Delivering Better Insight
Even the best-run labs can’t deliver perfect accuracy or precision in testing results. Ekidna’s patented technology delivers error rates of +/- 1% by weight, as accurate as a lab — and well within acceptable tolerances.
While error rates and error propagation are not a simple topic, we’ll provide an overview of the factors affecting accuracy and explain how we can control for them.
What is Error?
Measuring THC and CBD
There are two typical ways to measure the amount of cannabinoids in samples:
- Weight percent: the numbers that are quoted on packaging for THC and CBD. These are typically in “gram per gram” amounts, meaning that a 20% THC product has 0.2g of THC per gram total product. These weight percentages are the ones that everyone is most familiar with.
- Relative Error Percentage: when describing your test results, this is the percent off the true value of whatever you are testing. In this case, 5% error means +/- 5% off the true value.For example, if you had a 20% (weight percent) THC flower sample, and you tested this sample with a test that was 5% error (relative error percent) you would expect your test results to fall between 0.19g and 0.21g THC. This results in a weight percent reading between 19%-21% THC for a sample that has a true value of 20% THC. Ekidna promises to deliver +/- 1% (weight percent) accuracy with our devices.
The Current Testing Landscape
There is evidence that the variation in laboratory testing results can be as high as 19% weight percent¹.
This is a staggering error range. A study of a group of 84 different CBD products purchased from a variety of companies found that:
- 42.9% had less CBD than stated on the label.
- 26.2% had more CBD than stated.
- 21.5% had THC in them when they were only supposed to have CBD².
Combined, these results show that standard labs have relative percentage errors between 2-15%. Ekidna devices are at the low end of that range at about 5%, depending on the sample type.
There are many factors that can drive this variation across different labs. Some of these include:
- Sample preparation methods
- HPLC maintenance levels
- Calibration differences, standard
- Reagent variations
- Operator experimental error
- Sample storage
- HPLC equipment variations
- Sample complexity
- Chemistry volumes
- Column age
- Environmental conditions
- Large dilutions
- Sample instability over time
- Sample complexity
The main thing that can affect error range is sample complexity. For example, a pure extract of THC or CBD will always be the easiest to measure, as the sample is very “clean” (meaning it has no other or few other ingredients).
A flower extract is slightly less clean, as it also contains other cannabinoids, terpenes, etc. THC/CBD are still the largest component (by weight) in this case, and so there is still relatively little background noise.
On the other end of the complexity spectrum there are things like cannabinoid-infused skin creams or edibles. These products have many different compounds in them (e.g. essential oils, other plant extracts, flavouring, colouring, preservatives, etc.) that can all affect the level of noise in the signal and lead to less accurate readings.
For Hemp Growers:
Lower Limit of Detection
A limit of detection is the lowest amount that a test can identify is present, but not necessarily accurately quantify. The current lower limit of detection for our test kits is about a 1.5% (weight percent) THC flower sample.
The technology behind our sensors is capable of much lower limits, however, and we are currently tuning a new version of our test kits for hemp-specific analysis.
Please contact us to get more information.
Sample Types Tested
Accuracy in Testing —
Without the Need for a Lab
Ekidna is fast, accurate and decentralized. Learn more about how Ekidna can bring
efficiencies to your operation.
- Liz Wagner, Michael Bott, Mark Villarreal, M. H. Industry Insiders Warn of Fraud at Marijuana Testing Labs – NBC Bay Area. NBC Bay Area Investigative Unit (2017).
- Bonn-Miller, M. O. et al. Labeling accuracy of cannabidiol extracts sold online. JAMA – J. Am. Med. Assoc. 318, 1708–1709 (2017).