Big problem! Two Injections

[etaxene]

Good morning. I have a big doubt.
Lately I've been applying some RRSS methods where I have inject sample solution twice (2).

In some cases, an impurity value is very low in one injection and absent in the other.

According to your experience or guidelines can I mediate the two values (n and zero)?

Thank you

[tom jupille]

If this were my problem the first thing I would do is to verify the LOD (Limit of Detection) and LLOQ (Lower Limit of Quantitation) of the method as I am running it. If I have a noisy baseline (more noise than when the method was validated), it's quite possible that he impurity may be present at or just above the effective LOD. The same issue can arise if the detector response has diminished (different wavelength? different bandpass? lower plate count?).

Assuming the Signal/Noise ratio is adequate (and the fact that you had a "very low" value suggests that it is not!), I would look for a pattern in the results; if the first injection always higher than the second (or vice versa), then the issue may be one of sample stability. In that case, I would verify by doing a series of replicate injections and monitor the response over time.

[etaxene]

If this were my problem the first thing I would do is to verify the LOD (Limit of Detection) and LLOQ (Lower Limit of Quantitation) of the method as I am running it. If I have a noisy baseline (more noise than when the method was validated), it's quite possible that he impurity may be present at or just above the effective LOD. The same issue can arise if the detector response has diminished (different wavelength? different bandpass? lower plate count?).

Assuming the Signal/Noise ratio is adequate (and the fact that you had a "very low" value suggests that it is not!), I would look for a pattern in the results; if the first injection always higher than the second (or vice versa), then the issue may be one of sample stability. In that case, I would verify by doing a series of replicate injections and monitor the response over time.

Hi Tom. Obviously this is not a case that often occurs to me. Usually the injections are almost identical. Sometimes, however, a similar case could happen to me and I would like to write in the procedure the operating procedures to deal with this case.

For example:

Table 1
Batch 1234 Impurity %
Inj 1 <LOQ
Inj 2 0.04
Average 0.04


OR


Table 2
Batch 1234 Impurity %
Inj 1 <LOQ
Inj 2 0.04
Average 0.02

[tom jupille]

One of "Murphy's Laws" is Zymurgy's First Law of Evolving Systems Dynamics:
Once you open a can of worms, the only way to recan them is to use a bigger can.( http://sound.whsites.net/laws.htm)

Assuming there *was* a peak on the "less than LOQ" run, I'd be tempted to suggest that you assign the average: (LOD + LOQ)/2 to that run. So, if LOQ is 0.3 and LOD is 0.1, the average of those would be 0.2, and the overall average would be (0.2+0.4)/2 = 0.3

As with the two other choices, this is a kludge and *not* an adequate substitute for identifying and fixing the root cause of the problem.