Very unusual precision issue! Please help!

[a17thony]

We are having an unusual issue with the precision on one of our HPLC instruments. The injector precision is good for every other run. For example, if you ran six injections from one vial, injections 1, 3, and 5 would have good precision with each other (within 2%) and injections 2, 4, and 6 would have good precision with each other (also within 2%). If all six are compared with each other, the %CV might be as high as 10%.

The instrument is an Agilent 1290 UHPLC with the wellplate autosampler. If anyone has ever seen this and has an answer on how to fix it, please let me know. Agilent support is going to take almost two weeks to get here.

[Alexandre]

Looks to me as column is not conditioned enough after each run.

[Peter Apps]

Is this repeatable - how many sets of six have you run ?

What happens with six replicates from six different vials ?

Peter

[mkumar75]

Dear ,
Subjected to condition that only by changing the instrument , it has happened

Then check that sealwash & Injector wash are proper in the system .
& Last is Injector port leakage .

Thanks

[HPLCaddict]

You're referring to peak area RSD, right? What about retention times?
Isocratic or gradient? Is the chromatography "OK" (no overlapping peaks, enough retention of your analyte,...)

[a17thony]

Hi All,

Thank you for your input. The issue only happens on one instrument and is repeatable no matter how many runs you do. We have not noticed any leaks in the injector port and have changed the metering seal and the rotor seal in the injection valve. The method is a gradient method and the retention times are quite close for all runs. The precision problem I am talking about is regarding the RSD of the peak area.

Thanks!

[carls]

Take a very close look at every parameter in the injector settings (advanced injector settings as well) and injector program. Maybe someone mistakenly changed a setting on this instrument? Things to look for are:
1) automatic gradient delay (should be unchecked unless you specifically need this). If you need this feature then check/adjust the flush out factor. A minimum of 5 is recommended
2) Valve cleaning? turn off/uncheck if not needed
3) Injector program? turn off/uncheck if not needed. Delete all lines unless you must use a specific injection sequence, i.e. derivatization, etc.

[Peter Apps]

Hi All,

Thank you for your input. The issue only happens on one instrument and is repeatable no matter how many runs you do. We have not noticed any leaks in the injector port and have changed the metering seal and the rotor seal in the injection valve. The method is a gradient method and the retention times are quite close for all runs. The precision problem I am talking about is regarding the RSD of the peak area.

Thanks!

So, each even-numbered sample matches the other even numbered samples, and each odd numbered sample matches the other odd numbered samples, and this continues indefinitely through a long sequence. Is this still repeat injections from one vial ?

It is very hard to imagine what might cause a regular alternation of peak areas, unless it it something in the programmed sample sequence. What happens if you do six single runs, with a manual start for each one ?

And what are you analysing, and how ?

Peter

[a17thony]

So, each even-numbered sample matches the other even numbered samples, and each odd numbered sample matches the other odd numbered samples, and this continues indefinitely through a long sequence. Is this still repeat injections from one vial ?

It is very hard to imagine what might cause a regular alternation of peak areas, unless it it something in the programmed sample sequence. What happens if you do six single runs, with a manual start for each one ?

And what are you analysing, and how ?

Peter

Well, this happens indefinitely no matter how many vials you have so if you have five samples, for example, each with 3 injections, each of those samples will have two replicates that match up and one that doesn't. The six injections from one vial was an example to illustrate the point.

The idea of doing six single runs is a good idea. I will try that out in the lab.

The sample is a protein by reverse phase in a gradient method. One other thing is that the retention time of the protein does not change between injections and only one method is used throughout the entire sequence.

One thing that we noticed also is that while the baseline (which increases as ACN percentage increases) is steady through the section where the protein elutes, in a step in the gradient where there is a high ACN wash at the end of the run to clean the column, there is the same alternating pattern. I will post a picture of this in a few minutes when I get back into the lab.

Thanks everyone for your help.

[cody84]

Is the first injection area lower than the second? Is it possible you have very consistent carryover? Try a longer cleaning gradient in the run and then multiple injections again.
Actually I don't think that even makes sense...because the 3rd injection would be 'clean' again...hmm

[lmh]

Something in your system is capable of remembering whether it's on an odd run or an even run. Your system has two states! It could be something periodic that coincides only every second run (point in the pump's cycle where the injection gets made, or something). It's highly unusual behaviour, whatever it is, and possibly a weird side-effect of some other fault.

Can I suggest seeing what happens if you run the method but with a slight change in flow-rate?
Also, what happens if you change the injection volume (say, inject half as much)?

Can you replicate the effect on a short method injecting a single standard with no column, just some sort of restrictor to apply a realistic back-pressure? If so, that would speed up the diagnostic process, as well as excluding strange chromatographic effects.