Help me to reply the reviewers's question!!

[wxmwqr]

Recently, I subbmit a paper to a journal. This paper is about to used HPLC method to quantitate some compound in herb sample, I used external method to quantitate.
But the reviewer ask a question "Authors should comment on the appropriatenessof using external rather than internal standards".
because of the poor English of me, I cannot give him a satisfactory reply.
Now, I ask for a satisfactory and concise reply to the question that was asked by reviewers.

[Bill Tindall]

I understand your difficulty replying to this question. This reviewer is obviously not an experienced analytical chemist or they would not have asked this uninformed question, unless, there is some obvious problem with your method that results in poorer than expected precision. Does you method involve an extraction that is not compete? A properly chosen internal standard can compensate for an incomplete extraction step.

Internal standard methods were rarely used at the very large integrated chemical company I worked for. They involve more labor and they don't generally provide significant improvement in precision for typical quantitation methods.

An external standard is the preferred way to calibrate, unless, there is a need for better precision. If there is a need for better precision, then it must be proven that an internal standard results in better precision. It is never certain that an internal standard is better.There are many cases when an internal standard will provide worse precision. Which method is better is determined by the statistical correlation of various sources of error and the inherent poorer precision resulting from measuring 2 areas for the internal standard verses one area for external standard.

You might reply this way...
The precision of my method was adequate for this work. It is never certain that an internal standard method provides better precision. The additional work necessary to investigate an internal standard to show that it provides better precision was not justified.

[Kostas Petritis]

The reviewer question might be valid if you use mass spectrometry...

[james little]

Even in MS there is a lot of disagreement on if you should use an internal standard. However, most articles in literature use either unlabelled or labeled standard.

Last year there was quite a discussion on http://www.boomer.org/ (PharmaPK website) about this topic.

In my experience with IS on MS, they can be useful if the instrument is drifting some in response. The labeled is much better than the unlabeled for this use. But the labeled internal standard can suppress/enhance the response of the unlabeled analyte, so the concentration level of the internal standard must be adjusted correctly.

On my instrument when the instrument is not drifting some, I get same std deviation with or without IS. WHen it is drifting, the unlabeled IS does not always drift at the same rate as the analyte. THe labeled will drift at the same rate. What causes my instrument to drift, don't know..

[james little]

I found the link to PharmaPK site on internal standards, food for thought if interested in that topic..

http://www.boomer.org/pkin/PK04/PK2004313.html

[Bill Tindall]

I once was doing a GCMS analysis for PCB's and using an EPA recommended internal standard. In cases the response for the internal standard reverse correlated with the PCB response.

Recently we did an evaluation of using an internal standard for a LCMS analysis of a complex mixture. The internal standard resulted in slightly better precision for some componets and much worse precision for another.

The bottom line is that it should never be assumed that an internal standard will deliver better precision.

[tom jupille]

The bottom line is that it should never be assumed that an internal standard will deliver better precision.

Agreed.
But an internal standard can be a very useful troubleshooting tool. Simply compare the RSD with and without the internal standard.

If the precision is better with the IS, that suggests that the dominant error contributions to the IS and the analyte peaks were (highly) correlated. This implies that the "problem" occured when the peaks were still together: sample workup, dilution, injection, etc.

If the precision is worse with the IS, that suggests that the dominant error contributions to the two peaks were (highly) uncorrelated, implying problem occurence during/after separation: loss on column, tailing, detector response, integration, etc.

If you are not using an IS, you can get the same information by comparing the areas of any two (comparable size) peaks. Simply look at the precision of the area ratio vs. the precision of the area.

That said, I concur with Bill's original suggestion, which amounts to a version of Occam's Razor ("simpler is better"); if the precision without an IS is adequate for your purpose, then an IS is unnecessary.

