Right From the Grow: A Look at Sampling and Sample Preparation Methods of Solid Cannabis for Analysis

Scientific testing is perceived as a straightforward process that involves grabbing a pinch of a sample, running it through an instrument, and immediately out comes the exact answer. More often than not, laboratories are challenged with highly regulated and difficult sample schematics, sample preparation, extraction, and testing procedures that try to ensure accuracy and precision of testing. Accuracy in analytical testing starts at the very beginning with sampling and sample preparation prior to testing. If the initial sample collection and preparation are flawed then the final answers will be biased. The basis of accuracy of sampling and testing often rests on two inter-related and fundamental concepts: representative samples and homogeneity. Representative samples are selected to accurately reflect the larger group and should represent the characteristics of the group as a whole. Ideally representative samples are homogeneous or similar in nature, but when that is not possible the best attempts must be made to achieve samples that represent the majority of the characteristics of the larger grouping.

Agricultural samples can be some of the most difficult samples in the world to sample, prepare, and analyze because of their heterogeneous nature and complex matrices. Luckily for most of the agricultural testing world, the industry is equipped with detailed methods for operations, collection, and testing. Sampling for a crop farmer is a defined process of removing samples at designated intervals and testing those samples for the prescribed list of chemical and biological targets. Unfortunately, the cannabis grower has had limited guidance to refer to for managing operations, sampling, and testing. There is also difficulty in the fact that cannabis is a very complex plant. 

There have been more than 500 compounds identified in cannabis (many of which are unique to the Cannabaceae family). The distribution of these compounds is highly dependent on individual strains, the gender of the plant, and the location on the plant (1,2). The distribution and concentration of these compounds can also be effected by environmental conditions such as soil, water, and light (1–3). To further complicate the analysis of chemicals in cannabis is the fact that different amounts of compounds can occur in different locations within the plant. In some cases it has been reported that higher tetrahydrocannabinol (THC) concentrations are found in buds located high on the plant as opposed to buds located lower in the plant (2). Different growing conditions, seasons, environmental, and chemical exposure can also alter the chemical composition between growing cycles as well as the chemical distribution within an individual plant.

In addition to a lack of guidance and a complex heterogeneous nature, there is the added concern that the crop itself is a commodity of high economic value, which inherently forces the grower to limit the amount of samples submitted for testing while individual states have begun to mandate sampling minimums (4,5). Sampling of cannabis brings into question how much and what type of sample is enough to conduct representative sample testing, and what are the criteria for sample homogeneity. Many sample preparation and test methods depend on the foundation of representative samples and homogeneity to provide accurate results. If sampling schematics are not designed to ensure representation and homogeneity of the entire crop through to the final analytical sample, then the testing will be biased. 

The idea of a sample and sampling is commonplace to most of us. A sample is a small part of something which represents a larger whole or grouping. We take a sample of food and judge the taste of the dish based on one bite. We sample a few notes or bars of a song and decide if we like the music. Our minds often make an assumption of sameness or homogeneity based on a sample and the perception of the larger whole. We estimate the size of our sample based on our bias and perception. But what if the perception causes us to take too small of a sample to get all the different or heterogeneous elements? What happens when a sample is taken from a heterogeneous part and is not representative of the entire portion or population? Imagine taking a sample of some spicy chili and your first and only bite contains a whole hot pepper! Maybe you should have taken a bigger bite with more than just that pepper, or maybe you should have found a different bite to sample.

The concepts of big and little, same and different are introduced very early in all of our lives. We ask a child to pick out the big ball and they zero in on the largest ball in the group. We ask for the child to find which item in a set is different and they do. As we grow older, we start to understand that big and little, same and different are concepts that depend on focus, perspective, and often purpose. Those concepts change with our reasons for defining them and the potential bias that can be included in the choices made. It is important to look at sampling from the grow through to the laboratory to examine where the focus and perspective should be to ensure that cannabis sampling at every stage of the process represents the whole we are trying to characterize.

Into the Field

Figure 1: Examples of statistical sampling units.

