Laboratory testing, potency and the COALaboratory testing, potency and the COA

Purpose and scope

For each batch that you supply to a buyer, a testing laboratory gives a certificate of analysis (COA). In a system for medicinal cannabis, the laboratory also gives a COA for each batch that you release. The COA is a document of one page. It shows the materials that the flower contains (cannabinoids and terpenes) and the materials that the flower must not contain (mold, heavy metals, pesticides, and mycotoxins).

Most buyers read only one number on the COA. This number is total THC, and it changes the price. Thus some persons try to get a higher value for this number more frequently than for other numbers. Tests with peer review show a systematic error. The values on labels in the retail market are too high^[1][2].

One fact is important for all sections of this paper. A COA is not a property of your crop. A COA is the measurement of one small sample. A person collected the sample with one method. The laboratory prepared it with one method and measured it with one instrument on one day.

If one of these items changes, the number changes. Fraud is not necessary for this change. This fact is the cause of many problems in cannabis testing. Some persons do not know it, and some persons use it to get a higher price. Examples are the same product with a different number and labels with numbers that are too high. One more example is the selection of a laboratory for its high numbers.

Evidence and limitations

Primary items of laboratory testing

To read a COA in sixty seconds, examine these eight items in this sequence:

1. The laboratory. Make sure that the COA gives the name of the laboratory and an accreditation that you can examine (ISO/IEC 17025 or, in systems for medicinal cannabis, a GMP certificate)^[16].
2. The sample. Find the sample ID, the batch, the matrix, and the sample mass. Most important, find the person who collected the sample. ‘Client-submitted’ shows that the laboratory did not see your batch.
3. The basis. Find the basis (dry-weight or as-received) and the moisture content. The basis changes the potency by approximately 10 to 15%.
4. The potency table. A row for the acids (THCA) and a row for the neutral cannabinoids (THC) show an HPLC method. Then calculate the total THC: total THC = Δ9-THC + 0.877 × THCA. The result must agree with the total on the COA.
5. The units. The units % w/w and mg/g measure the same quantity (1% = 10 mg/g). Do not let a change of unit cause an error.
6. Each type of contaminant (microbes, metals, pesticides, mycotoxins, and solvents) has a different test and a different method. The result of each test is satisfactory or unsatisfactory. The potency does not give information about safety.
7. The footnotes. Find the LOQs, the definition of ‘ND’, and the references to the methods. If the COA has no LOQ column, you do not know the limit for the result ‘ND’.
8. The signature. Find a signature with a name and a date from a person in QA. In GMP systems, this signature makes the COA a decision to release the batch and not a tool for marketing^[15].

This paper also gives the facts about numbers that are too high. Tests with peer review show that the potency values on labels in some legal markets have a systematic error. In one test, the measured value of 70% of the Colorado flower samples was more than 15% less than the label value^[1]. In an audit of three states, the measured value of 70% of the samples was not in the range of ±20% of the label value^[12]. The number of products with a value that is a small quantity more than the price threshold of 20% THC is too high^[2].

When one number sets the price, some persons will try to change it. Your protection is to know how the laboratory makes the number. The remaining sections of this paper give this information.

Laboratory testing procedure

The COA is the last item of a sequence of steps that starts when a person collects the sample. Each step changes the numbers on the COA. The sampling and the sample preparation change the numbers more than the instrument does.

Each type of test uses different physics. Thus a laboratory has approximately six instruments:

How to read a certificate of analysis

This section has an example COA. It is from a laboratory with the example name Example Analytical Ltd. The COAs of most laboratories have the same blocks in the same sequence. The COA has eight areas with numbers. Examine the areas in the sequence of the numbers.

1. 1
   Name and accreditation of the laboratory

   A COA from a laboratory with accreditation gives the name, the address, and the accreditation number of the laboratory. Compare these data with the public register of the authority that gave the accreditation. A PDF with a logo and no accreditation number is only a claim.
2. 2
   Report ID and version

   A report has one version. A laboratory can make a new version (‘v2’) of a report for a correct cause. But if each new version only increases the THC value, tell the laboratory to give the cause.
3. 3
   Sample information

   This block gives the sample ID, the batch or lot, and the matrix. It also gives the mass that the laboratory received and the dates. It shows the person who collected the sample. ‘Client-submitted’ shows that the number is for the material in the bag that you sent. This result is very different from the result for a batch with a sample that the laboratory collected.
4. 4
   Basis and moisture

