The idea of internal quality control (IQC) was applied first in the industrial manufacturing process ( Shewart ). In clinical chemistry (= laboratory medicine), IQC was introduced by Levey & Jennings . Since the 70ies, IQC is very much related with the name of Westgard .
Before going into the details of IQC, some general aspects will be addressed, of:
Management of the analytical process
All activities of the overall management function that determine the quality policy, objectives and responsibilities, and implement them by means such as quality planning, quality control, quality assurance and quality improvement within the quality system.
While this text focuses on IQC, it shall be stressed that IQC should not be viewed isolated. IQC is embedded in the overall effort of the laboratory for quality. This relates to its technical competence as well as its management competence.
In the quality management system, IQC is part of the circle quality-planning, -assurance, -control, and -improvement (note: a glossary of terms is found in the Annex).
Thus, IQC is an integral part of the quality system.
Technical competence relates to all aspects of the analytical process. Naturally, any test in the medical laboratory has to be proven medically useful. Technical competence starts with the knowledge of the principles for the establishment of medically relevant analytical performance specifications. On that basis, the adequate test is selected, installed, and run in daily routine. However, routine performance of a test needs adequate analytical quality management, IQC being one part of it.
Thus, IQC is an integral part of the analytical process.
Shewart WA. Economic control of manufactured products. Van Nostrand: 1931.
Levey S, Jennings ER. The use of control charts in the clinical laboratory. Am J Clin Pathol 1950;20:1059-66.
Westgard JO, Groth T, Aronsson T, Falk H, de Verdier C-H. Performance characteristics of rules for internal quality control: probabilities for false rejection and error detection. Clin Chem 1977;23:1857-67.
IQC, an integral part of the analytical process ("Westgard")
An analytical process has two major parts:
… necessary to obtain a measurement
on a patient's sample.
… necessary to assess the validity of a measurement result.
In the words of Westgard , it is made absolutely clear that a measurement result that was obtained without IQC, "is no result". IQC is a "sine-qua-non" for reporting a result.
Thus, a well established IQC system is an important part of the technical competence of the laboratory.
Management of the analytical process
“ IQC should be imbedded in the overall quality philosophy of the laboratory ”
It is important that the laboratory does not elaborate "stand-alone" solutions for IQC. IQC is only one means of managing daily routine quality. For example, if it has chosen a robust test that easily fulfills the performance specifications, IQC may be quite easy. Also, work according to the motto: "prevention is better than curation". And, make use of the information available through external quality assessment (EQA).
Westgard JO, Barry PL. Cost-effective quality control. AACC Press, 1995
Art. 34. §1. The laboratory director has to organize IQC
in all disciplines.
§3. IQC consists of several procedures which allow, before the release of patient results, to detect all significant within- or between-day variations.
Art. 35. §1. The frequency of control measurements has to be such that it can guarantee a clinically acceptable imprecision. This frequency depends on the characteristics of the method and/or the instrument.
§2. The control material , … must be stable within a defined period of time. Different aliquots of the same lot must be homogeneous .
§3. For each new lot, the mean and the SD have to be determined. … IQC materials may, at the same time, not be used as calibrator and control material.
Practice guideline (Praktijkrichtlijn)
• A procedure for i nternal quality control (for every analyte)
nature of control samples
concentration , location in the run, number and frequency (concentration & location: additional to KB)
control rules used for start
control rules used for acceptance of a run
IQC at least at 2 occasions
Control of one and the same test with different instruments
In essence, the Belgian guidelines:
TELL: WHAT to do, but
NOT: HOW to do it
The Belgian guidelines in a nutshell:
"Suitable" IQC procedures
At least 2 IQC events: start/end
Determine mean & SD
1 Royal Decree from December 3 1999 regarding the authorization of clinical chemical laboratories. Moniteur Belge. December 30, 1999.
Materials for IQC should resemble the actually measured samples as far as possible.
Serum analysis should apply materials with a serum-like matrix
Urine analysis should apply materials with an urine-like matrix
Whole blood should apply materials based on a whole blood matrix
Usually, a compromise has to be made between stability and “nativity”.
Most commercial IQC materials exhibit artificial matrix effects. Therefore, they usually cannot be used for the assessment of trueness.
Analyte concentrations should be medically relevant (e.g., be in the mid, the upper, and the lower part of the reference range, or near decision points; see www.westgard.com for medical decision levels).
Be compatible with the test
Be stable & homogeneous (bottle to bottle)
Lyophilized samples are preferred for long-term stability.
Problem associated with lyophilization:
-Reconstitution accuracy (particular important for analytes that require tight control limits; e.g., Na, Cl)
Be available in large batches to allow their use over an extended period of time (e.g., two years).
Be purchased overlapping
The new material should be tested for some time together with the old material in order to have continuous experience. This prevents difficulties in problem-solving when they just occur at the moment a new IQC batch is introduced.
The target mean of a control material is particularly important.
Control materials should not be used as calibrators.
Target means should:
Have negligible uncertainty
Be provided with sufficient digits (adjusted to the precision of the method)
General problems with digits
Too few digits may give problems with target uncertainty (rounding problem), calculation of CV, violation of IQC rules & graphical display.
Lactate simulation (n = 20): Mean = 1,6 mmol/l; CV = 1,8% Red squares: 2 digits; Blue diamonds: 3 digits
Too few digits may give problems with target, IQC rules, calculation of CV, & graphic
Target setting (mean)
May be done by the laboratory itself
May be part of the control sample
Target part of the control sample: CAVE
Method dependent assigned values are valid only for homogeneous test-systems (instrument/reagent/-calibrator from the same manufacturer). In case that the laboratory uses, for example, reagent and calibrator from different sources (= heterogeneous test), it might obtain a value that is different from the original one. In that case, the laboratory has to determine the target with its own test procedure.
