I am curious if any analytical chemist have looked at the variability of their potency results when submerging pipette tips into the sample solution at various levels?
We recently submerged a pipette tip into a concentrate prep (60 mg mass + 10 mL solvent) at a low point, medium, and high point. Since this activity is conducted before diluting the sample, when applying the 50x multiplier at the instrument, the error is propagated out and we are seeing a 5% effect at the low point and a 10% at the mid-point of submersion. If we submerge the entire tip, bottom line is affected by 25%. We believe this can cause an enormous amount of discrepancy when running different samples between different analyst. So a difference between an 80 and an 85% sample could be caused from this simple lab step.
Our sample prep for concentrates uses 50 mg sample + 10 mL solvent + 50x dilution to get into the 50-100 ppm range at the instrument. Everything used to measure these values are verified and calibrated.
Are there any techniques used across other fields when measuring highly concentrated material using a super sensitive instrument? How do you minimize the overall error, while keeping costs, supplies amounts, etc. low? I mean as a lab, I could say that we have 20% error and live with that but the clients needs the error to be 5% within the target. Achieving that tight of an error range seems almost impossible, given our current technique. I am curious if any other chemists have ran into this same issue?
Do you not use something like a vortex mixer to mix your samples prior to pulling your sample? it sounds like all you need to do is add a step for mixing prior to pulling samples.
Are these variations all coming from the same prepared sample? How long is the sample being vortexed and/or sonicated? And is this happening before or after you prepare your dilution?
As for pipetting it can be tricky especially when you have multiple chemists preparing different samples, but if you can make sure that everyone follows their method of pipetting for every sample, you can avoid a lot of error. For example, when I go to transfer I will aspirate and dispense the pipette two times, then pull up liquid on the third time, wipe the excess on the lip, and dispense into the desired vial. I do that every single time with every single thing I pipette.
I would repeat this test with a reference standard, otherwise you are assuming every source of uncertainty and error is derived from one source. With a reference standard you will have less variables to deal with.
Personally, I would be quite shocked to see this large of a discrepancy in testing just due to the depth of the pipette alone.
Also this is what internal standards are helpful for.
Edit: You could even test this with some DI water on an analytical balance.
Yes, we are looking to add an ISTD but the logistics of that is challenging considering the wide range of matrices, concentrations, prep and dilution techniques.
The issue arises with the sample preparation before dilution. Lets assume the sample is fully homogenous for argument sake. For the test, I used the same sample and dipped the tip at 3 different levels.
Yeah that’s fine our pipette reads tighter than that. The problem is once you multiply the math out.
At 85% THC, the starting concentration of a 50 mg sample in 10 mL is 4250 ppm in the tube. I then will do a 50x dilution from that tube into 1.02 (1 mL + 0.0204mL) total mL. At the instrument this equals = 85 ppm. At this dilution step, if I just put the tip at low level 1, I get 7 ppm at the instrument. At the mid level I get 12 ppm. And at full tip submersion, I get about 28 ppm. I did not suck any liquid up into the pipette. I just dipped the tip into the sample and then dipped the tip into 1 mL solvent. these values are at 1x. So when i add the 50x dilution factor, that we normally do when running concentrates, these values are multiplied by 50. So by just back-calculating you can see how slight bias or error can be extrapolated out immensely.
It tells me that my starting concentration in the prep tube may be too high. My cal curve may be too sensitive.
I am curious in what other industry and what techniques are employed to measure 70%+ components accurately, involving a very sticky matrix. Is it common to wipe the tip?
Ideally, most of these issues would be taken care if we created in-matrix calibration curves. The bias from the tip would be exhibited in the calibration standards as well, but have minimal impact on the calculation of the unknowns. But we can’t make these
Also, yes we do the same thing. Prime the tip with the desired sample aliquot multiple times. Then we wipe excess off before dispensing into the vial. then we wash the inside of the tip multiple times.
My background comes from working with low concentration material where an excess of 3-5% generated from pipette error does not cause such a huge discrepancy in the final reading. I have never worked with very concentrated, sticky material, so I don’t know if techniques, such as wiping the outside before dispensing into the solvent is appropriate and/or recommended.
It definitely makes a difference, I calibrate pipets in my lab weekly, in doing so you understand where sources of error could occur. Pipeting solvents also requires priming as the evaporation rate in the tip can cause discrepancies.
Yes. When I do method validation pipette technique is always part of the training and its part of the analyst error when I’m doing my error calculations. There are automatic pipettors that remove this (every tip depth is the same using a robot). When I was at MUCH larger labs, handling hundreds if not thousands of samples a day, we had one of these robots. That got us so much closer to where we needed to be to satisfy our customers.
I’m going to assume that this is a similar amount of challenging to getting everyone prepping samples to uniformly pipette. There are methods (everyone does it the same way every time, using auto-pipettors, using ISTD so you can “see” the error).
I felt the need to point out that there is inherent error every single time you draw into the pipette - so you are adding that error with every “rinse”. There’s a pretty cool class by ANAB that really digs into how to calculate and handle measurement uncertainty. You’re questions about this topic might be answered by attending. NIST has a cool web app that helps calculate this (it really takes everything into account, each use of scale, pipette, person, instrument…). I’ve been having these battles with measurement uncertainty for the majority of my career. Its all about consistency and monitoring for when that consistency goes away.
But honestly - if your method is validated and you are monitoring your trend for your system suitability and standard checks then you are already doing the things you need to do to control this variability. What you have uncovered is the “why” behind the reason these are best practices for laboratory operations. Making sure every tech and analyst understands this - the impact it can have on sample results - how to identify if there was an impact on results - and what to do if that’s the case (AKA lab investigation and reprep) that’s the important part.
From your reply to Chaboes, if you’re putting in just the tip (lol) and seeing that big of a discrepancy, it seems that the concentrate is making its way to the bottom of your sample tube, which makes sense for a heavier compound. Is it sitting for a while before the final dilutions are prepped? It may just need to be shaken for ~20 seconds before you pull your aliquot.
What is the PPM range on your curve? I think you said above that your target is 50-100 ppm to read on your instrument so I am guessing that your range is 0.5 - 200 ppm or something close to that. And for a 50x final dilution, is that 20 ul of supernatant into 980 ul of solvent? That could definitely lead to some high variability, especially if your instrument is super sensitive.
Are your sample preparations methods set in stone? Do you have the ability to increase your initial solvent levels from 10 mL to 20 mL? If my math is right, then 50 mg @ 85% THC in 20 mL solvent is 2125 ppm. From there you can run a 25x dilution (40 ul s-natant in 960 ul solvent) and get down to 85 ppm. Or if you can exceed 100 ppm, you can increase that to a 20x dilution (50 ul s-natant in 950 ul solvent) and it comes out to ~106 ppm. If you start with 25 mL initial solvent with the same concentrate, that puts you at ~1700 ppm, and a 20x dilution can get you to the 85 ppm target as well.
I’ll test ourselves soon and see if we get a similar difference.
Our procedure is usually 350mg into 20mL then sonicate 20min. Then 1mL of that solution into 10mL of solvent. Shake and sample from that. Filter the sample into a vial with about 1mL solution for the instrument.
We’ll test the 11mL dilution for the top and bottom discrepancy.
Why is this sticky? If your solution is still sticky after extraction, you might want to increase your extraction volume to 20 mL and reduce your final dilution…
Yes, lowering the initial conc. in the tube would reduce the amount of excess sample sticking to the pipette tip. Most concentrate is homogenous, but some aren’t so much. But I don’t want to weigh up less than 20 mg of sample. And I don’t want to double my solvent due to increased cost in tubing + solvent…