The key concepts on the larger scale or the grow side are: population, sampling frame, and representative samples. A population is the entire possible group of objects that are being sampled or a subset of those objects. A sampling frame is a possible source material taken from the population where samples will be obtained (Figure 1). So if a single variety of cannabis in a crop of multiple varieties is to be harvested, then the population can be seen as the entire crop of all varieties and the sampling frame is all the harvested plants of a single variety or strain. The sampling frame is composed of primary samples which in turn are composed of sample units. Sample units are the smallest discrete portions that are taken to form the whole or part of a primary sample. For the cannabis industry, the population or sampling frame could be as large as an entire crop, just one variety within an operation, or as small as selected trimmed buds from specific plants, varieties, or just areas of growth, depending on the purpose for the sampling.

There are two basic types of sampling: probability sampling (random) and nonprobability sampling. Probability sampling is when each unit of a population or a whole has the same chance of being selected to make up a sample and the probability of being selected can be calculated. Nonprobability sampling is when samples are collected in a process where some samples are purposely selected and the selection processes do not give all the possible samples an equal chance of selection.  

Figure 2: Probability sampling methods: (a) Random sampling. (b) Systematic sampling where there is a method where every nth sample is chosen. (c) Stratified sampling in which samples are grouped and randomly sampled. (d) Cluster sampling is used when the population is very large and divided up by physical or geographical areas and all units within subset are sampled.

Probability sampling has four primary methods of selection (see Figure 2): simple random selection; systematic selection; stratified selection; or cluster selection. Simple random selection is, by its own title, a random process which sometimes can use a random number generator or table to obtain samples (Figure 2a). Systematic selection uses a collection method where every n^th member (that is, the sampling interval [κ] of a population or sampling frame is taken as a sample (Figure 2b). The sampling interval (κ) selected to use is dependent on the size of the population or sampling frame and the number of samples to be collected. To calculate the sampling interval (κ); the total population size (N) is divided by the targeted sample set size (n) or κ = N/n. For example, if a grower has 100 plants and wants to have a sample size of 20, then they would have to sample 100/20 = 5; or every 5^th plant.

The third type of probability sampling is stratified sampling which is when the population or sampling frame is divided into subsets or strata. This type of sampling can be used to differentiate between samples of different types such as different species or varieties within one population. Once the strata are established then another random selection process is used to select samples (Figure 2c). The final type of probability sampling is cluster sampling where physical areas or geography are designated into clusters that are then sampled. Cluster sampling is best used when there is an extremely large population, such as a national forest or a population of a state, which must be examined (Figure 2d). Focusing on reducing the sample frames down to smaller groups reduces the amount of time, energy, and money to represent the target population.

Figure 3: Nonprobability sampling methods: (a) convenience; (b) consecutive; (c) quota sampling where there are strata but the samples are not picked randomly; and (d) judgement sampling in which samples selected to be sampled to obtain a certain result or bias, for example, if plants in the northeast corner had the highest THC content historically, and the grower wanted to establish the highest THC of the most concentrated plants, then those northeast plants would be targeted for sampling.

In contrast to probability sampling, there is nonprobability sampling where a nonrandom selection process for the purpose of obtaining targeted data or results can intentionally or unintentionally create bias. The four applicable general types of nonprobability sampling methods that are most often used for obtaining analytical samples are seen in Figure 3. They include convenience sampling, consecutive sampling, quota sampling, and purposive or judgement sampling. Convenience sampling means that the sampler takes samples with easy access (Figure 3a). Convenience sampling is easy, cost effective, and fast but it can produce bias by under-representing the overall population. Similar to convenience sampling is consecutive sampling, where the samples are selected consecutively within or between units (Figure 3b), and in a typical crop, selection setting is essentially the same as convenience selection. Quota sampling is similar to stratified sampling by dividing into strata, but the selection is not random (Figure 3c). Finally, there is judgement or purposive sampling where the sample is chosen based on what the sampler or researcher thinks is needed for the study. This technique is used for research in a small group or field to create a specific population and is biased to the selected purpose. For example, if a grower wants to know the highest concentration of THC found in his crops, and the grower knows that a certain area of the crop, or certain parts of the plants, have traditionally the highest THC concentration because of variety, environment, water access, and other conditions, then sampling of only those areas have a purpose to create data on the population of the plants with the highest THC concentration (Figure 3d).

