Written and maintained by CASRAI Editorial Board
Last updated
The Westgard rules are a set of statistical decision criteria, applied together as a “multirule” procedure, for deciding whether a run of laboratory test results is in control or must be rejected and investigated. Developed by Dr. James O. Westgard and colleagues at the University of Wisconsin and published in 1981 (Clinical Chemistry 27:493-501), the multirule approach is still the dominant framework clinical, reference, and research laboratories use to interpret quality control (QC) data plotted on a Levey-Jennings chart. For anyone specifying, buying, or validating an analyzer, laboratory information system (LIS), or QC middleware, whether that platform implements Westgard multirule QC correctly and configurably is a real, checkable procurement criterion, not a marketing checkbox.
This guide covers what each rule detects, how the rules work together, where they sit inside CLIA and CLSI QC requirements, and what to actually evaluate when comparing instruments, control materials, and QC software that claim Westgard-rule support.
What problem the Westgard rules solve
Every quantitative test run includes control material with a known, previously characterized mean and standard deviation (SD). A single control result that falls outside the mean ± 2SD happens by chance in roughly 1 in 20 runs even when the analytical system is working correctly — using a single “2SD” rule alone (1-2s) as a hard reject criterion produces an unacceptably high false-rejection rate, historically estimated near 5% per control observation and higher across multiple control levels. Relying on a single 3SD rule (1-3s) alone, by contrast, misses many real errors because it is too permissive.
The Westgard multirule procedure resolves this trade-off by combining several rules of different strictness, applied together to two or more control levels, so that the lab gets both a low false-rejection rate and a high probability of detecting a genuine analytical problem. In practice, the 1-2s rule is used only as a warning that triggers inspection of the other rules, not as a rejection rule by itself.
The individual Westgard QC rules
Each rule is written in “NL” notation: N is the number of consecutive control observations being evaluated, and L is the SD limit. The core set most labs configure is:
| Rule | Definition | Primary use | Error type detected |
|---|---|---|---|
| 1-2s | One control result exceeds the mean ± 2SD | Warning/screening rule only — triggers a check of the rejection rules below | N/A (screen, not a reject rule) |
| 1-3s | One control result exceeds the mean ± 3SD | Rejection rule | Random error |
| 2-2s | Two consecutive control results exceed the mean + 2SD or both exceed mean − 2SD (same side, same or different run) | Rejection rule | Systematic error |
| R-4s | The range between two control results within the same run exceeds 4SD (one exceeds +2SD while the other exceeds −2SD) | Rejection rule | Random error |
| 4-1s | Four consecutive control results (within or across runs) exceed the same 1SD limit | Rejection rule | Systematic error |
| 10x (10-mean, sometimes extended to 8x/9x/12x) | Ten consecutive control results fall on the same side of the mean, regardless of SD magnitude | Rejection rule | Systematic error (drift/bias) |
The 2-2s Westgard rule specifically
The 2-2s rule flags a systematic shift: two consecutive control observations, on the same control material or across two different control levels run together, both exceeding the same 2SD limit on the same side of the mean. It is one of the more common Westgard flags to see in practice because it responds to gradual calibration drift, reagent lot changes, and other systematic effects — a real analytical problem worth investigating, not just statistical noise.
The R-4s Westgard rule specifically
The R-4s rule looks within a single run: if one control level’s result is above +2SD and the other control level’s result is below −2SD in that same run, the range between them exceeds 4SD, which is a strong indicator of random error — something like a bubble, clot, or pipetting inconsistency affecting that specific run rather than a systematic instrument or reagent problem.
How multirule QC is read on a Levey-Jennings chart
Control results are plotted sequentially on a Levey-Jennings chart with horizontal lines at the mean and at ±1SD, ±2SD, and ±3SD. When a new control result lands outside ±2SD, that triggers the 1-2s warning, and the analyst (or, in an automated system, the middleware) checks the other rules against the current and recent points before accepting or rejecting the run. If any rejection rule (1-3s, 2-2s, R-4s, 4-1s, or 10x) is violated, the run is rejected: patient results from that run are held, the cause is investigated (calibration, reagent, control material, instrument maintenance), and the run is repeated once the issue is resolved. This is the operational core of what “running Westgard QC” means day to day in a clinical or research lab.
Where Westgard rules sit inside CLIA and CLSI requirements
In the US, CLIA’s quality control regulations (42 CFR Part 493, Subpart K) require labs performing moderate- or high-complexity testing to run control material and have a defined procedure for evaluating and acting on out-of-control results; CLIA itself does not mandate the Westgard rules specifically, but the multirule approach is the most widely adopted method for meeting that requirement defensibly during a CAP or state survey. CLSI guideline C24 (“Statistical Quality Control for Quantitative Measurement Procedures”) is the primary consensus standard describing how to select and apply QC rules, including Westgard multirule design, and is the document most CAP and CLIA inspectors expect a lab’s QC policy to be built on.
