This article was written in partnership with Anresco Labs.
Potency stability continues to be a primary concern for the cannabis beverage industry, so it can be frustrating when results from lab testing come back out of spec. As advocates for safe infused products, it’s important to reflect on the impact of the variety of processes that may cause issues between labs, instruments, and samples. One of the most critical things to keep in mind about lab testing is that uncertainty at each level of the process compounds with each subsequent error-introducing step. Consider this when looking at any individual source of error.*
*Note: In California, the highest accepted deviation from label claims for Cannabis goods is 10%, so keep that in mind as the framework when discussing the impact of these variations.
Let’s start by understanding how the testing process works. High Performance Liquid Chromatography (HPLC) is the standard analytical technique for cannabinoid profiling. In this technique, cannabinoids are extracted in a solvent (usually Acetonitrile or Methanol) and a small quantity of this extract is injected into another stream of solvent. This sample is then pumped through a special tube, called a column, that’s packed with beads. These beads have high surface area and can have different molecules attached to them, depending on the characteristics of the analyte of interest.
Cannabinoids are generally non-polar molecules, and as such, non-polar carbon chains are the molecules of choice for the beads referenced above. When the cannabinoids are pushed through the column, they are attracted to the carbon chains through Van Der Waals interactions, with the strength of the interaction dependent on the structure of each cannabinoid. Cannabinoids that bind more strongly are pushed through at a slower rate.
Above: component A binds more strongly than component B, so component B will exit the column first.¹
HPLC pumps are very consistent, so the time it takes for a sample to pass through the system to the detector is extremely repeatable. This is commonly referred to as the retention time. So long as everything is operating as intended, each cannabinoid will have a consistent retention time that allows it to be distinguished from another.
After the column, the stream heads to a detector, a component that converts a physical property of the analyte of interest into a quantifiable electrical signal. Cannabinoids absorb UV light and each cannabinoid absorbs a different amount of each wavelength of light. Knowing this, scientists then use a diode array detector (DAD) to measure the absorbance of light across the UV and visible spectrum to create an absorbance spectrum that presents the unique “fingerprints” for each cannabinoid.
The signal is also concentration dependent. The Beer Lambert law dictates that absorbance is directly proportional to concentration, so not only is the signal unique for each cannabinoid, but the strength of the absorbance signal is directly related to its concentration, all else being equal. The resulting chromatogram is compared to a standard curve, the results of a sample purchased from a certified production facility.
Above: The spectra of THC, THCA, CBD, and CBDA, showing the absorbance (y-axis) across a range of wavelengths (x-axis). A lab may choose to select one singular wavelength where all cannabinoids have absorbance, or select different wavelengths specific to each cannabinoid depending on the sensitivity that is required for the analysis.²
Above: These two graphs show the chromatograms for a standard and a sample, respectively. Each of the peaks represents a different cannabinoid passing through the detector across time on the x axis. The height of the peak represents the absorbance of each analyte. The Standard Chromatogram shows each of the potential analytes (cannabinoids) of interest. The Sample Chromatogram is then compared to the standard run. In this case, one can see high amounts of CBDA and THCA in the sample, while some analytes (CBDVA, for instance) are not present at all in the sample. ³
Now that the groundwork for basic cannabinoid profiling has been laid, we can discuss the variables for error at each step of the process.
The most important part of HPLC analysis is the standards, or certified reference materials (CRM). These are samples that contain a known, verified concentration of singular or mixed compounds that must be purchased from an ISO 17034 certified production facility. This is where HPLC analysis begins.
The lab receives the standard at a known concentration. It is then diluted to various concentrations to create a calibration curve - a best fit line relating the absorbance of each dilution of the standard to the calculated concentration - the range of which is determined by the response of the instrument. In our experience, this is the largest factor in potency variation between labs.
Look at figure 4 for an example calibration curve that can be obtained by measuring the absorbances of certified reference materials and plotting them against the concentrations they are prepared at. The x axis represents the concentration of the standard levels (2.5, 5, 10, 40, 75, 100 ppm) and the y axis represents the absorbance of that analyte at that level. The absorbance of unknown samples can then be measured and the corresponding concentration can be calculated. For example, a sample with an absorbance of 100 would have a concentration of approximately 20 ppm.
At this point, calibration validation occurs. A second source of standards is run at a known concentration, and the result of the HPLC analysis is compared to the expected concentration from the dilution. If the secondary standard is within a certain range - +/- 30% via California Department of Cannabis Control text of regulations - then the curve is said to be validated and can be used for analysis.