   This block gives the basis (as-received or dry-weight) and the measured moisture. Without this row, you cannot compare two COAs. The section on the basis below gives more information.
5. 5
   The potency table

   This block has a row for THCA and a row for Δ9-THC (the two rows show an HPLC method). It also has the minor cannabinoids and a total with an asterisk (*). Calculate the total THC: total THC = Δ9-THC + 0.877 × THCA. In the example, 0.92 + 0.877 × 24.20 = 22.14%. This result agrees with the total on the COA. If the values are not the same, find the cause before you accept other data on the page.
6. 6
   Footnotes and LOQs

   ‘ND’ shows that the laboratory did not find the analyte in a quantity more than the limit in the report. It does not show zero. The LOQ column gives this limit, and thus you can use the result ‘ND’. If the report has no LOQ column, the report is defective.
7. 7
   Contaminant panels

   Each type (microbes, metals, pesticides, mycotoxins, and solvents) is a different test on a different part of the sample. A COA can show a very high THC number and an unsatisfactory result for Aspergillus.
8. 8
   Signature for release and the last clause

   A person from QA gives a name and a date. Then there is an ISO clause that limits the results: ‘Results relate only to the sample as received’. This clause shows the limit of all the data above it.

Total THC: how to calculate it, and the chemistry

The plant makes almost no THC. It makes THCA (tetrahydrocannabinolic acid). THCA is THC with an attached chemical group (–COOH). This group makes THCA a non-intoxicating cannabinoid. THCA is also approximately 13% heavier than THC.

Heat removes this group as CO₂ gas. This reaction is decarboxylation^[4]. After the reaction, the compound is chemically different. The heat of a flame, a vaporizer, or an oven causes this reaction. Thus this reaction makes most of the THC that a person uses.

The 0.877 factor is a ratio of molecular weights. It is not a correction. THC has a molecular weight of 314.5 g/mol, and THCA has a molecular weight of 358.5 g/mol. The ratio is 314.5 ÷ 358.5 = 0.877. Thus the standard formula for labels is:

A test with controlled conditions measured the speed of the conversion. In an open reaction vessel, all the THCA changed to THC in approximately 30 minutes at 110 °C (230 °F). The time was approximately 9 minutes at 130 °C (266 °F) and approximately 6 minutes at 145 °C (293 °F). When the test applied heat without light and in a vacuum, no important quantity of CBN (a product of oxidation) occurred^[4].

In air, with light, and at higher temperatures, the quantities of THCA and THC decrease more. The formula is correct only if all the THCA changes to THC. Thus the formula gives a maximum and not an estimate of the result.

Decarboxylation also occurs slowly at room temperature, for example during curing and in storage. Some THCA changes to THC, and some THC changes slowly to CBN with oxidation. Thus a previous COA and a new COA for the same batch can show different numbers, and the two numbers can be correct. The material changed.

HPLC and GC for potency testing

Two types of chromatography are the primary methods for potency testing. They do not measure the same molecules. HPLC (high-performance liquid chromatography) moves the extract through a column in a liquid at a temperature near room temperature. Different molecules move through the column at different speeds. Thus THCA and THC go to the detector at different times. They make two different peaks.

GC (gas chromatography) must change the sample to a gas in the hot inlet at approximately 250 to 300 °C (482 to 572 °F). At this temperature, THCA decarboxylates immediately. The acid does not go to the detector as THCA^[6].

Only a part of the THCA changes to THC in the inlet, and this part is not constant. A test with THCA without other compounds found that decarboxylation in GC conditions changed only approximately 70% of the acid. The test also showed that the only accurate method is to measure THCA and THC as two values and to add the two values. A measurement after decarboxylation gives a minimum and not the correct value^[5].

With GC, the laboratory cannot find the acids, unless it first does derivatization of the acids. Derivatization adds a chemical group to the acid. This group stays on the molecule at high temperature^[6].

Units and moisture basis

The units are easy. The number in mg/g is the number in % w/w with the decimal point moved by one position to the right. 1% w/w = 10 mg/g. For example, 22.14% = 221.4 mg/g.

Flower COAs usually give the result in %. COAs for oils and edible products frequently give the result in mg/g or in mg for each unit. The only problem is that you do not know the units.

