Westgard Rules Explained: 1-2s to 10x with Examples

Westgard rules are statistical rules for deciding whether an analytical run is in control, based on where internal quality control (QC) results fall against the mean and standard deviation (SD) of the control material. The core set is 1-2s, 1-3s, 2-2s, R-4s, 4-1s and 10x, and used together as a multirule they detect both random and systematic error while keeping false rejections low.

Getting this right matters daily. Too loose, and you release results from a faulty run. Too strict, and staff waste reagents and start ignoring alarms.

Levey-Jennings charts: the foundation

A Levey-Jennings chart plots each QC result in time order, with horizontal lines at the mean and at ±1, ±2 and ±3 SD. Westgard rules are simply patterns you look for on that chart.

The mean and SD must come from your own data on your own analyser. A common approach is to run at least 20 measurements of each control lot on separate runs or days, then calculate the mean and SD, refining them as more data accumulates. Manufacturer ranges printed on the control insert are usually much wider and are not a substitute.

In a normal distribution about 95% of results fall within ±2 SD and about 99.7% within ±3 SD. So a single result outside 2 SD happens by chance roughly once in every 20 control measurements, even when nothing is wrong.

The Westgard rules explained

Random vs systematic error

Random error is unpredictable scatter: a bubble in a probe, a poorly mixed sample, an electrical fluctuation. It widens the spread of results. Systematic error pushes results consistently in one direction: a calibration shift, a deteriorating reagent, a new lot or a lamp losing intensity. Rules such as 2-2s, 4-1s and 10x catch systematic error, while 1-3s and R-4s mainly catch random error.

How multirule QC works

In the classic Westgard multirule procedure, 1-2s acts as a warning that triggers a closer look rather than an automatic rejection:

1. If no control is outside 2 SD, accept the run.
2. If a control is outside 2 SD, apply 1-3s, 2-2s, R-4s, 4-1s and 10x in turn.
3. If none is violated, accept the run and record the warning.
4. If any is violated, reject the run, hold patient results and investigate.

Middleware or the LIS can apply the rules automatically, but a technologist who cannot explain why a run was rejected will struggle to find the cause.

What to do when a rule is violated

The worst response is to repeat the control until it passes. If the first result reflected a real problem, a lucky repeat hides it and patient results go out wrong. Instead:

• Stop and hold patient results from the affected run.
• Identify the error type from the rule that failed: random or systematic.
• Look for recent changes: reagent or calibrator lots, calibration, maintenance, control lot, operator.
• Fix the cause, then rerun controls to confirm the system is back in control.
• Assess patient impact: rerun patient samples since the last acceptable QC where needed, and recall reports if results changed significantly.
• Record the violation, cause, action and outcome.

Choosing rules with sigma metrics

The full multirule is not always needed. A precise, accurate assay can be controlled with simple rules, while a weaker one needs more rules, controls or QC frequency. Sigma metrics help you decide.

Sigma is calculated as (allowable total error minus absolute bias) divided by CV, with all three expressed as percentages. Allowable total error is a quality requirement your lab chooses from a recognised source, such as biological variation data or regulatory and EQA performance limits. Bias can come from EQA or comparison with a reference method, and CV from your IQC data.

For example, if your chosen allowable error for an analyte is 10%, bias is 2% and CV is 2%, sigma is (10 - 2) / 2 = 4. The Westgard Sigma Rules approach links sigma to rule selection: high-sigma methods can use a single wide rule such as 1-3s, while methods at lower sigma need the multirule with more control measurements. Below about 3 sigma, QC alone struggles and the method itself may need improvement.

Worked example: a 2-2s violation on glucose in Dubai

A hospital laboratory in Dubai runs two levels of glucose control at the start of each shift. From 20 days of its own data, level 1 has a mean of 5.0 mmol/L with an SD of 0.10, and level 2 a mean of 15.0 mmol/L with an SD of 0.30.

On Tuesday morning, level 1 reads 5.22 mmol/L (+2.2 SD) and level 2 reads 15.70 mmol/L (+2.3 SD). Neither exceeds 3 SD, so 1-3s is not violated, but both levels are beyond +2 SD in the same run. That is a 2-2s violation, pointing to systematic error, and the run is rejected.

The technologist does not simply rerun the controls. The reagent log shows a new glucose reagent lot was loaded that morning, and the lot-to-lot check planned for the previous week had not been done. After recalibration on the new lot, both controls return close to their means.

Because the lab had held patient results, none had been released. As a precaution, the technologist reruns a few retained samples from the previous day on the new lot and confirms agreement. The corrective action makes lot acceptance testing a mandatory step before any new reagent lot enters routine use.

Common mistakes

• Using manufacturer QC ranges instead of the lab's own mean and SD
• Treating 1-2s as a rejection rule, causing frequent false rejections
• Repeating controls until they pass instead of investigating
• Not reviewing patient results after a rejected run
• Applying the same rules to every analyte regardless of performance
• Changing control lots without overlapping old and new lots to set new targets
• Ignoring slow trends because no single point breaks a rule

Key takeaway: Westgard rules work only when targets come from your own data, each rule is read for what it says about random or systematic error, and every rejection leads to investigation, not a repeat.

Frequently Asked Questions

What are the Westgard rules?

Westgard rules are a set of statistical QC rules applied to Levey-Jennings charts to decide whether an analytical run is acceptable. The common set is 1-2s, 1-3s, 2-2s, R-4s, 4-1s and 10x. Used together as a multirule, they detect random and systematic error with fewer false rejections than a single tight rule.

Is 1-2s a rejection rule?

In the classic multirule procedure, 1-2s is a warning rule that prompts you to check the other rules, not an automatic reason to reject. Used as a rejection rule, it rejects many good runs, because about 1 in 20 results falls outside 2 SD by chance alone. Some labs drop it when software applies the other rules automatically.

What is the difference between the 2-2s and R-4s rules?

The 2-2s rule is violated when two consecutive results exceed the same 2 SD limit on the same side of the mean, which suggests systematic error. R-4s is violated when, within one run, one control is above +2 SD and another below -2 SD, which suggests random error because the spread has increased.

How many data points are needed to set QC limits?

A common approach is at least 20 measurements of each control level, on separate runs or days, to calculate an initial mean and SD. Limits are then refined as more data builds up. When changing control lots, run the new lot alongside the old one so you can establish new targets before switching over.

What are Westgard Sigma Rules?

Westgard Sigma Rules link the choice of QC rules to a method's sigma metric, calculated from allowable total error, bias and imprecision. Methods with high sigma can be controlled with simple rules and fewer controls, while lower-sigma methods need the full multirule, more control measurements or more frequent QC.

Next steps

If you want to apply IQC confidently and defend it during assessment, our ISO 15189 Advanced: Medical Laboratory Quality & Competence course covers quality control within the full standard. For related reading, see internal quality control in medical labs, EQA vs proficiency testing and measurement uncertainty in testing labs. For advice on which course fits your role, speak to a counsellor.