Six Sigma · 7 min read

Six Sigma in the Clinical Laboratory: How to Calculate and Use It for QC

Updated September 2026 · ValidToolsLab Academy

Most labs that report a Sigma metric treat it as a number to put on a validation summary and move on. That's a missed opportunity — Sigma isn't just a grade, it's meant to directly answer a very practical question: how many controls do I need to run, how often, and with which rules, to catch errors before they reach a patient result? This guide walks through the formula and how to actually use the answer.

The formula

The Sigma metric combines three things you likely already calculate for every method: the allowable total error (TEa), the method's bias, and its imprecision (CV%):

Six Sigma Metric σ = (TEa − |Bias|) ÷ CV

The higher the Sigma, the more room your method has between "typical performance" and "clinically unacceptable" — which is exactly what determines how much QC you need to safely catch a problem before it slips through.

Reading the score

Sigma (σ)GradeWhat it means
≥ 6.0World ClassExceptional performance; minimal QC needed to catch errors reliably.
5.0 – 5.9ExcellentVery good performance; standard QC is more than adequate.
4.0 – 4.9GoodSolid performance; standard Westgard multirule QC recommended.
3.0 – 3.9MarginalNeeds tighter QC and closer monitoring to catch errors reliably.
< 3.0PoorHigh risk of unreported error; investigate the method before routine use.

Turning Sigma into an actual QC plan

This is the part most labs skip. A Sigma score is only useful if it changes what you actually do at the bench. Here's how the score should translate into QC design, following the Westgard "Sigma-based QC design" approach used across the industry:

GradeRecommended RulesControls / Frequency
World Class (σ≥6)Single 1₃s rule2 levels, daily
Excellent (σ≥5)1₃s / 2₂s2 levels, daily
Good (σ≥4)1₂.₅s / 2₂s / R₄s / 4₁s / 10x̄2 levels, daily
Marginal (σ≥3)1₂s / 2₂s / R₄s / 4₁s / 10x̄3 levels, twice daily or per batch
Poor (σ<3)Full multirule + patient data QC (delta checks, average of normals)4+ levels, per batch
Why this matters practically: a low-Sigma method run with only a simple 1₃s rule will let real errors through undetected — that's the scenario regulators and accreditors worry about most. A high-Sigma method run with an unnecessarily complex multirule just burns reagent and staff time on QC that isn't adding meaningful error detection. Matching the rule set to the Sigma score gets both sides right.

Worked example

Troponin I, Low-concentration QC level

TEa (CLIA/biological variation limit): 20.0%
Observed bias vs. reference: 2.5%
Observed CV (from a precision study): 4.0%
σ = (20.0 − 2.5) ÷ 4.0 = 4.38
Grade: Good — recommended: standard Westgard multirule (1₂.₅s / 2₂s / R₄s / 4₁s / 10x̄), 2 control levels run daily.

Now run the same instrument's high-concentration level with a tighter CV of 2.8% and less bias (1.0%): σ = (20.0 − 1.0) ÷ 2.8 = 6.79 — World Class, needing only a simple 1₃s rule at that level. This is exactly why Sigma should be calculated per level, not just once for a method — and why your overall QC strategy for an assay should generally be driven by its worst-performing level, not its best.

This entire calculation, automatically

ValidToolsLab's Six Sigma / QC Rules Advisor calculates Sigma per level and recommends the exact Westgard rule set, control count, and frequency — the same logic used in this article — with a signed PDF report your director can approve in one click.

🚀 Start Free 5-Day Trial →

This guide reflects general, widely accepted Six Sigma QC design practice (Westgard-style) as of 2026. TEa values vary by analyte and source (CLIA vs. biological variation databases) and are updated periodically — always use the current TEa for your specific analyte and jurisdiction.