[Consumer Products Guy]

We rarely use internal standards any more, other than for cholesterol assay and ethanol assay by GC; the reason - today's GC and HPLC autosamplers are so good. All our pharmaceutical methods except ethanol assay use external standard. Oftentimes, when I come across an internal standard procedure, even from a large company, I find that it's a holdover from the days of manual injection. Similarly, when I see a procedure using UV detection at 254 nm, it's a red flag that the procedure was developed a while ago, and someone only had a single-254 nm wavelength detector available instead of a VWD.

[Victor]

Consumer Products guy- I am interested in why you only use internal standards for ethanol and cholesterol. Are these considered "difficult" analyses, or perhaps they are just "important" analyses for which another layer of safety is required?

[ravenwork]

We use internal standardization with a fluorescence detection method. This is useful as the fluorescence response degrades as the lamp ages. Furthermore, the method uses a post column reaction to generate the fluorescence response, and the post-column reagent also degrades as it ages. Both of these response changes are observable within a single batch of samples. I am thankful for the internal standardization in this case. Without it, I would probably have to run smaller batches and re-calibrate much more frequently.

[HW Mueller]

ravenwork, your batches must be huge?

To expound on what Tom (Jupille) said, one general example, mentioned before: If you do LC on a substance in hospital patient´s plasma/serum, you may be in for surprises regarding some patients´ LC. An IS can nicely show something went wrong. You don´t have to let the IS mess up your precision, just use the external standard (ES) for the calcs.
(see for instance: J Chrom B, 678, 137 (1996).

Regarding the assumption that an IS or ES is ok: Strictly speaking this is only possible if two independent and largely different analytical methods have been used to show this. Now, since this is not a general requirement I also think it´s odd that a referee would hit on this ES stuff as mentioned by wxmwqr, especially if he did what´s done in 100´reds of thousands of other articles that also did not use IS.

[Rafael Chust]

It is quite curious because I have this kind of discussion about once a year!

I must admit I have a natural tendency to use IS - my first HPLC teacher did not admit any method WITHOUT IS. Double trouble, as besides develop the method we had to find an adequate compound for IS!

Anyway, my rule of thumb is: when sample prep is simple (filtering and/or dilution) I do not care about IS, although I always inject the ES every 4 or 5 runs. When my sample prep is complex (SPE, solvent extraction, etc.) I always use IS or standand addition - another "crazy" but very effective method I convinced my teacher to accept!

[ravenwork]

For the method I discussed, we run batches of 20 samples. The total batch run time is about 16 hours. The degradation of the lamp is most noticeable when the lamp is brand new. When new, the response is highly sensitive and shows a measureable drop over the course of a batch. Once the lamp is about one-third used, this drop is less noticeable, more of a batch-to-batch change.

This is an environmental method operated at high sensitivity. It is a direct injection method, no sample prep.

My experience with this method has been a learning experience with IS. I would rather not operate a post-column reaction fluorescence method without IS. Unfortunately, I have another very similar method that doesn't have a good IS compound. For that method, I can predict when I will need to recalibrate based on the number of hours the lamp has run since the last cal.

I would probably never bother with IS for an HPLC-UV method where little or no sample preparation takes place.

[Consumer Products Guy]

Victor - good question. We use internal standard cholestane for GC cholesterol assay in foods because the "traditional" standard method uses it; however, we do not use benzene as solvent, we do use capillary GC, and we do add the internal standard early so it follows through the entire procedure, where "traditional" procedure adds it as a last step prior to GC (doesn't make sense, others have followed it through the entire extraction). For ethanol levels above 60%, we use n-propyl alcohol as internal standard mainly because we are staying close to an AOAC GC procedure, because how does one spike placebo product accurately with 60% of a volatile component? So we use the n-propyl alcohol internal standard. For lower-level competitive products, we would make up ethanol and sample in water and simply use external standard quantitation using the same GC conditions. We get great injection reproducibly either way, only inject 0.5 ul of the water solutions.

[Victor]

Consumer Products guy-thanks for your answer. Do you have any observations on the reproducibility of cholesterol injections without using cholestane as an internal standard? I have only ever done this manually using splitless injections in hexane or dichloromethane, but the reproducibility was never wonderful without cholestane. Perhaps it is o.k. with an autoinjector?