Clinlab Navigator

Quality Control

Quality control is designed to detect, reduce, and correct deficiencies in a laboratory’s internal analytical process prior to the release of patient results.Quality control samples are special specimens inserted into the testing process and treated as if they were patient samples by being exposed to the same operating conditions. The purpose of including quality control samples in analytical runs is to evaluate the reliability of a method by assaying a stable material that resembles patient samples.Quality control is a measure of precision or how well the measurement system reproduces the same result over time and under varying operating conditions.Pathologists need to be involved in development of quality control protocols, the selection of quality control materials, long term review of quality control data, and decisions about repeating patient samples after large runs are rejected.These quality control activities play an important part in assuring the quality of laboratory tests.

Quality control material is usually run at the beginning of each shift, after an instrument is serviced, when reagent lots are changed, after calibration, and when patient results seem inappropriate.A quality control scheme must be developed that minimizes reporting of erroneous results, but does not result in excessive repitition of analytical runs. The manufacturer should recommend in their product labeling the period of time within which the accuracy and precision of the instruments and reagents are expected to be stable.Each laboratory should use this information to determine their analytical run length, taking into consideration sample stability, reporting intervals of patient results, cost of reanalysis, work flow patterns, and operator characteristics.The user's defined run length should not exceed 24 hours or the manufacturer's recommended run length.Quality control samples must be analyzed at least once during each analytical run.Manufacturers should recommend the nature of quality control specimens and their placement within the run.Random placement of quality control samples yields a more valid estimate of analytical imprecision of patient data than fixed placement and is preferable.

Quality control materials should have the following characteristics.They should have the same matrix as patient specimens, including viscosity, turbidity, composition, and color.For example, a method that assays serum samples should be controlled with human serum based controls. Quality control material should be simple to use because complicated reconstitution procedures increase the chance of error.Liquid controls are more convenient than lyophilized controls because they do not have to be reconstituted.Controls should have minimal vial to vial variability, because variability could be misinterpreted as systematic error in the method or instrument.Quality control materials should be stable for long periods of time.Controls with short shelf lives necessitate frequent reordering and verification against the outgoing material, creating more unnecessary work.Quality control material should be available in large enough quantities to last at least one year. Purchasing a large batch decreases the number of times that control ranges have to be established.

Controls should have target values that are close to medical decision points.Quantitative tests should include a minimum of one control with a target value in the healthy person reference interval and a second control with a target value that would be seen in a sick patient.Examples include sodium controls of 140 and 115 mEq/L and glucose controls of 75 and 225 mg/dL. If three control levels are run, an abnormally low patient range should be included.Quality control levels for therapeutic drug monitoring should mirror therapeutic, toxic, and trough values.If a test is qualitative, giving either negative or positive results, a negative control and a weak positive control with a concentration at the lowest detectable level are recommended.Semi-quantitative tests should have controls at each graded level - trace, 1+, 2+, etc.

Both assayed and unassayed control material are available.Assayed controls are measured by a reference method and sold with published target values.They are more expensive than unassayed controls and are not cost effective for routine quality control in a hospital or reference laboratory.Assayed controls are recommended for physician office laboratories.Unassayed controls must be analyzed by the laboratory to determine the target value and acceptable range.Comparison studies need to be run between the current and new unassayed control materials.If the new control material is from the same manufacturer, only five samples of the new control material need to be run to establish a mean.If the mean is close to the mean of the outgoing quality control material, the new control material can be accepted.No data points should be excluded unless they are known to be result of operational errors.The standard deviation of the outgoing controls is adopted for use until enough data points are collected for calculation..

