steephill-express-shed-some-light-01.jpg

Diffuse Reflectance Near-Infrared (DRNIR) Spectroscopy for Potency Analysis of Cannabis

One of the most common requirements mandated by regulators of Cannabis products is Quality Assurance (QA) testing in the form of potency values for the major medicinal components, namely the cannabinoids THCA, THC, CBDA, and CBD.  Cannabis plants, depending on their genetics, produce mostly THCA, CBDA, or a combination of the two, ranging in amounts as small as ~0%, but sometimes exceeding 20% by weight of the flowers. Age or heat can convert some of the THCA and CBDA into THC and CBD, respectively. Of these four main components, only THC is psychoactive.  Accurate knowledge of these four analyte values allows the patient consumer to select medicine of appropriate potency with the preferred profile (THC vs CBD) to treat their individual symptoms with a reliable prediction of the desired medicinal and psychoactive effects (THC stimulates appetite, while CBD does not; THC causes psychoactivity, while CBD does not, etc.).  Accurate potency values also help producers and distributors better determine the value of any sample presented, since price points are driven predominantly by the amount of active ingredients present.  Additionally, if potency results can be obtained in a timely manner, it is possible to use them to guide process control activities, such as optimal harvest time.  Steep Hill has announced a new testing service, Steep Hill Express™. Steep Hill Express™ uses state-of-the-art rapid testing methods to produce Quality Assurance test results in a minimum of time using spectroscopic measurements combined with cloud-based calibration modeling technology. Once a flower sample is ground, packed into the sample cup, and loaded on the spectrometer (with internet access), potency results are obtained in approximately one minute.

NIR and the Quality of Goods

Regulatory guidelines exist for testing the quality of all food products, oils, and medicines.  Quality and labeling of goods are regulated to assure that consumers purchasing those goods are getting what is on the label, or that the goods aren’t spoiled or contaminated.  This quality assurance can be approached using conventional wet methods or by more modern rapid testing or “process analytical” testing methods.  While conventional wet methods, including chromatography, are very versatile at testing major, minor and trace constituents, process methods can offer rapid, solvent-less measurement with good accuracy on major constituents.  

Near infrared (NIR) spectroscopy has, over the last two decades, become a preferred method in cases where materials are heterogeneous and require multiple samples, or in cases where blending requires a more rapid assessment to control blend content on the fly.  Near infrared can be used on solids in a diffuse reflectance configuration or for liquids in a transmission configuration.  Industries that use near infrared extensively for quality monitoring include finished product testing of oil and gas, grains, cheese and meat (in the EU), and process control of pharmaceuticals.  The most prominent advantages of near infrared over chromatography include non-destructive and solvent-less testing, rapid turnaround time, much lower capital and maintenance costs, and a notably lower level of training required by the operator.  For the testing of bulk components, NIR can be very useful on heterogeneous materials where several different samples can be used to determine bulk contents as in the case of protein, fat, and water in grains.  The wide use of NIR in the use of measuring major constituents in agricultural applications suggests an application to Cannabis potency analysis.

Why NIR and Cannabis Potency?

Near infrared spectroscopy and Cannabis seem to be a natural fit since the method has been very successful in assessing bulk contents of other agricultural products where it is currently used extensively.  Since Cannabis flowers are another agricultural product, NIR is a natural choice for application development of a Cannabis potency analyzer.  Cannabis can be up to about 30% Cannabinoids (mainly THC and CBD, in their acid or neutral forms) and 5-25% water.  Cannabinoid content can easily vary by over 5% (e.g. from 13-18%) in the flowers of a single plant - more, if grown indoor.  This means that wet analytical methods, which are generally more accurate on a single sample, lose that accuracy advantage on a cost per analysis basis, since more frequent testing is advised for batches likely to have such variability.  Multiple NIR measurements are much more economical to estimate or correct for heterogeneity than are wet methods.  Additionally, the NIR analysis uses no polluting solvents to dissolve samples and has a speed advantage of at least 5-10 times over wet methods at similar accuracy on heterogeneous materials. 

Diffuse Reflectance Near Infrared Spectroscopy

Diffuse reflectance refers to the way light is diffusely scattered from a sample made of many small particulates, such as ground flowers.  In Diffuse Reflectance Near Infrared Spectroscopy (DRNIR), some of the light is reflected off of the surface and this light has little or no information about the sample.  Some of the light, however, passes into the sample and is absorbed and scattered multiple times by the sample particulates before exiting the sample in the general direction whence it came.   Diffuse reflectance spectroscopy is only useful when the sample scatters light sufficiently to return it in the direction of the sample/air interface so it can be detected.

Figure 1: Light from the spectrometer source (orange) is either reflected (stays orange) or transmitted into the sample and diffusely reflected (blue) and then emitted (red) where it can be detected and measured.  It is important to design spectrometer optics to limit the detection of orange light while detecting as much red light (which has probed the sample) as possible.

Figure 1: Light from the spectrometer source (orange) is either reflected (stays orange) or transmitted into the sample and diffusely reflected (blue) and then emitted (red) where it can be detected and measured.  It is important to design spectrometer optics to limit the detection of orange light while detecting as much red light (which has probed the sample) as possible.

