 96.2
80 101.9
90 113.1
100 124.3

2. Reference Intervals N = 120

lowest 5 values highest 5 values
72 112
76 112
78 118
78 120
80 124

r = .025(N + 1)
r = .025(120 + 1) = 3.03th samples from range

reference interval 78-118

range of values 124-72 = 52
is 124 outlier?

X(N) - X(N-1) 124-120

r = Hrrrr—^^rrrr 1 0.33 not an outlier
X(N) - X(L) > 0.33 outlier = ^124-72
rr^r = 0.07

3. Accuracy
specimen with known value of 100 mg/dL
N = 20

102 95
100 105
99.1
105 100
99 98
V
102 96 (X1-X)2
SD = 2 = 3.44
98 95 - n-1
99 105
100 102 M- - (10 99.8-100
t -- 0.26
105 98 Sd/(n)12 3.44/201's
95 97
t - 2.09 for 19DFat95%CL

0.26 < 2.09 thus acceptable

This is the final article in the current clinical chemistry CE Update series. An examination for CME
credit appears on page 450.

4 1 6 LABORATORY MEDICINE • VOL. 14, NO. 7, JULY 1983

Downloaded from https://academic.oup.com/labmed/article-abstract/14/7/411/2642914
by guest
on 29 March 2018