Target setting by the laboratory
This is done, for example, by parallel analysis of a new batch of unassigned material over 21 days under stable operation conditions.
Consider one performs 4 IQC measurements. Those can be done in block (all at once), or continuously.
Block: more patient samples are between the IQC events, but statistics are stronger.
Continuously: fewer patient samples are between the IQC events, but statistics are weaker in the beginning.
See also later Remedial actions
Checklist – Frequency and location
Minimum: 2 samples per run
Desirable: ~1-2% of patient samples
-Make a cost/benefit calculation
Frequency should be related to test stability: requires knowledge of instrument and test
Consider “dummy” measurements before 1st IQC
Frequency may depend on the control rule
Block: Maximizes chance of assignable cause of variability between subgroups
Continuous: Maximizes chance of assignable cause of variability within subgroups
IQC – practical aspects “ Block” IQC-events More samples between events, but stronger statistics Maximizes chance of assignable cause variability between subgroups Continuous IQC-events Fewer samples between events, weaker statistics in the beginning Maximizes chance of assignable cause variability within subgroups
Analytical procedures give results (x i ) that are independent from other results
x i comes from a Gaussian distribution with a mean µ
and a standard deviation
Coefficient of variation (CV)
>See: Basic statistics
The standard deviation, often, increases with increasing concentration of the analyte (see Figure).
the coefficient of variation (CV) is often more convenient for the description of random error (imprecision).
Gaussian distribution (standard and cumulated)
A Gaussian (= Normal ) distribution is characterized by
its mean and
standard deviation ( )
To understand the basis of statistical IQC, it is important to memorize the key characteristics of the Gaussian function, in particular, the expected location of single values that constitute the distribution.
We look at the percentage of observations that we expect in certain regions of the distribution.
x i comes from a Gaussian distribution with a mean µ
and a standard deviation
If we know the stable mean and SD of an analytical process
We can predict the location of future measurement with a certain probability
Statistical basis of IQC
Graphical presentation of the Gaussian distribution
The Gaussian distribution can be presented
In the normal way: "Bell-shaped" (similar to a histogram)
Cumulated & linearized = Normal probability plot
EXCEL® template from P Hyltoft Petersen
(note: not available in EXCEL ® itself)
These worksheets use the EXCEL NORMDIST function.
The "Print Screens" guide you through their application.
The graphs will appear automatically.
Gaussian distribution Gaussian D istribution (Worksheets "GaussBell"; "GaussCumul")
When data are Gaussian distributed, we can predict the frequencies (or probabilities) of their occurrence within or outside certain distances ( , or z-values) from the mean (see also Figures above).
These probabilities are used in parametric statistical calculations. They are listed in tables, but they also can be calculated with EXCEL ® . Of particular importance are probabilities that are used in statistical tests (95%, 99% probabilities).
2-sided and 1-sided probabilities
Statistics distinguish probabilities in
2-sided & 1-sided
2-sided probabilities: question is A different from B?
1-sided probabilities: question(s) is A > B (A < B)?
Of practical importance are probabilities
"Inside" & "Outside"
Outside probabilities, for example, are important in internal quality control.
We have seen that very few results can be found in certain regions of the Gaussian distribution. For example, it is highly unlikely to find results beyond a distance of 4 s from the mean. In that connection, a convention has been made about what we consider “unlikely” (“rare events”).
Values outside ±1.96 (2-sided view) are deemed “rare events”, they occur in 5% of the cases, only.
Values that are found outside ±1.96 , are not by chance .
It is assumed a non-statistical reason (e.g., systematic error) causes values to be found outside ±1.96 .
Monitoring “the outsides”: One IQC-principle
This gives an indication “whether something happened”.
REMARK on “rare events”
We called observations that happen in <5% of the cases rare events.
At the same time, when we observe them in daily practice, we suspect that their occurrence has non-statistical reasons.
BEWARE: Our judgement may be wrong in 5%, 1%, 0.3%, etc. of the cases!
This principle leads us to a simple family of IQC-rules. Namely, based on a known Gaussian distribution, we monitor in practice whether we observe an IQC result that falls, for example, out of the ±3 s limits of that population. In case that happens (note: the probability is less than 0.3%), we assume that the process became unstable.
the 1 3s -rule
The 1 3s -rule
The process is out-of-control when 1 IQC result is outside a distance of ± 3 s from the «true» mean.
It is a member of the family : n z • s
n: number of observations
z: certain number of standard deviations of the Gaussian distribution (= standard normal deviate)
s : stable (“true”) standard deviation
Basic monitoring principles of IQC-rules
“ The outsides” (for example 1 3s )
The location towards the mean (above or below)
A range (difference between results)
The mean (several results)
The imprecision, or variance (several results)
Selection of control rules – Fundamental problems
There are many different control rules
Different control rules have different power for error detection
Different control rules have different probabilities of false rejections
The selection of a particular control rule is always a trade-off between error detection and false rejection!
Power of control rules Statistical basis of IQC Power function graphs for the n 3s rule (various numbers of measurement) for Systematic error & Random error Notes - Power increases with n - P fr increases with n Usually, - Power for detection: RE < SE
The Figure compares power functions for mean rules with “simple” individual rules (e.g., the 1 3s rule ) (#).
Mean rules, usually , are more powerful than quality control rules based on individual values.
#Linnet K. Mean and variance rules are more powerful or selective than quality control rules based on individual values. Eur J Clin Chem Clin Biochem 1991;29:417-24.