Since CMS introduced the Individualized Quality Control Plan (IQCP) option in 2016, labs also have a risk-based alternative to the default “two levels of control, twice daily” schedule — but IQCP still requires a documented rule set for interpreting whatever control data the plan specifies, so Westgard-style multirule logic typically remains part of the design even under IQCP. See CASRAI’s guide to CLIA certification, certificate types, and complexity categories for how QC requirements fit into the broader compliance picture.
Evaluating instruments, LIS, and QC middleware for Westgard-rule support
For a procurement decision — a new analyzer, a laboratory information system, or standalone QC middleware — “supports Westgard rules” is a claim worth interrogating rather than accepting at face value. Concrete evaluation criteria:
- Configurable rule sets per analyte, not one global rule. Different assays have different clinical risk profiles and different total allowable error (TAE) budgets; a system that lets you assign a different rule combination (e.g., 1-3s/2-2s/R-4s for a high-risk analyte vs. a simpler rule set for a low-risk one) rather than applying one fixed rule set to every test is doing real multirule QC, not a marketing label.
- Automated rule evaluation across multiple control levels and runs, not just a single-level 1-2s flag — ask whether the system actually evaluates 2-2s and R-4s across control levels within a run, since that’s where those two rules do their real work.
- Sigma-metric-based rule selection. A more recent refinement, also developed by Westgard, uses the analytical Sigma-metric (derived from an assay’s bias, imprecision, and TAE) to select the simplest rule set that still meets the required error-detection probability — a high-Sigma assay may only need a simple 1-3s rule, while a low-Sigma assay needs the full multirule set. A platform or QC software package that supports Sigma-metric-driven rule design, rather than forcing one rule set on everything, gives a lab a real way to reduce unnecessary false rejections without weakening error detection on the assays that need it.
- Audit trail and documentation. Every out-of-control flag, the rule violated, the corrective action taken, and who approved releasing results after the corrective action should be time-stamped and retrievable — this is what a CAP or CLIA inspector will ask to see, and it’s a direct differentiator between QC modules that were built for compliance and ones that were bolted on.
- Peer-group and EQA data integration. Whether the system can import or export control data to a peer-comparison or external quality assessment (EQA) program (see below) for cross-lab comparison, not just within-lab trending.
- Control-material flexibility. Support for both manufacturer-matched assayed control material and third-party unassayed control material, and the ability to set your own mean/SD from your own data rather than relying solely on manufacturer-insert ranges.
Related CASRAI resources for this comparison: what a LIMS is and does, and the LIMS software comparison guide for labs evaluating whether QC rule management belongs in the analyzer, the middleware, or the LIMS layer.
Control materials and EQA/peer-comparison programs
Westgard rules are only as good as the control material and peer-group data behind them. When comparing control-material and proficiency-testing vendors, evaluate:
- Whether assayed control material’s stated mean/SD is method- and instrument-specific, or a generic cross-platform range that will produce misleading rule violations on your specific analyzer.
- Lot-to-lot consistency and the vendor’s process for transitioning control lots without disrupting an established mean/SD baseline.
- Participation in a real third-party peer-group program, so out-of-control flags can be checked against other labs running the same method and instrument, not just against your own historical data.
See CASRAI’s dedicated guide on proficiency testing and external quality assessment (EQA) for accredited labs for how EQA participation requirements interact with day-to-day statistical QC.
Frequently asked questions
What are the Westgard rules for QC?
They are a set of statistical criteria — 1-3s, 2-2s, R-4s, 4-1s, and 10x, with 1-2s used as a warning trigger — applied together to control results on a Levey-Jennings chart to decide whether a laboratory test run is in control or must be rejected and investigated.
What is the 22s Westgard rule?
The 2-2s rule rejects a run when two consecutive control observations, on the same side of the mean, both exceed the same 2SD limit. It flags systematic error, such as calibration drift or a reagent lot problem.
What is the R4s Westgard rule?
The R-4s rule rejects a run when the range between two control results within that run exceeds 4SD — one result above +2SD and the other below −2SD. It flags random error affecting that specific run.
Are the Westgard rules required by CLIA?
CLIA (42 CFR Part 493, Subpart K) requires a documented QC procedure and defined criteria for evaluating control results, but does not name the Westgard rules specifically. Multirule QC, built on CLSI C24, is the most widely used method for meeting that requirement, and is what most CAP and state inspectors expect to see documented.
What is the difference between Westgard rules and Westgard Sigma rules?
The classic multirule procedure applies a fixed combination of rules to every assay. Sigma-metric QC design uses each assay’s calculated Sigma-metric (from its bias, imprecision, and total allowable error) to select the simplest rule set that still achieves the required error-detection performance for that specific assay, rather than one rule set for everything.
This guide reflects publicly documented QC methodology and regulatory framework as of 2026 and is provided for general reference; it is not a substitute for your accrediting body’s current inspection checklist or your own validated QC policy.