The basis is the primary problem. Flower contains approximately 10 to 13% water after correct drying. The laboratory can calculate a potency percentage for the total mass with this water (as-received or ‘as-is’). The laboratory can also calculate it for the solids only (dry-weight). The flower and the chemistry are the same, but the two numbers are different:

Sampling and the accuracy of the result for the batch

The instrument can measure only the sample that the sampling procedure gives. A person does the sampling before the sample goes to the laboratory.

A batch can be 12 kg (26.5 lb). The composite sample is some tens of grams. The part that the laboratory extracts is approximately 0.5 to 1 g (0.02 to 0.04 oz). This one gram must show the batch. Thus the guidance of the pharmacopoeia identifies the sampling procedure as a quality attribute and not only as records^[7].

Sampling of cannabis is not easy because the analytes are in the trichomes. The quantity of trichomes is not the same in all parts of the plant. Top colas with more light have a higher potency than buds at the bottom in the shade. Resin falls from small buds when a person touches them. In ground material, the kief falls to the bottom.

A sample from only the best top buds is not a batch sample. It shows only the best part of the batch.

1. 1
   Set the limits of the batch first

   A batch has one cultivar, one room, one harvest, and one procedure. If the batch is not the same in all parts, no sampling procedure can correct this.
2. 2
   Collect many small samples and not a grab sample

   Collect many small samples from different containers, positions, and depths. Collect also from the middle and from the bottom. Many small samples are better than one large quantity from one position.
3. 3
   Make a composite sample and record

   Make a composite sample from the small samples. Record the person who collected each small sample, the position, and the time. This step starts the chain of custody.
4. 4
   Mix before you divide the sample

   Grind the sample. Then mix the sample. Do these steps before you remove a part of it. For potency, the laboratory grinds and mixes the part again.
5. 5
   Keep a second sample

   Divide the sample into two parts. Keep one part in storage. If you think that a number is incorrect, this part is the only sample that you can measure again.

Correct sampling also gives a variance that is not zero. When one laboratory measures two composite samples from one batch, the THC values frequently have a difference of approximately one percentage point. A difference of one to two percentage points is noise. It is not information.

Test methods for microbes

A test for microbes measures two types of item. The first type is the number of microbes on the sample. Counts measure this number. The counts are TAMC (total aerobic microbial count), TYM or TYMC (total yeast and mold count), and the count of bile-tolerant Gram-negative bacteria.

The second type is a dangerous microbe on the sample. A test with a yes or no result measures this type for specified organisms. The organisms are Salmonella, E. coli that are pathogens, and the four Aspergillus species that are pathogens in products for inhalation.

Plates and qPCR can give different results for the same sample. A test with metagenomic sequencing showed that the culture medium selects the organisms. The test sequenced the organisms that became colonies in standard culture tests for yeast and mold of cannabis. The colonies on the plates included bacteria. The plates did not show a sufficient quantity of the fungi that make toxins and that were on the flower. In some tests, the plates can find none of the organisms that are dangerous for patients^[9].

qPCR also counts DNA from killed cells. Thus a batch that had a remediation with heat or irradiation can have an unsatisfactory result with qPCR and a satisfactory result with plates.

Heavy-metal testing by ICP-MS

Cannabis has a high uptake of metals. Growers use hemp for the remediation of soil because of the same trait. The trait also causes the uptake of cadmium and lead from your substrate, fertilizer, and water. The plant keeps them in the tissue.

The panel for metals nearly always has four metals: arsenic, cadmium, lead, and mercury. Some systems use a test for a larger list of elements^[8].

The instrument is ICP-MS (inductively coupled plasma mass spectrometry). The laboratory mixes the sample with hot acid until only elements in solution stay. Then the laboratory puts a spray of the solution into an argon plasma.

The temperature of the plasma is thousands of degrees. The instrument finds the ions of each element from their mass and counts them. The instrument is very sensitive, in the range of ppb. Thus the LOQs for metals are, for example, 0.01 µg/g.

Pesticide panels and their limits

A test for pesticides is a panel. A panel is a specified list of compounds, and the laboratory compares each compound with an action limit. A satisfactory result shows that the laboratory found no compound on this list in a quantity more than these limits. It does not show that the product has no pesticide. It gives no information about compounds that are not on the panel.

The panels are very different in different legal markets. A survey of the regulations in the states of the US found 551 different pesticides with a limit in some state. The action limits for the same compound change by a maximum of four orders of magnitude from one state to a different state^[10].