Interpretation of quality control data involves both graphical and statistical methods. Quality control data is most easily visualized using a Levey-Jennings control chart. The dates of analyses are plotted along the X-axis and control values are plotted on the Y-axis.The mean and one, two,and three standard deviation limits are also marked on the Y-axis.Inspecting the pattern of plotted points provides a simple way to detect increased random error and shifts or trends in calibration.With a correctly operating system, repeat testing of the same control sample should produce a Gaussian distribution.That is, approximately 66% of values should fall between the +/- 1 s ranges and be evenly distributed on either side of mean.Ninety five percent of values should lie between the +/- 2 s ranges and 99% between the +/- 3 s limits.This means that 1 data point in 20 should fall between either of the 2 s and 3 s limits and 1 data point in 100 will fall outside the 3 s limits in a correctly operating system.In general, the +/- 2 s limits are considered to be warning limits.Values falling between 2 s and 3 s indicates the analysis should be repeated.The +/-3 s limits are rejection limits.When a value falls outside of these limits the analysis should stop,patient results held, and the test system investigated.

Normal Distribution of Control Values

+2s                         x              
+1s       x   x     x   x         x        
u   x     x     x   x         x     x   x
-1s x           x         x   x     x      
-2s     x                               x  
Day 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20

Reviewing the pattern of points plotted over time is useful in spotting shifts and trends in method calibration.A shift is a sudden change of values from one level of the control chart to another.A common cause of a shift is failure to recalibrate when changing lot numbers of reagents during an analytical run.A trend is a continuous movement of values in one direction over six or more analytical runs.Trends can start on one side of the mean and move across it or can occur entirely on one side of the mean.Trends can be caused by deterioration of reagents, tubing, or light sources.Shifts and trends can occur without loss of precision and can occur together or independently.The occurrence of shifts and trends on the Levey-Jennings control chart is the result of either proportional or constant error.

A Shift in Control Values

+3s                         x         x    
+2s               x x     x   x x   x   x  
+1s             x     x           x       x
u   x x               x                  
-1s x       x                              
-2s       x   x                            
Day 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20

A Trend in Control Values

+1s       x x                              
u   x       x                            
-1s x   x       x x                        
-2s                 x x       x   x   x    
-3s                     x x x   x   x   x x
Day 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20

Levy-Jennings charts can also demonstrate loss of precision by an increase in the dispersion of points on the control chart.Values can remain within the +/-2 s and

3 s limits, but be unevenly distributed outside of the +/-1 s limits.Random error is present if more than 1 in 20 values fall beyond the +/-2 s limits.

Increased Dispersion of Control Values

+3s             x           x             x
+2s       x                       x        
+1s   x               x         x          
u x         x     x     x           x    
-1s               x                 x      
-2s         x                 x            
-3s     x               x               x  
Day 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20

By running and evaluating the results of 2 controls together, trends and shifts can be detected much earlier.Westgard and associates have formulated a series of multi-rules to evaluate paired control runs.Westgard rules are:

  • 12s rule - one control value is outside a 2 standard deviation limit from the mean.Failure of this rule does not always indicate an analytical error has occurred.If the control falls outside of the 2s limit because of a normal Gaussian distribution, the next control result should pass this rule.
  • 13s rule - one control value falls outside a 3 standard deviation limit from the mean.This may be the result of random error and should be investigated.
  • 22s rule - two consecutive values have fallen outside of the same 2s limit.This rule can apply to a single level during 2 consecutive runs or both levels of control during the same run.Violation suggests systematic error.
  • R4s rule - the range between two consecutive values of the same control level is greater than or equal to 4 standard deviations.This rule also applies between control levels.One control is beyond the +2s limit and the other is beyond the -2s limit.Violation suggests random error.
  • 41s rule - four consecutive values are have fallen on the same side of the same 1s range.This rule can involve one or both control levels.Violation suggests systematic error.
  • 10x rule - ten consecutive values fall on the same side of the mean. This rule can apply within one control level or between control levels.Failure of this rule indicates a shift and the presence of systematic error.

Westgard rules are programmed in to automated analyzers to determine when an analytical run should be rejected.These rules need to be applied carefully so that true errors are detected while false rejections are minimized.The rules applied to high volume chemistry and hematology instruments should produce low false rejection rates.A single rule with a large limit, such as the 13s rule, is recommended.Batch analyzers and manual tests require control samples in each batch.If the test is error prone, the quality control protocol needs to have a high error detection rate.Error detection is improved by increasing the number of controls per run, narrowing control acceptance limits, and using Westgard rules with tighter limits.

