Quality Assurance (QA) Guidelines for International Laboratories

Author: Document Number: Pro61-05

Kurt Michael Effective (or Post) 2/12/09
Date:
Review History Date of last review: 6 May 2010
Reviewed by: Heidi Hanes
SMILE Comments: This document is provided as an example only. It must be
revised to accurately reflect your lab’s specific processes and/or specific protocol
requirements. Users are directed to countercheck facts when considering their
use in other applications. If you have any questions contact SMILE.

Chemistry Guideline for Establishing New Control Lot Means and Quality Control
(QC) Ranges Through Parallel Testing and Historic Coefficient of Variation (%CV h)
Authored by Kurt Michael and Paul Richardson

In order to optimize controls, it is important for clinical laboratories to establish their own
means and QC ranges for each new control lot number. Using the manufacturer’s QC
ranges will waste you QC money, as the ranges will be too wide to provide an
appropriate level of warning value. Assay ranges for new control lot numbers should be
confirmed prior to the expiration of the old lot. The means for all analytes of the new lot
should fall within the assay ranges provided by the manufacturer. If they do not, then
something in the system needs corrective action.

If your laboratory has been using the manufacturer’s ranges, but you want to improve
your QC effectiveness, then following the recommendations below will help. It is
strongly advised that you gradually implement changes. Pick one or two analytes to
change until you are comfortable with the results, and then add one or two more
analytes. Also, narrow the ranges gradually. The goal should be QC ranges based on
the Standard Deviation (SD) calculated from your historic CV h, but it might be better to
narrow a little at a time, until you reach this goal. This gradual approach will reduce
complications.

Three other side effects of this approach are worth mentioning. 1) Levey-Jennings
charts will show much more variation with narrow SDs. This is normal and will enable
the laboratory to spot problems earlier. A flat looking LJ chart is often an indication that
SDs are too wide. 2) The laboratory will likely consume more calibrator, reagent, QC
materials and other consumables, as more problems will be detected and corrected.
Inventories of these items should be monitored carefully. 3) The increased level of error
detection will give more out of control situations which will lead to more documentation
of corrective action taken.

PART I: Recommendations for establishing means through parallel testing of new

chemistry control lots

1. Parallel testing is running the current and new QC material together. Run the
current QC lot and verify that the QC material is within range. Then run the new
QC lot 2 to 3 times throughout the day ideally for a minimum of seven to ten days
before the old lot expires. Collect at least 20 data points.

407522934.doc SMILE & HPTN Document 26 June 2006

Page 1 of 6
 Quality Assurance (QA) Guidelines for International Laboratories

NOTE: If you observe any problems as you parallel test your controls, such as
a trend, stop running the new lot number and investigate to solve the problem.
A trend is here defined as six data points in one direction, i.e. steady increase
or steady decrease. If the old lot expires and you need to start the new lot,
then use the manufacturer’s ranges until you can establish your own.

2. Review the data to ensure that there are no trends and that the precision or CV is
acceptable according to the manufacturer’s recommendations (available in the
instrument or method documentation). If the CV is unacceptable then you must
investigate the cause, correct it and recollect the data.

3. Calculate the new means and experimental SDs for each chemistry analyte.
Utilize instrument software programs or EXCEL spreadsheets to perform the
calculations, if available. Or sum the values and divide by the number of values
to obtain the mean and calculate the SDs.

4. Reject any obvious outliers (data points more than 3 times the SD from the
mean). After any data rejection, recalculate. Do not include rejected data points
in any statistical calculations.

PART II: Establishing historical CV (CVh)

CVh is the historical instrument precision for that method measured as a percent (%CV
= SD/Mean X 100). It can be used to establish Standard Deviation (SD), QC ranges and
to monitor your instrument precision.

The CVh is best established over several lots (3 or more) and can either be the average
CV or more preferable, the highest acceptable CV over that period. CVs obtained during
times of instrument malfunction or significantly higher than the average CV should not
be used in calculations. If you have no previous CV data, then use the current CV as
the CVh basis and modify this value as you move on to subsequent new lots. This basis
value will need to be increased slightly (~1-2%) in order to establish QC ranges that are
not too narrow.