• A full panel uses two instruments. LC-MS/MS finds most residues of new pesticides. GC-MS/MS finds the volatile compounds and the halogenated compounds. A laboratory with a large panel uses the two instruments.
• Inhalation changes the toxicology. Residues in a quantity that is satisfactory on lettuce can break and make more dangerous compounds when a person smokes them. Reports show that some fungicides release hydrogen cyanide when they burn. Thus the limits for cannabis are frequently much lower than the limits for food.
• Previous results show the risk. Before regulation, tests of concentrates in California found pesticides in approximately one-third of the samples^[11].
• Drift and carryover are important. You can have an unsatisfactory result for a panel when you do not apply pesticide. Agriculture near your facility, used equipment with contamination, or a trim room that is not clean can put residues on the flower.

Tests for residual solvents and mycotoxins

Residual solvents are in extracts. A solvent removes the resin from the plant. The solvent can be butane, propane, ethanol, or CO₂ and, in a last step, ethanol. A small quantity of the solvent can stay in the extract, and headspace GC measures it in the product.

The limits are different for each solvent. They agree approximately with the types of solvent for pharmaceutical products. The limit is almost zero for solvents with a very high toxicity (benzene and toluene). Producers do not add them, but they can occur as impurities in gas with a low price. The limits are higher for the usual solvents of the procedure.

A test for solvents is also necessary for a solventless hash or rosin. There are three causes. First, in most legal markets the type of the product causes the test, for all procedures. Second, the test is the only method to make sure that the solventless claim is correct, and not to accept it without a test. Third, contamination can occur without an extraction step. Cleaning agents, fuels, and gases from materials in storage can supply volatile compounds.

A satisfactory result for the solvent panel on rosin shows, at a low cost, that your marketing claim is correct. First tests of concentrates found residual solvents in approximately 30% of the samples. Thus buyers started to examine this result^[11].

Mycotoxins are chemical compounds with toxicity that mold makes. They are the aflatoxins B1, B2, G1, and G2 (from Aspergillus flavus and related species) and ochratoxin A. LC-MS/MS measures them at limits in the range of ppb^[10]. Two facts make mycotoxins a different row on the COA and not a footnote to the section on microbes:

• The toxins stay after a kill step. Kill steps (heat, irradiation, and ozone) can decrease the TYM count a lot, but the toxins do not change. A batch can have a satisfactory result for microbes and an unsatisfactory result for mycotoxins. This occurs frequently in a product after remediation.
• They are very dangerous at very low doses. Aflatoxin B1 is one of the strongest natural carcinogens that persons know. Thus the limits are in the range of µg/kg (ppb) in systems for medicinal cannabis^[15].

Water activity and moisture content

A flower COA gives two numbers for water. The two numbers give different information. Moisture content (%) is the quantity of water in the sample. It is the mass of the water divided by the total mass. Water activity (a_w, scale 0 to 1) is how available that water is to microbes.

Water activity is the equilibrium relative humidity that the sample makes in a closed space. Two samples can have the same mass of water, but a different water activity. When the sample holds the water tightly, the water is not available to microbes. Mold can increase only with available water, and not only with a large quantity of water. Thus a_w is the important number for microbes. The pharmacopoeia for cannabis in storage has a specification of a maximum of 0.65 for water activity^[7].

In operation, dry the flower. Then do the curing until the water activity agrees with the target. Accept the moisture content that the drying gives. The two numbers on the COA also let you make sure that they agree. For example, a_w 0.75 with 11% moisture is not usual for a sample. Find the cause of this result.

COA values that are too high: data and warning signs

When one number sets the price, there is pressure on the number. Tests with peer review record this problem in legal markets for cannabis. Each grower that selects a laboratory must know this fully.

The data of states show the cause. The potency values for chemotype-I flower from the six largest laboratories in Washington show a systematic difference. The median total THC was 17.7% at the laboratory with the lowest values and 23.2% at the laboratory with the highest values. This difference of 5.5 percentage points stayed when the cultivar and the producer were the same^[3].

Values a small quantity more than the price threshold of 20% are also too frequent in the reports. At 20%, the frequency of products increases suddenly, by 43% in Nevada and by 17% in Washington. At some laboratories, the frequency increases more. At the two laboratories that the authority suspended subsequently, the frequency increases by 47%. At the largest laboratory of the state, the frequency increases by 1%^[2]. The value of 20% is a price threshold and has no cause in biology.

