Total Allowable Error and Sigma Calculator

Total allowable error (TEa) is how far a single result may sit from the true value and still be good enough for the test's clinical use. Westgard calls defining it, “how good a test needs to be for its intended clinical use”, the hard part of the whole exercise (Westgard Sigma Rules). Everything else on this page is arithmetic once TEa is chosen.

In a US lab the usual starting point is the CLIA proficiency testing limit: the band a PT program uses to grade your result against its target (how PT works is in CLIA proficiency testing rules). CMS rewrote those limits in the final rule of July 11, 2022 (87 FR 41194), and the new analytes and limits took effect on July 11, 2024. The calculator offers 13 of them as presets, each criterion as 42 CFR 493 prints it:

A limit with an absolute part has no single percentage. “±8% or ±6 mg/dL (greater)” for glucose means 6 mg/dL wins below 75 mg/dL and 8% wins above it, so the calculator asks for the concentration you care about (the decision level, or the control's mean) and works the percentage out there:

CLIA's limits are not the only goals. Westgard notes that the 2015 Milan consensus accepted that no single set of TEa goals fits every test, and that the EFLM biological-variation goals some reviewers prefer are often beyond what today's instruments can reach. His practical advice for the US is blunt: “your laboratory is bound by CLIA and CAP goals, so implement your QC design based on metrics derived from those goals” (Westgard). For an analyte without a preset, or a tighter goal your director has set, choose “My own goal” and type it.

On Westgard's scale, sigma values “typically range from 2 to 6, where the goal for ‘world class quality’ is 6”; below 3 a process is “so unreliable it shouldn't be used for routine production” (lesson 78). The higher the sigma, the less QC it takes to catch an error that matters. The Westgard Sigma Rules turn that into rules and control counts, read off lines at 6, 5 and 4 sigma (so 5.4 uses the 5-sigma row):

N is the number of control measurements in a run, R the number of runs a rule reads across. N = 2, R = 2 means splitting the day into two runs with two controls each, so the 4-1s rule can look across both. The diagram drops the old 1-2s warning rule altogether; what each remaining rule catches is in Westgard rules explained.

Westgard's view of the real world: on many highly automated systems most tests run at 5 to 6 sigma, where a Levey-Jennings chart with ±3 SD limits, or a short multirule, does the job. Point-of-care devices often do worse. Seven HbA1c point-of-care devices evaluated in a 2014 Clinical Chemistry study came out at roughly 0.0 to 4.5 sigma, and for the best of them Westgard's best case was the four-rule procedure with four controls.

Westgard's October 2026 lesson reworks a published study of albumin and ALT that kept the minus sign on a negative bias. The study's TEa goals were 10% and 20%. Kept signed, the bias inflated both metrics; with the bias's size, albumin falls from 3.13 to 1.26 sigma and ALT from 6.01 to 3.47. The last column is computed by this calculator's code when the page is built.

The goal matters as much as the data. Those 10% and 20% goals are CLIA's 1992 limits. Today's albumin limit is “Target value ±8%” (§ 493.931), and with the same bias and CV the albumin method comes out at 0.82 sigma: the bias alone uses more than half of the allowance.