Quality control data is the laboratory’s early warning system. Every day we generate hundreds of control values, and most of the time we glance at them only long enough to see “pass.” But those numbers carry a lot of information about how well our methods are performing, if we know how to read them.
The three numbers behind every chart
- Mean. The average of your control results, your method’s “center.” It should come from your own laboratory’s data (CLSI and Westgard recommend at least 20 data points from separate runs), not just the manufacturer’s range.
- Standard deviation (SD). How spread out the results are. A small SD means the method is precise.
- Coefficient of variation (CV). The SD expressed as a percentage of the mean: CV = (SD ÷ mean) × 100. Because it is a percentage, CV lets you compare precision across analytes and across levels.
Example: if your glucose control has a mean of 100 mg/dL and an SD of 2 mg/dL, the CV is 2%.
Reading the Levey-Jennings chart
A stable method produces a random scatter around the mean. Patterns are what matter:
- A shift: results suddenly settle at a new level and stay there. Typical causes are a new reagent or calibrator lot, a recalibration, or a changed component.
- A trend: results drift gradually in one direction over days. Think deteriorating reagent, an aging lamp, evaporation, or slowly changing temperature.
- Increased scatter: the points spread wider than before. Suspect precision problems such as pipetting, sampling, or electrode issues.
Many shifts and trends are visible on the chart before any Westgard rejection rule fails. Reviewing charts weekly, not only when the analyzer flags, lets you fix problems early.
Bias: are you accurate, not just precise?
Precision is not accuracy. A method can produce tightly clustered results that are all consistently wrong. Bias is the difference between your results and the true or target value, usually estimated from proficiency testing or external quality assessment (EQA) results, or a peer group comparison. In the Philippines, participation in an external quality assessment program is part of what keeps a clinical laboratory’s results comparable to others.
Putting it together: the Sigma metric
The Sigma metric combines precision and accuracy into one number, judged against how much error is clinically acceptable for that test (allowable total error, TEa):
Sigma = (TEa − |bias|) ÷ CV, with all values in percent.
Example: if TEa for an analyte is 10%, your bias is 2% and your CV is 2%, then Sigma = (10 − 2) ÷ 2 = 4.
| Sigma | What it means in practice |
|---|---|
| 6 or more | World-class. Simple QC (a 1-3s rule) is enough |
| 4–6 | Good. Standard multirule QC |
| 3–4 | Marginal. Needs more controls and stricter rules |
| Below 3 | Poor. Improve the method before relying on QC to catch errors |
One caution: Sigma values depend heavily on which TEa source you use, so pick one source and use it consistently, and calculate over a long enough period (at least several months) to be meaningful.
QC is not paperwork. It is how we earn the right to release a patient’s result.
A short checklist for your next QC review
- Are your means and SDs based on your own data, and recalculated after lot changes?
- Is any analyte’s CV creeping up compared with last month?
- Do any charts show shifts or trends that have not triggered a rule yet?
- What did your last EQA or proficiency testing report say about bias?
- Are corrective actions documented for every rejected run?
References
- Westgard QC, Westgard Sigma Rules: westgard.com
- “Sigma metrics in laboratory medicine: a call for harmonization”: researchportal.unamur.be
- Biochemia Medica, Sigma metrics in clinical chemistry: biochemia-medica.com
- Practical application of Westgard Sigma rules with run size, PMC: ncbi.nlm.nih.gov