PART III: Recommendations for establishing QC ranges using CV h

1. From the data created by the parallel testing in Part I above, calculate the CV
and SD for the new control. Standard Deviation is “s” in the following example
formula. It is much easier to use a calculator or an EXCEL spreadsheet for this
calculation.

407522934.doc SMILE & HPTN Document 26 June 2006

Page 2 of 6
 Quality Assurance (QA) Guidelines for International Laboratories

2. The calculated new lot CV must be comparable to the manufacturer’s instrument-

method CV and should approximate the CVh. If it is not, then there is a precision
problem with your instrument.

3. If you use the SD calculated directly from the new control data to establish your
new QC ranges, you may find that these ranges are too narrow. Because they
were collected over a short time period and with limited data, they do not
represent enough variation in a system. It is better practice to use the CV h, as
this will accommodate lot-to-lot variation.

4. Calculate the new control range by using the new mean with the SD determined
by the equation (SD = CVh /100 X Mean). The number of SDs on either side of
the mean can vary, but in many laboratories the QC range will be +/- 2 SD of the
mean. See the examples of how to calculate these parameters below.

QC ranges established by this method will be more sensitive to changes and will thus
alert laboratory personnel, when there are changes in a testing system. The goal is to
detect these changes before they become clinically significant and affect patient results.

NOTE: Tracking CV from lot-to-lot and month-to-month will alert staff to

instrument problems. The precision is a very good indicator of the
instrument’s overall function. Increased CV must be corrected as soon as
possible through instrument maintenance or servicing. The manufacturer’s
representative will not be inclined to service an instrument until the CV
exceeds the manufacturer’s claims, but your laboratory maybe able to correct
problems before then.

PART IV: Comparing your CVh with Manufacturer’s CVs and Peer CVs

It is important to compare your CVh to other sources. Too large a CV will not allow
the QC material to be the warning signal you want it to be. Using smaller CVs will
improve survey performance. Too small a CV will cause more QC errors wasting
material and time unnecessarily.

1. The first comparison of your CVh should be made to the specifications of the
instrument manufacturer. This information is generally contained in the
instrument manual or the method documentation. If your CV h is higher than those
specified, then service needs to be called. CVs are generally higher when the
mean of the analyte is at a low value. This can lead to misinterpretation so some
analyzer manufacturers publish acceptable SDs at the lower levels rather than
CV.

2. Some labs like to compare their CVh with the average CV from CAP or other
External Quality Control (EQA) Provider peer for each analyte. Generally, the
laboratory’s CV should be less than the EQA CV, since the peer group will
encompass many more variables. EQA CVs are calculated for each sample by

407522934.doc SMILE & HPTN Document 26 June 2006

Page 3 of 6
 Quality Assurance (QA) Guidelines for International Laboratories

dividing the survey peer sample SD by the sample mean and multiplying by 100
to arrive at the sample %CV. This is done for each analyte sample on the survey.
Then all samples are either averaged or the highest sample CV for the analyte is
compared with your laboratory’s CV. Ideally, your lab should be less that the EQA
CV. If it is not, then you should investigate why your system is unable to recover
the precision of your peers.

3. Some Quality Control material companies offer a peer group program. This is a
program in which your lab submits their mean, SD and CV data each month.
The company then compiles the data with other labs using the same QC
material, instrument and method and publishes this data. This peer data is very
helpful in allowing you to judge your instrument’s performance on a monthly
basis. You would prefer to have your CV less than the peer’s CV. Contact your
QC material supplier to identify what might be available.

PART V: Example

New normal and abnormal controls have arrived a month before the old lot expires or
runs out. The laboratory was able to obtain enough new control to last until its
expiration date. The lab staff ran the new lot of controls in parallel with the old lot each
day until they had 20 data points. They then calculated the new mean for each control
and for each analyte.

Glucose Normal Control Mean

The manufacturer in their control package insert stated that the glucose normal control
mean should be between 80 -92. The lab recovered a mean of 85 from the parallel
testing above. The mean is within the manufacturer QC range and this became the new
normal control’s glucose mean.

Glucose Normal Control SD

The manufacturer of Analyzer XYZ states that their instrument’s precision for this
glucose method is 4.0%. The last three lots of the normal control had CVs of 2.2, 1.8
and 2.5. These numbers show that the instrument-method precision is performing as
designed.

