Article in press - uncorrected proof
Clin Chem Lab Med 2006;44(4):387–390 2006 by Walter de Gruyter • Berlin • New York. DOI 10.1515/CCLM.2006.066 2006/366
Opinion Paper
Modified Levey-Jennings charts for calculated laboratory
tests
Abdurrahman Coskun* tive procedure in which a stable reference sample
was analyzed repeatedly and individual measure-
Abant Izzet Baysal University, Duzce School of
ments were plotted directly on a control chart. This
Medicine, Department of Clinical Biochemistry,
reference sample type of quality control (QC) in which
Duzce, Turkey
individual values are plotted directly on a control
chart is commonly known today as a Levey-Jennings
Abstract (LJ) chart. LJ charts are visual materials and clearly
display quality control data over time. The control lim-
Background: In clinical laboratories, many test results its of LJ charts have usually been calculated using
such as low-density lipoproteins can be calculated standard deviation (SD) and mean (X) derived from
using the results of specifically measured tests. How- repeated measurements of control serum for which
ever, quality control programs deal with only meas- the concentrations are known.
ured tests and there is no quality control procedure Many test results, such as low-density lipoprotein
for calculated tests. I aimed to modify the Levey-Jen- (LDL), total globulins, unconjugated bilirubin, correct-
nings chart for calculated tests as part of a total qual- ed total calcium, international normalized ratio (INR),
ity system. creatinine clearance, plasma osmolality, anion gap,
Methods: I developed a method using a Taylor series and many hematological and different kinds of tests,
expansion and alternative equations to obtain the can be obtained using the results of specifically meas-
standard deviation of calculated tests and then pre- ured tests. However, checking only measured tests by
pared Levey-Jennings charts for these tests. QC procedures may not be sufficient for good labo-
Results: Preparation of Levey-Jennings charts for ratory management (3). In clinical laboratories, QC
low-density lipoproteins (and other calculated tests) programs deal only with measured tests, and unfor-
was simple, and using this chart increased the relia- tunately there is no QC procedure for any calculated
bility of calculated test results. tests. We check only the components of calculated
Conclusion: Checking only measured tests by quality tests, not the test itself (3). Thus, I believe that an
control procedures before reporting patient results approach to develop quality assessment procedures
may be inadequate. Preparation of Levey-Jennings for calculated tests would be beneficial to obtain more
charts for calculated tests as part of total quality man- accurate patient results. We should apply quality pro-
agement will increase the reliability of test results. cedures to all calculated tests and modify the soft-
ware in clinical laboratories to check the reliability of
Keywords: laboratory management; Levey-Jennings calculated tests as part of a daily routine QC proce-
charts; low-density lipoprotein (LDL); quality control; dure. Application of QC programs to calculated labo-
reliability. ratory data is introduced for the first time in this
paper. Today we use LJ charts only for measured
tests and there is no application of LJ charts to cal-
Introduction culated tests. In this study I discuss how to identify a
general way to apply LJ charts to all calculated tests.
Quality assessment is required not only for measured I also present an LJ chart for calculated LDL-choles-
tests, but also for all kinds of data obtained in the terol as a special example.
clinical laboratory, including calculated tests.
In 1950 Levey and Jennings introduced statistical
control methods in clinical laboratories (1). They pro- Method development
posed making duplicate measurements on a patient
specimen. This procedure was more difficult in clini- In clinical laboratories we use at least two different
cal application, since the actual concentration of the commercial control sera that have concentrations
measured constituent varied from specimen to speci- near important medical decision levels. To prepare LJ
men. Henry and Segalove (2) developed an alterna- charts, these materials are analyzed once per day for
a period of 20 days. X and SD are calculated from
*Corresponding author: Abdurrahman Coskun, MD, Abant these data (4). Control limits can be determined as
Izzet Baysal University, Duzce School of Medicine, X"2 SD or X"3 SD, according to the type of test and
Department of Clinical Biochemistry, Konuralp, Duzce,
the control serum.
Turkey
Phone: q90-542-2380639, Fax: q90-380-5414105, For calculated tests, preparation of LJ charts may
E-mail: coskun_a2002@[Link] be a complicated process, depending on the number
Brought to you by | Karolinska Institute
Authenticated
Download Date | 5/27/15 7:56 AM
Article in press - uncorrected proof
388 Coskun: Levey-Jennings chart for calculated tests
Figure 1 Flow diagram for preparation of an LJ chart for calculated tests (gsf(x,y,z)).
of components measured; before drawing the charts n
2
8Žgiyg¯¯.
we have to calculate the SD of the equation using the
SD for each of its components.