Cannabis measurement with DRNIR is performed by grinding the flower and packing the sample into a sample cup, then placing the sample cup on the instrument measurement window over the light source.  A white reference is measured just prior to placement of the sample, and these two scans are compared to mathematically assess the light absorbed by the sample across the wavelengths collected.  Molecular bond types can be differentiated by their spectral absorbance/wavelength signatures, which makes molecular selectivity and quantitation possible in many cases.

Figure 2: Orange arrows represent light launched by the spectrometer to probe the sample.  Blue arrows represent randomly scattered and absorbed light after it enters the sample, and red arrows represent the light that is emitted from the surface and can be collected and analyzed.

Figure 2: Orange arrows represent light launched by the spectrometer to probe the sample.  Blue arrows represent randomly scattered and absorbed light after it enters the sample, and red arrows represent the light that is emitted from the surface and can be collected and analyzed.

Making of a NIR-based Cannabis Analyzer

The use of a NIR based analyzer system is notably simpler compared with wet methods like chromatography.  Developing a NIR calibration does require knowledge and experience in the multivariate modeling with these analyzers.  Model development is achieved by collecting NIR spectral measurements on a wide array of samples and wet method analysis for comparison on those same samples.  Once an adequate number of samples is collected, a data processing of these “training” samples allows the creation of a mathematical model that relates the NIR spectral signatures to the concentrations of these calibration samples determined by wet methods (HPLC).  For example, spectra are collected on Cannabis samples of various Cannabis strains and Cannabinoid concentrations.  Next a model is computed -using the spectra and concentration information- that uses the spectrum to predict constituent concentrations in “test set only” samples.  The NIR model is validated using the regression solution and new samples not used in calibration to predict potency.  NIR predicted potency is examined vs the wet method to statistically validate the model.  Future samples can be measured on the NIR and cannabinoids for those new samples can be predicted using the spectrum and the NIR model.   While the synopsis above sounds relatively simple, success depends on gathering enough of the appropriate data and understanding the use of the mathematical methods used in model building which is quite challenging.  The discipline of “Chemometrics”, also called “Process Analytical Technology”, is focused on the development of these models.  A good understanding of the mathematical method, physics of the spectrometer measurement, chemistry of the samples, and preferably experience in practice is desired.  But once the model is built, a person with a high school education can be trained to operate the analyzer.

The Steep Hill Solution: Steep Hill Express™

The Cannabis flower testing capabilities of Steep Hill Express™ currently use Steep Hill’s proprietary QuantaCann2™ NIR Cannabis analyzer. This device uses chemometric modeling of NIR spectroscopy and calibration data using wet chemistry (HPLC) to estimate the content of cured Cannabis flowers in percent by weight content of four cannabinoids: THCA, D9-THC, CBDA, and CBD. These are, by far, the most prevalent cannabinoids found in commercial Cannabis. (The next highest in content would be CBGA and CBG.)  Once an instrument is “calibrated” using the training data, many additional samples (samples not used in the training data calibration) are subsequently run by both NIR using the QuantaCann2™, and also by HPLC. The four plots that follow show a comparison of the NIR predicted value on the Y-axis, plotted against the HPLC-measured data for the same sample on the X-axis. Ideally, a straight line with slope = 1 is obtained. Statistical scatter is, as expected for all measured data, observed and is evident on the plots by deviations between the actual points and the line of slope 1.  The absolute percent errors are all within about 1%. That means, if the Steep Hill Express™ predicted value obtained for THCA is 18%, then HPLC analysis of the same sample would likely yield a value in the range 17-19%. In fact, this approaches the precision of HPLC for repeated measurements of the same homogeneous plant sample.

High Times 2016 NorCal Medical Cannabis Cup™

Steep Hill Express™ was used to perform the flower testing at the High Times 2016 NorCal Medical Cannabis Cup™ in San Francisco in June, 2016.  Flower samples were also tested using Steep Hill’s HPLC wet method for comparison.  The NIR analyzer was previously calibrated against HPLC results with 250 widely varying samples of Cannabis collected for this purpose.  The High Times Cup samples under test included CBD class, Sativa class, Indica class, and Hybrid class flower samples.  HPLC analysis helped determine that 13 samples had in excess of 25% THCA, a very potent group of samples.  These samples exceed the highest unadulterated flower sample in the calibration set.  Also, too few samples showed the presence of CBDA or CBD for meaningful graphical representations, thus, we show the CBDA and CBD predictions vs HPLC data for another set of test samples not included n the calibration data set (not from the High Times Cup).

Figure 3: THCA comparison: NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match.  Data shown are for High Times Cup samples not included in the calibration.  Note the red datum, representing a spectrum that mathematically lies outside the calibration set.  Samples lying outside the calibration set are rejected as outliers, since quantitation for these can be unreliable.  Outliers can be detected and removed using objective statistical data analysis procedures.