Even though we will not be using the CV from the parallel studies, it is important to
check that it is comparable to (less than) that of the historic CV.

The historic average CV is 2.2 and the historic highest CV is 2.5. The EQA peer group
in this control range had a CV of 3.1. The laboratory also participates in a peer group
program with the control vendor. The peer group for the new control and instrument-
method combination runs a CV of approximately 2.7. It was decided to use 2.5 as the
historic CVh.

The lab used this formula to calculate their new SD goal

407522934.doc SMILE & HPTN Document 26 June 2006

Page 4 of 6
 Quality Assurance (QA) Guidelines for International Laboratories

SD = CVh X mean
100

Glucose Normal Lot 001 Mean = 85 mg/dL, CVh = 2.5

SD= 2.5 X 85 = 2.1

100

The laboratory uses a +/- 2 SD range for this assay, so the control range became:
Range = Mean +/- 2 SD (85.0 +/- 4.2 or 80.8 – 89.2). This range was monitored and
adjusted, if needed, as more data was accumulated. A record of all changes to the QC
settings in the instrument is kept in a log, and any aliquoting of QC is also logged to
monitor stability of the control.

The laboratory would then go through the same steps for the abnormal control. The
CVs will be different for each control and each analyte.

Appendix I –Roche Description of how to run parallel testing

or precontrols on the the Cobas Integra. Instructions
obtained from Roche website.

How do I define and measure precontrols?

The QC parameters for precision controls are assigned mean and standard
deviation by the laboratory. These values have to be established beforehand
in a precontrol period before the precision control can be used in the daily
routine. Over a period of time, the precontrol is measured in regular
intervals. At the end of the precontrol period, the measured results are used
to calculate the assigned mean and standard deviation for all tests that have
been included in the precontrol measurements.
The process of measuring precontrols and calculating QC values is as follows:
1. At the beginning of the precontrol period, the control ID and the control lot are
entered, and a precontrol request is created in the Create Precontrol Request
window.
2. During the precontrol period, regular precontrol measurements are requested in
the Append Pre-Control Request window. These requests include all tests for
which QC values have to be established.
3. At the end of the precontrol period, the QC values are calculated for all measured
tests and read-into the Control Lot window.

To create a precontrol request for a new control lot

1. Open the Quality Control work area.

407522934.doc SMILE & HPTN Document 26 June 2006

Page 5 of 6
 Quality Assurance (QA) Guidelines for International Laboratories

2. Choose Request > Pre-Control > Create. The Create Pre-Control Request
window is displayed.
3. Enter the control ID (short name), the lot number of the precontrol, and its rack
and cup position.
4. Select the tests for which you want to establish the QC values.
5. Click OK. A precontrol request is scheduled, including all selected tests.
6. Place the precontrol in the rack and cup position defined in the Create Pre-
Control Request window.

All precontrol requests have the prefix “$P_” in the order ID (for example
$P_RON) and are listed in the Orders work area in the View By Orders and View
By Patient lists. Precontrols are not listed in the View Cal/QC list of the Orders
work area.

To continue precontrol requests during the precontrol period

1. Open the Quality Control work area.

2. Choose View > By Pre-Controls and select the precontrol you want to order.
3. Choose Request > Pre-Control > Continue. The Append Pre-Control Request
window is displayed.
4. Enter the rack and cup position of the precontrol.
5. Click OK. A new precontrol request is scheduled.
6. Place the precontrol in the rack and cup position defined in the Append Pre-
Control Request window.

To read-in QC values from precontrols at the end of the precontrol

period

1. Open the Configuration work area.

2. Choose Definitions > Controls.
3. Select the control from which you want to read-in precontrol results.
4. Choose Modify > Pre-QC. The Control Lot window is displayed.
5. Click Read Pre QC to read-in all QC values of all tests listed in the Test List of the
Control Lot window. The assigned mean and assigned standard deviation values
are automatically calculated and read-in for all tests included in the precontrol
measurements.

To read-in only selected tests from the precontrol measurement, choose Modify >
Lot, select the test in the Test List and click Read Pre QC.

407522934.doc SMILE & HPTN Document 26 June 2006

Page 6 of 6