Consider a calculated variable g that is a function
SDgs y is1
Žny1.
w3x
of the random variables x, y, and z, i.e., gsf(x,y,z). If
the variances (SDi2) of x, y, and z are known, the var- This equation is more useful and gives more accu-
iance of g can be approximately obtained using a Tay- rate results than Eqs. w1x and w2x. Using SDg and g, it
lor series expansion of the function (5). If gsf(x,y,z) is easy to prepare LJ charts. Before preparation of LJ
is approximately linear with respect to x, y, and z in charts for g, we should check the LJ charts of x, y,
the region of interest, the approximate variance of g and z. All components of the equation should be with-
is obtained as: in their control limits. In addition to control limits, the
CV of g should be equal to or lower than the maxi-
mum allowable CV value recommended by national
SDg2((≠g/≠x)2SDx2q(≠g/≠y)2SDy2q(≠g/≠z)2SDz2 w1x or international laboratory organizations; otherwise,
LJ charts should not be prepared for g (Figure 1) and
the requested test should be measured using chemi-
where x, y and z are the individual means of each test. cal methods instead of calculation.
If the test result is calculated only by addition or
subtraction, we can choose a simple alternative meth-
od, as described below, to obtain the approximate
variance of calculated laboratory tests (6). Results and discussion
If gsxqyqz or gsx–y–z, then:
We prepared an LJ chart for calculated LDL-choles-
terol as an example. LJ charts could equally well be
SDg2sSDx2qSDy2qSDz2 w2x prepared and applied to other calculated tests.
Clinicians usually prefer a total lipid profile, includ-
ing triglycerides, total cholesterol, high-density lipo-
However, in clinical practice these equations might protein (HDL)-cholesterol and LDL-cholesterol. In this
not be reliable under all conditions, since they are val- situation, if serum triglycerides are lower than
id only if the variables (measured components of the 4000 mg/L (4520 mmol/L), obtaining an LDL-choles-
equation) are independent; if the variables are signif- terol result by calculation, using the relationship
icantly correlated, the SD of the components of the between total cholesterol, HDL-cholesterol, VLDL-cho-
equation cannot be used to calculate the SD of the lesterol and LDL-cholesterol, involves no cost. How-
equation. To overcome this problem, an alternative ever, if only LDL-cholesterol has to be measured, an
equation can be used to calculate the SD of the equa- enzymatic method is preferable to calculation. In our
tion using all data for each component of the equa- laboratory, the triglyceride levels of approximately
tion, not only the SD of the components. 98.5% of patients are under 4000 mg/L. Thus, we use
If gsf(x,y,z) and g is the mean, the SD of g can be enzymatic methods to measure LDL-cholesterol for
calculated as follows: only 1.5% of the total LDL-cholesterol test samples.
Brought to you by | Karolinska Institute
Authenticated
Download Date | 5/27/15 7:56 AM
Article in press - uncorrected proof
Coskun: Levey-Jennings chart for calculated tests 389
Table 1 Total cholesterol, VLDL-cholesterol (triglyceride/2.17), HDL-cholesterol and calculated LDL-cholesterol (using Frie-
dewald equation) levels of commercial internal control serum (level 1, Olympus, Japan).
Day Total cholesterol, VLDL-cholesterol, HDL-cholesterol, LDL-cholesterol,
mmol/L mmol/L mmol/L mmol/L
1 4558 684 1217 2657
2 4662 694 1191 2777
3 4636 699 1191 2746
4 4688 736 1243 2709
5 4636 694 1217 2725
6 4740 699 1191 2850
7 4584 699 1217 2668
8 4610 704 1191 2715
9 4766 699 1243 2824
10 4740 741 1243 2756
11 4766 704 1243 2819
12 4714 694 1269 2751
13 4766 715 1243 2808
14 4610 704 1243 2663
15 4662 689 1269 2704
16 4740 699 1269 2772
17 4766 689 1243 2834
18 4688 684 1243 2761
19 4610 673 1217 2720
20 4714 725 1269 2720
Mean (X) 4683 701 1233 2749
SD 67.3 16.9 27.2 57.3
CV, % 1.44 2.41 2.21 2.08
Serum LDL-cholesterol concentrations can be To prepare an LJ chart for calculated LDL-choles-
obtained indirectly using the Friedewald equation (7) terol, I measured total cholesterol (A), HDL-cholester-
as shown below. If we neglect serum intermediate ol (B) and triglyceride (C) using commercial internal
density lipoprotein (IDL) and lipoprotein(a) wLp(a)x control sera (level 1, Olympus, Tokyo, Japan) for
concentrations, then we can write: 20 days. All data are given in Table 1.