Each method of sampling we have examined so far has both advantages and disadvantages. There are places for each type of sampling method and the choice of sampling method often comes down to purpose and intent. In some cases, different methods of sampling may be combined at different points in a process to achieve either an overall representative sample, or a sample with a specific purpose and intent. Most commercial agricultural operations do not perform sample collection for testing directly from the field but wait until harvest to partition and sample their batches or lots. The methods for post-harvest sampling remain the same as pre-harvest sampling, just some of the terminology changes. Instead of defining population and sampling frame, there is a harvest of all or portions of a crop. That harvest yields bulk material to be processed either at one time or in separate manufacturing events.

The U.S. Food and Drug Administration (FDA) defines a batch as a specific quantity of a material that is intended to have uniform character and quality produced during a single manufacturing cycle (6). The California Department of Food and Agriculture’s Cannabis Cultivation Branch defines a cannabis batch as “a specifically identified quantity of dried flower or trim, leaves, and other cannabis plant matter that is uniform in strain, harvested in whole, or in part, at the same time, and, if applicable, cultivated using the same pesticides and other agricultural chemicals” (7). A lot is a very similar and somewhat interchangeable term which is defined by the FDA as a batch or specific portion of a batch that has uniform character and quality produced by a continuous process within a unit of time of quantity that assures having uniform character and quantity (6). However, in Oregon, the terms of batch and lot are reversed in which the batch can be a defined portion of a lot and there are designations between a harvest lot and a process lot. A harvest lot is “a specifically identified quantity of marijuana that is cultivated utilizing the same growing practices, harvested within a 72-hour period at the same location and cured under uniform conditions.” A batch is a quantity from a lot (4).

In the harvest and processing of cannabis material, the samples are no longer grouped by which plant is sampled or area is sampled, but by which parts of the harvested and processed materials are sampled and the method by which the samples and units are selected. The same methods of sampling still apply, for example, if systematic sampling is used-instead of sampling every 5^th plant, every fifth container of harvested buds has to be sampled, or every 5^th bud processed is taken, depending on the sampling plan design.

Figure 4: Sample divisions for analytical testing.

Once the lot is sent for testing, the material is subdivided many times according to the laboratory or state regulations to create representative samples for testing (Figure 4). Primary samples can be taken from the lot and composited together to create a bulk testing sample. The laboratory samples can either be processed at this point-or extracted, ground, and so forth, depending on final testing-and purpose or the material can be further subdivided, processed, and sampled for individual analytical samples, such as portions and aliquots. In each step where there is further subdivision of the material, an appropriate sampling method must be used to try and achieve the homogeneous representative sample of the lot.

Sample Material Processing Challenges for Cannabis

Laboratory or analytical samples must be processed into a form that allows for analytical, instrumental, and chemical testing. Most of the time, this process involves grinding the samples for homogeneity and extraction into a suitable matrix for analysis. At each subdivision of material, the samples get smaller and smaller until by the final step only a small amount of material is actually tested. This fact raises the questions of how much and what type of sample is enough to conduct representative sample testing, and what are the criteria for sample homogeneity. It can be difficult in the cannabis industry to determine if what is thought to be representative by the state regulations actually fits the bill scientifically for accurate results. In the January/February 2019 issue of Cannabis Science and Technology, Dr. Brian C. Smith warned of representative samples and testing that, while some states dictate minimum representative samples for analysis, that does not mean that the number is analytically sufficient for accurate results (8). If sampling schematics are not designed to ensure scientifically valid sample representation and homogeneity of the lot, then the end testing will be biased.

An important issue in sample preparation methods for cannabis is homogeneity. By its very nature, dried plant material is not particularly homogeneous. In sampling methods, where the sample being tested is a high-value commodity, sample size matters, and laboratories are often tasked with preparing smaller samples to meet all the testing demands. Small samples, however, increase potential bias and error unless it can be reasonably assured that the samples are homogeneous.