Note: The 30% variance allowed likely came from contaminant testing, which is not really an appropriate perspective for cannabinoid testing. Take pesticides for example. A vast majority of samples run for pesticide analysis are not near the action limit (total concentration allowed by regulations) - either they are ND (none detected) or low level, or they are significantly higher than the action limit. The only ones that would be impacted by the 30% tolerance are those with concentrations near the action limit, which represent a very small fraction of the total number of samples. For cannabinoids, however, even a 10% difference can mean massive financial ramifications - from relabeling a finished good or massive changes to the value of flower or distillate.
The most alarming piece of information regarding CRMs is the variance of cannabinoid concentrations between different suppliers of CRMs (or even different lots) despite the fact that they should contain the exact same concentration.
The chart in figure 6 shows the variance in concentration in CRMs from different vendors. Let’s examine D9 THC, the second to last. The lowest is coming in around 96.5%, while the highest is coming in around 103%. This is a 6.5% difference between the two reference materials. For sample analysis, you are comparing the absorbance of your analyte of interest to the response from a known concentration. Therefore, a lower concentration in the calibration standard will yield a larger calculated concentration in the sample. If there is only 10% variance to work with regarding potency label claims, 6.5% is a large portion of that without even touching the sample material yet.
CRM’s arrive in the lab in 1 mL ampules of known concentration, typically 1 mg/mL. This initial concentration is often too concentrated for the HPLC to generate a good response (too much analyte can saturate the detector). In order to create a calibration curve, at least 5 dilutions of this 1 mg/mL starting material must be made. Any analytical tool, even operated by a well trained technician, will impart some error. Weighing and dispensing tools all have tolerances for what margin of error is deemed acceptable.
Syringes are typically used over autopipettes for calibration standard dilution as they tend to introduce less error when operated correctly. Imagine you have your 1 mg/mL CRM, but need to dilute it to 0.5 mg/mL for the top of your calibration curve. One might add 500 uL of diluent to 500 uL of the original CRM to do a 1:1 dilution. However the chart above indicates that using a syringe can still impart an average error of 1.14%, meaning the actual concentration could vary between 0.495 and 0.505 mg/mL.
Again, one point doesn’t make the curve, so this material is likely going to be further diluted (possibly serially diluted, where one makes a dilution, then uses the diluted material to make a subsequent dilution). At each of these dilutions, more error is introduced.
Cannabinoids CRMs typically come in and are diluted in Methanol - a volatile solvent. Volatile solvents provide greater difficulties for volumetric measurements and leaving a vial of either the CRM or its dilution open to atmospheric conditions can cause the evaporation of methanol, which would increase the potency of the calibration standard.
After the instrument is calibrated, samples need to be prepared. Sample preparation is as important as the standard preparation methods outlined above.
In most markets, sampling procedures for in-process material (i.e. distillate that will be later emulsified at Vertosa’s facility, or Vertosa’s emulsion that will be turned into a finished product) are not very heavily regulated. During the sampling processes, it is possible to get a portion of the material that is not representative of the whole batch.
Example: A flower producer may select the best looking buds to submit for testing, so as to get a higher potency value, though it may not be reflective of the overall THC content in their entire harvest.
For finished goods, there are often stricter guidelines, though there are undoubtedly market pressures on manufacturers, especially of flower and distillate, to select portions that will yield greater potency results. It is very important to ensure that the material tested is fully representative of the lot of material in larger pass through manufacturing operations, where deviations in potency may have large impacts on the final product. Guidelines as in Figure 8, from the NY Office of Cannabis Management, can help determine what a representative sample may be, but internal Quality Control Programs should establish specific sampling requirements based on the makeup of the product.
The first step after sample selection is homogenization. Given the irregular makeup of some cannabis products, homogenization is necessary to ensure that the sample tested is representative of the bulk of the material. For flower, this means grinding the entirety of the plant, including its leaves, stems, seeds, etc.. For edibles, this can mean melting or cryoblending (using a liquid nitrogen or dry ice to freeze the material before blending). For concentrates or emulsions, this generally means mixing (using heat to loosen distillate). Homogenization is crucial to ensure consistent testing results. After homogenization, the sample is weighed out to a specific mass appropriate for its sample type.
Some product types are inherently non-homogeneous. Chocolate covered espresso beans, for instance, have their cannabinoid content in the chocolate only, so it is important that the lab homogenize the material so that the quantity weighed out for testing contains the same ratio of chocolate:espresso bean as the final product would, otherwise they potency would not be reflective of the submitted material.
