When can you exclude calibration points?

Ask the Expert Question:
When is it permissible to exclude calibration points?

Expert Response:
Removal or replacement of calibration points (point(s)) from the middle of a calibration curve (i.e., points other than the highest or lowest) is not permitted unless an injection or instrument problem confined to that point was directly observed and can be clearly documented. In this case, the failed standard must be re-run within 24 hours and before any samples and inserted into the initial calibration. If that new point is not useful, recalibration is required.

Removal of points for individual analytes from levels other than the highest and lowest is not permitted in any event.

If the analyst can document that a calibration point is not valid because of an injection or instrument problem confined to that run, the point may be excluded if the curve still has sufficient points, or the run may be repeated once only. The whole point, for all compounds, must be removed or replaced. The curve is then evaluated with the point removed or replaced. If the curve still fails to meet criteria, then corrective action must be taken and the whole curve reanalyzed. Corrective action may include, but is not limited to, instrument maintenance and/or re-preparation of standards

One of the following four (4) conditions must be satisfied to allow removal or replacement of a calibration point:

• The data file is corrupted and unusable or the run is interrupted before completion.
• The analyst observes and documents a problem such as leaking of a purge vessel.
• For internal standard methods, if the recovery of the internal standard is less than 70% or greater than 130% of the recovery in the other standards (all internals show the same bias for the standard in question), or the amount of analyte recovered is less than 70% or greater than 130% of the expected values; indicating an improperly made up standard (all analytes in a spike mix must show the bias).
• For external standard methods, if the unit response of the analyte is less than 70% or greater than 130% of the average unit response for the analyte in the other calibration standards; indicating an improperly prepared standard or bad injection. (all analytes in spike mix must show the bias to demonstrate a standard is bad, all analytes in calibration standard must show bias to demonstrate a bad injection)

When using autosamplers with discrete sample pathways for different samples (such as 16 port purge and trap autosamplers) the point to be replaced must be reanalyzed on the same port or that port must be excluded from sample analysis until corrective action is performed and verified by successful analysis of a continuing calibration standard on that port.

The reason for replacing the point must be documented in the run log. The fact that the curve passes criteria with the point removed is not, alone, sufficient evidence to document an injection or instrument problem confined to the point.

Removal of the highest or lowest calibration point is permitted, but the calibration range must be adjusted accordingly. If the lowest point is removed then the reporting limit is raised to be equivalent to the lowest point used in the calibration curve. In any event the number of points remaining in the calibration must be at least that required by the method.

Removal of the highest or lowest point is permitted on a compound specific basis. This may be necessary when strongly responding and poorly responding analytes are included in the same standard mix at the same level. Each compound must have at least the minimum number of calibration points required by the method.

View Ray Frederici's expert profile.