Table I: Effect of particle size on amount of material needed to ensure various uncertainty levels for representative sample (9)

The most common method for obtaining a homogeneous sample is grinding or comminution. Grinding samples has many benefits for sample preparation since it increases homogeneity, increases surface area, and decreases particle size, which improve extraction efficiency. Grinding samples also allows for a reduced sample size to increase accuracy and decrease uncertainty. In a study by Thiex and colleagues, it was shown that the smaller the particle size, the less sample was needed to achieve a lower amount of uncertainty in a sample (9). A particle size of 5 mm is about the size of a pencil eraser. If a particular laboratory needed to have results within 5% uncertainty, they would have to use 500 g of material for testing. But, if the laboratory reduced the particle size to less than 0.5 mm (the size of a fine point pen tip), the amount of sample needed to ensure 5% uncertainty would drop to less than 0.5 g (Table I) (9).

The grinding of cannabis plants and products produce a fair amount of challenges in regards to the physical state and efficient grinding. The plants are very fibrous and resist methods that use cutting or they may clog filters and screens. The bud material contains high amounts of waxes and oils which would stick to grinding media. Cannabis edibles could also be sticky and have difficulty in most grinding apparatus. The moisture content can vary greatly with the type of product as would its melting or softening temperature.

The glass state of a material or the glass-transition temperature (T_g) is the range of temperatures over which amorphous materials or semi-crystalline materials transition from a viscous or rubbery state to a hard and brittle glassy state. The process of a viscous liquid or semi-solid transitioning to a glass state through super cooling is often referred to as vitrification. Moisture level in products affects a material’s glass-transition temperature. Tg decreases with increased moisture levels. A study of food products including cassia showed that water in the food had a plasticizing effect, which resulted in needing lower temperatures to achieve the glass-transition temperature in food items with higher water content (10,11). In cannabis products with high moisture content, it becomes especially important to negate the effect of the moisture to ensure efficient grinding.

The second area where cryogenic applications to sample preparation can aid in laboratory analysis is in the stability of materials and the retention of important labile or volatile compounds or elemental species. As has been discussed previously, cannabis products have a multitude of volatile compounds that must be retained during sample processing. The approach to the sample preparation and grinding of cannabis should mimic another similarly economically valuable group of products-spices. Spices are full of the same highly aromatic compounds as cannabis (terpenes, volatile oils) which contribute to taste, aroma, and medicinal attributes. Some spices and cannabis can be degraded by high temperatures and oxidation. In ambient temperature spice grinding processes, heat and energy are generated, which can raise the temperature of spices to almost 100 °C and cause loss of critical aromatic components (12).

Studies of ground spices showed that spices ground under cryogenic grinding conditions contained almost 40% more volatile compounds and essential oils than ambient grinding (13). The low temperatures prohibited the breakdown of volatile compounds. In one study, it was found that grinding black pepper under cryogenic conditions showed better retention of monoterpenes (myrcene, limonene, and pinene) than grinding at ambient temperature. These monoterpenes are the same primary monoterpenes in many cannabis varieties (14,15). Cryogenic grinding reduces compound loss and the vaporization of the liquid nitrogen creates an inert environment to reduce oxidation (6). In addition, the extreme low temperatures generated by solutions, such as liquid nitrogen, solidified the fats and oils, allowing for a more finely ground sample of consistent particle size with increased surface area allowing for better extraction and increased homogeneity during further sample processing.

Another group of compounds that can be damaged by heat and oxidation are pesticides. While most growers would be happy that pesticides were degraded during sample preparation and analysis, the practice of good science cannot allow critical areas of analysis to be damaged by sample processing. Many pesticides that are commonly used and monitored for cannabis analysis are easily degraded by high temperatures and oxidation. In cases where potentially important compounds (that is, terpenes, THC, volatile oils, or pesticides) could be lost to processing, it becomes necessary to be able to prevent the loss and calculate for the loss by using standards.