Just as the calibration curve is impacted by the error introduced by analytical tools, sample extraction procedures will introduce error. Training is a very important part of this aspect of the process. Typically, lab technicians (as opposed to analysts) are performing this task, who are often lesser experienced members of a lab team operation.
At this stage, an accurate sample weight must be taken. This can be tricky when dealing with concentrate samples, for instance, as the material is typically heated before weigh out, which can produce large strands of distillate that can stick to the outside of the weigh vessel. Anything on the outside of the vessel will not be extracted, giving an inflated weight. Similarly, balance calibration and lab airflow (impactful for sensitive analytical balances) can lead to inaccurate measurements.
Once the sample is weighed, the cannabinoids need to be extracted from its sample matrix, which is essentially anything that is not an analyte of interest. Using infused chocolate as an example, the cannabinoids are the analytes of interest that must be extracted from the chocolate matrix which includes fats, proteins, sugars, and anything else used to make the chocolate. A known volume of solvent (typically Methanol or Acetonitrile, but depending on the product type/matrix other solvents may be needed) is added. The samples are then typically sonicated and/or heated, with some agitation, until the cannabinoids have been pulled from the starting material into the extraction solvent.
Again, any equipment used for this process will impart some error, so an analysis into the accuracy of the uncertainty caused by whichever dispensing system the lab has chosen is important. Is the 50mL of solvent added exactly 50 mL? Any deviation contributes to the compounding error.
At this stage, the extracts, if heated, are allowed to cool to room temperature before the crucial step of dilution. Dilution is utilized to bring the concentration of cannabinoids in the extract within quantitation range. The quantitation range is generally defined as the range between the lowest concentration calibrator and highest concentration calibrator, though method validation procedures determine this range. Dilution can be performed manually (using volumetric flasks and syringes for best technique, using autopipettes for low cost and reasonable throughput with reasonable accuracy, or using automated dilution systems).
Following the first extraction, further dilution may be required. Similarly to the calibration curve dilution discussion, this will always add some quantity of error.
The cannabis and hemp landscapes are constantly evolving and new product types are regularly created. For standard products (flower, concentrates, tinctures, and many edibles), extraction methods have been validated and many labs will use similar techniques to ensure complete extraction of cannabinoids. However, new products may require different extraction methods, or modifications to existing methods, in order to ensure accurate testing. Pulp containing beverages, for instance, can impact final product potency as the vegetal pulp matter can hold cannabinoids in a manner that means the product isn’t truly homogenous, and therefore adaptations to the method must be made in order to get an accurate cannabinoid profile.
A typical way around this problem is overdosing. If a novel product only shows 80% recovery (showing 4 mg/serving instead of 5), many companies will simply overdose the product; dosing to 6.25 mg/serving, which with 80% recovery, would give a COA showing 5 mg/serving. This is often the easier route for manufacturers that need to get product on the shelf, instead of delving into which part of the process is responsible for the potency loss, especially with the pressure to get product to market. This is definitely a regulatory blindspot, as the COA is shown as proof of the cannabinoid concentration, though the material may contain more than the COA indicates.
In a marketplace where nearly every actor is incentivized to have higher potencies, one cannot discuss lab variance without discussing potency inflation - intentionally adjusting one or more steps in the process to result in higher reported potency values. This can happen at nearly any step in the testing process, though some methods are more nefarious than others. Choosing a certified reference material with a lower concentration (as shown in Figure 6) will result in a higher potency value. Even something as simple as this (choosing to buy a CRM from one vendor over another) must be examined with the guiding principle of accuracy and objectivity.
More intentional methods are also easy and undetectable, though they would typically rely on falsifying some aspect of the analysis. Adding more solvent during CRM dilution/calibration standard generation, for example, would again result in higher potency results, though this would mean falsifying standard generation documentation. Adding more material, or less solvent during extraction would have the same effect and again require falsifying documentation.
The hemp and regulated cannabis space is at an important nexus. If we wish to be taken seriously as a category, we’ll need to earn and keep the consumer’s trust. Potency stability is the foundation of building that trust. We understand that there’s a lot to consider in the cannabis beverage supply chain, let alone just the labs. If you’re interested in learning more, contact a member of the Vertosa team!
² Figure 2 - Spectra of a variety of cannabinoids
³ Figure 3 - Mixed Calibration Standard and Sample Chromatogram
⁴ Figure 5- Cayman Scientific Analysis of 4 Different Single CRM Vendors