OR WAIT null SECS
Brian C. Smith, PhD, is Founder, CEO, and Chief Technical Officer of Big Sur Scientific. He is the inventor of the BSS series of patented mid-infrared based cannabis analyzers. Dr. Smith has done pioneering research and published numerous peer-reviewed papers on the application of mid-infrared spectroscopy to cannabis analysis, and sits on the editorial board of Cannabis Science and Technology. He has worked as a laboratory director for a cannabis extractor, as an analytical chemist for Waters Associates and PerkinElmer, and as an analytical instrument salesperson. He has more than 30 years of experience in chemical analysis and has written three books on the subject. Dr. Smith earned his PhD on physical chemistry from Dartmouth College. Direct correspondence to: email@example.com
Inter-laboratory variation has been studied extensively. Here, ideas on how to improve the situation are presented, particularly standardizing sample preparation methods.
The problem of different cannabis laboratories obtaining different results on the same samples, inter-laboratory variation, has been studied extensively. One of the causes of the problem is different laboratories prepare their samples differently. Another issue is how varying amounts of moisture in plant material can impact potency measurements. Ideas on how to improve the situation are given.
Inter-laboratory variation is the problem in cannabis and hemp analysis where different laboratories obtain significantly different results on the same samples. Scientifically the problem has been well documented (1–6). Anecdotally I have found that many people in this industry have experienced the problem. For example, in one study (6) the tetrahydrocannabinol (THC) readings on a homogeneous marijuana distillate sample sent to five different California state licensed, ISO certified laboratories, came back with a range from 77 wt.% to 94 wt.%. These results are alarming because a uniform sample containing a high concentration of the target analyte should be one of the easiest samples to analyze correctly. In the same study a distillate sample known to be free of pesticides was spiked with known amounts of six pesticides, some well above the state of California legal limits (7). Again five laboratories were involved, and none of them detected the presence of all six pesticides, and two of them said the sample contained no pesticides at all!
The inter-laboratory variation problem isn’t found just on the THC side of the business. Significant variations in cannabinoid readings are found when the same hemp sample is sent to multiple laboratories as well. Some supporting data are shown in Table I.
These data were generated by sending seven of the same hemp samples to two state licensed, ISO certified laboratories in different parts of the country. A comparison of the cannabidiolic acid (CBDA) results show huge differences in the weight percent values for the samples, the standard deviation being greater than 4.5 wt.%. There is also a small amount of bias between the two data sets, indicating that most of the error here is random.
These results are troubling since hemp farmers are paid on the total cannabidiol (CBD) content in their crop, most of which is in the form of CBDA. If a test at 10% CBDA is used to set a purchase price, and the crop really contained 15%, then the farmer is being ripped off for hundreds of thousands of dollars. Similarly, if an extractor purchases hemp biomass that is supposedly 15% CBDA, but it’s really 10%, they are overpaying by hundreds of thousands of dollars for CBD that is not there. The inter-laboratory variation problem threatens the simple act of buying and selling product in the cannabis and hemp industries.
Table II contains an even more alarming example of inter-laboratory variation. Since the passage of the 2018 U.S. Farm Bill (8), and the promulgation by the U.S. Department of Agriculture (USDA) in October 2019 of an interim final rule (9), hemp is legally defined as having not more than 0.3 wt.% total THC. Table II shows the results of sending the same six hemp samples to two different laboratories for total THC measurements.
There is almost zero bias in the data set, indicating almost all the error in these measurements is random (10). The standard deviation in the data is almost 0.2 wt.% total THC. This means one laboratory could test a hemp sample and find 0.2 wt.% total THC and declare a crop legal, whereas a different laboratory could find 0.4 wt.% total THC in the same sample and declare the crop illegal. People’s livelihoods and hundreds of thousands of dollars are on the line when hemp is tested for total THC. With data like this, how can we honestly assure hemp farmers that their test results are fair? Clearly, in terms of potency and pesticide testing there is an inter-laboratory variation problem. Other analyses performed by cannabis laboratories, such as terpenes, heavy metals, biocontaminants, and mycotoxins may be just as variable.
The inter-laboratory variation problem threatens the very existence of our industry. If there is no consistency across laboratories, how can regulators know which materials are legal and which are not? If there is no consistency between laboratories, how can cannabis businesses rationally run their businesses in the face of unreliable data? This is a problem that needs to be solved sooner rather than later. In a previous column (2), I gave a laundry list of potential solutions to the problem. One of them was standardizing sample preparation methods. In this column I discuss this idea in detail.
