TNI, NELAP and the 2016 standard.

An auditor opens the ICAL record for one analyte. The r² is excellent, and there is no relative error measure anywhere in the file. Under the 2016 TNI standard, that curve does not meet the rule.

The NELAC Institute (TNI) writes the consensus standard used to accredit environmental labs in the United States. Its accreditation program is NELAP, the National Environmental Laboratory Accreditation Program. State agencies that TNI recognizes grant accreditation through it [1]. Volume 1 is modeled on ISO/IEC 17025, so much of it will look familiar from Insights 10.

The current lab standard is the 2016 edition. The NELAP Accreditation Council set its formal adoption for January 31, 2020. The 2009 edition now applies only in Kansas [2][3]. Module 2 covers general quality system rules. Module 4, Quality Systems for Chemical Testing, holds the calibration rules. It is the module to know line by line.

TNI calibration requirements for the ICAL.

Section 1.7.1.1 of Module 4 sets the ICAL rules. Four of its items are the ones most often reviewed in an audit [4]:

1. Minimum number of standards, 1.7.1.1(f). Four for an average response model, five for linear and six for quadratic, unless the method asks for more.
2. The range is the standards, 1.7.1.1(g) and (h). The lowest and highest standards bound the range for quantitative results. A result outside that range is not bracketed by the curve.
3. A measure of relative error, 1.7.1.1(k). The lab must use one and document it. If you use percent relative error, check it at one standard at the midpoint and one at the lowest level. RSE is the alternative.
4. Second-source verification, 1.7.1.1(n). Verify every ICAL with a standard from a second maker, or from a separate lot shown to be independently prepared.

Item (k) is the one that moves a lab past the correlation coefficient. A curve accepted on r² alone, with no relative error measure in the record, does not meet the standard. How good the r² is does not matter. Insights 01 explains why the rule exists.

TNI calibration requirements for the CCV.

Section 1.7.1.2 sets the rules for the CCV, the continuing calibration verification [4]. A verification runs at the start and end of each analytical batch, with exceptions for certain internal standard methods.

When a verification fails, the lab takes corrective action and runs a second verification right away. If that one also fails, the lab runs a new ICAL before it analyzes new samples.

Data reported against a failed check must be qualified. The standard allows two exceptions, where the bias cannot change the conclusion. The first is a non-detect when the verification is biased high. The second is a result above a regulatory limit when it is biased low [4]. Every other result goes out with a qualifier.

The DoD and DOE layer.

Labs that test for the Department of Defense or Department of Energy have a second rulebook. They are also audited against the DoD/DOE Consolidated Quality Systems Manual, now version 6.0. Its Module 4 mirrors the TNI chemistry module. Its Appendix B tables add numeric limits by method type [5][6].

TNI requires the measure, and your method or program sets the limit.

For GC/MS organics, Table B-4 accepts two model types. One is an average response calibration with RSD of 20% or less. The other is a linear or quadratic fit with r² of at least 0.99.

Table B-4 then adds relative error limits: 20% at or near the midpoint and 50% at the low level. The RSE may not exceed 30%. Continuing verification must fall within ±20% of the true value, or ±50% for the end-of-batch check [6]. Where QSM 6.0 applies, these are your pass and fail lines.

Three programs, one curve.

The programs agree on what must be measured. They differ on where the line sits. Apply all three to the unweighted curve from Insights 01 and the gap is plain.

Values come from the eight-level illustrative data set in Insights 01, with an unweighted linear fit. The midpoint is the 10 ppb standard. Method 8000D’s 50% allowance for the lowest standard applies only for some data uses.

At −36.7% error at the lowest standard, the same curve passes the DoD program and fails Method 8000D’s general limit. Under TNI alone, the outcome depends on the limit your method and procedures set. TNI requires the measure, and your method or program sets the limit.

So a lab that runs one method for several clients needs to know which program applies to each data set. Its calibration record must show the check against that program’s limits.

A TNI audit readiness checklist.

1. Confirm the edition. Know which edition each accrediting state applies, and whether QSM 6.0 applies to the client.
2. Count standards against 1.7.1.1(f). Four, five or six by model, or more if the method requires it.
3. Record relative error for every analyte. Use RSE, or relative error at the lowest and midpoint standards, against a documented limit.
4. Verify every ICAL with a second source. Keep the lot and maker records that show independence.
5. Bracket each batch. Verify at the start and end, unless the method’s exception applies.
6. Handle failures the same way every time. Take corrective action, run one second check right away, then recalibrate. Qualify any data you report.
7. Evaluate against the right program. Where QSM 6.0 applies, test the low and mid relative error and RSE limits in its tables.

Where WLTR fits.

A multi-program lab checks the same curve against different limits for different clients. Done by hand for every analyte, this is where the relative error rule most often exists only on paper.

WLTR is calibration data integrity infrastructure for regulated analytical laboratories. For each analyte, it derives nine calibration regression models: average response factor, plus linear and quadratic fits across four weightings each. It scores every model against fifteen ICAL parameters, including RSE. All of it is computed orthogonally to the CDS, under U.S. Patent 12,038,448 B2 [P].