Analytes with large biological (intra-individual) variation do not require as much analytical accuracy as analytes with small biological variations.One recommendation is that total analytical variation should be less than half the biological variation (see Appendix A for a complete listing of biological variation).For example, the biological variation of fasting triglycerides is ~20%; therefore, analytical variation can be as high as 10% without significantly affecting medical decision making.Examples of recommended goals for imprecision (%CV) for some commonly ordered chemistry tests at their medical decision points are listed below.

Examples of Analytical Goals for Method Precision (CV %)

Analyte Method CV % Analyte Method CV %
Albumin 1.4 IgA 2.2
ALT 13.6 IgG 1.9
ALP 3.4 IgM 2.3
AST 7.2 Iron 15.9
Amylase 3.7 Phosphate 4.0
Bicarbonate 2.3 Potassium 2.4
Bilirubin 11.3 Proteins 1.4
Calcium 0.9 Sodium 0.3
Chloride 0.7 T4 3.3
Cholesterol 2.7 TSH 8.1
Creatinine 2.2 Urate 4.2
Glucose 2.2 Urea 6.3

Some of the most common problems causing quality control samples to shift are summarized in the following table.

Shift within range Shift out of range Trending Excessive scatter
Improper mixing of controls Any of column 1 Change in instrument reaction temperature Improper mixing in instrument
Controls left at room temperature too long Improper reconstitution of control Instrument sampling problem Contamination during testing
Vial to vial variation Concentration of control in error Instrument reagent delivery problem  
Change in reagent lot number (especially with enzymes) Contaminated reagent    
  Control deteriorated    
  Instrument malfunction    


When these problems are identified the following corrective actions can be taken.

  • Check expiration date of the control.
  • Check expiration date of the reagent.
  • If a new control was used, make sure it was reconstituted properly.
  • Retest the control.If the new value is within acceptable limits, record both values and proceed with patient testing.The problem with the first value was probably random error, which is expected in one of every 20 values.
  • If the repeat value is still out of range, run a new vial of control.If the new control value is within acceptable limits, record the values and proceed with patient testing.The problem with the first set of controls was probably specimen deterioration.
  • If the new control value is out of control, troubleshoot the instrument (check sampling, reagent delivery, mixing, lamp integrity, and reaction temperature).
  • Recalibrate the method, especially if two or more controls have shifted.
  • If controls shift after a new reagent lot number has been introduced, rerun some normal and abnormal patient samples.If patient correlations are good, control shifts are probably acceptable.If they are poor, reagent may be bad.
  • Try a new lot number of reagent.If the problem is corrected, check with the manufacturer to find out if anyone else has reported problems.

It is important to record every quality control value, including those that are out of control.The object of quality control is not to produce beautiful control charts by trying to keep all results within +/- 2 SD.Five percent of control values are expected to be out of range.If quality control samples are routinely rerun until they fall within the current control limits and the outliers are not recorded, the acceptable range will become smaller each time they are recalculated.Eventually they will approach zero standard deviation and become useless and unattainable.

When an analytical run is rejected because quality control is out of acceptable limits, it is often necessary to determine if patient results reported between the last acceptable run and the rejected run need to be repeated.This decision should be based on the nature and size of the error.A 5% bias has no clinical significance for most patients, but a 25% bias is unacceptable for nearly all patients.Biases in between these extremes need to be examined on a case by case basis.If analytical errors have clinical significance, then some of the patient specimens should be re-tested, starting with the samples analyzed just before the rejected control samples.

For example, suppose all glucose results in a run had a 10% negative bias.A patient with a true blood glucose of 82 mg/dL would have been reported as 74 mg/dL.Both values are normal and the results do not need to be corrected.However, an error of this magnitude would be significant for patient samples near medical decision points such as values below 60 mg/dL, fasting glucose values near 140 mg/dL, and glucose tolerance tests near 180 mg/dL.An acceptable strategy would be to repeat patient samples that were less than 60 or greater than 140 mg/dL, starting with samples analyzed just before the quality control failure.Repeat testing should be done in reverse chronological order until the new glucose results closely match the original results.

AddThis Social Bookmark Button