Chromatographic analyses are used to measure cannabinoids (11), terpenes (11), and pesticides in cannabis samples. For these analyses to work properly, the sample must be homogenized and extracted. This involves significant manual sample preparation including grinding, mixing, extracting, and filtering.
I have been involved in cannabis analysis since marijuana first become legal in Colorado in 2013. In that time, I have visited dozens of cannabis laboratories around the U.S. and have observed that no two laboratories prepare their samples for chromatographic analysis the same way.
I have seen laboratories use mortars and pestles, handheld grinders, coffee grinders, spice grinders, herb grinders, and cryo-grinders to prepare samples for analysis. These different methods will produce samples with differing surface area, meaning different amounts of cannabinoids will be extracted from the same sample. Here is a list of the variables that are not controlled and why they matter.
This is not relevant for liquid samples, but is very important for plant material such as buds, trim, and biomass. Work by myself and others have clearly shown that cannabis is an inhomogenous, naturally variable material (11). This means plant material must be homogenized by grinding prior to analysis. Grinding will produce a sample with a specific particle size distribution, particle shape distribution, and surface area. The latter matters because when performing a solid-liquid extraction, the surface area of the sample effects the rate at which molecules are extracted. Typically, the greater the surface area the faster analytes will be extracted.
2. Moisture Content
Given the nature of cannabis based materials, analyte concentrations are reported as weight percents rather than in moles/liter. In fact, Federal law requires total THC in hemp be measured as a weight percent (8,9). When plant material is harvested it contains significant amounts of moisture, and must be quickly dried to prevent the growth of mold. Since we are using weight percent measurements, and the denominator of these calculations is the weight of the original sample, variations in moisture content can affect the final results. Properly dried plant material contains about 10% moisture, however changes in temperature and humidity can affect the amount of adsorbed moisture on plant material, alter the sample’s weight, and then cause variation in analyte weight percent values. In a perfect world, third party laboratories would measure the moisture content of every plant sample before analysis, and take this into account when calculating weight percents. Some laboratories do not routinely measure the moisture content of incoming plant material, trust the sample submitter to have dried the sample properly, and assume ambient conditions will not affect moisture content. One cannot assume any of this. One way to dry biomass is to perform a loss on drying experiment. The weight of the sample is determined, the sample is thrown in an oven, heated for some time at some temperature, and then the weight is determined again. The assumption is made that all of the loss is from moisture evaporating. Let me be crystal clear: This measurement is useless for marijuana and hemp. We all know that in addition to moisture cannabis contains volatiles such as terpenes. Part of the loss on drying in addition to moisture is loss of terpenes. Also, at the temperatures often used for loss on drying the acid cannabinoids can decarboxylate resulting in additional weight loss. Work by myself and others has shown that only about half of the weight loss upon drying is moisture, the rest is most likely terpenes and other volatiles (12). Thus, the loss of drying values is not a measure of moisture content, but a measure of the loss of total volatiles upon heating. These values should not be used to correct analyte weight percent measurements. A solution to this problem is to measure the moisture content of samples prior to analysis using near infrared absorbance. Moisture meters based on this technology exist that are fast, accurate, and affordable (Google the term “near infrared moisture analyzers” to find a number of manufacturers). These systems measure moisture, not loss on drying. In my opinion, all laboratories should have one of these systems, measure the moisture content of submitted plant material, and then take this into account when calculating analyte weight percent values.
3. Extraction Solvent
Standardizing the solvent used to extract samples is a no-brainer. Chemistry 101 tells us that the same solutes dissolve in different solvents at different rates and in different amounts. I have seen methanol, ethanol, isopropanol, and chlorinated solvents used to extract cannabis samples. These extracts will vary in composition when applied to the same sample because of the variation in properties of these solvents. The industry needs to standardize this.
4. Varying Amounts
I have seen different laboratories use anywhere from 1–5 g of sample in an extraction. I have also seen variable amounts of extraction solvent used, from a few to many milliliters. Of course this affects the composition of the analytes extracted from a material. The more sample used, the more will be extracted. The more solvent perhaps the more efficient the extraction, but the analyte concentrations will vary with the amount of solvent used.
5. Mixing and Vortexing
Once the sample and extraction solvent are in intimate contact in a sample vial, they must be stirred up and mixed together via some mechanism. I have seen people shake vials, use vortexers, shakers, mixers, and other contraptions. The variables here are the size and shape of the container, the amount of sample and solvent, and the time and manner of mixing. The amount of time spent and the manner in which mixing occurs will affect the efficiency of the analyte extraction.