The mean for LDL-cholesterol (X) can be calculated
Total cholesterol as follows:
sHDL-cholesterolqLDL-cholesterol
qVLDL-cholesterol n
LDL-cholesterol 8ŽAiyŽBiqCiy2.17..
stotal cholesterol–(HDL-cholesterol Xs
is1
w6x
qVLDL-cholesterol) n
LDL-cholesterol
stotal cholesterol–(HDL-cholesterol
qtriglyceride/5), w4x Since the results of serial measurements of HDL-
cholesterol and total cholesterol presented in Table 1
where the factor triglyceride/5 is an estimate of the are significantly correlated (rs0.455; ps0.044), I used
VLDL-cholesterol concentration (7). It is based on the Eq. w3x instead of Eq. w2x to obtain the SD for calculat-
average ratio of triglyceride to cholesterol in VLDL ed LDL-cholesterol:
and cannot be used unless the serum triglyceride con-
centration is less than 4000 mg/L.
When cholesterol and tryglycerides are expressed n
2
8wŽAiyŽBiqCiy2.17..yXx
in SI units, VLDL-cholesterol is estimated as trigly-
ceride/2.17.
Using suitable abbreviations, the equation can be
SDLDLs y is1
Žny1.
. w7x
written as follows:
Now it is easy to obtain the SD for calculated LDL-
LDL-cholesterolsA–(BqC/2.17) w5x cholesterol and prepare the LJ chart for level 1 control
serum. Similarly, we can prepare LJ charts for differ-
To calculate the LDL-cholesterol concentration, we ent levels of control sera. To make inspection easier,
have to measure the serum concentration of total cho- I included the charts for all components of the equa-
lesterol (A), HDL-cholesterol (B) and triglyceride (C) tion on a single chart (chart for equation). This sim-
in the clinical laboratory using specific chemical plified decision-making for LDL-cholesterol status
methods. means that the data are available to inspect at a
Brought to you by | Karolinska Institute
Authenticated
Download Date | 5/27/15 7:56 AM
Article in press - uncorrected proof
390 Coskun: Levey-Jennings chart for calculated tests
Figure 2 LJ chart for total cholesterol (A), HDL-cholesterol (B), VLDL-cholesterol (C) and calculated LDL-cholesterol (D*) on
a single chart for level 1 control serum (data from Table 1 used to calculate means and SDs). Measured A, B, C and calculated
D* are within control limits. TC, total cholesterol.
glance (Figure 2). Using this chart as part of the daily References
QC procedure with the LJ charts for measured tests,
we can report calculated laboratory results for pa- 1. Levey S, Jennings ER. The use of control charts in the
tients with confidence. clinical laboratory. Am J Clin Pathol 1950;20:1059–66.
Before introducing an equation into clinical prac- 2. Henry RJ, Segalove M. The running of standards in clin-
tice, the reliability of its results should be evaluated ical chemistry and the use of the control chart. J Clin
Pathol 1952;27:493–501.
by QC procedures such as LJ charts. If the QC results
3. Coskun A. The reliability of calculated laboratory tests.
for the calculated test are outside the acceptable lim- Clin Chem Lab Med 2005;43:880–2.
its, the appropriate course is not to use this equation. 4. Westgard JO, Klee GG. Quality management. In: Burtis
In this context, using an LJ chart (chart for equation) CA, Ashwood ER, Bruns DE, editors. Tietz textbook of clin-
is of great importance for detecting systematic or ran- ical chemistry and molecular diagnostics. St. Louis, MO:
dom errors in calculated tests. If a systematic or ran- Elsevier Saunders, 2006:485–529.
5. Kringle RO, Bogovich M. Statistical procedures. In: Burtis
dom error related to a calculated test itself or one of
CA, Ashwood ER, editors. Tietz textbook of clinical chem-
its components is detected, the test should be carried istry. Philadelphia, PA: Saunders, 1999:265–309.
out using chemical methods instead of calculation 6. Fraser CG. Biological variation: from principles to practice.
until the problem has been resolved. Washington, DC: AACC Press, 2001:151 pp.
In conclusion, this method is easy, reliable and can 7. Rifai N, Warnick GR. Lipids, lipoproteins, apolipoproteins
increase laboratory effectiveness for calculated tests. and other cardiovascular risk factors. In: Burtis CA, Ash-
When the results of calculated tests are reliable, it is wood ER, Bruns DE, editors. Tietz textbook of clinical
chemistry and molecular diagnostics. St. Louis, MO: Else-
more convenient and cost-effective to calculate such
vier Saunders, 2006:903–81.
results than to measure samples using reagents and
equipment. Received October 6, 2005, accepted January 5, 2006
Brought to you by | Karolinska Institute
Authenticated
Download Date | 5/27/15 7:56 AM