Selection of a laboratory for its high numbers causes this problem. A grower divides one batch into three parts and sends the parts to three laboratories. The grower keeps the highest number. Then the grower gives the work to that laboratory. The laboratories know this. A laboratory that gives correct values gets less work than a laboratory that gives high values.

The methods to make the values too high range from small changes (a calibration bias for flower only, a rounding to a higher value, and sampling with a large tolerance) to fraud. In 2024, seven of the eleven laboratories with accreditation in Oregon had an enforcement action from the authority of the state. The cause was THC results that were too high. The enforcement action included the claim that personnel at three laboratories added kief to samples from clients before the test^[14]. Enforcement actions about licenses and lawsuits by other laboratories followed in California and Massachusetts. The causes were potency values that were too high and products with contamination that had a satisfactory result.

We recommend that a grower does these steps. Select a laboratory because its accreditation includes your type of test and its COA shows the method.

Do not select a laboratory because of its averages. At intervals, divide a sample into two parts. Send the parts to two laboratories. The usual variation is approximately 1 to 2 percentage points. Keep samples in storage. A person at a laboratory who tells you the numbers before the test is a risk for your license.

In systems for medicinal cannabis with GMP, the incentive is the opposite. The laboratory gives data for the decision to release the batch and not for marketing. Thus the numbers are more stable in these systems^[15][16].

Limits of one result

A COA gives information that you can use in its limits. One COA can show three items. The first item is the potency group of the sample (a batch with 15% and a batch with 25% are very different).

The second item is the result of the sample for the panel (satisfactory or unsatisfactory). The third item is a trend that you can use to control your crop. For this trend, many batches from your room are necessary, with the same sampling procedure each time. One COA cannot show these items:

• The number for your room. The COA gives information about the sample. The batch has the same number only if your sampling is accurate.
• Differences of one or two percentage points. The variation of the sampling and the variation of the results from different laboratories are larger than these differences. In a test, the systematic difference from one laboratory to a different laboratory was 5.5 percentage points, with no other cause^[3].
• Quality or effect. The THC percentage has a weak correlation with the effect of a product when a person uses it. Terpenes, minor cannabinoids, curing, and freshness cause most of the effect. A high number on the COA is not the correct target.
• The next batch. A COA is a record of one batch. It does not show the result for the next batch. The genetics, the environment, and the procedure will change the next batch.

When the number is not usual

Signs of a problem with a COA

Testing for release in NZ and Australia

In the systems of Australia and New Zealand for medicinal cannabis, the COA has a different function from a COA in the retail market. In Australia, a medicinal cannabis product without approval must agree with TGO 93 (Therapeutic Goods (Standard for Medicinal Cannabis) Order 2017). The measured quantity of cannabinoids must be in the range of 90.0 to 110.0% of the label claim. The limits for contaminants (for example aflatoxins and pesticide residues) apply. The authority can collect product and do a test on it at all times^[15].

In New Zealand, products must agree with the minimum quality standard in the Misuse of Drugs (Medicinal Cannabis) Regulations 2019. Facilities with a GMP certificate must do the critical tests. For the other tests, the authority accepts accreditation to ISO/IEC 17025^[16].

The primary procedure is release testing. The laboratory does a test of the batch and compares the results with a registered specification. A qualified person examines all the data. Then this person releases the batch or does not release it. The COA is one source of data for a decision that is in a document. This person makes the decision, gives a signature, and has liability for it.

In the retail market, the COA has a primary task: marketing. Thus the record of values that are too high, in a previous section, is a result of the different incentive. The document is the same, but the incentive is the opposite.

• A range of 90 to 110% of the label claim shows that a batch can be unsatisfactory because the potency is too high. The target is accuracy and not a high value^[15].
• The data on stability and the claims on shelf life use the same measurements. Subsequent measurements examine again the COA that releases the batch. Thus the first numbers must be accurate.
• Testing in GMP uses methods with validation, instruments with qualification, and audit trails. The laboratory shows the basis of each result in a system of documents. It does not give a result without data.

The primary fact is the same for growers in other legal markets. Use a testing procedure with the properties of release testing. The procedure has an SOP for sampling that does not change, and one laboratory with accreditation. It has samples that you keep, trend charts, and numbers with no incentive for a high value. When your procedure agrees more with release testing, your COAs have a higher value for you and for each auditor.