Procedures for Detection and Quantitation

Federal Advisory Committee on Detection and Quantitation for Uses in Clean Water Act Programs

Meeting #10

Verification of Detection and Quantitation Limits

(Prepared by Michael Murray, based in part on previous FACDQ material and additional suggestions by David Kimbrough and Richard Burrows. The DQFAC Single Lab DL QL Procedure as currently written contains a verification process within it. The proposal is that EPA adopt the single lab procedure, along with the verification process embedded within it, with the caveat noted below. Two alternative recommendations on verification are proposed for the scenario where EPA does not adopt the DQFAC Single Lab DL QL Procedure. These recommendations are followed by a discussion section that could be an additional attachment to the Uses document.)

Recommendation

It is recommended that EPA adopt the verification process embedded within the DQFAC Single Lab DL QL Procedure, with the additional proviso that DL_lab and QL_lab determined with the procedure are not greater than DL_nat and QL_nat, respectively (assuming DL_nat and QL_nat exist).

In the case where EPA does not adopt the DQFAC Single Lab DL QL Procedure (and the verification process contained within it),

Alternative 1

The FACDQ recommends that EPA address verification of detection limits and quantitation limits as part of the process of implementing procedures for these parameters, in a way that balances rigor and practicality. Among other outcomes, the verification process will confirm the validity of DL_lab and QL_lab determined with the procedures (i.e., that MQOs are met), and that these values are not greater than DL_nat and QL_nat, respectively (assuming DL_nat and QL_nat exist).

Alternative 2

The process should cover specific issues/components of verification, including the details on how verification would be carried out; steps for validation of initial DL and QL values (and indication of when new limits should be obtained - e.g., major changes to an instrument) as well as steps for verifying those limits on an ongoing basis; description of the frequency of steps undertaken in the ongoing verification process (e.g., number of samples over a given period); and implications of failure to meet verification criteria (e.g., invalidation of a set of samples run over particular period).

Discussion

Verification of DLs and QLs is tied in to the MQO process; MQOs are needed both to establish DLs and QLs, and to verify that they can in fact be achieved. The practical challenge is that for small error rates (e.g., one percent false positive rate), a large number of samples or blanks are required in order to verify that a rate is in fact not being exceeded. In lieu of this requirement, statistical estimates are used to estimate DLs and QLs. The verification process can either be written into a procedure or be part of each analytical method. The overall verification process can be thought of as containing two components - a step validating that initial estimates of DL and QL are reasonable (this could be part of an initial demonstration of capability for the method overall), and ongoing verification of the reasonableness of the DL and QL values obtained initially.

Given the practical challenges in verifying low error rates, two general approaches for verifying DL and QL values with time are the following:

Batch-wise: Some type of verification is done with each batch (e.g., if a blank is greater than the estimated DL, a decision (potentially invalidation) is made regarding samples run in that batch).
Ongoing collection of data: Verification would entail periodic examination of a series of quality assurance data (e.g., obtained from blanks, low-level spikes, etc.) to determine whether estimated DLs and QLs appear reasonable.

The approaches are not mutually exclusive - both would typically be pursued in a given laboratory, leading to decisions confirming the validity of sample results in individual batches as well as the reasonableness of DLs and QLs in light of longer-term data collection, respectively.

The second approach (ongoing collection of data) is followed in the DQFAC Single Lab DL QL Procedure (Ver. 2.3 8-15-07), involving an assessment (annually or more frequently) of method blanks and QL spikes for verifying DL and QL estimates, respectively. This type of approach (but on a shorter time frame, and with generally more blanks/samples) could also be pursued in validation of initial DL and QL estimates.

This second approach can also be used in a re-estimation process—i.e., if a larger number of blanks shows that the provisional estimate of DL appears to be inappropriate, a new DL can be calculated, based on the additional data.

Other potential verification approaches are discussed in more detail in the Verification & Remediation document (FACDQ7-11) (PDF) (4 pages, 18K) by David Kimbrough.

Top of Page