Ha! The space bar can be your friend.
I guess Im not saying that I cant be at fault but its also not my first day and I think that our kitchen may play a bigger part in this and hence driving me crazy.
We use purchased isolate that is kept in an opaque,plastic container in a refrigerator.
My original thought for this thread was to see if others were having the same issues and who do you believe?
Maybe it was answered earlier by just sending a sample around and see who gives you the number closest to yours?
1 Like
I would say its human error/precision problem.
I also recommend making 3-5 samples exactly how you normally go about it. take a little break between each sample and then go back to it. If you have someone else that is experience have them do the same, compare your results and see how much they vary. It shouldny vary much unless there is something wrong with your method or glassware has some variances.
Has anyone ever used a Hamilton microlab 600? I have been thinking about trying it out. Its supposed to be as accurate as grade a glassware but much less chance of human error.
Go crazy and calculate the linear retention index for each compound of interest, then have the lab do the same. You will find the discrepancy. Probably standards or instrument calibration. Any instrument tech will tell you the vast majority of all the instruments they have in the field are not being operated within their defined parameters.
1 Like
What are your R^2 values for your target analytes?
Using microliter syringes from Hamilton instead of pipettes for dilution of standards got me better calibration curves. No weighting, no removed points (9 total; 0.5, 1, 5, 7.5, 10, 25, 50, 75, and 100ug/mL), and at least 0.99997 or greater R^2 for all analytes.
How many digits does your scale display? Your 3rd party lab may have, for example, a scale that reads to the 0.001 mg instead of your 0.1 mg scale. More digits leads to greater accuracy and precision. Less digits leads to rounding and less repeatable results.
4 Likes
You went in the software and updated the sample weights for each one, correct?
Ah, seems like you have a 1 mg balance. That’s almost certainly your issue, or at least one of them.
Spend a few grand on a 0.01 mg balance and you’ll probably see your results fall more in line with the third party lab’s.
0.09567g will always give better results than 0.100g. Make sense?
1 Like
Sure did.
2 Likes
Why half a gram? The typical protocol for PeakSimple with an SRI GC is a tenth of a gram, or as close to a tenth as you can.
Guessing it’s because that’s what grandma’s recipe says to use with one of them new fangled hplc thingymabobs 
1 Like
I’ve seen this before - at a previous facility in NY, we were consistently 10% lower than the state lab. We had to be within 5% in order to not have to make a new lot number for the product. At the time (not sure if this has changed since), the state lab was the only lab so I held bi-weekly meeting with them to try and discern the real issue at hand.
While there are a number of factors that can influence the lab-to-lab accuracy, in my experience it really comes down to sample prep.
The difference between sonicating for 15 min vs 30 min, methanol vs ethanol as a diluent, accuracy in the sample prep tools… they contribute significantly.
If you’re consistently lower, then it could be that your sample prep extraction is inefficient. It could also be your pipettes, although I haven’t seen this contribute to more than a 5% difference, especially if they are 3rd party calibrated (if yours are not, just bite the bullet and do it). Cannabinoids like to stick to plastic, so glass is always better for greater accuracy with these guys.
My advice to you would be to run a split lab accuracy study and run several samples back to back to look at the true variance between your results and theirs. Basically perform a mini validation with the lab. This won’t help you fix anything in particular, but knowing if the issue is actually repeatable will tell you a lot. If it’s not repeatable, then it’s an instrument issue. If it is repeatable, then it’s most likely a sample prep issue.
Keep in mind that in clinical toxicology, you can actually have up to 25% variance on proficiency tests. So, albeit annoying, 10% variance is actually pretty good.
3 Likes
Thank you, took me a while to do your math.
I was glad to see your math too.
I was going with 1000ng/ul == 40% from memory.
Or more correctly from the memory of making a combined 3 cannabinoid standard giving a 13.33% reading.
I would have typed it out more fully if I wasn’t on my phone grocery shopping at the time 
1 Like
I will echo the sentiment regarding the influence sample preparation has on accuracy. At the pharma company, when an OOS was generated root cause analysis would often demonstrate some mishap with sample preparation. Someone once got an OOS on a produced called sorbitol-mannitol. The solutions dept. would batch this product hot, and when it was delivered to the lab we had to cool the sample bottle to room temp before we diluted our solutions. I think she pipetted while it was still hot and got an OOS for low potency.
Another product was an amino acid mixture : l-tyrosine, l-tryptophan, n-acetyl-l-tyrosine. Always had issues with one Of those compounds being OOS, due to standard prep. issues (weighing the standard and sufficiently dissolving in a sonicating water bath. I think it was tyrosine that was so stubborn.) it was exceptionally rare that the solutions department actually made a mistake batching (and these were at times 80,000 kg batches of drug solution)
OOS = out of speification
4 Likes
I think that was the problem with my wacky results. Sample prep.
Do you think its better to tell the lab – hey this sample is xyz in MCT - so there’s a better chance the sample is prepped properly?
1 Like
Giving them an approximate potency is also helpful.
I’ve had a lab return 10x more THC in my CBD tincture because they injected too little…despite me telling them the target range. Took them three weeks to return 20:1 CBD rather than 2:1, despite having tested the concentrate from which it was made.
2 Likes
IMO 10% variance is accepted as internal error. I have heard of different companies reporting false numbers (lower than actual on the order of 10). Have you checked that you are properly calibrated for the standard you bought? To really test it out you could run several tests on another HPLC machine or through a GC-MS and compare those. If you note that the variance on different machines vs standard is larger with your machine, maybe there are calibration issues. 10% is fine though, even if you get consistently higher.
Standard deviation is used to describe precision. For accuracy you really need a certified reference material with a known purity, not just a random isolate.
2 Likes
I often confuse precision and accuracy… 
The certified standards are good to start with. But in fact I find much more convenient working with isolates. Not random isolates of course, ones that I purified myself and characterized well. Even better if using a range of isolates with various levels of impurities inside. Certified standards are too diluted, plus if using GC, one need to further dilute the 1ml solution with the internal std… in addition, once in solutin, the stavilty over extended time is lowered. 
I only see such necessary if doing certified procedures. But not for doing research and routine tests.
2 Likes
Isolate is isolate. Someone did original method development. Most cannabis standards are garbage and can be improved upon by a large number of people. In food and beverage in house is always trusted above 3rd party.
2 Likes