Conclusion

Everything about the new cannabis industry is challenging from the actual structure and chemistry of the plant itself down to how samples are chosen and transported for testing. Despite all the complexity, the approach to sample collection, processing, and testing still must remain along the same lines as other similarly complex agricultural products (such as hops and spices). The rational for sampling, processing, grinding, extraction, and testing of these similar products can be used as a roadmap for the cannabis sampling and processing methods. The simplistic early sampling approaches of just grabbing a sample for the laboratory must be discarded in favor of good science. Good science can be difficult to achieve in an industry faced with uncertainties and lack of validated methodologies for a complex and valuable commodity. Sometimes the pursuit of a good scientific result (or adequate samples for representative sampling, sample volumes for homogeneity, and quality certified reference materials and standards) is hindered by physical, legal, and economic stumbling blocks which seem to fight against good science. Accurate sample analysis starts with the two fundamental concepts of representative samples and homogeneity. In the end, the development of sampling plans-sample preparation methods that understand the importance of these concepts-will ultimately be the most important step in achieving good analyses.

References

1. B. Zana, I. Papoutsis, S. Athanaselis, P. Nikolaou, E. Pazari, C. Spiliopoulou, and G. Vishal, Forensic Sci. Int. 222(1–3), 40–46 (2012). https://doi.org/10.1016/j.forsciint.2012.04.032.
2. M. Sexton and J. Ziskind, “Sampling Cannabis for Analytical Purposes.” Steep Hill Laboratory, November 15, 2013. https://lcb.wa.gov/publications/Marijuana/BOTEC%20reports/1e-Sampling-Lots-Final.pdf.
3. D.J. Potter and P. Duncombe, J. Forensic Sci. 57(3), 618–22. https://doi.org/10.1111/j.1556-4029.2011.02024.x (May 2012).
4. Oregon Liquor Control Commission, “Sampling and Testing Metrc Guide,” October 30, 2018. https://www.oregon.gov/olcc/marijuana/Documents/CTS/SamplingandTestingGuide.pdf.
5. Bureau of Cannabis Control, “Text of Regulations California Code of Regulations Title 16 Division 42. Bureau of Cannabis Control Order of Adoption,” n.d. https://bcc.ca.gov/law_reg/cannabis_order_of_adoption.pdf.
6. “CFR - Code of Federal Regulations Title 21,” Accessed March 1, 2019. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfCF/CFRSearch.cfm?fr=210.3.
7. Bureau of Cannabis Control, “California Code of Regulations Title 3. Food and Agriculture Division 8. Cannabis Cultivation Chapter 1. Cannabis Cultivation Program,”n.d. https://static.cdfa.ca.gov/MCCP/document/060418%20CalCannabis%20Text%20of%20Proposed%20Emergency%20Regulations%20Readopt.pdf.
8. B.C. Smith, Cannabis Science and Technology 2(1), 14–19 (2019).
9. N. Thiex, L. Novotny, C. Ramsey, G. Latimer, L. Torma, and R. Beine, “Guidelines for Preparing Laboratory Samples,” (Association of American Feed Association Officials, 2008).
10. B.M. Ghodki and T.K. Goswami, Powder Technology 299, Supplement C,168–77 (2016). https://doi.org/10.1016/j.powtec.2016.05.042.
11. B.M. Ghodki, A.D. Srikanth Tangirala, and T.K. Goswami, Cogent Food & Agriculture 2(1), 1227110 (2016). https://doi.org/10.1080/23311932.2016.1227110.
12. K.K. Singh and T.K. Goswami, J. Food Eng. 39, 359–68 (1999). https://doi.org/10.1016/S0260-8774(98)00172-1.
S.
13. Shailendra, Y. Sharma, S.S. Rathore, K. Singh, P. Barnwal, R. Saxena, P. Upadhyaya, and M M. Anwer, J. Food Sci. Technol. 52, https://doi.org/10.1007/s13197-013-1004-0 (2013).
14. V.V. Sreenarayanan and S.M. Mathew, J. Spices Aromatic Crops 16(2), 82–87 (2007).
15. C.T. Murthy and S. Bhattacharya, J. Food Eng. 85(1), 18–28 (2008). https://doi.org/10.1016/j.jfoodeng.2007.06.020. 

About the Columnist

Patricia Atkins is a Senior Applications Scientist with SPEX CertiPrep and a member of both the AOAC and ASTM committees for cannabis.

How to Cite This Article

P. Atkins, Cannabis Science and Technology 2(2), 26-34 (2019).