Figure 3: THCA comparison: NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match.  Data shown are for High Times Cup samples not included in the calibration.  Note the red datum, representing a spectrum that mathematically lies outside the calibration set.  Samples lying outside the calibration set are rejected as outliers, since quantitation for these can be unreliable.  Outliers can be detected and removed using objective statistical data analysis procedures.

Figure 4: Δ9-THC comparison High Times Cup data. NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match and dash dot lines are +/- 0.5%.  Data shown are for samples not included in the calibration. Cannabis plants produce mostly acid cannabinoids, so the range of THC neutral in flowers is much smaller (0-5%) than for THCA.

Figure 4: Δ9-THC comparison High Times Cup data. NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match and dash dot lines are +/- 0.5%.  Data shown are for samples not included in the calibration. Cannabis plants produce mostly acid cannabinoids, so the range of THC neutral in flowers is much smaller (0-5%) than for THCA.

Figure 5: CBDA comparison from non-High Times Cup data (insufficient sample count). NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match.  Data shown are for samples not included in the calibration. Most flowers that produce significant THCA do not produce much CBDA. Flowers that produce large amounts of CBDA are rarer, so there are fewer samples with high CBDA, but lots with very low CBDA.

Figure 5: CBDA comparison from non-High Times Cup data (insufficient sample count). NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match.  Data shown are for samples not included in the calibration. Most flowers that produce significant THCA do not produce much CBDA. Flowers that produce large amounts of CBDA are rarer, so there are fewer samples with high CBDA, but lots with very low CBDA.

Figure 6: CBD comparison from non-High Times Cup data (insufficient sample count). NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match.  Data shown are for samples not included in the calibration. Cannabis plants produce mostly acid cannabinoids, so the range of CBD neutral in flowers is much smaller than for CBDA. Many samples have almost none and this was true of the High Times Cup entrants: no CBD was detected above 0.5% by QuantaCann.

Figure 6: CBD comparison from non-High Times Cup data (insufficient sample count). NIR prediction (Y) vs HPLC measurement (X). The solid line indicates a perfect match.  Data shown are for samples not included in the calibration. Cannabis plants produce mostly acid cannabinoids, so the range of CBD neutral in flowers is much smaller than for CBDA. Many samples have almost none and this was true of the High Times Cup entrants: no CBD was detected above 0.5% by QuantaCann.

Scientific Personnel

Steep Hill’s QuantaCann2™ development has been conducted by two experienced PhD Analytical chemists.  Dr. Donald Land is Professor of Chemistry at the University of California, Davis where he builds and uses spectrometers of many types and was also a co-founder of Halent Laboratories, a Cannabis testing firm that merged with Steep Hill in 2013.  Dr. Land pioneered the development of wet lab Cannabis testing methods for over 6 years, first at Halent and more recently at Steep Hill.  Dr. Land is also a professor of forensic chemistry with special expertise in infrared spectroscopy and mass spectrometry.  Development of a Cannabis analyzer requires high quality reference data in modeling and Dr. Land’s experience in Cannabis lab analytics enables the collection of the highest quality reference data.

Dr. Thomas Blank is an analytical chemist with 20 years’ spectroscopic applications development experience including NIR/chemometrics modeling in the oil industry where he was a post-doctoral fellow at Exxon Research and Engineering.  Dr. Blank has deployed and tested more than 50 refinery models for gasoline and diesel blending using mid/NIR analyzers while at Exxon.  Dr. Blank also has 12 years in medical device testing of NIR and Raman spectroscopic instruments and air quality testing analyzers based on Near and Mid IR spectroscopy.

References

NIR in the Oil and gas industry

http://www.asdi.com/getmedia/111f2a9e-45e5-4bc4-849a-cbb281df6a14/Near-Infrared-Spectroscopy-finding-use-in-energy-industry-Oil-Gas-Product-News1.pdf.aspx

https://www.osapublishing.org/as/abstract.cfm?uri=as-55-2-197

http://www.ogj.com/articles/print/volume-91/issue-18/in-this-issue/refining/near-infrared-offers-benefits-and-challenges-in-gasoline-analysis.html

http://www.ogj.com/articles/print/volume-92/issue-26/in-this-issue/general-interest/experience-leads-to-accurate-design-of-nir-gasoline-analysis-systems.html

NIR observed degradation of grains

http://naldc.nal.usda.gov/download/56644/PDF

http://naldc.nal.usda.gov/download/12144/PDF

http://pubag.nal.usda.gov/pubag/downloadPDF.xhtml?id=23127&content=PDF

Use of NIR in Agriculture, book

https://dl.sciencesocieties.org/publications/books/articles/agronomymonogra/nearinfraredspe/frontmatter

NIR in nutrition content of meals

http://pubag.nal.usda.gov/pubag/downloadPDF.xhtml?id=17538&content=PDF

NIR and applications, general overview

http://www.jsac.or.jp/analsci/data/pdf/28/06/a28_0545.pdf

http://www.spectroscopynow.com/details/education/sepspec1881education/an-introduction-to-near-infrared-spectroscopy.html?tzcheck=1,1,1,1,1,1&&tzcheck=1