In house vs. 3rd party testing results

Forgive me if this has been covered somewhere else but…
Im having a tough time lining up our in house potency testing results with 3rd party testers.
I operate the HPLC here and I seem to get consistently higher results than when we send them out for verification.
Our only fix seems to be to try and formulate 10% higher and cross our fingers.
I know that theres production issues in house and variances there but…
What is everyone elses experiences and what do you do to combat it?

TIA

Sometimes Lab give especially on THC result on the lower end… In order to not scary customer, and lose potential future business with analisys…

It could be a difference in the quality of the standards you’re using or they’re using, poorly calibrated pipettes, math error, etc. What are you testing, flower/edibles/topicals/concentrates?

When I have my internal analysis properly calibrated and tuned up I’m generally only 0.5%-1% off from the state lab

Im testing primarily flower,chewables and tinctures.
I use check standards to check my curve daily(for CBD) and routinely get within 10% which even the instrument vendor was satisfied with.
Ive calibrated my pipettes in house with water and use standards from Cerilliant.

10% off from your standard curve? Well, there’s your problem.
Your curve went close to zero right? Y and x intercept both very very close to zero? And the curve is straight?
If so you can get by with single point calibration and then you can easily calibrate often, every day or more.

I’d have my head on a stake if I got results that were 10% off. Hell, the other people at my company get grumpy about 2% variability

2 Likes

Have you run any stats on your standard injections? Are they consistently above or below the expected value based on the calibration?
What about injection to injection variability of the same sample? Stats on those?

1 Like

Is this accuracy vs precision?!?

a single operator who gives a damn should be able to beat the statistics of multiple operators who are merely doing it for a paycheck.

Ie you should have less sample to sample variation that your third party lab. I could certainly tell once the PhD chemist handed the sample prep off to paid employees with the local lab I used to favor.

If your numbers are always high, adjust them down. Or find another lab.

There is actually good data that in WA labs tend to drift upwards over time.

Customers buy based on numbers, higher numbers sell more flower, labs that artificially inflate numbers sell more tests…

Have you talked to your 3rd party lab director? Who are they buying their standards from?

Will they inject your standards for you?

4 Likes

We’ve had to move to two injection volumes (2ul and 10ul) in order to target the center of our cal curve. We also moved to all class A glassware for dilutions.

Would a better curve help?

It took me 3 injections each at 7 concentrations to build a calibration curve for my SRI-GC that covers three orders of magnitude (30ng => 2000ng)

Haven’t tested this one against my local lab yet…

(Three replicates and seven levels is where peaksimple maxes out. Single point calibration “curves” are not ideal)

Would you mind breaking your math down for me?

1 Like

Injected 2ul at 1000ng/ul

1ul, and five serial 2x dilutions.

So 31.25ng to 2000ng

Not quite 3 orders…

1 Like

You can prepare samples in simple PP centrifuge tube and be fine with that.

The first thing to check is the quality of your preparation, i.e. you accuracy. You have to run replicate analysis (different samples of the same homogenized material) and calculate the standard deviation of the results. This value should fall well below 5% of the measured value, even below 1% if you are good. If using small volumes (below 50ml), using gravimetric dilution instead of volumetric dilution should also improve your accuracy.

Then you also of to check your precision. The best way is to analyse cbd isolates for this. Even better if you get a set of them ranging betwwen 97 and 99.7%. The calibration should be checked twice a day ( if you don’t turn the machine ar noght, then 1 time should be fine as long as you are not checking isolates…).

If you are systematically 10% above, this sound like calibration, or maybe calculation issues.

1 Like

Send your same sample to 5-10 different labs.

Get rid of the outliers.

Average the remaining.

Calibrate your machine to that number.

5 Likes

So…
I run all of my analyses in duplicate and in the beginning I would do a replicate injection out of the same vial.
I stopped doing that because THOSE replicates fell almost on top of each other.
Duplicate extractions are another story…
Tinctures can vary by 100mg/bottle(based on calculations)
A little background on me:Im a 25 year analytical chemist specializing in elemental analysis by ICP-MS.
Ive been in this industry about 7 months and this is my first go around with organics and HPLC and cannabinoids.

2 Likes

What do you mean here ?

Ive used the isolate I have here to create a spiking solution to mirror my analyses to EPA style.
I make 100 mg/ml and check it against my curve and get 97% recovery.

Sorry…
Let me express it this way(this is today’s struggle,btw)
I weighed out 4 different aliquots of the same tincture: 0.510g,0.504g,0.501g and 0.505g.
The CBD result from my curve in ppm = 40.7,36.2,39.5,37.6

how do replicate injections spread?

If you’re 10x tighter on repeat injections than replicate samples, it suggests your balance or liquid handling is the issue.

1 Like

I’m having a hard time reading these results, its all smooshed together. You need to hit the space bar lol

Are you using house made isolate? If so are you keeping it dry after your purifications? Its a bit hygroscopic and will take on moisture from the air, enough to mess up analytical results.

Replicates on top of each other, as in the results were similar? Whats the stdv of say, 10 replicates?
Of 10 duplicates?

If you’re getting good precision and your results when using a new vial of purchased standards are making good sense then I would start to think its the external lab thats the issue.