When extracting plant material, it must be filtered before proceeding with analyzing the extraction obtained from the sample. Depending upon how the solution is filtered analytes may be left behind on the filtration medium. This can be confirmed by filtering an extract, and then passing pure solvent through the filtration medium and analyzing it for the analytes. When filtering the extract is often open to the environment. Different amounts of solvent may evaporate from different samples during filtration, further adding concentration variation to the results.
For chromatographic analyses some samples may be too concentrated to be analyzed and need to be diluted before analysis. This is particularly true of samples high in cannabinoids, such as extracts and distillates. At times dilution factors of 100 or more may be needed. This means that volume measurement error of less than a drop can have a huge impact on results. Laboratories vary in how they dilute and how much they dilute contributing to the inter-laboratory variation problem.
8. Variation in Extraction Efficiency and Not Extracting to Exhaustion
For any extraction technique to be accurate it should pull out all of the analytes in a sample not just a fraction. Even if your extraction technique gets reproducible numbers, it will not compare to other laboratories who may pull out more or less of an analyte from the same sample because of variations in extraction efficiency. This variation is caused by the variables discussed above. The solution to this problem is for all laboratories to extract to exhaustion to make sure all the analyte is extracted from all samples. This means performing studies by extracting the same samples over and over again and measuring analyte concentrations until they reach zero. Then, a method needs to be developed for all samples to insure all analytes are extracted all the time. If all laboratories extracted to exhaustion they would each pull the same amount of analytes from the same samples, and much of the inter-laboratory variation problem would go away.
9. Human Error
As we have seen, the extraction methods for preparing samples for chromatography involve a significant amount of manual sample preparation. Being who we are, human beings are bound to make mistakes when preparing samples, which contributes to the inter-laboratory variation problem. The solution to this problem is better training, or use of analytical methods that require less manual sample preparation (13).
The definition of accuracy is how far away one is from the true value (10). In cannabis analysis there are currently no standard reference materials such as a standard marijuana bud or hemp distillate. Recently I learned NIST is working on standard reference materials for hemp analysis but that might not be available for years (13). On the marijuana side there will not be standard reference materials until Federal legalization occurs. Hence, there are no agreed upon true values in this industry and accuracy is illusory. The best we can do is run the same sample over and over again and calculate a precision value.
What are we to do then? We have to figuratively put a “stake in the ground” and decide upon a standard method and instrument, and calibrate all instruments with respect to this “golden” one. For example, for potency analysis high performance liquid chromatography (HPLC) should be preferred (12), and we need to standardize on a sample preparation method, calibration method, and instrument. Once we have a “golden” HPLC and method, other chromatograph and spectrometer instruments could be calibrated to the golden one, and then used for analysis going forward (13). The advantages of spectroscopy are very little sample preparation, removing the human error associated with extensive manual sample preparation. Also, spectroscopic methods of cannabis analysis are fast and inexpensive, allowing for more representative sampling to be performed.
As I have pointed out in the past, the method of Giese, Lewis, Giese, and Smith (12) should be the method the industry adopts in terms of sample preparation, HPLC to determine cannabinoids, and gas chromatography with a flame ionization detector (GC-FID) to determine terpenes. I have witnessed this method in action at multiple third party laboratories. As part of my own work, I have correlated HPLC results to mid-infrared (IR) results on the same sample set at several laboratories using different HPLC potency methods. I always obtain the best correlation to laboratories using the method of Giese and colleagues. Assuming mid-IR is an unbiased, independent, orthogonal method, this indicates the Giese and colleagues method is the best.
The phenomenon of inter-laboratory variation is a real problem in cannabis analysis. Part of the problem is variation in sample preparation across laboratories. Variables such as grinding, moisture content, extraction solvent, varying amounts of sample and solvent, and human error contribute to the problem. The industry needs to decide on a standard sample preparation protocol and analysis method, or else the inter-laboratory variation problem will continue to plague us.
Brian C. Smith, PhD, is Founder, CEO, and Chief Technical Officer of Big Sur Scientific in Capitola, California. Dr. Smith has more than 40 years of experience as an industrial analytical chemist having worked for such companies as Xeros, IBM, Waters Associates, and Princeton Instruments. For 20 years he ran Spectros Associates, an analytical chemistry training and consulting firm where he improved their chemical analyses. Dr. Smith has written three books on infrared spectroscopy, and earned a PhD in physical chemistry from Dartmouth College.
B.C. Smith, Cannabis Science and Technology3(2), 10–15 (2020).